Recommendation 1

Primary care practitioners should take the opportunity, whenever possible, to identify inactive adults and advise them to aim for 30 minutes of moderate activity on 5 days of the week (or more)^*. They should use their judgement to determine when this would be inappropriate (for example, because of medical conditions or personal circumstances). They should use a validated tool, such as the Department of Health’s forthcoming general practitioner physical activity questionnaire (GPPAQ), to identify inactive individuals.

^* The practitioner may be a GP or another professional with specific responsibility for providing encouragement or advice. This will depend on local conditions, professional interest and resources. Health trainers are likely to have a role in offering brief advice. ‘Inactive’ is used as shorthand for those failing to reach the CMO’s recommendation. ‘Advise’ is used as shorthand for ‘encourage, advise, discuss, negotiate’.

Recommendation 2

When providing physical activity advice, primary care practitioners should take into account the individual’s needs, preferences and circumstances. They should agree goals with them. They should also provide written information about the benefits of activity and the local opportunities to be active. They should follow them up at appropriate intervals over a 3 to 6 month period.

The NICE public health intervention advisory committee determined that there was insufficient evidence to recommend the use of exercise referral schemes to promote physical activity, other than as part of research studies where their effectiveness can be evaluated.

This guidance aims to help practitioners deliver effective interventions that will increase people’s physical activity levels and therefore benefit their health.

The use of physical activity as part of a non-pharmacological therapy for IBS is described as “reasonable” despite the relationship between exercise and gastrointestinal system being unclear (Bi and Triadafilopoulos 2003). For example, moderate physical activity (e.g. brisk walking) is reported to improve gut transit time, whereas vigorous physical activity (e.g. running) can result in “runners trots” (Oettle, 1991). Physical activity has been associated with improved outcomes in uncontrolled studies (Colwell et al, 1998).

Types of studies

For intervention studies, randomised trials (RCTs) examining the use of physical activity for the treatment or management of IBS were to be included. In the absence of randomised trials, quasi randomised studies were to be considered. Crossover trials with a washout period of less than 2 weeks were to be excluded. All study designs were to be included for adverse effects, but specific searches for adverse effects will not be carried out. Studies were restricted to the English language.

Types of intervention

Studies were included if they had one or more of the following interventions:

• The use of physical activity alone or in combination with other therapies
• 12 weeks minimum length of intervention

A physical activity intervention is defined as the use of physical activity or exercise as a therapeutic and/or preventative medical procedure used to support the management and treatment of IBS. Physical activity is usually defined as any force exerted by skeletal muscles that results in energy expenditure above resting level whereas exercise is defined as a subset of physical activity, which is volitional, planned, structured, repetitive and aimed at improvement or maintenance of any aspects of fitness or health (Casperson et al, 1985). The GDG defined the minimum acceptable dose of physical activity to be at least 30 minutes per week of at least moderate intensity physical activity. The duration of the intervention is also considered important, and the minimum duration of intervention was set at twelve weeks.

Types of comparisons

The following comparisons were to be included

• Physical activity versus attention control
• Combination of physical activity with another non-pharmacological intervention (e.g. diet advice) versus control.

Types of participants

Studies were to be included if the participants were:

• Adults (18 years and over)
• Had symptoms of IBS
• No serious diseases (e.g. cancer, heart disease) other than IBS
• Did not have a single symptom of IBS only (e.g. not constipation only)

In the absence of studies in patients with IBS, we extended the review to cover studies in people with single symptoms such as constipation or diarrhoea. Studies in these participants were regarded as indirect as far as the population was concerned.

Subgroup analyses

Subgroup analyses were proposed to examine any heterogeneity as follows:

• Dose
• Type of physical activity
• Symptom severity.

Sensitivity analyses

The following sensitivity analyses may be considered:

• Setting (primary/secondary care).

Evidence from Systematic Reviews

Bi and Triadafilopoulos (2003) reviewed the relationship between exercise and gastrointestinal function for eight disease types. The authors described their review as systematic but provided no methods in the paper.

1. Gastroesophageal reflux disease
2. Gastric emptying and gastric acid production
3. Peptic ulcer disease
4. Inflammatory bowel disease
5. Constipation and gastrointestinal motility
6. Colorectal cancer
7. Gastrointestinal bleeding
8. Liver disease

The authors attempted to identify if there were any differential effects of physical activity (by intensity or type) on gastrointestinal function. IBS was not identified as a separate class, but it may be possible to extrapolate from the indirect evidence in section 5 of the Bi and Triadafilopoulos (2003) review. Participants in these studies tended to be young, fit and active males, rather than typical clinical populations. The review found that physical activity could improve gastric emptying and lower the risk of bowel cancer. However, there was insufficient evidence to suggest that exercise can relieve chronic constipation. The authors also noted consistent improvements in aerobic fitness and general health for all subjects participating in regular physical activity programmes and that this outcome is a notable behavioural goal for sedentary patients. The majority of risks to gastrointestinal organs relate to very high levels of sustained physical activity (performed at elite levels). However these risks do not outweigh the benefits of light and moderate physical activity.

Evidence from intervention studies

One randomised trial was included as an indirect study as it examined the impact of physical activity upon adults with chronic constipation only (De Schryver 2005). However this study was not included in the Bi and Triadafilopoulos (2003) review.

The aim of the study was to investigate the effect of regular physical activity on colonic transit time and defecation in middle aged inactive patients suffering from chronic idiopathic constipation. Forty three adults aged 51 to 61 were recruited from general practice lists and pharmacies. Using Rome I criteria all were categorised as suffering from constipation, with IBS patients excluded. Participants’ physical activity levels were also assessed using a self report measure and all the participants were categorised as sedentary if they failed to reach the current physical activity recommendation (under 30 minutes or more of moderate physical activity on most days of the week).

Other baseline measures included food consumption, assessed by self report using a diary, in order to determine the average daily fibre and water intake. Defecations patterns were recorded in a 7-day diary, at the start of the study and at 12 weeks follow up. Colonic transit time was measured using radiopaque markers and x-rays. Transit time was calculated based on the number of markers visible in the colon, segmented into three (right colon, left colon and rectosigmoid).

Participants were randomised in to two groups, physical activity versus waiting list control. Group A maintained their normal lifestyle for 12 weeks, and then started their 12 week physical activity programme. Group B started their physical activity programme immediately after randomisation. Both groups were given dietary advice by a dietician concerning the consumption of fluid and fibre at the start of the study. Group A received a second dietary advice after 12 weeks, before starting the physical activity programme. This programme consisted of both aerobic and strength/flexibility exercises. Brisk walking was chosen for aerobic training and strength/ flexibility exercises were chosen for a home based programme. Brisk walking was performed at least twice a week for at least 30 minutes per session, performed at 70 to 80% of the subject’s maximal heart rate. Participants were able to monitor their heart rates using a Polar sports tester (a heart rate monitor worn on the wrist, in conjunction with a chest sensor). Maximal heart rate was assessed for all patients at baseline using a maximal heart test performed on a cycle ergometer. Participants were also asked to perform a walking test on a treadmill at 70% of their heart rate for 5 minutes to establish an average heart rate for their brisk walking.

