Nasal Symptoms

Of three identified trials, one⁸³ (N=209) reported nasal symptom outcomes (TNSS). This trial was rated fair quality, and reported a non-statistically significant treatment effect of 0.3 on a 0-12 point scale (3 percent of maximum score) favoring oral selective antihistamine. Risk of bias was considered moderate based on trial quality. Because consistency of the observed effect cannot be assessed with a single trial and because the effect was imprecise, the evidence was insufficient to support the use of one treatment over the other.

Quality of Life

Of three identified trials, one⁸² (N=86) reported quality of life outcomes. This trial was rated poor quality due to noncomparable groups at baseline and inappropriate analysis of results (unadjusted for baseline group differences). The treatment effect was 12.9 on a 0-168 point scale favoring oral selective antihistamine and was statistically significant. Extrapolating the anchor-based MCID (0.5) for the RQLQ 0-6 point scale yields an MCID of 14 points. Risk of bias for this outcome was considered high based on both trial quality and the use of quality of life measures in an unblinded trial population. Consistency is unknown with a single trial, and the treatment effect was imprecise. The evidence was therefore insufficient to support the use of one treatment over the other for this outcome.

Nasal Symptoms

Four trials^84-87 (N=1022) assessed congestion after 2 weeks of treatment and reported greater improvement with nasal antihistamine than with oral selective antihistamine. Of three trials that reported p-values, this result was statistically significant in two.^{85, 86} One⁸⁵ was a good quality trial of 360 patients (35 percent of patients reporting this outcome) that did not report the magnitude of the treatment effect. The other trial⁸⁶ (n=136) was rated poor quality due to noncomparable groups at baseline and inappropriate analysis of results (unadjusted for baseline group differences). The magnitude of the treatment effect was not reported. Treatment effects of 0.08 and 0.17 on a 0-6 point scale (both less than 3 percent of maximum score) were reported by two trials^{84, 87} that were rated good quality (51 percent of patients reporting). Statistical significance of the former result was not assessed. The latter result was not statistically significant.

For the outcome of congestion at 2 weeks, the risk of bias was rated as low. Eighty-seven percent of patients assessed for this outcome were in good quality trials. All four trials were consistent in favoring nasal antihistamine, but treatment effects were imprecise. The evidence was therefore insufficient to support the use of one treatment over the other for this outcome.

Three trials^{84, 86, 87} (N=662) assessed rhinorrhea after 2 weeks of treatment and reported greater improvement with nasal antihistamine than with oral selective antihistamine. Of two trials that reported p-values, this result was statistically significant in both.^{86, 87} One⁸⁷ was a good quality trial of 307 patients (46 percent of patients reporting this outcome) that reported a treatment effect of 0.46 on a 0-6 point rating scale (8 percent of maximum score). The other⁸⁶ was the poor quality trial identified above. The magnitude of the treatment effect was not reported. A treatment effect of 0.22 on a 0-6 point scale (4 percent of maximum score) was reported by one trial⁸⁴ that was rated good quality (33 percent of patients reporting). Statistical significance was not assessed.

For the outcome of rhinorrhea at 2 weeks, the risk of bias was rated as low. Seventy-nine percent of patients assessed for this outcome were in good quality trials. All three trials were consistent in favoring nasal antihistamine, but treatment effects were imprecise. The evidence was therefore insufficient to support the use of one treatment over the other for this outcome.

Four trials^84-87 (N=1022) assessed sneezing after 2 weeks of treatment and reported greater improvement with nasal antihistamine than with oral selective antihistamine. Of three trials that reported p-values, this result was statistically significant in one.⁸⁵ This was a good quality trial of 360 patients (35 percent of patients reporting this outcome) that did not report the magnitude of the treatment effect. A statistically nonsignificant treatment effect of 0.29 on a 0-6 point scale (5 percent of maximum score) was reported by another good quality trial⁸⁷ of 307 patients (30 percent of patients reporting). The magnitude of the statistically nonsignificant treatment effect in the poor quality trial⁸⁶ identified above was not reported. A treatment effect of 0.23 on a 0-6 point scale (4 percent of maximum score) was reported by another good quality trial⁸⁴ of 219 patients (21 percent of patients reporting). Statistical significance was not assessed.

For the outcome of sneezing at 2 weeks, the risk of bias was rated as low. Eighty-seven percent of patients assessed for this outcome were in good quality trials. All four trials were consistent in favoring nasal antihistamine, but treatment effects were imprecise. The evidence was therefore insufficient to support the use of one treatment over the other for this outcome.

Three trials^{84, 86, 87} (N=662) assessed nasal itch after 2 weeks of treatment and reported greater improvement with nasal antihistamine than with oral selective antihistamine. Of two trials^{84, 87} that reported p-values, results were not statistically significant in either. One⁸⁷ was a good quality trial of 307 patients (46 percent of patients reporting this outcome) that reported a treatment effect of 0.30 on a 0-6 point scale (5 percent of maximum score). The other⁸⁶ was the poor quality trial previously identified. No treatment effect was reported. A treatment effect of 0.25 on a 0-6 point scale (4 percent of maximum score) was reported by another good quality trial⁸⁴ of 219 patients (33 percent of patients reporting). Statistical significance was not assessed.

For the outcome of nasal itch at 2 weeks, the risk of bias was rated as low. Seventy-nine percent of patients assessed for this outcome were in good quality trials. All three trials were consistent in favoring nasal antihistamine, but treatment effects were imprecise. The evidence was therefore insufficient to support the use of one treatment over the other for this outcome.

Three trials^{84, 85, 87} (N=886) assessed TNSS at 2 weeks and reported greater improvement with nasal antihistamine than with oral selective antihistamine. Of three trials^{84, 85, 87} that reported p-values, this result was statistically significant in one trial.⁸⁷ This was a good quality trial of 307 patients (35 percent of patients reporting this outcome) that reported a treatment effect of 1.24 on a 0-24 point scale (5 percent of maximum score). Treatment effects of 0.78 and 0.70 on a 0-6 point scale (13 percent and 12 percent of maximum score, respectively) were reported by two trials^{84, 85} that were rated good quality (65 percent of patients reporting). Statistical significance of the former result was not assessed. The latter result was not statistically significant.

For the outcome of TNSS at 2 weeks, the risk of bias was rated as low. All three trials reporting this outcome were rated as good quality. All three trials also were consistent in favoring nasal antihistamine, but treatment effects were imprecise. The evidence was therefore insufficient to support the use of one treatment over the other for this outcome.

Quality of Life

Two trials^{85, 87} (N=667) assessed quality of life at 2 weeks using the RQLQ. Both were good quality trials that reported statistically significant reductions in RQLQ with nasal antihistamine compared to oral selective antihistamine. Treatment effects on a 0-6 scale were 0.4 points in one trial⁸⁵ and 0.3 points in the other.⁸⁷ One poor quality trial⁸⁶ of 136 patients reported a statistically nonsignificant difference in the proportion of patients who reported an excellent or good response to treatment rather than a fair or poor response, with a treatment difference of 0.6 percent favoring nasal antihistamine.

For the outcome of quality of life as measured by the RQLQ at 2 weeks, the risk of bias was rated as low. Both trials reporting this outcome were rated good quality. Trials were consistent in favoring nasal antihistamine over oral selective antihistamine, but neither treatment effect exceeded the MCID of 0.5 and was therefore imprecise. The evidence was therefore insufficient to support the use of one treatment over the other for this outcome.

Nasal Symptoms

The one trial⁹⁰ that reported on nasal congestion at 2 weeks (N=341) reported a treatment effect of 0.3 on a 0-6 point scale (5 percent of maximum score) favoring intranasal corticosteroid. This trial was rated poor quality, and the result was not statistically significant. Evidence for the outcome of congestion at 2 weeks is therefore insufficient to support the use of one treatment over the other. One poor quality trial⁹⁰ with high risk of bias reported an imprecise treatment effect.

Six trials^{91-93, 95, 96, 99} assessed congestion at 4 weeks (N=1600). All six showed statistically significant improvements in congestion with intranasal corticosteroid. Two^{91, 95} were good quality trials of 558 patients total (35 percent of patients reporting this outcome). One⁹⁵ reported results using a 0-3 point scale but did not report the magnitude of the treatment effect. The other⁹¹ reported a treatment effect of 10.3 on a 0-100 VAS (10 percent of maximum score). Four trials^{92, 93, 96, 99} were rated poor quality due to noncomparable groups at baseline^{92, 93, 96} and inappropriate analysis of results (unadjusted for baseline group differences^{93, 96} and not intention to treat⁹⁹). Two^{92, 99} of these trials reported treatment effects of 0.3 and 0.46 on a 0-3 point scale (10 percent and 15 percent of maximum score, respectively).

Three trials^{91, 92, 99} (N=938; 59 percent of patients reporting this outcome) were pooled in a meta-analysis (Figure 5). Because trials used different symptom rating scales (0-3 and 0-100), the standardized mean difference was calculated. The pooled effect estimate was 0.45 (95 percent confidence interval (CI): 0.28 to 0.61), a statistically significant result that favored intranasal corticosteroid and was consistent with the direction of effect reported by individual trials. Effect estimates in the pooled trials were in the same direction, and their 95 percent CIs did not touch the “no effect” line. The magnitude of the pooled effect estimate could not be compared with estimates from individual trials not included in the meta-analysis because the latter were not reported.

Figure 5

Congestion at 4 weeks: meta-analysis of 3 trials–oral selective antihistamine versus intranasal corticosteroid.

For the outcome of congestion at 4 weeks, the risk of bias was rated as high. Sixty-five percent of patients were in poor quality trials, and 35 percent were in good quality trials. All six trials^{91-93, 95, 96, 99} were consistent in finding statistically significant treatment effects favoring intranasal corticosteroid, and this finding was confirmed in a meta-analysis of three of these trials.^{91, 92, 99} Because reported treatment effects were less than an MCID of 30 percent maximum score, and because the magnitude of effects in three trials^{93, 95, 96} representing 40 percent of patients reporting this outcome were not reported, the body of evidence was considered imprecise. The evidence was therefore insufficient to form a conclusion about the comparative effectiveness of oral selective antihistamine and intranasal corticosteroid for this outcome.

The one trial⁹⁰ that assessed rhinorrhea at 2 weeks (N=341) reported a treatment effect of 0.3 on a 0-3 point scale (10 percent of maximum score) favoring intranasal corticosteroid. This trial was rated poor quality, and the result was not statistically significant. Five trials^{92, 93, 95, 96, 99} assessed rhinorrhea at 4 weeks. One poor quality trial⁹³ reported neither the magnitude nor the direction of the treatment effect. This trial was excluded from analysis of this outcome, reducing the total number of patients assessed from 1284 to 979. The remaining four trials all favored intranasal corticosteroid over oral selective antihistamine. One trial⁹⁵ was a good quality trial of 242 patients (25 percent of patients reporting this outcome) that demonstrated a statistically significant improvement in rhinorrhea with intranasal corticosteroid. The magnitude of the treatment effect was not reported. The remaining three trials^{92, 96, 99} were rated poor quality due to noncomparable groups at baseline and inappropriate analysis of results, as described above. Two^{96, 99} of these reported statistically significant treatment effects favoring intranasal corticosteroid. One⁹⁹ reported a treatment effect of 0.55 on a 0-3 point scale (18 percent of maximum score). The other⁹⁶ did not report the magnitude of the treatment effect. The third poor quality trial⁹² reported a statistically nonsignificant treatment effect of 0.2 on a 0-3 scale (7 percent of maximum score).

Evidence for the outcome of rhinorrhea at 2 weeks was insufficient to support the use of one treatment over the other. One poor quality trial⁹⁰ with high risk of bias reported an imprecise treatment effect. For the outcome of rhinorrhea at 4 weeks, the risk of bias was rated as high. Seventy-five percent of patients were in poor quality trials. Effect estimates consistently favored intranasal corticosteroid. However, none exceeded an MCID of 30 percent maximum score, and the body of evidence was considered imprecise. Evidence was therefore insufficient to support the use of one treatment over the other for this outcome.

The one trial⁹⁰ that assessed sneezing at 2 weeks (N=341) reported a treatment effect of 0.1 on a 0-3 point scale (3 percent of maximum score) favoring intranasal corticosteroid. This trial was rated poor quality, and the result was not statistically significant. Five trials^{92, 93, 95, 96, 99} assessed sneezing at 4 weeks (N=1284). All five showed statistically significant improvements in sneezing with intranasal corticosteroid. One of these was a good quality trial⁹⁵ of 242 patients (19 percent of patients reporting this outcome) that did not report the magnitude of the treatment effect. The remaining four trials^{92, 93, 96, 99} were rated poor quality for noncomparable groups at baseline and inappropriate analysis of results, as described above. Two of these trials^{92, 99} reported treatment effects of 0.3 and 0.45 on a 0-3 point scale (10 percent and 15 percent of maximum score, respectively). The other two^{93, 96} did not report the magnitude of the treatment effects.

Evidence for the outcome of sneezing at 2 weeks was insufficient to support the use of one treatment over the other. One poor quality trial⁹⁰ with high risk of bias reported an imprecise treatment effect. For the outcome of sneezing at 4 weeks, the risk of bias was rated as high. Eighty-one percent of patients were in poor quality trials. All five trials^{92, 93, 95, 96, 99} were consistent in finding statistically significant treatment differences favoring intranasal corticosteroid. However, reported treatment effects were imprecise. The evidence was therefore insufficient to support the use of one treatment over the other for this outcome.

The one trial⁹⁰ that assessed nasal itch at 2 weeks (N=341) reported no difference in effect between oral selective antihistamine and intranasal corticosteroid. This trial was rated poor quality. Four trials^{92, 93, 95, 99} assessed nasal itch at 4 weeks (N=1196). All four reported statistically significant improvement with intranasal corticosteroid compared with oral selective antihistamine. One of these was a good quality trial⁹⁵ of 242 patients (20 percent of patients reporting this outcome). Treatment effect was not reported. The remaining three trials^{92, 93, 99} were rated poor quality for noncomparable groups at baseline^{92, 93} and inappropriate analysis of results (unadjusted for baseline group differences⁹³ and not intention to treat⁹⁹). Two of these trials^{92, 99} reported treatment effects of 0.2 and 0.29 points on a 0-3 scale (7 percent and 10 percent of maximum score, respectively). The third⁹³ did not report the magnitude of the treatment effect.