The number of defecations did not change in either of the study groups (Table 1). However in Group B the percentage of incomplete stools decreased significantly, compared to Group A at 12 weeks (Group A from 58.8% to 39.5% whereas in Group B from 54.3% to 27.4%).

Table 1

Defecation patterns at baseline and after 12 week physical activity programme for 41 adult participants aged 51–61 years old (De Schryver et al, 2005).

Despite randomisation, there were considerable differences between right and total colonic transit times at baseline between groups (Table 2). No significant changes in right or left colonic transit time were observed in either group at the end of the physical activity programme. In Group B there was an observed acceleration in rectosigmoid mean transit time compared to Group A. Total colonic transit time also improved with a significant reduction in Group B. The authors reported that there was no correlation between fibre intake and improvements in defecation patterns and colonic transit times.

Table 2

Colonic transit times (hours) at baseline and after 12 week physical activity programme for 41 adult participants aged 51–61 years old (De Schryver et al, 2005).

The GDG noted that the normal total colonic transit time is 72 hours and concluded that group B was significantly different from group A, so that the study was considered to be at least partially confounded.

The evidence from this study was assessed to be low, using GRADE criteria. Limitations included (i) the study was conducted in secondary care (ii) there were important differences in baseline characteristics, (iii) IBS patients were excluded. This study was also limited because the participants had relatively high levels of baseline physical activity, which equates to over 2 hours of walking per week, and may not be representative. The study did show that moderate physical activity could deliver a consistent reduction in total colonic transit time and improvements in ROME I symptoms amongst older adults with chronic constipation.

Studies used to aid GDG discussions

One pre-post intervention study examined the impact of a lifestyle education programme upon IBS symptoms. This study design was judged inadequate to make recommendations on interventions, but was considered useful to inform GDG discussions, and does illustrate a suitable approach for evaluating a lifestyle intervention for IBS patients.

Colwell et al (1998) assessed the impact at one and six months of a patient education class, that included exercise, on 52 adult outpatients with IBS (definition of IBS not stated). Patients were advised to increase their physical activity by walking or basic stretching exercises during one 3 hour structured class, delivered by a specialist nurse. Pre-class data was compared with results for physical activity levels at follow up. Exercise scores increased significantly at one month but not at 6 months, compared with baseline, using a self-rating scale. It is difficult to assess if this increase was clinically significant because the physical activity variable was assessed using a categorical scale, and so the physical activity change scores were not adjusted for baseline values. Pain scores at 1 and 6 months reduced significantly (see Table 3). The Manning score also decreased significantly, on a scale of 0 to 6 using Manning criteria.

Table 3

Symptom scores at 1 and 6 months for 57 adult participants in an IBS educational class aged 21–79 years old (Colwell et al, 1998, page 903).

Evidence from Epidemiological studies

Three observational studies reported the prevalence and association between IBS and physical activity. In a case-control study, Kim and Ban (2006) reported a small, non-significant difference in the mean number of hours of exercise per week for students with IBS compared to students without IBS (defined by ROME 2 criteria) (students with IBS 2.38 h/week, SD 3.2 versus students with non-IBS: 2.69 h/week, SD 6.3).

Figure 1

Lustyk et al (2001) compared prevalence and severity of IBS symptoms between active and sedentary women with IBS (defined by ROME I criteria). They found that active and inactive women reported the same level of recalled psychological and somatic symptoms as well as daily reports of GI and psychological distress. Active women (those who took at least 2.5 hours per week of moderate physical activity and meeting recommended physical activity levels) reported significantly less fatigue than sedentary women. This outcome was assessed by combining frequency and severity of fatigue using categorical scale. No differences were observed for other somatic symptoms including backache, headache and insomnia between active and sedentary women with IBS.

Dancey et al (2002) examined gender differences in the prevalence and severity of IBS symptoms. They used a cross sectional survey to compare the prevalence and impact of IBS symptoms between 117 male and female IBS patients. IBS was assessed by self report measure with respondents rating severity of abdominal pain, constipation, diarrhoea, incomplete evacuation after bowel movement, bloating and flatulence on a 7 point severity scale (0 = no symptoms to 7 = extremely severe). Illness intrusiveness ratings were assessed across 13 life domains, using a 7 category Likert scale. Respondents were asked to rate the extent to which their illness interfered with each life domain important to quality of life (e.g. health, diet, financial situation, relationship with partner etc.). One life domain was active recreation (e.g. sports). The authors reported that in response to this item, men and women scored the interference of IBS similarly (i.e. moderate interference), with no significant differences between genders. They found that IBS inference was higher in diet, health and self expression domains than active recreation. Other domains reporting less interference than active recreation were social relations, work, community/civic life, sex life, relationship with spouse, family relations, financial situation, passive recreation and religious expression.

Two studies examined the relationship between physical activity and bowel frequency in the general population. In a cohort study, Sanjoaquin et al (2004) investigated the association between mean number of bowel movements and physical activity, adjusted for other confounding variables (e.g. age, BMI, diet, fibre intake) in 20,630 EPIC-Oxford cohort participants. The EPIC-Oxford cohort is a cohort study forming part of the European Prospective Investigation into Cancer and Nutrition (EPIC). Participants were recruited from general practice surgeries, vegetarian and health food magazines, the Vegetarian Society, the Vegan Society and from friends and relatives of participants. In a follow up study a short questionnaire was sent to all participants and included two questions relating to bowel movements, (i) “About how many bowel movements do you have each week? And (ii) How often do you take laxatives?” The number of bowel movements was counted for each participant. Respondents were then dichotomised into one of two groups, either above of below 7 movements per week.

The authors reported a positive association between increasing amounts of vigorous physical activity and mean number of bowel movements per week for both men and women. However only highly active women (more than 7 hours per week of vigorous physical activity) had a greater likelihood of reporting more than 7 bowel movements per week (OR; 1.70 [95%CI 1.42, 2.03]) compared to women who reported no vigorous physical activity. Curtin et al (1996) conducted a population survey of bowel habits in urban Swiss men but found no relationship between physical activity status and bowel habits.

Types of studies

The GDG decided that the washout period for crossover studies in this review should be at least 4 weeks. Trials with shorter washout periods were not to be included in the analysis.

Types of intervention

Studies were to include the following interventions:

• Insoluble fibre (corn, wheat, fruit and vegetables)
• Soluble fibre (pectins, fruit and vegetables, oats, nuts and seeds, psyllium, ispaghula)
• Bran.

It was to be noted if the fibre was provided as a food or as a capsule/supplement. In addition, the total amount of fibre in the diet for each intervention was to be recorded where possible.

The following comparisons were to be included:

• Fibre + normal diet versus normal diet (fibre versus nothing)
• Fibre versus low fibre diet or placebo (fibre versus placebo)
• Bran versus placebo
• Insoluble fibre versus soluble fibre
• Insoluble fibre + soluble fibre versus soluble fibre
• Insoluble fibre + soluble fibre versus insoluble fibre
• Fibre level 1 versus fibre level 2
• Duration of treatment 1 versus duration 2
• Fibre versus another type of intervention
• Fibre plus another type of intervention versus another type of intervention.

In spite of the large placebo effect associated with IBS, comparisons with no treatment were to be included.

The fibre review was to be concerned only with longer-term maintenance treatment. The GDG decided that there should be a minimum duration of treatment of four weeks for this review. Studies of shorter durations were to be excluded.

Outcomes

In addition to the outcomes discussed in the general methods section, the GDG were interested in the number of people with global deterioration, other than those who withdrew because of the treatment.

Data extraction

In addition to the items given in the general section, we also extracted information on the total amount of fibre (i.e. the sum of the intervention and the fibre in the diet).

Subgroup analyses

We planned to carry out subgroup analyses by type of fibre (soluble, insoluble, mixed), dose (both intervention and total amount), duration of intervention, and, post-hoc, by means of ingestion (supplement or dietary).

Study Design

Setting: The majority of studies took place in secondary care; one was in primary care (Dettmar 1999) and one study did not report the setting (Tarpila 2004).

There were two crossover studies (Cook 1990; Lucey 1987) in which participants were allocated to receive both the intervention and control treatments during the course of the study, in a random order. The GDG defined the minimum washout period to be four weeks for crossover studies in this review, so the only crossover study eligible was Cook (1990). However, a second crossover study (Lucey 1987) became eligible because individual patient data were reported, allowing calculation of first period results. This gave the study a ‘pseudo-parallel’ design, although the power was reduced. The remaining studies had a parallel design. One study had more than two arms: Kruis (1986) compared bran with mebeverine (anti-spasmodic) and placebo.

Population

The definition of IBS varied between studies: two used the Manning criteria (Chapman 1990; Cook 1990); two used Rome I (Parisi 2002; Rees 2005); two used Rome II (Aller 2004; Parisi 2005) and two met criteria defined by the authors that were similar to the above (Fielding 1984; Tarpila 2004). In five studies, the authors stated that the participants had IBS, with no further explanation (Lucey 1987; Manning 1977; Ritchie 1979; Ritchie 1980; Vilagrasa 1991). The remaining seven studies (Arthurs 1983; Dettmar 1999; Fowlie 1992; Kruis 1986; Longstreth 1981; Prior and Whorwell 1987; Søltoft 1976) did not use a formal definition but described a range of symptoms consistent with IBS.

Most studies included a combination of IBS types. Four specified constipation-predominant IBS (Cook 1990; Fielding 1984, Rees 2005; Tarpila 2004) and three were unclear (Arthurs 1963; Dettmar 1999; Fowlie 1992).

None of the studies stated that any participants had IBS as result of gastrointestinal infection. The majority of studies (13) did not state the number of participants with bloating. Four studies reported that some people had bloating (Aller 2004; Kruis 1986; Longstreth 1981; Vilagrasa 1991). Two studies (Prior and Whorwell 1987; Tarpila 2004) stated that all people had bloating.

Most of the studies did not describe symptom severity. Six studies stated that participants had symptoms of mixed severity (Dettmar 1999; Fowlie 1992; Longstreth 1981; Parisi 2002; Parisi 2005; Prior and Whorwell 1987).

The age range of participants across studies was 14 to 82 years, with the mean age (where given) ranging from 25.8 to 45 years. No study particularly identified elderly people. All studies had more women than men.

Interventions

The studies varied in the type of fibre used: six had insoluble fibre (wheatbran); eight had soluble fibre (six ispaghula, one partially hydrolysed guar gum [‘PHGG’], one psyllium); five had mixed fibres: studies used a combination of fruit, vegetables and cereal.

One study gave the fibre in a capsule form (Fowlie 1992), eight gave the fibre as a supplement (Arthurs 1983; Chapman 1990; Dettmar 1999; Fowlie 1992; Longstreth 1981; Prior and Whorwell 1987; Ritchie 1979; Ritchie 1980); and the rest added fibre to the diet with food (e.g. bran-containing biscuits).

A fibre level of 18g per day is regarded as a threshold dose. When assessing dose we considered both the amount of additional fibre and the amount of total fibre (intervention plus that in the diet). The amount of additional fibre ranged from 7g per day (Dettmar 1999), although a third 3.5g sachet could be added if needed, to 40g per day (Fielding 1984). Ten studies gave additional fibre as amounts of less than 18g (Chapman 1990; Dettmar 1999; Fowlie 1992; Kruis 1986; Lucey 1987; Parisi 2005; Prior and Whorwell 1987; Rees 2005; Ritchie 1979; Ritchie 1980). Nine studies gave more than 18g (Aller 2004; Arthurs 1983; Cook 1990; Fielding 1984; Longstreth 1981; Parisi 2002; Manning 1977; Søltoft 1976; Villagrasa 1991). One study (Tarpila 2004) gave 12 to 24g daily.

Eight studies reported the total fibre in the intervention arm (Aller 2004; Arthurs 1983; Cook 1990; Fielding 1984; Fowlie 1992; Prior and Whorwell 1987; Tarpila 2004; Villagrasa 1991).

The duration of the intervention ranged from four weeks (Arthurs 1983; Dettmar 1999; Fielding 1984; Parisi 2002) to two years (Villagrasa 1991). One study reported follow-up after the end of the trial (Parisi 2005; 3 months follow-up).