Evidence for the outcome of nasal itch at 2 weeks is insufficient to support the use of one treatment over the other. One poor quality trial⁹⁰ with high risk of bias reported no difference between treatments. For the outcome of nasal itch at 4 weeks, the risk of bias was rated as high. Eighty percent of patients were in poor quality trials. All four trials^{92, 93, 95, 99} were consistent in finding statistically significant treatment effects favoring intranasal corticosteroid. However, the effects were imprecise. The evidence was insufficient to support the use of one treatment over the other for this outcome.

Six trials^{89, 90, 93, 97, 98} assessed TNSS at 2 weeks (N=2756). Three^{89, 98} of these showed statistically significant improvements with intranasal corticosteroid compared with oral selective antihistamine. All three were rated fair quality. Two of these trials⁸⁹ reported treatment effects of 1.0 and 1.3 using a 0-12 point scale (8 percent and 11 percent of maximum score, respectively). The other three trials^{90, 93, 97} were rated poor quality due to noncomparable groups at baseline⁹³ and inappropriate analysis of results (unadjusted for baseline group differences⁹³ and not intention to treat^{90, 97}). Two of these trials^{90, 93} reported statistically nonsignificant treatment effects of 0.8 and 1.0 using a 0-12 point scale (7 percent and 8 percent of maximum score, respectively). The third⁹⁷ reported a treatment effect of 0.17 (1 percent of maximum score) but did not assess statistical significance.

Five trials^{92-94, 99, 100} assessed TNSS after 2 weeks, that is, at 3 or 4 weeks. One poor quality trial⁹⁹ reported neither the magnitude nor the direction of the treatment effect at 4 weeks. This trial was excluded from analysis of this outcome, reducing the total number of patients assessed from 1306 to 1008. The four remaining trials^{92-94, 100} reported improvement in TNSS with intranasal corticosteroid at 3 weeks^{93, 100} and at 4 weeks.^92-94 All four trials were rated poor quality due to noncomparable groups at baseline^92-94 and inappropriate analysis of results (unadjusted for baseline group differences⁹³ and not intention to treat¹⁰⁰). Treatment effects at 3 weeks were 1.2 on a 0-12 point scale⁹³ (10 percent of maximum score) and 2.17 on a 0-15 point scale¹⁰⁰ (14 percent of maximum score). At 4 weeks, treatment effects of 0.8 on a 0-12 point scale (7 percent of maximum score) were reported by two trials.^{92, 93}

For the outcome of TNSS at 2 weeks, the risk of bias was rated as medium. Forty-five percent of patients reporting this outcome were in poor quality trials, and 55 percent were in fair quality trials. Treatment effects consistently favored intranasal corticosteroid, although effects were imprecise. The evidence was therefore insufficient to support the use of one treatment over the other for this outcome.

For TNSS at 3 to 4 weeks, the risk of bias was rated as high. All four trials^{92-94, 100} reporting this outcome were rated poor quality. Treatment effects consistently favored intranasal corticosteroid but were imprecise. The evidence was therefore insufficient to support the use of one treatment over the other for this outcome.

Eye Symptoms

Eye symptoms were reported using a variety of measurement scales and varied definitions. Most treatment effects favored intranasal corticosteroid over oral selective antihistamine.

Four trials^{89, 90, 93} that assessed eye symptoms at 2 weeks reported greater improvement with intranasal corticosteroid than with oral selective antihistamine (N=1905). For TOSS, statistically significant treatment effects of 0.3 and 0.6 on a 0-9 point scale (3 percent and 7 percent of maximum score, respectively) were reported in two fair quality trials⁸⁹ (N=1074). The other two trials^{90, 93} were rated poor quality. One⁹⁰ reported a statistically significant treatment effect of unknown magnitude for the single symptom of tearing. The other⁹³ reported a statistically nonsignificant treatment effect of unknown magnitude for undefined symptoms.

For eye symptoms at 2 weeks, the risk of bias was rated as medium. Forty-four percent of patients were in poor quality trials, and 56 percent were in fair quality trials. All four trials^{89, 90, 93} were consistent in favoring intranasal corticosteroid. Reported treatment effects did not exceed an MCID of 30 percent maximum score and were considered imprecise. The evidence was therefore insufficient to support the use of one treatment over the other for eye symptoms at 2 weeks.

Four trials^{91-93, 99} assessed TOSS^{91, 92} and unspecified eye symptoms^{93, 99} at 4 weeks (N=1270). Three^{91, 93, 99} of these reported treatment effects that favored intranasal corticosteroid. One⁹¹ was a good quality trial of 316 patients (25 percent of patients reporting this outcome) that showed a statistically significant treatment effect of 16.2 on a 0-300 point scale (5 percent of maximum score). The other two trials^{93, 99} were rated poor quality. One⁹³ reported a statistically significant treatment effect of unknown magnitude for undefined eye symptoms, and the other⁹⁹ reported a statistically nonsignificant treatment effect of 0.11 on a 0-9 point scale (1 percent of maximum score). The fourth trial⁹² was rated poor quality and reported no difference in effect for TOSS. A meta-analysis of three of these trials^{91, 92, 99} was conducted (N=938 [74% of patients reporting this outcome]; Figure 6). Because trials used different symptom rating scales (0-9 and 0-300), the standardized mean difference was calculated. The pooled effect estimate was 0.13 (95 percent CI: -0.02 to 0.27), a statistically nonsignificant result that favored intranasal corticosteroid. Treatment effects in two^{91, 99} of the pooled trials favored intranasal corticosteroid, and in the third,⁹² showed no treatment difference. The meta-analysis excluded one trial⁹³ that showed a statistically significant treatment effect of unknown magnitude favoring intranasal corticosteroid.

Figure 6

Eye symptoms at 4 weeks: meta-analysis of 3 trials–oral selective antihistamine versus intranasal corticosteroid.

For eye symptoms at 4 weeks, the risk of bias was rated as high. Seventy-five percent of patients were in poor quality trials.^{92, 93, 99} Treatment effects at 4 weeks were not consistent across individual trials, with three^{91, 93, 99} of four trials reporting effects in favor of intranasal corticosteroid and the fourth⁹² (28 percent of patients reporting this outcome) showing no treatment difference. Because all reported effects were less than an MCID of 30 percent maximum score, and because one trial⁹³ (24 percent of patients reporting this outcome) did not report the magnitude of effect, the body of evidence was considered imprecise. The evidence was therefore insufficient to form a conclusion about the comparative effectiveness of oral selective antihistamine and intranasal corticosteroid for this outcome.

Quality of Life

All three trials^{92, 96, 98} that used the RQLQ to assess quality of life at 2 weeks (N=889) reported statistically significant treatment effects with intranasal corticosteroid compared to oral selective antihistamine. In two^{96, 98} of these trials, treatment effects of 1.0 and 0.9 on a 0-6 point scale exceeded the MCID of 0.5. The larger of these⁹⁸ was a fair quality trial of 450 patients (51 percent of patients reporting this outcome), and the other⁹⁶ was rated poor due to lack of blinding (n=88). The treatment effect in the third trial⁹² (n=351) was 0.25 on a 0-6 scale. This trial was rated poor due to noncomparable groups at baseline.

For quality of life outcomes measured using the RQLQ at 2 weeks, the risk of bias was rated as medium. Forty-nine percent of patients were in poor quality trials, and 51 percent were in the fair quality trial. All three trials^{92, 96, 98} were consistent in finding statistically significant treatment differences favoring intranasal corticosteroid. Treatment effects were larger than the MCID in two trials^{96, 98} but smaller in one trial⁹² that accounted for 39 percent of patients reporting. The body of evidence was therefore considered imprecise. Evidence was insufficient to support the use of one treatment over the other for this outcome.

All four trials^{91, 92, 94, 96} that assessed quality of life using the RQLQ at 4 weeks (N=869) reported statistically significant treatment effects favoring intranasal corticosteroid. One⁹¹ of these was a good quality trial of 316 patients (36 percent of patients reporting this outcome). The magnitude of effect was not reported. The remaining three trials were rated poor quality. Of two trials^{92, 96} that reported the magnitude of treatment effects, the effect in one⁹⁶ (0.9) exceeded the MCID. This was a trial of 88 patients⁹⁶ that was rated poor quality due to lack of blinding. The other⁹² reported a treatment effect of 0.25. The fourth trial⁹⁴ did not report the magnitude of the treatment effect.

For quality of life outcomes measured using the RQLQ at 4 weeks, the risk of bias was rated as high. Sixty-four percent of patients were in the poor quality trials. All four trials^{91, 92, 94, 96} were consistent in finding statistically significant treatment differences favoring intranasal corticosteroid. However, reported treatment effects exceeded the MCID in only one trial⁹⁶ representing 10 percent of patients reporting this outcome. The body of evidence was therefore considered imprecise. Evidence was insufficient to support the use of one treatment over the other for this outcome.

Two fair quality trials⁸⁹ (N=1074) that used the Nocturnal RQLQ at 2 weeks reported statistically significant treatment effects of 0.5 and 0.7 on a 0-6 point scale (8 percent and 12 percent of maximum score, respectively). For quality of life outcomes measured using the Nocturnal RQLQ at 2 weeks, the risk of bias was rated as medium. Both trials⁸⁹ from the same published article were rated fair quality. Both also were consistent in finding statistically significant treatment differences favoring intranasal corticosteroid. However, effect estimates were imprecise. Evidence was therefore insufficient to support one treatment over the other for this outcome.

Four trials^{91, 94, 98, 100} reported PGA scores at 2 weeks,⁹⁸ 3 weeks,¹⁰⁰ and 4 weeks.^{91, 94} Results supported the quality of life findings described above (intranasal corticosteroid favored), but statistical significance of effect estimates was variable.

Nasal Symptoms

All seven trials^101-107 assessed nasal congestion at 2 weeks, and all reported greater improvement with oral decongestant than with oral selective antihistamine (N=3592). Treatment effects of 0.1 to 0.17 on a 0-3 point scale (3 percent and 6 percent of maximum score, respectively) were reported in three^{101, 103, 107} good quality trials of 1327 total patients (37 percent of patients reporting this outcome). Statistical significance was not reported. Three trials^{102, 104, 105} were rated poor quality due to inappropriate analysis of results (not intention to treat; N=1583). Treatment effects reported by these trials and by one trial¹⁰⁶ (n=682) rated fair quality ranged from 0.05 to 0.1 on a 0-3 point scale (all less than 3 percent of maximum score). In the two trials^{105, 106} that reported p-values, results were not statistically significant.

For congestion at 2 weeks, the risk of bias was rated as medium. Forty-four percent of patients were in poor quality trials, and 37 percent were in good quality trials. All seven trials were consistent in finding treatment effects that favored oral decongestant. However, none of the effects exceeded an MCID of 30 percent maximum score, and all were considered imprecise. The evidence was therefore insufficient to support the use of one treatment over the other for the treatment of congestion.

Three^{101, 103, 107} good quality trials assessed other nasal symptoms at 2 weeks (rhinorrhea, sneezing, nasal itch, TNSS). Grosclaude (1997)¹⁰³ and Sussman (1999)¹⁰⁷ reported treatment effects favoring oral selective antihistamine for rhinorrhea and sneezing (N=890). Grosclaude (1997)¹⁰³ reported treatment effects of 0.21 for rhinorrhea (7 percent of maximum score) and 0.32 for sneezing (11 percent of maximum score) using a 0-3 point scale. Sussman (1999)¹⁰⁷ reported treatment effects of 0.1 for rhinorrhea (3 percent of maximum score) and 0.2 for sneezing (5 percent of maximum score) using a 0-4 point scale. Grosclaude (1997; n=454) also assessed nasal itch at 2 weeks and reported a treatment effect of 0.13 on a 0-3 point scale (4 percent of maximum score). Bronsky (1995)¹⁰¹ assessed TNSS at 2 weeks and reported a treatment effect of 0.1 on a 0-3 scale (3 percent of maximum score) favoring oral decongestant. Because the direct comparison of interest in these trials involved the combination treatment arm, p-values for comparative effects of the two components were not reported.

For other nasal symptoms at 2 weeks, the risk of bias is rated as low. Results come from good quality trials.^{101, 103, 107} For rhinorrhea and sneezing, trials^{103, 107} were consistent in showing treatment effects that favored oral selective antihistamine. For nasal itch and TNSS, results are from single trials,^{101, 103} and consistency is unknown. All effects were imprecise, and evidence was insufficient to support the use of one treatment over the other for rhinorrhea, sneezing, nasal itch, or TNSS.

Eye Symptoms

Two^{103, 107} of seven trials assessed ocular outcomes (N=890). Both were good quality trials. Oral selective antihistamine was favored for both ocular itching¹⁰³ and total ocular symptoms (itching, tearing, and redness).¹⁰⁷ Treatment effects were 0.08 on a 0-3 point scale¹⁰³ and 0.1 on a 0-4 point scale¹⁰⁷ (both 3 percent of maximum score). P-values were not reported.

For ocular outcomes at 2 weeks, the risk of bias was rated as low. Both trials^{103, 107} reporting ocular outcomes were rated good quality. Trials were consistent in showing treatment effects that favored oral selective antihistamine. However, effects did not exceed an MCID of 30 percent maximum score and were considered imprecise. Evidence was therefore insufficient to support the use of one treatment over the other for ocular outcomes at 2 weeks.

Nasal Symptoms

Individual nasal symptom scores were assessed in three trials^110-112 at 2 weeks (N=1593). Two of these^{110, 111} were good quality trials of 643 patients total (40 percent of all patients reporting this outcome). The third trial¹¹² (n=950) was rated poor quality due to the inappropriate analysis of results (not intention to treat). P-values were not reported for any outcome. For congestion and rhinorrhea, the good quality trials showed treatment effects ranging from 0.02 to 0.08 on a 0-3 point scale (both less than 3 percent of maximum score) favoring leukotriene receptor antagonist over oral selective antihistamine. For sneezing and itching, treatment effects in these trials favored oral selective antihistamine over leukotriene receptor antagonist and ranged from 0.02 to 0.12 (1 percent and 4 percent of maximum score, respectively). Treatment effects in the poor quality trial favored oral selective antihistamine for all four nasal symptoms and ranged from 0.10 to 0.18 (3 percent to 6 percent of maximum score).