Comparisons

The included studies covered the following comparisons:

• Eleven comparisons of fibre versus placebo, including one versus usual diet (Kruis 1986); and one versus reduced fibre (Manning 1977):
  • Four gave soluble fibre (Arthurs 1983; Longstreth 1981; Prior and Whorwell 1987; Ritchie 1979)
  • Six gave insoluble fibre (Cook 1990; Kruis 1986; Lucey 1987; Manning 1977; Rees 2005; Søltoft 1976)
  • One gave mixed fibre (Fowlie 1992);
• Three studies compared different classes of fibre:
  • Two studies compared soluble versus insoluble fibre

    ▪

    PHGG versus bran (Parisi 2002)

    ▪

    Ispaghula versus bran (Ritchie 1980)

  • One study compared mixed versus soluble fibre

    ▪

    Ground flax seed (containing 20% flaxseed oil) versus psyllium (Tarpila 2004);

• One study compared different types of fibre in the same class (mixed):
  • One study compared different combinations of fruit and cereal fibre (Fielding 1984);
• Two studies compared different doses of fibre:
  • One compared 30.5g with 10.4g of mixed fibre. However, the proportion of soluble fibre differed between the two groups (13% versus 19%) (Aller 2004)
  • One study compared 5 and 10g of PHGG (Parisi 2005);
• Two studies compared fibre + mebeverine versus mebeverine + dietary advice (Chapman 1990; Dettmar 1999)
• Two studies compared fibre with an antispasmodic (Kruis 1987; Villagrasa 1991).

1. Global symptoms

2. Individual symptoms

A. Fibre versus Placebo

There were eleven studies that compared fibre with placebo (Arthurs 1983; Cook 1990; Fowlie 1992; Kruis 1986; Longstreth 1981; Lucey 1987 first period only; Manning 1977; Prior and Whorwell 1987; Rees 2005; Ritchie 1979; Søltoft 1976). Two of these studies were in people with constipation-predominant IBS (Cook 1990; Rees 2005); three did not specify the type of IBS (Arthurs 1983; Fowlie 1992; Ritchie 1979) and the remainder had mixed IBS types. Therefore the studies were not stratified by IBS type. Similarly, there was too little information to separate by severity, post-infective cause or bloating status.

Where outcomes were measured at different times during the study, we took the end-study results unless there were significant numbers of withdrawals or problems with compliance. Therefore, for the Kruis (1986) study we took the values at four weeks. The results in Rees (2005) were collected between week 8 and week 12 (11 people were assessed at week 8; six at week 9; three at week 10; one at week 11, and; one at week 12).

1. Global symptoms

a. Number of people with improvement in global symptoms

Nine studies with 545 participants reported this outcome. Overall the relative risk was 1.18 (95% CI 1.03 to 1.35), i.e. statistically significant, in favour of fibre.

Figure 1

Subgroup analysis into soluble and insoluble fibres (Figure 2) gave some suggestion that soluble fibre was more effective than insoluble, however, this conclusion was fairly reliant on the Prior and Whorwell (1987) study, which had some attrition bias and was analysed using the last measurement carried forward method. A sensitivity analysis without Prior and Whorwell (1987), Longstreth (1981), Rees (2005 - attrition bias) and Kruis (1986 - which did not have a placebo comparator) showed little difference in global improvement between fibre and placebo overall, although the results for soluble fibre were still significant (Figure 3a).

Figure 2

Figure 3

Sensitivity analysis.

Sensitivity analysis by method of ingestion

A further sensitivity analysis was carried out on the studies that were not at risk of bias, to investigate if there was an effect of supplementary fibre compared with dietary fibre. This was examined in a subgroup analysis (Figure 3b). There was heterogeneity (I²=58%, p=0.09) in the supplement group, which was probably caused by different types of fibre.

Figure 3b

b. Number of people with deterioration in global symptoms

Three studies reported this outcome, and included 140 participants (Figure 4). The numbers of events were few and there was too much uncertainty (wide confidence interval) to draw conclusions.

Figure 4

c. Global symptom score (mean)

This outcome was recorded by four studies (Cook 1990; Fowlie 1992; Longstreth 1981; Lucey 1987), and different scales were used. Fowlie (1992) did not give scores for the two groups and Cook (1990) was a crossover design (and had some attrition bias). In view of the different scales it was not possible to meta-analyse the parallel and crossover studies using the generic inverse variance method, so the two remaining parallel studies and the crossover study were analysed separately using the standardised mean difference. The results were inconclusive (Figure 5).

Figure 5

2. Individual symptoms

a. Pain

The following studies measured pain:

1. Number of people with no pain: one study (Prior and Whorwell 1987)
2. Number of people with less pain: three studies (Kruis 1986)
3. Pain score (final): four studies (Cook 1990; Fowlie et al 1992; Longstreth et al 1981; Manning 1977);
4. Two studies reported pain severity at the end of the study (Cook 1990; Fowlie 1992)
5. Two studies reported pain severity from daily diary readings (Longstreth 1981; Manning 1977).

Figure 6 shows the number of people with less pain and the number of people with no pain, in two single studies. The confidence intervals were too wide to draw conclusions.

Figure 6

Fowlie (1992) only gave the difference in the mean change score from baseline and its 95%CI, which was 1 (95%CI −1.5, 4), i.e. not statistically significant.

Combining the other three studies recording pain score, using the standardised mean difference (Figure 7), showed little difference between fibre and placebo, but the data was limited.

Figure 7

b. Bloating

Only one study (Prior and Whorwell 1987) reported bloating (Figure 8). This showed that statistically significantly more people had bloating when they took fibre (soluble) compared with placebo. It should be noted that this was a last measurement carried forward analysis, but that a large proportion withdrew from the study in the ispaghula group.

Figure 8

c. Combined bloating and flatulence score

One study reported this outcome (Longstreth 1981). The results showed a small non-significant difference (0.31 on a scale of 0 to 4) in favour of placebo. We noted that this study had attrition bias.

Figure 9

d. Bowel habits

i. Number of people with improved bowel habits

Two studies, with 106 participants, recorded the number of people with improved bowel habits (Kruis 1986; Manning 1977). Meta-analysis showed some heterogeneity between studies and a wide confidence interval. Each study was a comparison with a non-placebo comparator (low fibre or usual diet).

Figure 10

ii. Stool score (aggregate)

Longstreth (1981) reported the number of normal stools per week. The confidence interval was fairly wide (−2.0 to 2.6), but there was little difference between fibre and placebo. We noted that this study had attrition bias.

Figure 11

B. Fibre type 1 versus Fibre type 2

B1. Insoluble versus soluble fibre

Two studies compared insoluble and soluble fibre: Parisi (2002) compared wheat bran (insoluble; 30g/day) with guar gum (soluble; 5g/day) in people with a mixture of IBS types; Ritchie (1980) compared coarse natural bran (insoluble; 20g/day) with ispaghula (soluble; Fibogel 7g/day).

1. Global outcomes

a. Global improvement of symptoms

Meta-analysis of two studies (Parisi 2002; Ritchie 1980) in 281 people, found a statistically significant increase in the number of people reporting improved global symptoms in favour of the soluble fibre (RR 0.61, 95% CI 0.51 to 0.73), with no heterogeneity. This corresponded to a number needed to harm of 3 (95%CI 2, 4), for a soluble group rate of 69 to 88%.