For these outcomes at 2 weeks, the risk of bias was rated as medium. Sixty percent of patients reporting were in the poor quality trial,¹¹² and 40 percent were in good quality trials.^{110, 111} Findings were not consistent across trials for congestion and rhinorrhea, but were consistent for sneezing and nasal itch. Treatment effects were imprecise. Evidence was therefore insufficient to support the use of one treatment over the other for these outcomes.

TNSS was assessed at 2 weeks in seven trials,^{97, 110-114} one of which also reported 4-week outcomes.¹¹⁴ An additional trial¹⁰⁹ reported 4-week results only (total N=3609). Two trials^{110, 111} were good quality trials of 643 patients total (18 percent of patients reporting this outcome) and two^{113, 114} were rated fair (1321 patients total; 37 percent of patients reporting). No trial reported p-values. All but two trials^{110, 111} favored oral selective antihistamine over leukotriene receptor antagonist at 2 or 4 weeks, with treatment effects ranging from 0.01 to 0.17 on a 0-3 point scale (all less than 6 percent of maximum score). The two good quality trials^{110, 111} favored leukotriene receptor antagonist at 2 weeks with treatment effects of 0.02 and 0.04 on a 0-3 point scale (both less than 2 percent of maximum score). A meta-analysis of seven^{97, 109-111, 113, 114} of these eight trials was performed (total N=2648; 73 percent of patients reporting this outcome). The pooled treatment effect was 0.06 on a 0-3 point scale (95 percent CI: 0.00 to 0.12) favoring oral selective antihistamine (Figure 7). This was a statistically significant result. The larger bound of the 95 percent CI represented 4 percent of maximum score. Statistical heterogeneity was low to moderate but not statistically significant (I²=39 percent, p=0.13) and likely due to variation in treatment effect direction and precision. An eighth trial¹¹² that was rated poor quality was not included in the meta-analysis due to lack of variance reporting (n=950). This trial reported a treatment effect of 0.1 on a 0-3 point scale (3 percent of maximum score) favoring oral selective antihistamine. A p-value was not provided. Finally, 4-week results from van Adelsberg (2003)¹¹⁴ were not included in the meta-analysis because 2-week results were the identified primary outcome. The treatment effect at 4 weeks was 0.07 on a 0-3 point scale (2 percent of maximum score) favoring oral selective antihistamine. A p-value was not reported.

Figure 7

Total nasal symptom score at 2 to 4 weeks: meta-analysis of 7 trials–oral selective antihistamine versus leukotriene receptor antagonist.

For TNSS at 2 to 4 weeks, the risk of bias was rated as medium. Eighteen percent of patients reporting this outcome were in good quality trials, and 45 percent were in poor quality trials. Although findings at 2 weeks were not consistent across individual trials, statistical heterogeneity of a meta-analysis that included trials with conflicting results was low to moderate. Further, the pooled treatment effect (0.06) favoring oral selective antihistamine was consistent with treatment effects reported by the one trial (Philip [2002]) not included in the meta-analysis¹¹² and by another trial¹¹⁴ included in the meta-analysis that reported results at an additional time point (4 weeks). The 95 percent CI for the pooled estimate (0.00 to 0.12) fell within an interval bounded by −MCID and +MCID (-0.9 and +0.9 on the 0-3 point scale used). The Philip (2002) trial¹¹² (26 percent of patients reporting this outcome) showed a treatment effect (0.1 on a 0-3 point scale) that was larger than the pooled effect (0.06) but smaller than the MCID (0.9). To determine the impact of this trial on the pooled estimate, we added it to the meta-analysis with an assumed standard deviation equal to half the mean change in score in each treatment group. Under this assumption, the pooled effect increased from 0.06 to 0.08 on a 0-3 point scale (95 percent CI: 0.03 to 0.12) favoring intranasal corticosteroid. The larger bound of the 95 percent CI represented 4 percent of maximum score. Based on this analysis, it is unlikely that the 95 percent confidence interval of a meta-analysis including the Philip (2002) trial¹¹² would contain the MCID. The body of evidence supporting a conclusion of equivalence of oral selective antihistamine and leukotriene receptor antagonist for this outcome is therefore precise. The overall strength of evidence for this conclusion is rated as moderate based on these considerations.

Eye Symptoms

Four trials^{109, 110, 113, 114} assessed a four symptom TOSS comprising eye tearing, itching, redness, and puffiness (N=1708) at 2 or 4 weeks. One of these¹¹⁰ was a good quality trial of 187 patients (11 percent of patients reporting this outcome) that showed a treatment effect of 0.03 on a 0-3 point scale (1 percent of maximum score) favoring leukotriene receptor antagonist. A p-value was not reported. All other assessments favored oral selective antihistamine, including two fair quality trials^{113, 114} of 1321 patients (77 percent of patients reporting) and one trial¹⁰⁹ that was rated poor quality due to noncomparable groups at baseline and inappropriate analysis of results (unadjusted for baseline group differences). Treatment effects were 0.05 and 0.12 in the fair quality trials (2 percent and 4 percent of maximum score, respectively) and 0.16 in the poor quality trial (5 percent of maximum score). All four trials were included in a meta-analysis (Figure 8). The pooled treatment effect was 0.08 on a 0-3 point scale (95 percent CI: 0.02 to 0.14) favoring oral selective antihistamine over leukotriene receptor antagonist. The larger bound of the 95 percent CI represented 5 percent of maximum score. Statistical heterogeneity was low (I²=14 percent, p=0.32). Four-week results from one trial¹¹⁴ were not included in the meta-analysis because 2-week results were the identified primary outcome. At 4 weeks, the treatment effect was 0.02 on a 0-3 point scale (1 percent of maximum score) favoring oral selective antihistamine. A p-value was not reported.

Figure 8

Total ocular symptom score at 2 to 4 weeks: meta-analysis of 4 trials–oral selective antihistamine versus leukotriene receptor antagonist.

For TOSS at 2 to 4 weeks, the risk of bias was rated as medium. Eleven percent of patients were in the good quality trial, and 12 percent were in the poor quality trial. Although results across individual trials were inconsistent at 2 weeks, statistical heterogeneity for the pooled treatment effect was low. The 95 percent CI for the pooled estimate fell within an interval bounded by −MCID and +MCID (-0.9 and +0.9 on the 0-3 point scale used). The body of evidence supporting a conclusion of equivalence of oral selective antihistamine and leukotriene receptor antagonist for this outcome is therefore precise. The strength of evidence for this conclusion is rated as moderate based on these considerations.

Asthma Symptoms

One good quality trial¹⁰⁸ (N=622) reported individual and TASS scores in addition to rescue medication use and FEV₁ at 2 and 4 weeks. Patients had moderate asthma symptoms at baseline. All outcomes had greater improvements with leukotriene receptor antagonist than with oral selective antihistamine, but no statistically significant differences between treatment groups were observed for any outcome during the 4 weeks of the trial. For all outcomes, the risk of bias was rated as low, and consistency could not be assessed with a single trial.

Treatment effects at 2 and 4 weeks were:

• Total asthma symptoms on a 0-9 point scale: 0.09 at 2 weeks and 0.16 at 4 weeks (1 percent and 2 percent of maximum score, respectively).
• Rescue medication use: 2.4 puffs per day at 2 weeks and 3.8 puffs per day at 4 weeks (both greater than an MCID of 1 puff per day)

Treatment effects at 4 weeks were:

• Individual asthma symptoms on a 0-3 point scale: 0.02 for cough, 0.04 for wheeze, and 0.06 for difficulty breathing (1 percent, 1 percent and 2 percent of maximum score, respectively)
• FEV1: 0.03 percent predicted (less than an MCID of 10 percent)

For rescue medication use at 2 and 4 weeks, the treatment effect is precise, and there is moderate strength evidence to support the use of oral leukotriene receptor antagonist. For all other outcomes, evidence was insufficient to support the use of one treatment over the other.

Quality of Life

All six trials^109-114 that assessed quality of life at 2 and 4 weeks used the RQLQ (N=3114). Two of these were good quality trials^{110, 111} of 643 patients (21 percent of patients reporting this outcome), two^{113, 114} were fair quality trials of 1321 patients (42 percent of patients reporting), and two^{109, 112} were poor quality trials of 1150 patients (37 percent of patients reporting). P-values were not reported in any trial. Treatment effects exceeded the MCID of 0.5 in three trials^{109, 112, 113}, two^{112, 113} at 2 weeks (0.10 and 0.08 in a poor and fair quality trial, respectively) and one¹⁰⁹ at 4 weeks (0.13 in a poor quality trial). All three results favored oral selective antihistamine over leukotriene receptor antagonist. A meta-analysis of four trials^111-114 that reported 2-week RQLQ results was performed (N=2723; Figure 9). The pooled treatment effect favored oral selective antihistamine (mean difference 0.06; 95 percent CI: -0.03 to 0.15) but was not statistically significant. Statistical heterogeneity was low (I²=0 percent, p=0.84). One¹¹⁰ of two trials not included in the meta-analysis represented 6 percent of patients reporting this outcome and found no treatment difference between groups at 2 weeks. The other trial,¹⁰⁹ also representing 6 percent of patients reporting this outcome, showed a treatment effect of 0.13 favoring oral selective antihistamine at 4 weeks. One of the trials¹¹⁴ included in the meta-analysis reported 4-week results, which were not included because 2-week results were the identified primary outcome. In contrast to the 2-week result, the treatment effect at 4 weeks favored leukotriene receptor antagonist. The effect was 0.04, which did not exceed the MCID.

Figure 9

Rhinoconjunctivitis quality of life at 2 weeks: meta-analysis of 4 trials–oral selective antihistamine versus leukotriene receptor antagonist.

For quality of life as assessed by the RQLQ, the risk of bias was rated as medium. Twenty-one percent of patients were in good quality trials, and 37 percent were in poor quality trials. Statistical heterogeneity for the pooled effect favoring oral selective antihistamine was low, and one trial¹¹⁰ not included in the meta-analysis that showed a treatment difference of zero represented only 6 percent of patients reporting this outcome. The 95 percent CI for the pooled estimate (-0.03 to 0.15) fell within an interval bounded by −MCID and +MCID (-0.5 and +0.5). Of two trials not included in the meta-analysis,^{109, 110} one¹⁰⁹ (6 percent of patients reporting this outcome) showed a treatment effect (0.13) favoring oral selective antihistamine that was smaller than the MCID but larger than the pooled effect (0.06). The other¹¹⁰ (6 percent of patients reporting this outcome) showed a treatment difference of zero that was, therefore, smaller than both the MCID and the pooled effect. If these trials were included in the meta-analysis, the pooled effect would change very little, and it is unlikely that the 95 percent confidence interval would contain the MCID. One trial,¹¹⁴ a large trial representing 20 percent of patients reporting this outcome, was included in the meta-analysis of results at 2 weeks and reported an additional treatment effect of 0.04 at 4 weeks favoring leukotriene receptor antagonist. If this result were included in the meta-analysis, the effect estimate favoring oral selective antihistamine would decrease and possibly cross the “no effect” line to favor leukotriene receptor antagonist, but it is unlikely that the MCID would lie within the 95 percent confidence interval. Based on these considerations, the body of evidence supporting a conclusion of equivalence of oral selective antihistamine and leukotriene receptor antagonist for this outcome is therefore considered precise. The strength of evidence for this conclusion is rated as moderate.

Nasal Symptoms

Eight^{115-119, 121} of nine trials assessed congestion after 2 weeks of treatment (N=2443 of 2473). Seven of these^115-119 reported treatment effects favoring intranasal corticosteroid, although none were reported to be statistically significant. In the eighth trial,¹²¹ representing 4 percent of patients reporting this outcome, the treatment difference was zero. The USPSTF quality rating was good in five of these trials^{115, 117, 121} (85 percent of patients reporting this outcome) and poor in three trials.^{116, 118, 119} Poor USPSTF ratings were assigned for noncomparable groups at baseline,¹¹⁶ inadequate blinding,^{116, 118, 119} and inappropriate analysis of results (unadjusted for baseline group differences¹¹⁶ and not intention to treat¹¹⁹). A meta-analysis of four good quality trials^{115, 121} (N=1791; 73 percent of patients reporting this outcome) yielded a statistically significant pooled effect of 0.14 on a 0-6 point scale (95 percent CI: 0.02 to 0.26) favoring intranasal corticosteroid (Figure 10). The larger bound of the 95 percent CI represented 4 percent of maximum score. Statistical heterogeneity was low (I²=0 percent, p=0.54). Treatment effects for four trials^116-119 not included in the meta-analysis favored intranasal corticosteroid with a range of 0.11¹¹⁷ on a 0-6 point scale (2 percent of maximum score) to 0.7¹¹⁶ on a 0-3 point scale (23 percent of maximum score).

Figure 10

Congestion at 2 weeks: meta-analysis of 4 trials–intranasal corticosteroid versus nasal antihistamine.