Figure 12

b. Global deterioration in symptoms

One study (Parisi 2002) showed a wide confidence interval for this outcome and conclusions could not be drawn.

Figure 13

2. Individual symptoms

a. Pain

One study (Parisi 2002) showed little difference between the interventions for the number of people with pain.

Figure 14

b. Bowel habits

There was no significant difference in the number of people with improved bowel habits.

Figure 15

B2. Mixed fibre versus soluble fibre

Tarpila (2004) compared 6 to 24g/day flax seed (mixed fibre: 33% insoluble, 11% soluble, 20% flaxseed oil) with 6 to 24g/day psyllium (soluble), in 55 people with IBS-C.

a. Bloating

There were significantly more people with a reduction in bloating for the mixed fibres (flax seeds) group, compared to psyllium.

Figure 16a

The number of people with more bloating also significantly favoured the mixed fibre, although the confidence interval was very wide.

Figure 16b

b. Bowel habits

There was no significant difference in the number of people with improved bowel habits.

Figure 17

C. Mixed fibre 1 versus mixed fibre 2

One study (Fielding 1984) compared 40g of mixed fibre diet with different proportions of cereal and fruit/vegetables 75% cereal versus 25% cereal. The study recorded a state of well being score and individual symptom outcomes.

1. Number of people with an improved state of well being

There was little difference between interventions, although the confidence interval was fairly wide.

Figure 18

2. Individual symptoms

a. Number of people with less pain

There was little difference in pain incidence between the two types of mixed fibre.

Figure 19

b. Number of people with improved bowel habit

There was little difference between interventions.

Figure 20

D. Fibre dose 1 versus fibre dose 2

Two studies compared different doses of fibre (Aller 2004; Parisi 2005). In the former, the comparison was 30.5 versus 10.4g/day of mixed fibre over 12 weeks (i.e. above versus below the 18g/day threshold). The latter compared 10 and 5g/day of partially hydrolysed guar gum over 12 weeks, which was then followed up for a further 12 weeks.

a. Global symptom score

There was little difference between interventions in a single study in 96 patients, and the further 12 weeks follow-up did not change this conclusion

Figure 21

b. Pain score

There was a small, non-significant difference between interventions, favouring the lower dose of soluble fibre in Parisi (2005) at 12 weeks, which decreases to zero after a further 12 weeks. There was no significant difference in the two doses (above and below the threshold) for the Aller (2004) study.

Figure 22

c. General bloating and flatus score

There is little difference between dose levels in either study.

Figure 23

d. Bowel scores

There is little difference between doses for the Aller (2004) study.

Figure 24

E. Fibre plus another intervention versus another intervention alone

Two studies (Dettmar 1999; Chapman 1990) assessed ispaghula plus mebeverine (antispasmodic) versus mebeverine plus high fibre dietary advice. Each study reported the number of people improved in terms of abdominal pain, and in terms of improvements in bowel habit, at 4 weeks.

Figure 25

Figure 26

F. Protective effects of fibre for the prevention of colorectal adenomas and carcinomas, coronary heart disease and breast cancer

Types of studies

The GDG decided that crossover studies should not be included in this review because it was unclear whether probiotics effected longer term changes or how long they were retained in the gut.

Types of intervention

Studies should include the following interventions:

• Single probiotics
• Combination probiotics
• Single prebiotics
• Synbiotics.

Probiotics may be given as a food or as an enteric coated capsule. Prebiotics should fulfil three criteria: (a) resistance to gastric acidity, hydrolysis by mammalian enzymes and gastrointestinal absorption; (b) fermentation by intestinal microflora; (c) selective stimulation of the growth and/or activity of intestinal bacteria associated with health and well-being. Acceptable prebiotics are mainly fructo-oligosaccharides, galacto-oligosaccharides and lactulose.

The following comparisons were included:

• Single probiotic versus placebo
• Combination probiotic versus placebo
• Single prebiotic versus placebo
• Synbiotics versus placebo
• Probiotic 1 versus probiotic 2
• Probiotic dose 1 versus dose 2
• Intervention duration 1 versus duration 2.

The probiotics review was concerned only with longer-term maintenance treatment.

In spite of the large placebo effect associated with IBS, comparisons with no treatment were included, and the minimum duration of treatment was four weeks.

Stratification and Subgroup analyses

Pre and probiotics were to be treated separately. We planned to carry out subgroup analyses as follows:

• Type of probiotic (single, combination)
• Nature of bacteria, including the strain (e.g. Lactobacillus salivarius, Bifidobacterium infantis, Streptococcus faecium)
• Dose (above and below 1 x 10⁶ bacteria per day; this was later revised to 10⁶, 10⁸, 10¹⁰ subgroups and the GDG later excluded studies with levels below 1 x 10⁶)
• Duration of intervention (5–8, 9–12, 13–16, 16+ weeks).

We also planned to investigate the effect of enteric coated capsules compared with the addition of probiotics as a food.

Study Design

Setting: The majority of studies took place in secondary care, but three were carried out in primary care (Bittner 2005; Gade 1989; Whorwell 2006) and one was assumed to be primary care (Nobaek 2000; recruited by newspaper advertisement). One study had patients from both primary and secondary care (O’Mahony 2004; patients from gastroenterology clinics and newspaper advertisement).

All the studies included in the review had a parallel design.

One study (O’Mahony 2004) had 3 arms comparing Lactobacillus salivarius UCC4431, Bifidobacterium infantis 35624 and placebo malted milk. Whorwell (2006) compared three different doses of encapsulated Bifidobacterium infantis 35624 with placebo in women with IBS. This gave a total of 20 comparisons in the review.

Population

The definition of IBS varied between studies: two used the Manning criteria (Niedzielin 2001; Olesen 2000); one used the Rome I criteria (Kajander 2005); one met criteria defined by the authors that were similar to the above (Gade 1989); and the rest used the Rome II criteria.

All studies but one included patients who had a range of IBS types; the other study specified diarrhoea predominant IBS symptoms (Kim 2003). Only one study (Niv 2005) stated that the participants had IBS as result of gastrointestinal infection.

The majority of studies (12) did not state the number of participants with bloating. Five studies had some patients with bloating measured as an outcome (Kajander 2005; ^{Kim 2003}; Olesen 2000; Tsuchiya 2004; Whorwell 2006). Two studies (^{Kim 2005}; Niedzielin 2001) identified all patients as having bloating.