For nasal congestion at 2 weeks, the risk of bias was rated as low. Eighty-five percent of patients were in good quality trials. Treatment effects consistently favored intranasal corticosteroid in 96 percent of patients reporting on this outcome. This finding was consistent with results of a meta-analysis of four^{115, 121} of these trials (73 percent of patients reporting this outcome), including the one trial¹²¹ that reported a treatment effect of zero, and statistical heterogeneity was low. The 95 percent CI for the pooled effect (0.02 to 0.26) fell within an interval bounded by −MCID and +MCID (-1.8 and +1.8 on the 0-6 point scale used). Of four trials^116-119 not included in the meta-analysis, one¹¹⁸ reported a treatment effect favoring intranasal corticosteroid but did not report the magnitude of the effect. Because this trial represented 5 percent of patients reporting this outcome, its impact on the pooled estimate if this trial were included in the meta-analysis likely would be minimal. One trial¹¹⁷ (12 percent of patients reporting this outcome) showed a treatment effect (0.11 on a 0-6 point scale) that was smaller than both the pooled effect (0.14 on a 0-6 point scale) and the MCID (1.8). If this trial were included in the meta-analysis, the pooled effect estimate would decrease slightly, and it is unlikely that the 95 percent confidence interval would include the MCID. Two trials^{116, 119} (9 percent of patients reporting this outcome ) showed treatment effects of 0.5¹¹⁹ and 0.7¹¹⁶ on a 0-3 point scale (1.0 and 1.4 on a 0-6 point scale, respectively). These effects were substantially larger than the pooled effect (0.14 on a 0-6 point scale) but smaller than the MCID (1.8). To determine the impact of these two trials^{116, 119} on the pooled estimate, we added both to the meta-analysis with assumed standard deviations equal to half the mean change in score in each treatment group. Under this assumption, the pooled effect increased from 0.14 to 0.45 on a 0-6 point scale (95 percent CI: 0.06 to 0.85) favoring intranasal corticosteroid. The larger bound of the 95 percent CI represented 14 percent of maximum score. Based on this analysis, it is unlikely that the 95 percent confidence interval of a meta-analysis including these two trials^{116, 119} would contain the MCID. The body of evidence in support of a conclusion of equivalence of intranasal corticosteroid and nasal antihistamine for this outcome is therefore considered precise. The overall strength of evidence for this conclusion is rated as high.

Eight^{115-119, 121} of nine trials assessed rhinorrhea after 2 weeks of treatment (N=2443 of 2473 patients). Seven^{115, 117-119, 121} of eight reported treatment effects in favor of intranasal corticosteroid, although none were reported to be statistically significant. The eighth trial¹¹⁶ (n=50; 2 percent of patients reporting this outcome) reported a treatment difference of zero. The USPSTF quality rating was good in five trials^{115, 117, 121} (85 percent of patients reporting this outcome) and poor in three trials.^{116, 118, 119} Poor USPSTF ratings were assigned for noncomparable groups at baseline,¹¹⁶ inadequate blinding,^{116, 118, 119} and inappropriate analysis of results (unadjusted for baseline group differences¹¹⁶ and not intention to treat¹¹⁹). A meta-analysis of four trials^{115, 121} (N=1791; 73 percent of patients reporting this outcome) yielded a statistically significant pooled effect estimate of 0.17 on a 0-6 point scale (95 percent CI: 0.04 to 0.30) favoring intranasal corticosteroid (Figure 11). The larger bound of the 95 percent CI represented 5 percent of maximum score. Statistical heterogeneity was low (I²=0 percent, p=0.70). Of four trials^116-119 not included in the meta-analysis, three^117-119 reported treatment effects favoring intranasal corticosteroid. Reported effect sizes were 0.28¹¹⁷ on a 0-6 point scale (5 percent of maximum score) and 0.4¹¹⁹ on a 0-3 point scale (13 percent of maximum score, respectively). One trial¹¹⁸ did not report the magnitude of effect. The fourth trial¹¹⁶ reported no treatment difference.

Figure 11

Rhinorrhea at 2 weeks: meta-analysis of 4 trials–intranasal corticosteroid versus nasal antihistamine.

For rhinorrhea at 2 weeks, the risk of bias was rated as low. Eighty-five percent of patients were in good quality trials. Treatment effects consistently favored intranasal corticosteroid in 98 percent of patients reporting this outcome, and statistical heterogeneity in a meta-analysis of four trials^{115, 121} (73 percent of patients reporting this outcome) was low. The 95 percent CI for the pooled effect (0.04 to 0.30) fell within an interval bounded by −MCID and +MCID (-1.8 and +1.8 on the 0-6 point scale used). Three^117-119 of four trials^116-119 not included in the meta-analysis reported treatment effects favoring intranasal corticosteroid. One¹¹⁸ of these did not report the magnitude of effect, but this trial represented 5 percent of patients reporting this outcome. Two trials^{117, 119} (19 percent of patients reporting this outcome) reported treatment effects of 0.28¹¹⁷ on a 0-6 point scale and 0.4¹¹⁹ on a 0-3 point scale (0.8 on a 0-6 point scale) that were larger than the pooled effect but smaller than the MCID. To determine the impact of these two trials^{117, 119} on the pooled estimate, we added both to the meta-analysis with assumed standard deviations equal to half the mean change in score in each treatment group. Under this assumption, the pooled effect increased from 0.17 to 0.34 on a 0-6 point scale (95 percent CI: 0.10 to 0.57) favoring intranasal corticosteroid. The larger bound of the 95 percent CI represented 10 percent of maximum score. Based on this analysis, it is unlikely that the 95 percent confidence interval of a meta-analysis including these two trials^{117, 119} would contain the MCID. The fourth trial¹¹⁶ reported a treatment effect of zero (and therefore smaller than both the pooled estimate [0.17] and the MCID), but this trial represented 2 percent of patients reporting this outcome. The body of evidence in support of a conclusion of equivalence of intranasal corticosteroid and nasal antihistamine for this outcome is therefore considered precise. The overall strength of evidence for this conclusion is rated as high.

Eight^{115-119, 121} of nine trials assessed sneezing after 2 weeks of treatment (N=2443 of 2473). Six^{115-117, 119} of these reported treatment effects favoring intranasal corticosteroid, although none were reported to be statistically significant. One trial¹²¹(4 percent of patients reporting this outcome) showed no treatment difference, and the eighth trial¹¹⁸ (5 percent of patients reporting this outcome) reported a treatment effect favoring nasal antihistamine. This treatment effect was not statistically significant and its magnitude was not reported. The USPSTF quality rating was good in five trials ^{115, 117, 121} (85 percent of patients assessed for this outcome) and poor in three trials.^{116, 118, 119} Poor USPSTF ratings were assigned for noncomparable groups at baseline,¹¹⁶ inadequate blinding,^{116, 118, 119} and inappropriate analysis of results (unadjusted for baseline group differences¹¹⁶ and not intention to treat¹¹⁹). A meta-analysis of four good quality trials^{115, 121} (N=1791; 73 percent of patients reporting this outcome) yielded a statistically nonsignificant pooled effect estimate of 0.12 on a 0-6 point scale (95 percent CI: -0.01 to 0.25) favoring intranasal corticosteroid (Figure 12). The larger bound of the 95 percent CI represented 4 percent of maximum score. Statistical heterogeneity was low (I²=0 percent, p=0.89). Of four trials^116-119 not included in the meta-analysis, treatment effects favoring intranasal corticosteroid were reported by three.^{116, 117, 119} Reported effect sizes were 0.1¹¹⁶ and 0.4¹¹⁹ on a 0-3 point scale (3 percent and 13 percent of maximum score, respectively), and 0.12¹¹⁷ on a 0-6 point scale (2 percent of maximum score). One trial¹¹⁸ reported a statistically nonsignificant treatment effect favoring nasal antihistamine but did not report the magnitude of effect.

Figure 12

Sneezing at 2 weeks: meta-analysis of 4 trials–intranasal corticosteroid versus nasal antihistamine.

For the outcome of sneezing at 2 weeks, the risk of bias was rated as low. Eighty-five percent of patients were in good quality trials. Statistical heterogeneity in a meta-analysis of four trials^{115, 121} (73 percent of patients reporting this outcome) that favored intranasal corticosteroid was low. This included one trial¹²¹ (4 percent of patients reporting this outcome) that reported a treatment difference of zero. The pooled effect was consistent with three.^{116, 117, 119}of four trials not included in the meta-analysis. The fourth trial¹¹⁸ reported a treatment effect of unknown size favoring nasal antihistamine, but this trial represented 5 percent of patients reporting this outcome. The body of evidence was therefore considered consistent. The 95 percent CI for the pooled estimate (0.01 to 0.25) fell within an interval bounded by −MCID and +MCID (-1.8 and +1.8 on the 0-6 point scale used). Of four trials^116-119 not included in the meta-analysis, one¹¹⁸ showed a treatment effect in the opposite direction (favoring nasal antihistamine) but did not report the magnitude of effect. Because this trial represented 5 percent of patients reporting this outcome, the reduction in the pooled estimate if this trial were included in the meta-analysis likely would be minimal, unless the treatment effect was unexpectedly large. One trial¹¹⁷ (12 percent of patients reporting this outcome) showed a treatment effect (0.12 on a 0-6 point scale) equal to the pooled effect. Change in the pooled estimate likely would be minimal if this trial were included in the meta-analysis. Two trials^{116, 119} reported treatment effects of 0.1¹¹⁶and 0.4¹¹⁹ on a 0-3 point scale (0.2 and 0.8 on a 0-6 point scale, respectively). These effects were larger than the pooled effect (0.12 on a 0-6 point scale) but smaller than the MCID of 1.8. To determine the impact of these two trials^{116, 119} on the pooled estimate, we added both to the meta-analysis with assumed standard deviations equal to half the mean change in score in each treatment group. Under this assumption, the pooled effect increased from 0.12 to 0.25 on a 0-6 point scale (95 percent CI: 0.03 to 0.54) favoring intranasal corticosteroid. The larger bound of the 95 percent CI represented 9 percent of maximum score. Based on this analysis, it is unlikely that the 95 percent confidence interval of a meta-analysis including these two trials^{116, 119} would contain the MCID. The body of evidence in support of a conclusion of equivalence of intranasal corticosteroid and nasal antihistamine for this outcome is therefore considered precise. The overall strength of evidence for this conclusion is rated as high.

Seven^{115, 117-119, 121} of nine trials assessed nasal itch after 2 weeks of treatment (N=2393 of 2473), all of which reported treatment effects favoring intranasal corticosteroid, although none were reported to be statistically significant. The USPSTF quality rating was good in five trials^{115, 117, 121} (88 percent of patients reporting this outcome) and poor in two trials.^{118, 119} Poor USPSTF ratings were assigned for inadequate blinding^{118, 119} and inappropriate analysis of results (not intention to treat¹¹⁹). A meta-analysis of four good quality trials^{115, 121} (N=1791; 75 percent of patients reporting this outcome) yielded a statistically significant pooled effect of 0.19 on a 0-6 point scale (95 percent CI: 0.03 to 0.34) favoring intranasal corticosteroid (Figure 13). The larger bound of the 95 percent CI represented 11 percent of maximum score. Statistical heterogeneity was low (I²=27 percent, p=0.25). Of three trials^117-119 not included in the meta-analysis, two^{117, 119} reported treatment effects of 0.09¹¹⁷ on a 0-6 point scale (2 percent of maximum score) and 0.4¹¹⁹ on a 0-3 point scale (13 percent of maximum score). The third trial¹¹⁸ did not report the magnitude of the treatment effect.

Figure 13

Nasal itch at 2 weeks: meta-analysis of 4 trials–intranasal corticosteroid versus nasal antihistamine.

For nasal itch at 2 weeks, the risk of bias was rated as low. Eighty-eight percent of patients reporting this outcome were in good quality trials. Treatment effects consistently favored intranasal corticosteroid in all trials, and statistical heterogeneity of a meta-analysis of four trials^{115, 121} (75 percent of patients reporting this outcome) was low. The 95 percent CI for the pooled estimate (0.03 to 0.34) fell within an interval bounded by −MCID and +MCID (-1.8 and +1.8 on the 0-6 point scale used). Of three trials^117-119 not included in the meta-analysis, one¹¹⁸ did not report the magnitude of the treatment effect, but this trial represented 5 percent of patients reporting this outcome. One trial¹¹⁷ (13 percent of patients reporting this outcome) showed a treatment effect (0.09) that was less than both the pooled effect (0.19) and the MCID. If this trial were included in the meta-analysis, the pooled effect would decrease, and it is unlikely that the 95 percent CI would include an MCID of 30 percent maximum score. The third trial¹¹⁹ (7 percent of patients reporting this outcome) showed a treatment effect of 0.4 on a 0-3 point scale (0.8 on a 0-6 point scale). This effect was larger than the pooled effect but smaller than the MCID. To determine the impact of this trial¹¹⁹ on the pooled estimate, we added it to the meta-analysis with an assumed standard deviation equal to half the mean change in score in each treatment group. Under this assumption, the pooled effect increased from 0.19 to 0.35 on a 0-6 point scale (95 percent CI: 0.02 to 0.68) favoring intranasal corticosteroid. The larger bound of the 95 percent CI represented 11 percent of maximum score. Based on this analysis, it is unlikely that the 95 percent confidence interval of a meta-analysis including this trial¹¹⁹ would contain the MCID. The body of evidence in support of a conclusion of equivalence of intranasal corticosteroid and nasal antihistamine for this outcome is therefore considered precise. The overall strength of evidence for this conclusion is rated as high.

Seven^{115, 117, 118, 120, 121} of nine trials assessed TNSS after 2 weeks of treatment (N=2257 of 2473). Five^{115, 117, 121} reported treatment effects favoring intranasal corticosteroid, and two^{118, 120} reported treatment effects favoring nasal antihistamine. None were reported to be statistically significant. The USPSTF quality rating was good in five trials^{115, 117, 121} (93 percent of patients reporting this outcome) and poor in two trials.^{118, 120} Poor USPSTF ratings were assigned for noncomparable groups at baseline,¹²⁰ inadequate blinding,¹¹⁸ and inappropriate analysis of results (unadjusted for baseline group differences¹²⁰). A meta-analysis of five good quality trials^{115, 117, 121} (N=2097; 93 percent of patients reporting this outcome) yielded a statistically significant pooled effect estimate of 0.65 on a 0-24 point scale (95 percent CI: 0.25 to 1.05) favoring intranasal corticosteroid (Figure 14). The larger bound of the 95 percent CI represented 4 percent of maximum score. Statistical heterogeneity was low (I²=0 percent, p=0.98). Two trials^{118, 120} not included in the meta-analysis reported treatment effects favoring nasal antihistamine. Effect sizes were not reported.

Figure 14

Total nasal symptom score at 2 weeks: meta-analysis of 5 trials–intranasal corticosteroid versus nasal antihistamine.

For TNSS at 2 weeks, the risk of bias was rated as low. Ninety-three percent of patients were in good quality trials. Treatment effects consistently favored intranasal corticosteroid in 93 percent of patients reporting this outcome. The 95 percent CI for the pooled estimate (0.25 to 1.05) fell within an interval bounded by −MCID and +MCID (-7.2 and +7.2 on the 0-24 point scale used). The two trials^{118, 120} not included in the meta-analysis did not report treatment effect magnitudes, but both favored nasal antihistamine. If these trials were included in the meta-analysis, the pooled effect would decrease. Because these trials represented only 7 percent of patients reporting this outcome, it is unlikely that the 95 percent confidence interval would include the MCID. The body of evidence in support of a conclusion of equivalence of intranasal corticosteroid and nasal antihistamine for this outcome is therefore considered precise. The overall strength of evidence for this conclusion is rated as high.

Individual nasal symptoms of congestion, rhinorrhea, and sneezing were assessed at 3 weeks and 4 weeks in one trial¹¹⁶ (N=50). For congestion, a treatment effect of 0.5 on a 0-3 point scale (17 percent of maximum score) favored nasal antihistamine at both 3 weeks and 4 weeks. For rhinorrhea and sneezing, treatment differences were zero at both 3 weeks and 4 weeks. Statistical significance was not reported for any outcome. This trial was rated poor quality due to noncomparable groups at baseline, inadequate blinding, and inappropriate analysis of results (unadjusted for baseline group differences). Consistency could not be assessed in a single trial, and no observed result at any assessment period exceeded an MCID of 0.9, representing 30 percent of maximum score. The evidence was therefore insufficient to support the use of one treatment over the other for these outcomes.

At 3, 4 and 5 weeks, one trial¹²⁰ (N=30) reported improvement in TNSS with nasal antihistamine. Treatment effects ranged from 1.0 to 1.4 on a 0-12 point scale (from 8 percent to 12 percent of maximum score) and were statistically nonsignificant. This trial was rated poor quality due to noncomparable groups at baseline and inappropriate analysis of results (unadjusted for baseline group differences). Risk of bias is therefore high. Consistency could not be assessed in a single trial, and no observed result at any assessment period exceeded an MCID of 4, representing 30 percent of maximum score. The evidence was therefore insufficient to support the use of one treatment over the other for this outcome.

Eye Symptoms

Five^{115, 117, 118} of nine trials (N=2128 of 2473 patients) assessed eye symptoms at 2 weeks. Four^{115, 117} were rated good quality (94 percent of patients reporting this outcome). Treatment effects in these trials favored nasal antihistamine and ranged from 0.2 to 0.45 on a 0-18 point scale (from 1 percent to 3 percent of maximum score). A meta-analysis of three¹¹⁵ of these trials (N=1697; 80 percent of patients reporting this outcome) yielded a statistically nonsignificant pooled effect estimate of 0.22 on a 0-18 point scale (95 percent CI: -0.12 to 0.57) favoring nasal antihistamine (Figure 15). The larger bound of the 95 percent CI represented 3 percent of maximum score. Statistical heterogeneity was low (I²=0 percent, p=0.97). The fourth good quality trial¹¹⁷ (n=305; 14 percent of patients reporting this outcome) showed a treatment effect of 0.45 on a 0-18 point scale (11 percent of maximum score). The fifth trial¹¹⁸ (n=130; 7 percent of patients reporting this outcome) showed a statistically nonsignificant treatment effect of 0.6 on a 0-9 point scale (7 percent of maximum score) favoring intranasal corticosteroid. This trial was rated poor quality due to inadequate blinding.

Figure 15

Total ocular symptom score at 2 weeks: meta-analysis of 4 trials–intranasal corticosteroid versus nasal antihistamine.

For the outcome of eye symptoms, the risk of bias was rated as low. Ninety-four percent of patients were in good quality trials. Treatment effects consistently favored nasal antihistamine in 94 percent of patients reporting this outcome, and statistical heterogeneity of a meta-analysis of 80 percent of patients was low. The 95 percent CI for the pooled estimate (-0.12 to 0.57) fell within an interval bounded by −MCID and +MCID (-5.4 and +5.4 on the 0-18 point scale used). Of two trials^{117, 118} not included in the meta-analysis, one¹¹⁸ reported a treatment effect of 0.06 on a 0-9 point scale (0.12 on a 0-18 point scale) favoring intranasal corticosteroid. If this trial were included in the meta-analysis, the pooled effect estimate (0.22 favoring nasal antihistamine) would decrease. Because this trial represented 7 percent of patients reporting this outcome, it is unlikely that the 95 percent CI of the pooled effect would include the MCID of 5.4. The other trial¹¹⁷ (14 percent of patients reporting this outcome) showed a treatment effect of 0.45 on a 0-18 point scale favoring nasal antihistamine. If this trial were included in the meta-analysis, the pooled effect estimate would increase, but it is unlikely that the 95 percent CI would include the MCID. The body of evidence to support a conclusion of equivalence of intranasal corticosteroid and nasal antihistamine for this outcome was therefore considered precise. The overall strength of evidence supporting this conclusion is rated as high.

Quality of Life

Two trials^{117, 121} (N=404; 16 percent of the total patient sample for this comparison) assessed quality of life using the RQLQ instrument. Both were good quality trials, and both observed statistically nonsignificant treatment effects in favor of intranasal corticosteroid (0.26 for both).

For RQLQ at 2 weeks, the risk of bias is rated as low based in the good quality of the trials. Results consistently favored intranasal corticosteroid, but neither exceeded the MCID of 0.5 points. Evidence was therefore insufficient to support the use of one treatment over the other for this outcome.

A meta-analysis of 2-week RQLQ outcomes from three good quality trials¹¹⁵ (N=1693) yielded a pooled effect estimate of 0.1 favoring intranasal corticosteroid. This result is consistent with the treatment effects reported in two trials^{117, 121} described above. Because the published meta-analysis lacked details about the how the analysis was conducted, this result could not be replicated and was not included in the formal evidence assessment.

Nasal Symptoms

One¹²² of three trials that assessed nasal symptoms reported outcomes (congestion, rhinorrhea, sneezing, nasal itch, and TNSS) at 2 weeks (N=43). The trial was rated poor quality due to noncomparable groups at baseline. Statistically significant treatment effects favoring intranasal corticosteroid were reported for four nasal symptoms and for TNSS. For individual symptoms, treatment effects ranged from 0.21 for nasal itch to 0.59 for rhinorrhea on a 0-3 point scale (from 7 percent to 20 percent of maximum score). For TNSS, the treatment effect was 1.53 on a 0-12 point scale (13 percent of maximum score).

For nasal symptoms at 2 weeks, the risk of bias was considered high because the trial¹²² was small and of poor quality. Consistency cannot be assessed for a single trial. Effect estimates were imprecise because none exceeded an MCID of 30 percent maximum score. Evidence was insufficient to support the use of one treatment over the other for these outcomes.

Three trials (total N=344) assessed nasal symptoms beyond 2 weeks: Bjerrum (1985)¹²² at 3 weeks, Lange (2005)¹²⁴ at 4 weeks, and Bousquet (1993)¹²³ at 6 weeks. Trial quality ratings were poor due to noncomparable groups at baseline,^{122, 123} lack of blinding,¹²⁴ and inappropriate analysis of results (unadjusted for baseline group differences¹²³).

At 3 weeks,¹²² statistically significant treatment effects were shown for rhinorrhea, sneeze, nasal itch, and TNSS. All favored intranasal corticosteroid. Treatment effect magnitudes were comparable to those seen at 2 weeks¹²² and ranged from 0.15 for nasal itch to 0.49 for rhinorrhea on a 0-3 point scale (from 5 percent to 16 percent of maximum score). The treatment effect for TNSS was 1.19 on a 12-point scale (10 percent of maximum score). Nasal congestion was the only symptom for which a statistically nonsignificant treatment effect was reported (0.28 on a 0-3 point scale [9 percent of maximum score]).

At 4 weeks¹²⁴ and 6 weeks¹²³, statistically significant treatment effects favoring intranasal corticosteroid were reported for four individual nasal symptoms. At 4 weeks,¹²⁴ there was a statistically significant treatment effect favoring intranasal corticosteroid for TNSS. TNSS was not assessed at 6 weeks. Magnitude of effects at 4 and 6 weeks were not reported.

For nasal symptoms at 3 to 6 weeks, the risk of bias was considered high. All three trials were rated poor quality. Treatment effects consistently favored intranasal corticosteroid. Most treatment effects were not reported. None of the reported treatment effects exceeded an MCID of 30 percent maximum score. The body of evidence was therefore imprecise, and evidence to support the use of one treatment over the other for these outcomes is insufficient.

Two trials reported PGA of treatment efficacy at 4 weeks¹²⁴ and 8 weeks¹²⁵ (N=173). Both trials were rated poor quality due to lack of blinding^{124, 125} and lack of maintenance of comparable groups.¹²⁵ Both reported statistically significant results favoring intranasal corticosteroid.

Nasal Symptoms

Three^{126, 127, 129} of five trials (2014 of 2328 patients, 87 percent) assessed individual nasal symptoms (congestion, rhinorrhea, sneezing, and nasal itch) at 2 weeks. All three trials were rated good quality. For each symptom, the treatment effect favored intranasal corticosteroid over oral leukotriene receptor antagonist and was statistically significant. Meta-analyses of the three trials for each symptom favored intranasal corticosteroid with statistically significant treatment effects ranging from 7.9 to 8.7 on a 100 VAS (Figure 16 through Figure 19). The larger bound of the 95 percent CIs ranged from 10.08 to 10.76 on a 100-point VAS (from 10 percent to 11 percent of maximum score, respectively). Statistical heterogeneity was low (I²=0 percent for all four analyses, p=0.33 to 0.88).

Figure 16

Congestion at 2 weeks: meta-analysis of 3 trials–intranasal corticosteroid versus oral leukotriene receptor antagonist.

Figure 17

Rhinorrhea at 2 weeks: meta-analysis of 3 trials–intranasal corticosteroid versus oral leukotriene receptor antagonist.

Figure 18

Sneezing at 2 weeks: meta-analysis of 3 trials–intranasal corticosteroid versus oral leukotriene receptor antagonist.

Figure 19

Nasal itch at 2 weeks: meta-analysis of 3 trials–intranasal corticosteroid versus oral leukotriene receptor antagonist.

For individual nasal symptoms at 2 weeks, the risk of bias was rated as low. All three trials were good quality. Treatment effects consistently favored intranasal corticosteroid in all three trials. The 95 percent CIs for the pooled estimates fell within an interval bounded by −MCID and +MCID (-30 and +30 on the 100-point VAS used). The body of evidence to support a conclusion of equivalence of intranasal corticosteroid and oral leukotriene receptor antagonist for each of these outcomes is therefore precise. The strength of evidence supporting these conclusions is high.

All five trials assessed TNSS at 2 weeks (N=2038). Three good quality trials^{126, 127, 129} of 2014 patients represented 87 percent of patients reporting this outcome. Thirteen percent of patients were in two trials^{97, 128} that were rated poor quality due to inappropriate analysis of results (not intention to treat). Treatment effects favored intranasal corticosteroid over oral leukotriene receptor antagonist and were statistically significant in all but one trial.¹²⁸ All three good quality trials^{126, 127, 129} assessed TNSS using a 0-400 point VAS. Treatment effects were 33.6 (95 percent CI: 20.6 to 46.5);¹²⁶ 26.1 (95 percent CI: 9.7 to 42.5);¹²⁷ and 34.4 (95 percent CI: 23.4 to 49.3)¹²⁹ The larger bounds of the 95 percent CIs were 12 percent, 11 percent and 12 percent of maximum score, respectively. Of two poor quality trials reporting on this outcome using an interval rating scale, one¹²⁸ (n=29) reported a statistically nonsignificant effect of 0.8 on a 0-16 point scale (5 percent of maximum score), and the other (Trial 1 in Lu [2009];⁹⁷ n=285) reported a statistically significant effect of 0.34 on a 0-3 point scale (11 percent of maximum score).

A meta-analysis of four trials^{97, 126, 127, 129} was performed. The fifth trial¹²⁸ was excluded due to lack of a variance estimate for the treatment effect. The meta-analysis yielded a statistically significant pooled effect (standardized mean difference) of 0.40 (95 percent CI: 0.27 to 0.52) favoring intranasal corticosteroid (Figure 20). Effect estimates in the pooled trials were in the same direction, and their 95 percent CIs did not touch the “no effect” line.

Figure 20

Total nasal symptom score at 2 weeks: meta-analysis of 4 trials–intranasal corticosteroid versus oral leukotriene receptor antagonist.

For TNSS at 2 weeks, the risk of bias was rated as low. Eighty-seven percent of patients were in good quality trials. Treatment effects consistently favored intranasal corticosteroid for all patients reporting this outcome. Ninety-five percent CIs of pooled treatment effects fell within an interval bounded by −MCID and +MCID (-120 and +120^{126, 127, 129} or -1.8 and +1.8⁹⁷). These effects were therefore considered precise. The one trial¹²⁸ excluded from the meta-analysis did not alter the precision assessment because this trial represented 1 percent of patients reporting this outcome. The body of evidence supporting a conclusion of equivalence of intranasal corticosteroid and leukotriene receptor antagonist for this outcome is therefore considered precise. The overall strength of evidence for this conclusion is rated as high.

One good quality trial¹²⁷ (N=573) assessed four individual nasal symptoms (congestion, rhinorrhea, sneezing, and nasal itch) at 4 weeks. All comparisons favored intranasal corticosteroid and were statistically significant. Using a 100-point VAS, treatment effects ranged from 6.0 for nasal itch to 8.3 for congestion (from 6 percent to 8 percent of maximum score).

The risk of bias for this outcome was rated as low based on the good quality of the trial reporting. Consistency cannot be addressed for a single trial. Results were imprecise because none exceeded an MCID of 30 percent maximum score. Evidence was therefore insufficient to support the use of one treatment over the other for this outcome.