Most of the studies described symptom severity as mixed; one study described the symptoms as mild (Nobaek 2000). Two studies did not state symptom severity (Kajander 2005; O’Mahony 2004). Three studies suggested that the patients had refractory IBS: Saggioro (2004) reported that the patients had been treated with drugs without success; Niedzielin (2001) stated that all patients had been referred to secondary care because of problems with management; Tsuchiya (2004) reported that all patients had undergone a number of treatments without significant and lasting benefit.

The age range of participants across studies was 19 to 78 years, with the mean age (where given) ranging from 34 to 48 years. No study particularly identified elderly participants. All studies had a ratio of women to men greater than one. Whorwell (2006) included only women participants.

Interventions

The studies varied in the type of probiotics used:

• Six used a single probiotic (Gade 1989; Niedzielin 2001; ^{Niv 2005}; Nobaek 2000; O’Mahony 2004; Whorwell 2006)
• Five used a combination of probiotics (Kajander 2005; ^{Kim 2003; Kim 2005}; Saggioro 2004; Tsuchiya 2004)
• One study used a prebiotic (Olesen 2000)
• One gave a pre/probiotic combination (Bittner 2005).

A range of different bacteria was used, and included various strains of Lactobacillus, Bifidobacterium and Streptococcus. Further details are available in the Appendix.

Three studies gave the probiotic in a capsule form (Kajander 2005; Whorwell 2006; Bittner 2005), two gave a tablet (Gade and Thorn 1989; ^{Niv 2005}) and the remainder used a food source or solution as the means of ingestion. The food sources used included milk or milk products, yoghurt, oatmeal soup and fruit drinks. The GDG considered the medium in which the probiotics were ingested to be an important difference and decided to consider, as subgroups, capsules versus other delivery routes. Whether the intervention was given with food was also considered important because of the increased levels of bile salts, as a result of digestive process, which are a serious obstacle to probiotic survival. However only three studies gave details as to when the probiotics were taken: Niedzielin (2001) directed that they should be taken before breakfast and two hours after the evening meal; in Gade (1989) the dose was given in the morning and evening with meals; and Olesen (2000) required the dose of prebiotics be taken with breakfast.

The GDG defined the minimum dose of probiotic as 1 x10⁶. The doses of probiotic varied considerably, and ranged from 8x10⁶ (Gade 1989) to 4x10¹⁰ (Niedzielin 2001) for single probiotics, and 5x10⁹ to 5x10¹¹ for combination probiotics. It is noted that the activity of the probiotics vary according to strain.

The duration of the intervention ranged from four weeks (Gade 1989; Niedzielin 2001; Nobaek 2000; Saggioro 2004; Whorwell 2006) to six months (Kajander 2005; ^{Niv 2005}). Three studies had durations of eight weeks (^{Kim 2003; Kim 2005}; O’Mahony 2004), and two studies had interventions lasting 12 weeks (Olesen 2000; Tsuchiya 2004). One study followed the patients for 12 months (Nobaek 2000).

Comparisons

The included studies covered the following comparisons:

• Nine comparisons of a single probiotic versus placebo (Gade 1989; Niedzielin 2001; ^{Niv 2005}; Nobaek 2000; O’Mahony 2004 x 2; Whorwell 2006 x3)
  • Lactobacillus salivarius UCC4331 (1x10¹⁰) in malted milk drink (O’Mahony 2004)
  • Lactobacillus plantarum DSM 9843 (5x10⁷ CFU) in oatmeal soup (Nobaek 2000)
  • Lactobacillus plantarum 299V (5x10⁷) in oatmeal soup (Niedzielin 2001)
  • Lactobacillus reuteri ATCC 55730 (1x10⁸) tablet (Niv 2005)
  • Two used Bifidobacterium infantis 35624 (O’Mahony 2004 (1x10¹⁰) in malted milk drink; Whorwell (2006) 1x10⁶, 1x10⁸, 1x10¹⁰ CFU in capsule)
  • Streptococcus faecium (dose estimated as 8x10⁶) tablet (Gade 1989);
• Five comparisons of a combination of probiotics versus placebo:
  • Two studies used VSL3 powder sachet (Bifidobacterium 3 strains, Lactobacillus 4 strains, Streptococcus 1 strain) (^{Kim 2003; Kim 2005})
  • SCM-III solution (Lactobacillus acidophilus1.25x10⁶ CFU; Lactobacillus helveticus 1.3x10⁹; bifidobacterium 4.95x10^9) (Tsuchiya 2004)
  • Lactobacillus rhamnosus GG, L. rhamnosus LC705, Bacillus breve Bb99, P.freudenreichii ssp.shermanii JS capsule (Kajander 2005)
  • Lactobacillus plantarum LPO1 & Bifidobacterium Breve BRO 5x10⁹ CFU sachet dissolved in water (Saggioro 2004);
• One study compared two different probiotics (Lactobacillus salivarius UCC4331 versus Bifidobacterium infantis 35624) (O’Mahony 2004)
• One study compared three doses of probiotics (Whorwell 2006; 3 comparisons)
• One comparison of Prebiotic versus placebo (Oleson and Hoyer 2000)
• One comparison of a pre/probiotic capsule versus placebo, but this study contained no analysable data (Bittner 2005).

1. Global symptoms

2. Individual symptoms

3. Quality of Life

• Two studies reported quality of life as an outcome (^{Niv 2005}; Whorwell 2006).

A. Probiotics versus placebo

Ten studies compared probiotics (singly or in combination) with placebo (Gade 1989; Kajander 2005; ^{Kim 2003; Kim 2005}; Nobaek 2000; Niedzielin 2001; ^{Niv 2005}; O’Mahony 2004; Tsuchiya 2004; Whorwell 2006).

1. Global symptoms

a. Number of patients with an improvement in global symptoms

Six studies (eight comparisons), with 629 patients, recorded the patients’ assessment of global improvement (Gade 1989; Kajander 2005, ^{Kim 2003}, Niedzielin 2001, Tsuchiya 2004, Whorwell 2007). Meta-analysis showed significant heterogeneity (I²=65%; p=0.006).

Figure 1

This heterogeneity was investigated in terms of the pre-specified subgroup analyses: by type of probiotic (single, combination), by duration and by dose and strain of bacterium (Figures 2 to 3).

Figure 2

Subgroup analysis by type of probiotic.

Figure 3

By type and dose of bacterium.

Type of probiotic (Figure 2)

There was still significant heterogeneity in the single probiotic group, but it was not significant in the combination probiotic group (I²=31%, p=0.23). Meta-analysis of three studies in 173 patients showed a statistically significant improvement in global symptoms. This corresponded to an NNT of 3 (95%CI 3, 5), for a control group rate of 42 to 47%. We noted there was significant heterogeneity for the risk difference (I²=74%, p=0.02), which may have been an indication that the particular combination of probiotics was important.