Two trials^{127, 128} assessed TNSS at time points beyond 2 weeks (N=602). One¹²⁷ was a good quality trial in 573 patients (95 percent of patients reporting this outcome) that reported 4-week outcomes. A statistically significant treatment effect of 28 points on a 0-400 scale (7 percent of maximum score) favored intranasal corticosteroid. The other was a poor quality trial¹²⁸ that reported outcomes (mean results during the previous 2 weeks) at 5 and 8 weeks. At 5 weeks, a statistically nonsignificant treatment effect of 1.4 on a 0-16 point scale (9 percent of maximum score) favored intranasal corticosteroid. At 8 weeks, a statistically significant treatment effect of 0.7 on a 0-16 point scale (4 percent of maximum score) favored intranasal corticosteroid.

The risk of bias for these outcomes was rated as low. Ninety-five percent of patients were in the good quality trial.¹²⁷ Treatment effects consistently favored intranasal corticosteroid, but none exceeded an MCID of 30 percent maximum score. The body of evidence for these outcomes was therefore imprecise. Evidence was insufficient to support the use of one treatment over the other for this outcome.

Asthma Symptoms

One good quality trial¹²⁷(N=573) assessed symptoms and objective measures of asthma over 4 weeks of treatment. There were no statistically significant differences between treatment groups in any outcome, nor were there differences when treatment groups were stratified by baseline asthma severity. For all outcomes, the risk of bias was rated as low, and consistency could not be assessed with a single trial.

Treatment effects favoring oral leukotriene receptor antagonist were:

• Proportion of symptom-free days: 1.3 percent difference between groups
• Proportion of albuterol-free days: 0.7 percent difference between groups

MCIDs for these outcomes have not been established. Because neither result was statistically significant, evidence was insufficient to support the use of one treatment over the other for these outcomes.

Treatment effects favoring intranasal corticosteroid were:

• Morning peak expiratory flow (PEF): 2.4 L/min
• Evening PEF: 1.8 L/min
• Proportion of patients experiencing an asthma exacerbation, defined as any asthma-related event that required treatment with asthma medications beyond study medications: less than 1 percent

For morning PEF, the treatment effect was less than an MCID of 25 L/min and therefore imprecise. For evening PEF, which has no well-defined MCID, the treatment effect was statistically nonsignificant and therefore imprecise. Evidence was insufficient to support the use of one treatment over the other for these outcomes.

For asthma exacerbations, any reduction in severe exacerbations may be considered clinically significant.^{70, 135} Because the definition of “asthma exacerbation” used in this trial is broad, the severity of exacerbations observed is unclear. Further, the outcome measure reported patients rather than number of exacerbations; it is unclear whether exacerbations were in fact reduced. The effect is therefore considered imprecise and the evidence insufficient to support the use of one treatment over the other for this outcome.

Nasal Symptoms

One⁹⁰ of three trials^{90, 98, 130} (350 of 677 patients) assessed individual nasal symptoms at 2 weeks. Statistically significant improvements in all four symptoms (congestion, rhinorrhea, sneezing, and itch) with combination therapy were shown. Treatment effects on a 0-3 point scale ranged from 0.1 for nasal itch to 0.3 for congestion (from 3 percent to 10 percent of maximum score). This trial was rated poor quality due to inappropriate analysis of results (not intention to treat).

For individual nasal symptoms at 2 weeks, the risk of bias was rated as high based on the poor quality rating of the trial.⁹⁰ Consistency cannot be assessed with a single trial. Estimates of treatment effects were imprecise. Evidence was insufficient to support the use of one treatment over the other for this outcome.

All three trials^{90, 98, 130} assessed TNSS at 2 weeks (total N=677). All showed improvements in TNSS with combination therapy. In two trials, treatment effects reached statistical significance. One⁹⁸ of these was a fair quality trial of 300 patients (44 percent of total patients reporting this outcome) that showed a treatment effect of 90 using a 0-400 VAS (22 percent of maximum score). The other⁹⁰ was a poor quality trial in 350 patients (52 percent of patients reporting) that reported a treatment effect of 1.1 on a 0-12 point scale (9 percent of maximum score). The third trial¹³⁰ assessed TNSS at both 2 weeks and 4 weeks. This was a fair quality trial of 27 patients. Treatment effects on a 0-12 point scale were 1.2 at 2 weeks and 0.9 at 4 weeks (10 percent and 8 percent of maximum score, respectively). Both favored combination therapy, but neither was statistically significant.

For the outcome of TNSS at both 2 weeks and 4 weeks, evidence was insufficient to support the use of one treatment over the other. At 2 weeks, the risk of bias was rated as high. Fifty-two percent of patients reporting this outcome were in the poor quality trial, and neither of the other two trials were rated good quality. Results were consistent across trials but also imprecise. At 4 weeks, the risk of bias was rated as high based on the small size and fair quality rating of the trial. Consistency of results could not be assessed in a single trial, and the effect estimate was imprecise.

Eye Symptoms

Two^{90, 130} of three trials (377 of 677 patients) assessed eye symptoms. One trial⁹⁰ reported statistically significant improvements in individual symptoms of eye itching, tearing, and redness at 2 weeks with combination therapy. This was a trial of 350 patients that was rated poor quality due to inappropriate analysis of results (not intention to treat). Treatment effects were not reported. The other trial¹³⁰ assessed TOSS at 2 weeks and 4 weeks. The treatment effect at both time points was 0.2 on a 0-9 point scale (2 percent of maximum score). These were statistically nonsignificant effects that favored oral selective antihistamine monotherapy. This was a fair quality trial of 27 patients.

For individual eye symptoms at 2 weeks, the risk of bias was rated as high based on the poor quality rating of the trial.⁹⁰ Consistency of results could not be assessed in a single trial. Because the magnitude of the treatment effects is unknown, they are considered imprecise. Evidence was insufficient to support the use of one treatment over the other for this outcome.

For TOSS at 2 and 4 weeks, the evidence was insufficient to support the use of one treatment over the other. The risk of bias is medium based on the fair quality rating of the trial.¹³⁰ Consistency could not be assessed, and effect estimates at both time points were imprecise.

Quality of Life

One trial⁹⁸ assessed quality of life at 2 weeks using the RQLQ. This was a fair quality trial in 300 patients. A statistically significant treatment effect of 1.0 favoring combination therapy was shown. This exceeded the MCID of 0.5 points for the RQLQ. This trial⁹⁸ also reported a PGA of treatment response. A statistically significant treatment effect favoring combination therapy was reported (32-percentage point increase in the proportion of patients reporting moderate or significant improvement in the combination therapy arm using a 7-point Likert scale).

For quality of life at 2 weeks, the risk of bias was rated as medium based on the quality rating of the trial.⁹⁸ Although the effect estimate is precise, consistency cannot be assessed. The overall strength of evidence was low.

Nasal Symptoms

Evidence for the assessment of individual nasal symptoms comes from two^{62, 90} of five trials (407 of 1201 patients) that assessed nasal symptoms at 2 weeks, and one trial¹³² that assessed nasal symptoms at 6 weeks. All three trials were rated poor quality due to noncomparable groups at baseline¹³² and inappropriate analysis of results (not intention to treat^{62, 90}).

At 2 weeks, Anolik (2008)⁹⁰ showed no difference between treatments for congestion (treatment effect = 0), and Barnes (2006)⁶² showed a statistically significant treatment effect of 0.11 on a 0-3 point scale (4 percent of maximum score) favoring intranasal corticosteroid monotherapy. At 6 weeks, Di Lorenzo (2004)¹³² showed a statistically nonsignificant treatment effect of 0.04 on a 0-3 point scale (1 percent of maximum score).

Evidence for the outcome of congestion at 2 weeks was insufficient to support the use of one treatment over the other. Two poor quality trials^{62, 90} with high risk of bias reported inconsistent and imprecise treatment effects. For the outcome of congestion at 6 weeks, evidence also is insufficient to support the use of one treatment over the other. One poor quality trial¹³² with a high risk of bias reported an imprecise treatment effect.

At 2 weeks, Anolik (2008)⁹⁰ showed no difference between treatments for rhinorrhea (treatment effect = 0), and Barnes (2006)⁶² showed a statistically nonsignificant treatment effect of 0.04 on a 0-3 point scale (1 percent of maximum score) favoring combination therapy. At 6 weeks, Di Lorenzo (2004)¹³² also showed a statistically nonsignificant treatment effect of 0.04 on a 0-3 point scale (1 percent of maximum score), favoring intranasal corticosteroid monotherapy.

Evidence for the outcome of rhinorrhea at 2 weeks was insufficient to support the use of one treatment over the other. Two poor quality trials^{62, 90} with high risk of bias reported inconsistent and imprecise treatment effects. For the outcome of rhinorrhea at 6 weeks, evidence also is insufficient to support the use of one treatment over the other. One poor quality trial¹³² with a high risk of bias reported an imprecise treatment effect.

At 2 weeks, Anolik (2008)⁹⁰ and Barnes (2006)⁶² both showed greater improvements in sneezing with combination therapy than with intranasal corticosteroid monotherapy. Treatment effects were 0.1 and 0.15 on a 0-3 point scale (3 percent and 5 percent of maximum score, respectively). Neither was statistically significant. At 6 weeks, Di Lorenzo (2004)¹³² showed a statistically nonsignificant treatment effect of 0.08 on a 0-3 point scale (3 percent of maximum score) favoring combination therapy.

Evidence for the outcome of sneezing at 2 weeks was insufficient to support the use of one treatment over the other. Two poor quality trials^{62, 90} with high risk of bias reported consistent but imprecise treatment effects. For the outcome of rhinorrhea at 6 weeks, evidence also is insufficient to support the use of one treatment over the other. One poor quality trial¹³² with a high risk of bias reported an imprecise treatment effect.

At 2 weeks, Anolik (2008)⁹⁰ and Barnes (2006)⁶² both showed greater improvements in nasal itch with combination therapy than with intranasal corticosteroid monotherapy. Treatment effects were 0.1 and 0.03 on a 0-3 point scale (3 percent and 1 percent of maximum score, respectively). Neither was statistically significant. At 6 weeks, Di Lorenzo (2004)¹³² reported a statistically significant treatment effect of 0.1 on a 0-3 point scale (3 percent of maximum score) favoring combination therapy.

Evidence for the outcome of nasal itch at 2 weeks was insufficient to support the use of one treatment over the other. Two poor quality trials^{62, 90} with high risk of bias reported consistent but imprecise treatment effects. For the outcome of nasal itch at 6 weeks, evidence also is insufficient to support the use of one treatment over the other. One poor quality trial¹³² with a high risk of bias reported an imprecise treatment effect.

Three^{62, 90, 98} of five trials (707 of 1201 patients) assessed TNSS at 2 weeks. All three trials showed greater improvement in TNSS with combination therapy than with intranasal corticosteroid monotherapy. In two^{62, 98} of these, this finding was statistically significant. One⁹⁸ of these was a fair quality trial of 300 patients (42 percent of patients reporting this outcome). The other ⁶² was rated poor quality due to inappropriate analysis of results (not intention to treat). This trial⁶² reported results using a 0-12 point scale and showed a treatment effect of 0.11 (1 percent of maximum score). The fair quality trial⁹⁸ reported a treatment effect of 30 on a 0-400 VAS (8 percent of maximum score). The third trial⁹⁰ reported a statistically nonsignificant treatment effect of 0.3 on a 0-12 point scale (3 percent of maximum score). This was a poor quality trial of 345 patients (49 percent of patients reporting this outcome).

For the outcome of TNSS at 2 weeks, the risk of bias was rated as high. Fifty-eight percent of patients were in poor quality trials, and 42 percent were in a fair quality trial. Treatment effects consistently favored combination therapy in all three trials. However, treatment effects were imprecise. The evidence was therefore insufficient to support the use of one treatment over the other for this outcome.

Two trials^{131, 132} assessed TNSS at time points beyond 2 weeks (N=494). The larger of these¹³¹ (92 percent of patients reporting this outcome) was rated good quality, and the smaller¹³² (n=40) was rated poor quality due to noncomparable groups at baseline. At 6 weeks, the latter trial¹³² showed a statistically nonsignificant treatment effect of 0.2 on a 0-3 point scale (7 percent of maximum score) favoring combination therapy. At 8 weeks, the larger trial¹³¹ reported a treatment difference of zero.

For the outcome of TNSS at 6 to 8 weeks, evidence also is insufficient to support the use of one treatment over the other. The risk of bias was considered low; 92 percent of patients reporting this outcome were in the good quality trial. However, reported treatment effects were inconsistent and imprecise.

Eye Symptoms

Two trials assessed eye symptoms, one⁹⁰ at 2 weeks (N=345) and one¹³¹ at 8 weeks (N=454). At 2 weeks, Anolik (2008)⁹⁰ reported statistically nonsignificant treatment effects favoring combination therapy for itchy eyes, watery eyes, and red eyes. Effect sizes were not reported. This trial was rated poor quality due to inappropriate analysis of results (not intention to treat). At 8 weeks, Benincasa (1994)¹³¹ reported a statistically nonsignificant treatment effect favoring combination therapy for unspecified eye symptoms. Effect size was not reported. This trial¹³¹ was rated good quality.

Evidence for the outcome of eye symptoms at 2 weeks was insufficient to support the use of one treatment over the other. One trial⁹⁰ with a high risk of bias reported imprecise results. At 8 weeks, the evidence also is insufficient to support the use of one treatment over the other. One trial¹³¹ with a low risk of bias reported imprecise results.

Quality of Life

Two trials^{62, 98} assessed quality of life at 2 weeks using different measures. The larger trial⁹⁸ (83 percent of patients reporting this outcome) was rated fair quality and showed a treatment effect of 0.1 on the 0-6 point RQLQ scale favoring combination therapy. The smaller trial⁶² was rated poor quality and showed a treatment effect of 0.1 on the 0-6 point mini-RQLQ scale favoring intranasal corticosteroid monotherapy. Neither result was statistically significant, and neither exceeded the MCID. The larger trial⁹⁸ also assessed PGA of treatment. More patients treated with combination therapy reported moderate to significant improvement using a 7-point Likert scale (significantly worse to significantly improved) than patients treated with intranasal corticosteroid monotherapy. This result was not statistically significant.

Evidence for quality of life outcomes at 2 weeks is insufficient to support one treatment over the other. Two trials^{62, 98} with medium risk of bias reported inconsistent and imprecise results.

Nasal Symptoms

All five trials^{115, 117, 121} assessed four individual nasal symptoms and TNSS at 2 weeks (total N=2102). Four trials^{115, 121} (85 percent of patients reporting this outcome) were included in meta-analyses for each nasal outcome. Variance estimates necessary for pooling were not reported by Hampel (2010),¹¹⁷ preventing inclusion of this trial in the meta-analyses. All five trials were rated good quality. For each outcome, results were consistent across trials.

All five trials showed greater improvement in congestion with combination therapy than with intranasal corticosteroid monotherapy. In three trials, including Hampel (2010),¹¹⁷ treatment effects were statistically significant and ranged from 0.3 to 0.6 on a 0-6 point scale (from 5 percent to 10 percent of maximum score). The pooled effect was 0.16 on a 0-6 point scale (95 percent CI: 0.02 to 0.30), a statistically significant result favoring combination therapy (Figure 21). The larger bound of the 95 percent CI represented 5 percent of maximum score. Statistical heterogeneity was low (I²=17 percent, p=0.31).

Figure 21

Congestion at 2 weeks meta-analysis: combination intranasal corticosteroid plus nasal antihistamine versus intranasal corticosteroid.

For the outcome of congestion, the risk of bias was rated as low based on the quality of the trials. Statistical heterogeneity of a meta-analysis of four trials^{115, 121} was low, and the pooled effect was consistent with the effect reported in the one trial¹¹⁷ not included in the meta-analysis. The 95 percent CI for the pooled effect (0.02 to 0.30) fell within an interval bounded by −MCID and +MCID (-1.8 and +1.8 on the 0-6 point scale used). The Hampel (2010) trial¹¹⁷ reported a treatment effect of 0.38 on a 0-6 point scale (6 percent of maximum score) favoring combination therapy. If this trial were included in the meta-analysis, the pooled effect (0.16; 3 percent of maximum score) would increase. Because the trial represented only 15 percent of patients reporting this outcome, it is unlikely that the 95 percent CI of the pooled effect would include an MCID of 30 percent maximum score. The body of evidence supporting a conclusion of equivalence of combination therapy and intranasal corticosteroid for this outcome was therefore considered precise. The overall strength of evidence for this conclusion is high.

All five trials showed greater improvement in rhinorrhea with combination therapy than with intranasal corticosteroid monotherapy. The treatment effect (0.25 on a 0-6 point scale; 4 percent of maximum score) was statistically significant in only one trial (Carr, Trial 1¹¹⁵). The pooled effect was 0.14 on a 0-6 point scale (95 percent CI: 0.01 to 0.28), a statistically significant result favoring combination therapy (Figure 22). The larger bound of the 95 percent CI represented 5 percent of maximum score. Statistical heterogeneity was low (I²=0 percent, p=0.74).

Figure 22

Rhinorrhea at 2 weeks meta-analysis: combination intranasal corticosteroid plus nasal antihistamine versus intranasal corticosteroid.

For the outcome of rhinorrhea, the risk of bias was rated as low based on the quality of the trials. Statistical heterogeneity of a meta-analysis of four trials^{115, 121} was low, and the pooled effect was consistent with the effect reported in the one trial¹¹⁷ not included in the meta-analysis. The 95 percent CI for the pooled effect (0.01 to 0.28) fell within an interval bounded by −MCID and +MCID (-1.8 and +1.8 on the 0-6 point scale used). The Hampel (2010) trial¹¹⁷ reported a treatment effect of 0.27 on a 0-6 point scale (5 percent of maximum score) favoring combination therapy. If this trial were included in the meta-analysis, the pooled effect (0.14; 2 percent of maximum score) would increase. Because the trial represented only 15 percent of patients reporting this outcome, it is unlikely that the 95 percent CI of the pooled effect would include an MCID of 30 percent maximum score. The body of evidence supporting a conclusion of equivalence of combination therapy and intranasal corticosteroid for this outcome was therefore considered precise. The overall strength of evidence for this conclusion is high.

All five trials showed greater improvement in sneezing with combination therapy than with intranasal corticosteroid monotherapy. In four trials, including Hampel (2010),¹¹⁷ treatment effects were statistically significant and ranged from 0.2 to 0.6 on a 0-6 point scale (from 3 percent to 10 percent of maximum score). The pooled effect was 0.22 on a 0-6 point scale (95 percent CI: 0.07 to 0.36), a statistically significant result favoring combination therapy (Figure 23). The larger bound of the 95 percent CI represented 6 percent of maximum score. Statistical heterogeneity was low (I²=0 percent, p=0.50).

Figure 23

Sneezing at 2 weeks meta-analysis: combination intranasal corticosteroid plus nasal antihistamine versus intranasal corticosteroid.

For the outcome of sneezing, the risk of bias was rated as low based on the quality of the trials. Statistical heterogeneity of a meta-analysis of four trials^{115, 121} was low, and the pooled effect was consistent with the effect reported in the one trial¹¹⁷ not included in the meta-analysis. The 95 percent CI for the pooled effect (0.07 to 0.36) fell within an interval bounded by −MCID and +MCID (-1.8 and +1.8 on the 0-6 point scale used). The Hampel (2010) trial¹¹⁷ reported a treatment effect of 0.49 on a 0-6 point scale (8 percent of maximum score) favoring combination therapy. If this trial were included in the meta-analysis, the pooled effect (0.22; 4 percent of maximum score) would increase. Because the trial represented only 15 percent of patients reporting this outcome, it is unlikely that the 95 percent CI of the pooled effect would include an MCID of 30 percent maximum score. The body of evidence supporting a conclusion of equivalence of combination therapy and intranasal corticosteroid for this outcome was therefore considered precise. The overall strength of evidence for this conclusion is high.

All five trials showed greater improvement in nasal itch with combination therapy than with intranasal corticosteroid monotherapy. In two trials, one of which was Hampel (2010),¹¹⁷ treatment effects of 0.31 and 0.6 on a 0-6 point scale (5 percent and 10 percent of maximum score, respectively) were statistically significant. The pooled effect was 0.10 on a 0-6 point scale (95 percent CI: -0.03 to 0.23), a statistically nonsignificant result favoring combination therapy (Figure 24). The larger bound of the 95 percent CI represented 4 percent of maximum score. Statistical heterogeneity was low (I²=0 percent; p=0.39).

Figure 24

Nasal itch at 2 weeks meta-analysis: combination intranasal corticosteroid plus nasal antihistamine versus intranasal corticosteroid.

For the outcome of nasal itch, the risk of bias was rated as low based on the quality of the trials. Statistical heterogeneity of a meta-analysis of four trials^{115, 121} was low, and the pooled effect was consistent with the effect reported in the one trial¹¹⁷ not included in the meta-analysis. The 95 percent CI for the pooled effect (-0.03 to 0.23) fell within an interval bounded by −MCID and +MCID (-1.8 and +1.8 on the 0-6 point scale used). The Hampel (2010) trial¹¹⁷ reported a treatment effect of 0.31 on a 0-6 point scale (5 percent of maximum score) favoring combination therapy. If this trial were included in the meta-analysis, the pooled effect (0.10; 2 percent of maximum score) would increase. Because the trial represented only 15 percent of patients reporting this outcome, it is unlikely that the 95 percent CI of the pooled effect would include an MCID of 30 percent maximum score. The body of evidence supporting a conclusion of equivalence of combination therapy and intranasal corticosteroid for this outcome was therefore considered precise. The overall strength of evidence for this conclusion is high.

All five trials showed statistically significant improvements in TNSS with combination therapy. Treatment effects ranged from 0.6 to 2.2 on a 0-24 point scale (from 3 percent to 9 percent of maximum score). The pooled effect was 0.61 on a 0-24 point scale (95 percent CI: 0.15 to 1.08), a statistically significant result favoring combination therapy (Figure 25). The larger bound of the 95 percent CI represented 5 percent of maximum score. Statistical heterogeneity was low (I²=0 percent, p=0.46).

Figure 25

Total nasal symptom score at 2 weeks meta-analysis: combination intranasal corticosteroid plus nasal antihistamine versus intranasal corticosteroid.

For TNSS, the risk of bias was rated as low based on the quality of the trials. Effect estimates were precise. Statistical heterogeneity of a meta-analysis of four trials^{115, 121} was low, and the pooled effect was consistent with the effect reported in the one trial¹¹⁷ not included in the meta-analysis. The 95 percent CI for the pooled effect (0.15 to 1.08) fell within an interval bounded by −MCID and +MCID (-7.2 and +7.2 on the 0-24 point scale used). The Hampel (2010) trial¹¹⁷ reported a treatment effect of 1.47 on a 0-24 point scale (6 percent of maximum score) favoring combination therapy. If this trial were included in the meta-analysis, the pooled effect (0.61; 3 percent of maximum score) would increase. Because the trial represented only 15 percent of patients reporting this outcome, it is unlikely that the 95 percent CI of the pooled effect would include an MCID of 30 percent maximum score. The body of evidence supporting a conclusion of equivalence of combination therapy and intranasal corticosteroid for this outcome was therefore considered precise. The overall strength of evidence for this conclusion is high.

Eye Symptoms

Four^{115, 117} trials that assessed eye symptoms at 2 weeks (total N=2000) showed greater improvements in TOSS with combination therapy than with intranasal corticosteroid monotherapy. In two trials, one of which was Hampel (2010),¹¹⁷ treatment effects of 0.45 and 0.88 on a 0-18 point scale (3 percent and 5 percent of maximum score, respectively) were statistically significant. The pooled effect from a meta-analysis of three trials¹¹⁵ (85 percent of patients reporting this outcome; Hampel [2010]¹¹⁷ excluded) was 0.48 on a 0-18 point scale (95 percent CI: 0.07 to 0.90), a statistically significant result favoring combination therapy (Figure 26). The larger bound of the 95 percent CI represented 5 percent of maximum score. Statistical heterogeneity was low (I²=21%, p=0.28).

Figure 26

Total ocular symptom score at 2 weeks meta-analysis: combination intranasal corticosteroid plus nasal antihistamine versus intranasal corticosteroid.

For TOSS, the risk of bias was rated as low based on the quality of the trials. Statistical heterogeneity of a meta-analysis of three trials¹¹⁵ was low, and the pooled effect was consistent with the effect reported in the one trial¹¹⁷ not included in the meta-analysis. The 95 percent CI for the pooled effect (0.07 to 0.90) fell within an interval bounded by −MCID and +MCID (-5.4 and +5.4 on the 0-18 point scale used). The Hampel (2010) trial¹¹⁷ reported a treatment effect of 0.45 on a 0-18 point scale (3 percent of maximum score) favoring combination therapy. If this trial were included in the meta-analysis, the pooled effect (0.48; 3 percent of maximum score) would decrease slightly. Because the trial represented only 15 percent of patients reporting this outcome, it is unlikely that the 95 percent CI of the pooled effect would include an MCID of 30 percent maximum score. The body of evidence supporting a conclusion of equivalence of combination therapy and intranasal corticosteroid for this outcome was therefore considered precise. The overall strength of evidence for this conclusion is high.

Quality of Life

Both trials^{117, 121} that assessed quality of life (total N=408) showed greater improvement in RQLQ scores with combination therapy than with intranasal corticosteroid monotherapy. Treatment effects were 0.17 and 0.45, neither of which was statistically significant. Treatment effects did not exceed the MCID of 0.5 points.

For RQLQ, the risk of bias was rated low based on the quality of the trials. Results were consistent across trials, but effects were statistically and clinically nonsignificant, that is, imprecise. The evidence was insufficient to support the use of one treatment over the other for this outcome.

Nasal Symptoms

All five trials^{115, 117, 121} assessed four individual nasal symptoms and TNSS at 2 weeks (total N=2101). Four trials^{115, 121} (85 percent of patients reporting this outcome) were included in meta-analyses for each nasal outcome. Variance estimates necessary for pooling were not reported by Hampel (2010),¹¹⁷ preventing inclusion of this trial in the meta-analyses. All five trials were rated good quality.

All five trials showed statistically significant improvements in congestion with combination therapy compared to nasal antihistamine monotherapy. Treatment effects ranged from 0.2 to 0.6 on a 0-6 point scale (from 3 percent to 10 percent of maximum score). The pooled effect was 0.28 on a 0-6 point scale (95 percent CI: 0.16 to 0.41), a statistically significant result favoring combination therapy (Figure 27). The larger bound of the 95 percent CI represented 7 percent of maximum score. Statistical heterogeneity was low (I²=0 percent, p=0.41).

Figure 27

Congestion at 2 weeks: meta-analysis of 4 trials–combination intranasal corticosteroid plus nasal antihistamine versus nasal antihistamine.

For the outcome of congestion, the risk of bias was rated as low based on the quality of the trials. Statistical heterogeneity of a meta-analysis of four trials^{115, 121} was low, and the pooled effect was consistent with the effect reported in the one trial¹¹⁷ not included in the meta-analysis. The 95 percent CI for the pooled effect (0.16 to 0.41) fell within an interval bounded by −MCID and +MCID (-1.8 and +1.8 on the 0-6 point scale used). The Hampel (2010) trial¹¹⁷ reported a treatment effect of 0.49 on a 0-6 point scale (8 percent of maximum score) favoring combination therapy. If this trial were included in the meta-analysis, the pooled effect (0.28; 5 percent of maximum score) would increase. Because the trial represented only 15 percent of patients reporting this outcome, it is unlikely that the 95 percent CI of the pooled effect would include an MCID of 30 percent maximum score. The body of evidence supporting a conclusion of equivalence of combination therapy and nasal antihistamine for this outcome was therefore considered precise. The overall strength of evidence for this conclusion is high.

All five trials showed greater improvement in rhinorrhea with combination therapy than with nasal antihistamine monotherapy. In four trials,^{115, 117, 121} including Hampel (2010)¹¹⁷ whose results were not pooled, treatment effects were statistically significant and ranged from 0.2 to 0.6 on a 0-6 point scale (3 percent to 10 percent of maximum score). The pooled effect was 0.31 on a 0-6 point scale (95 percent CI: 0.18 to 0.45), a statistically significant result favoring combination therapy (Figure 28). The larger bound of the 95 percent CI represented 8 percent of maximum score. Statistical heterogeneity was low (I²=1 percent, p=0.39).