Duration

The Kajander (2005) study was six months duration, Tsuchiya (2004) was 12 weeks and the others were four or eight weeks. This did not account for the heterogeneity amongst studies.

Strain and dose of probiotic (Figure 3)

All the studies had different strains and/or doses, and the confidence intervals are wide in some cases. The heterogeneity may be indicative of different efficacies of the different probiotics; most of the probiotics tested in the trials gave a greater improvement in symptoms than placebo, but there were some exceptions. Whorwell (2006) showed a maximum in the improvement of global symptoms with increasing dose, with only the 10⁸ dose of Bifidobacterium infantis being significant at four weeks. The authors attributed this effect to dissolution problems of the capsule for particular concentrations (see GDG discussion at the end of this review).

b. Number of patients with deterioration in global symptoms

This outcome was recorded by two studies (Gade 1989; Tsuchiya 2004). The confidence intervals were very wide, although Tsuchiya (2004) was statistically significantly in favour of probiotics.

Figure 4(NB 0.01 to 100 scale)

c. Global symptom score (mean)

This outcome was recorded by seven studies (Kajander 2005; ^{Kim 2003; Niv 2005}; Nobaek 2000; O’Mahony 2004; Saggioro 2004; Whorwell 2006). One study (Saggioro 2004) reported percentage change scores, with p values, so these results are given separately. The other studies reported the global symptom score, but on different scales, so the standardised mean difference was used to analyse the data (Figure 5). The Niv (2005) values were taken from a graph and it was assumed that the standard error was given. This study also had some attrition bias, so a sensitivity analysis was repeated excluding this study (Figure 6).

Figure 5

Figure 6

Sensitivity analysis without Niv (2005).

Meta-analysis of eight comparisons in 624 patients showed no significant difference between probiotics and placebo overall, with little heterogeneity (I²=4%, p=0.41). However, there was a statistically significant difference for the combined probiotic subgroup in 105 patients. Meta-analysis of seven single probiotics showed no significant difference between probiotic and placebo, with no heterogeneity. Sensitivity analysis without Niv (2005) made a small difference.

The Saggioro (2004) study in 40 patients reported a statistically significant difference in the percentage change in IBS symptom severity (−44% versus −8.5% after 28 days; p<0.001 for the combined probiotic versus placebo).

Subgroup analyses were carried out by strain and dose of bacterium because the GDG was uncertain whether studies using different bacteria should be combined. Figure 7 shows the studies by bacterium and dose. Most comparisons showed no significant difference compared with placebo, including the meta-analysis of two studies in 232 patients, receiving Bifidobacterium infantis at a dose of 1x10¹⁰ CFU; there was no significant heterogeneity for these two studies (I²=6%, p=0.30). There were only two statistically significant comparisons:

Figure 7a

Global symptom score.

• Encapsulated Bifidobacterium infantis at a dose of 1x10⁸ CFU versus placebo, in 182 patients. This had a mean difference of −0.33 (95%CI −0.59, −0.07) on a scale of 0 to 15 (i.e. a fairly small effect)
• Encapsulated combined probiotic versus placebo in 81 patients. This had a mean difference of −6.48 (95%CI −12.56, −0.40) on a scale of 0 to 112.

Figure 7bGlobal symptom score – 5/6 weeks and 12 month follow-up

This may, however, be a size effect; most of the non-significant studies had around 50 patients or fewer.

Nobaek (2000) also reported 12 month follow-up, but it was unclear if the patients continued to modify their diet after the trial had ended. There was a borderline significant effect of probiotics at 12 months, but not at 5 to 6 weeks. The scale was 0 to 10.

e. Global improvement in symptoms score (mean)

This outcome was recorded by one study (Kim 2003), which showed too much uncertainty to draw conclusions. It is unclear what scale is used.

Figure 8

2. Individual symptoms

a. Pain

i. Number of patients with pain

Two studies measured this outcome (Gade 1989; Niedzielin 2001) and both separately were statistically significantly in favour of probiotic. The relative risk ranged from 0.03 to 0.2 (i.e. 5 to 33 times less risk of pain with the probiotic). However, combining the studies gave heterogeneity (I²=80%; p=0.03).

Figure 9

ii. Pain score

Eight studies (Kajander 2005; ^{Kim 2003; Kim 2005}; Nobaek 2000; O’Mahony 2004; Saggioro 2004; Tsuchiya 2004; Whorwell 2006) reported a pain score. Saggioro (2004) reported percentage change scores, with p values, so these results are given separately.

This outcome showed no significant difference between interventions for the single probiotic group (although it is difficult to estimate the width of the confidence interval because the standardised mean difference was used) and a highly heterogeneous result for the combined probiotics group (I²=93%, p=0.00001), attributable to the Tsuchiya (2004) study, from which data were extracted from a graph, which may not have been to scale for the standard deviations. In the absence of this study the meta-analysis of three studies gave a statistically significant reduction in pain for the combined probiotic group, and no heterogeneity (I²=0%; p=0.94).

Saggioro (2004) reported a statistically significant difference in the percentage change in pain score (−38% versus −18% after 28 days; p<0.05 for the combined probiotic versus placebo).

Figure 10a

Figure 10bWithout Tsuchiya 2004

Nobaek (2000) also reported 12 month follow-up data, shown in Figure 10c. The scale is a visual analogue scale of 0 to 10. There was no significant difference at 5 to 6 weeks, but a statistically significant difference after 12 months. It was unclear if the patients in the intervention group changed their dietary habits following the trial or if there was a long-term effect.

Figure 10c

iii. Number of patients with less pain

Two studies compared Lactobacillus planetarum, given in food, with placebo (Niedzielin 2001; Nobaek 2000) and reported the number of patients with reduced pain. There was a statistically significant reduction in pain; RR 1.67 (95% CI 1.09, 2.56), with no heterogeneity (I²=0, p=0.64). This corresponded to a number needed to treat of 5 (95% CI 3, 20) for a control group rate of 19 to 55%.

Figure 11

b. Bloating

i. Number of patients with less bloating (Kim 2005)

In a single study in 48 patients, there was no significant difference between probiotics and placebo in the number of patients with less bloating, although the confidence interval was fairly wide.