Figure 28

Rhinorrhea at 2 weeks: meta-analysis of 4 trials–combination intranasal corticosteroid plus nasal antihistamine versus nasal antihistamine.

For the outcome of rhinorrhea, the risk of bias was rated as low based on the quality of the trials. Statistical heterogeneity of a meta-analysis of four trials^{115, 121} was low, and the pooled effect was consistent with the effect reported in the one trial¹¹⁷ not included in the meta-analysis. The 95 percent CI for the pooled effect (0.18 to 0.45) fell within an interval bounded by −MCID and +MCID (-1.8 and +1.8 on the 0-6 point scale used). The Hampel (2010) trial¹¹⁷ reported a treatment effect of 0.55 on a 0-6 point scale (9 percent of maximum score) favoring combination therapy. If this trial were included in the meta-analysis, the pooled effect (0.31; 5 percent of maximum score) would increase. Because the trial represented only 15 percent of patients reporting this outcome, it is unlikely that the 95 percent CI of the pooled effect would include an MCID of 30 percent maximum score. The body of evidence supporting a conclusion of equivalence of combination therapy and nasal antihistamine for this outcome was therefore considered precise. The overall strength of evidence for this conclusion is high.

All five trials showed greater improvement in sneezing with combination therapy than with nasal antihistamine monotherapy. In four trials, including Hampel (2010),¹¹⁷ treatment effects were statistically significant and ranged from 0.2 to 0.61 on a 0-6 point scale (from 3 percent to 10 percent of maximum score). The pooled effect was 0.34 on a 0-6 point scale (95 percent CI: 0.20 to 0.48), a statistically significant result favoring combination therapy (Figure 29). The larger bound of the 95 percent CI represented 6 percent of maximum score. Statistical heterogeneity was low (I²=0 percent, p=0.74).

Figure 29

Sneezing at 2 weeks: meta-analysis of 4 trials–combination intranasal corticosteroid plus nasal antihistamine versus nasal antihistamine.

For the outcome of sneezing, the risk of bias was rated as low based on the quality of the trials. Statistical heterogeneity of a meta-analysis of four trials^{115, 121} was low, and the pooled effect was consistent with the effect reported in the one trial¹¹⁷ not included in the meta-analysis. The 95 percent CI for the pooled effect (0.20 to 0.48) fell within an interval bounded by −MCID and +MCID (-1.8 and +1.8 on the 0-6 point scale used). The Hampel (2010) trial¹¹⁷ reported a treatment effect of 0.61 on a 0-6 point scale (10 percent of maximum score) favoring combination therapy. If this trial were included in the meta-analysis, the pooled effect (0.34; 6 percent of maximum score) would increase. Because the trial represented only 15 percent of patients reporting this outcome, it is unlikely that the 95 percent CI of the pooled effect would include an MCID of 30 percent maximum score. The body of evidence supporting a conclusion of equivalence of combination therapy and nasal antihistamine for this outcome was therefore considered precise. The overall strength of evidence for this conclusion is high.

All five trials showed greater improvement in nasal itch with combination therapy than with nasal antihistamine monotherapy. In three trials, including Hampel (2010),¹¹⁷ treatment effects were statistically significant and ranged from 0.3 to 0.8 on a 0-6 point scale (from 5 percent to 13 percent of maximum score). The pooled effect was 0.30 on a 0-6 point scale (95 percent CI: 0.12 to 0.48), a statistically significant result favoring combination therapy (Figure 30). The larger bound of the 95 percent CI represented 8 percent of maximum score. Statistical heterogeneity was low to moderate (I²=34%) but not statistically significant (p=0.21). The 95 percent CI of one of the trials (21 percent of the pooled sample) included zero (-0.48 to 0.08).

Figure 30

Nasal itch at 2 weeks: meta-analysis of 4 trials–combination intranasal corticosteroid plus nasal antihistamine versus nasal antihistamine.

For the outcome of nasal itch, the risk of bias was rated as low based on the quality of the trials. Statistical heterogeneity of a meta-analysis of four trials^{115, 121} was low to moderate, and the pooled effect was consistent with the effect reported in the one trial¹¹⁷ not included in the meta-analysis. The 95 percent CI for the pooled effect (0.12 to 0.48) fell within an interval bounded by −MCID and +MCID (-1.8 and +1.8 on the 0-6 point scale used). The Hampel (2010) trial¹¹⁷ reported a treatment effect of 0.40 on a 0-6 point scale (7 percent of maximum score) favoring combination therapy. If this trial were included in the meta-analysis, the pooled effect (0.30; 5 percent of maximum score) would increase. Because the trial represented only 15 percent of patients reporting this outcome, it is unlikely that the 95 percent CI of the pooled effect would include an MCID of 30 percent maximum score. The body of evidence supporting a conclusion of equivalence of combination therapy and nasal antihistamine for this outcome was therefore considered precise. The overall strength of evidence for this conclusion is high.

All five trials showed statistically significant improvements in TNSS with combination therapy. Treatment effects ranged from 0.7 to 2.6 on a 0-24 point scale (from 3 percent to 11 percent of maximum score). The pooled effect was 1.28 on a 0-24 point scale (95 percent CI: 0.82 to 1.74), a statistically significant result favoring combination therapy (Figure 31). The larger bound of the 95 percent CI represented 7 percent of maximum score. Statistical heterogeneity was low (I²=0 percent, p=0.54).

Figure 31

Total nasal symptom score at 2 weeks: meta-analysis of 4 trials–combination intranasal corticosteroid plus nasal antihistamine versus nasal antihistamine.

For TNSS, the risk of bias was rated as low based on the quality of the trials. Statistical heterogeneity of a meta-analysis of four trials^{115, 121} was low, and the pooled effect was consistent with the effect reported in the one trial¹¹⁷ not included in the meta-analysis. The 95 percent CI for the pooled effect (0.82 to 1.74) fell within an interval bounded by −MCID and +MCID (-7.2 and +7.2 on the 0-24 point scale used). The Hampel (2010) trial¹¹⁷ reported a treatment effect of 2.06 on a 0-24 point scale (9 percent of maximum score) favoring combination therapy. If this trial were included in the meta-analysis, the pooled effect (1.28; 5 percent of maximum score) would increase. Because the trial represented only 15 percent of patients reporting this outcome, it is unlikely that the 95 percent CI of the pooled effect would include an MCID of 30 percent maximum score. The body of evidence supporting a conclusion of equivalence of combination therapy and nasal antihistamine for this outcome was therefore considered precise. The overall strength of evidence for this conclusion is high.

Eye Symptoms

Four^{115, 117} trials that assessed eye symptoms at 2 weeks (total N=2000) showed greater improvements in TOSS with combination therapy than with nasal antihistamine monotherapy. Treatment effects ranged from 0.03 to 0.71 on a 0-18 point scale (from less than 1 percent to 4 percent of maximum score), but effects were either statistically nonsignificant or statistical significance was not reported. The pooled effect from a meta-analysis of three trials (85 percent of patients reporting this outcome; Hampel [2010]¹¹⁷ excluded) was 0.25 on a 0-18 point scale (95 percent CI: -0.12 to 0.61), a statistically nonsignificant result favoring combination therapy (Figure 32). The larger bound of the 95 percent CI represented 3 percent of maximum score. Statistical heterogeneity was low (I²=0 percent, p=0.37).

Figure 32

Total ocular symptom score at 2 weeks: meta-analysis of 4 trials–combination intranasal corticosteroid plus nasal antihistamine versus nasal antihistamine.

For TOSS, the risk of bias was rated as low based on the quality of the trials. Statistical heterogeneity of a meta-analysis of three trials¹¹⁵ was low, and the pooled effect was consistent with the effect reported in the one trial¹¹⁷ not included in the meta-analysis. The 95 percent CI for the pooled effect (-0.12 to 0.61) fell within an interval bounded by −MCID and +MCID (-5.4 and +5.4 on the 0-18 point scale used). The Hampel (2010) trial¹¹⁷ reported a treatment effect of 0.71 on a 0-18 point scale (4 percent of maximum score) favoring combination therapy. If this trial were included in the meta-analysis, the pooled effect (0.25; 1 percent of maximum score) would increase. Because the trial represented only 15 percent of patients reporting this outcome, it is unlikely that the 95 percent CI of the pooled effect would include an MCID of 30 percent maximum score. The body of evidence supporting a conclusion of equivalence of combination therapy and nasal antihistamine for this outcome was therefore considered precise. The overall strength of evidence for this conclusion is high.

Quality of Life

Both trials^{117, 121} that assessed quality of life showed statistically significant improvement in RQLQ scores with combination therapy. The larger trial¹¹⁷ (n=459; 75 percent of patients reporting this outcome) showed a treatment effect of 0.43. The smaller trial¹²¹ showed a treatment effect of 0.71. The latter result exceeds the MCID for the RQLQ of 0.5 points.

For the outcome of quality of life, the risk of bias was rated as low based on the quality of the trials. Effect estimates were consistent across trials but not precise. Evidence to support the use of one treatment over the other for this outcome is insufficient.

Nasal Symptoms

All seven trials^101-107 assessed congestion at 2 weeks (total N=3575). All seven showed statistically significant improvements in nasal congestion with combination therapy. Three^{101, 103, 107} were good quality trials of 1329 patients total (37 percent of patients reporting this outcome). Two^{101, 103} showed treatment effects of 0.2 and 0.25 on a 0-3 point scale (7 percent and 8 percent of maximum score, respectively). One fair quality trial¹⁰⁶ (n=676, 19 percent of patients reporting) showed a treatment effect of 0.2 on a 0-3 point scale (7 percent of maximum score). Three trials^{102, 104, 105} were rated poor quality due to inappropriate analysis of results (not intention to treat). Treatment effects reported by these trials ranged from 0.16 to 0.27 on a 0-3 point scale (from 5 percent to 9 percent of maximum score).

For the outcome of nasal congestion at 2 weeks, the risk of bias was assessed as medium. Forty-four percent of patients were in poor quality trials, and 37 percent were in good quality trials. Treatment effects consistently favored combination therapy in all trials. Although statistically significant, no treatment effect exceeded an MCID of 30 percent maximum score. The body of evidence was therefore considered imprecise. Evidence was insufficient to support the use of one treatment over the other for the treatment of congestion.

Two^{103, 107} of seven trials assessed rhinorrhea at 2 weeks (total N=891). Both trials were large (approximately 450 patients in each), and both were rated good quality. Both favored combination therapy over oral selective antihistamine monotherapy for this outcome. Treatment effects were 0.1 and 0.13 on a 0-3 point scale (3 percent and 4 percent of maximum score, respectively); the latter was statistically significant.

For the outcome of rhinorrhea at 2 weeks, the risk of bias was assessed as low based on the quality of the trials. Treatment effects were consistent but imprecise. The evidence was insufficient to support the use of one treatment over the other for this outcome.

Two^{103, 107} of seven trials assessed sneezing at 2 weeks (total N=891). Both trials were large (approximately 450 patients in each), and both were rated good quality. Both favored combination therapy over oral selective antihistamine monotherapy. Treatment effects were 0.08 and 0.1 on a 0-3 point scale (both 3 percent of maximum score); the former was statistically significant.

For the outcome of sneezing at 2 weeks, the risk of bias was assessed as low based on the quality of the trials. Treatment effects were consistent but imprecise. The evidence was insufficient to support the use of one treatment over the other for this outcome.

One good quality trial¹⁰³ assessed nasal itch at 2 weeks (N=458). The treatment effect (0.1 on a 0-3 point scale; 3 percent of maximum score) favored combination therapy and was statistically significant.

For the outcome of nasal itch at 2 weeks, the risk of bias was rated as low based on the quality of the trial. Consistency of results could not be assessed in a single trial, and the effect estimate was imprecise. The evidence was insufficient to support the use of one treatment over the other for this outcome.

One good quality trial¹⁰¹ assessed TNSS at 2 weeks (N=438). The treatment effect (0.6 on a 0-3 point scale; 20 percent of maximum score) favored combination therapy and was statistically nonsignificant.

For TNSS at two weeks, the risk of bias was rated as low based on the quality of the trial. Consistency could not be assessed in a single trial, and the effect estimate was imprecise. Evidence was insufficient to support the use of one treatment over the other for this outcome.

Eye Symptoms

Two good quality trials^{103, 107} assessed eye symptoms at 2 weeks (total N=891). One trial¹⁰³ assessed ocular itching, and the other¹⁰⁷ assessed TOSS comprising ocular itching, tearing, and redness. The treatment effect for ocular itch¹⁰³ was 0.01 on a 0-3 point scale (less than 1 percent of maximum score), a statistically nonsignificant result that favored oral selective antihistamine monotherapy. The treatment effect for TOSS¹⁰⁷ was 0.1 on a 0-4 point scale (3 percent of maximum score), favoring combination therapy. Statistical significance was not reported.

For eye symptoms at 2 weeks, the risk of bias was low based on the quality of the trials. Treatment effect estimates were inconsistent and imprecise. Evidence was insufficient to support the use of one treatment over the other for this outcome.

Nasal Symptoms

One¹³⁴ of two trials (N=126) assessed nasal congestion and sneezing at 2 weeks. For nasal congestion, there was a statistically nonsignificant treatment effect of 0.1 on a 0-3 point scale (3 percent of maximum score) that favored nonselective antihistamine. For sneezing, no treatment difference was reported. The trial was rated poor quality due to lack of blinding; therefore, risk of bias was high. Treatment effects for both nasal symptoms were imprecise. The evidence was insufficient to support the use of one treatment over the other for either outcome.

Eye Symptoms

One¹³⁴ of two trials (N=126) assessed ocular itching and tearing. For both outcomes, treatment effects were 0.1 on a 0-3 point scale (3 percent of maximum score). Both favored nonselective antihistamine, but neither was statistically significant. The trial was rated poor quality due to lack of blinding; therefore, risk of bias was high. Treatment effects for both ocular symptoms were imprecise. The evidence was insufficient to support the use of one treatment over the other for either outcome.