Figure 12

ii. Bloating score (Kajander 2005; ^{Kim 2003; Kim 2005}; O’Mahony 2004; Tsuchiya 2004; Whorwell 2006)

This outcome showed no significant difference between interventions for the single probiotic group and a highly heterogeneous result for the combined probiotics group, attributable to the Tsuchiya (2004) study, from which data were extracted from a graph, which may not be to scale for the standard deviations. The authors of this study reported that there was no significant difference between groups, which belies the data on the graph, suggesting that the standard deviations on the graph were inaccurate. In the absence of this study, meta-analysis of the three studies in 73 patients gave a statistically significant reduction in bloating score for the combined probiotic group, MD −0.42 (95% CI −0.73, −0.10), with no heterogeneity (I²=0%; p=0.87).

Figure 13aBloating score (final scores)

Figure 13bSensitivity analysis without Tsuchiya (2004)

c. Bowel habits

i. Stool frequency

Only one study specified the type of IBS (Kim 2003), which was in patients with diarrhoea predominant IBS. For this study, the frequency was seen as a negative outcome and there was no significant difference between probiotic and placebo.

Figure 14

ii. Stool score

This was aggregated to include stool frequency, consistency, ease of passage and completeness of evacuation (^{Kim 2003; Kim 2005}; O’Mahony 2004; Nobaek 2000; Whorwell 2006 – bowel habit satisfaction). Tsuchiya (2004) also reported assessment of bowel habits, but these values were not included in the meta-analysis in view of the uncertainties in the standard deviation described above.

In the meta-analysis of eight comparisons (562 patients) there was no significant difference between probiotics and placebo for this outcome, either overall, or for single or combined probiotics, and there was no significant heterogeneity.

Figure 15

3. Quality of Life

Only one study reported quality of life as an outcome (Niv 2005). This showed no significant difference in the quality of life score. The scale used was unclear, however: the study reported that there were 26 questions, each rated from mild (1) to severe (7), and the sum of all of them yielded the total QoL score, but the baseline scores for the total were similar to the individual components and were about 4 to 5 points.

Figure 16

Examination of the two studies using Bifidobacterium infantis 35624

Two studies compared Bifidobacterium infantis 35624, at a dose of 1×10¹⁰ CFU per day, versus placebo. One study (Whorwell 2006) gave the probiotic in a capsule and the other (O’Mahony 2004) in a malted drink.

The outcomes are summarised in Figure 17. There was some heterogeneity between studies for the outcomes of pain and stool score, with the encapsulated probiotic having less effect. This is discussed further in the next section.

Figure 17

B. Probiotic dose 1 versus probiotic dose 2

One study (Whorwell 2006) compared three doses of Bifidobacterium infantis 35624, 1×10⁶, 1×10⁸, 1×10¹⁰, with approximately 90 patients in each arm. The outcomes compared are reported in Figure 18.

Figure 18

Head-to-head comparison of the doses 1×10⁸ and 1×10⁶ showed there was a significant difference in global symptoms, pain and bloating scores and a borderline difference in stool score, favouring the 1×10⁸ dose. However, there was no significant difference between the 1×10¹⁰ and 1×10⁶ doses, an unexpected dose effect.

Whorwell (2006) explained this using in-vitro dissolution experiments, showing that the highest concentration of probiotic coagulated on exposure to moisture, making dissolution very difficult, such that the probiotic was not bioavailable to the patient.

This effect also explains the differences between Whorwell (2006) and O’Mahony (2004); in the latter, the probiotic was bioavailable because it was present in a drink.

C. Probiotic 1 versus probiotic 2

One study compared two strains of bacteria, Lactobacillus salivarius UCC4331 versus Bifidobacterium infantis 35624 (O’Mahony 2004) directly in a randomised trial of 50 patients. The results are presented below for the different outcome measures.

1. Global symptom score

There was no significant difference in global symptoms at eight weeks, but the Bifidobacterium is favoured. A 10 cm visual analogue scale was used for individual symptoms and combined to give a global score (maximum 30).

Figure 19

2. Individual symptoms

There was no significant difference between the two types of bacteria for pain, bloating or stool score. Likert scales were used for each component with a maximum of 7.

Figure 20

D. Prebiotics versus placebo

One study (Olesen 2000) compared a prebiotic (Fructooliogsaccharide given as a 10g sachet for 2 weeks then 20g for 10 weeks) with placebo in 98 patients. The results are given below and generally showed no significant differences between prebiotics and placebo, in either global symptoms or bloating (although the confidence interval was fairly wide in the latter). The confidence interval was too wide to determine if there was a difference for the pain outcome. We noted that there was some attrition bias for this study.

Figure 21

Figure 22

Types of Studies

Two randomised trials were included (Odes and Madar 1991; Davis 2006) and two excluded studies are listed in Appendix E, along with reasons for exclusion.

Types of participants

All participants in Davis (2006) had IBS (28% IBS-C, 52% IBS-D, 20% IBS-A); participants had to be between 18 and 65 years, have no other co-morbidities and had to have previously failed conventional management of IBS defined as antispasmodics, bulking agents and dietary interventions. Constipation was defined as per Rome II criteria.

Odes and Madar (1991) had 11/32 people with IBS-C (the rest had simple constipation); people with IBS-D or IBS-A were excluded; participants had to have been receiving laxative therapy for constipation for a minimum of two years as an indication of severity. Participants had previously received other treatments including diet and enemas, but it is not stated if they were refractory to treatment.

Types of intervention

Davis (2006) used aloe vera gel made up in a pink mango flavoured syrup. The dose was 50 ml taken four times a day for one month. The placebo was a matching pink mango flavoured inert syrup.

Odes and Madar (1991) used a capsule laxative preparation made up of celandin, aloe vera and psyllium in ratio of 6:3:1 (total fibre content 47%) given for 28 days. The aloe vera fibre was derived from leaves of Aloe socotrine and contains anthraquinone. The dose was one 500 mg capsule per day taken with water at bedtime increasing to a maximum of three capsules a day. The placebo capsule was of identical appearance but contained no active ingredient. Participants in Odes and Madar (1991) were given no dietary modification advice. No additional medication was prescribed throughout the treatment period but people could continue with prescribed laxative medication, provided that the dose and frequency were recorded in the study data sheet. Davis (2006) did not state if other medications could be continued.

A. Aloe vera gel versus placebo

One study (Davis 2006) in compared aloe vera gel with placebo in 58 people with IBS.

1. Global improvement of symptoms

The primary outcome was the number of people with an improvement in global symptom score (pain; distension; bowel habit, and; quality of life). The symptom score was derived by adding the scores of individual symptoms and the proportion of days symptoms occurred with a maximum score of 500. A reduction of 50 points was defined as improvement. Participants were assessed at one month and at three months post-intervention. The forest plots below illustrate that there was no significant difference between the active and placebo treatment for global improvement of symptoms, although the confidence interval was fairly wide.

Figure 1aGlobal improvement of symptoms at one month

Figure 1bGlobal improvement of symptoms at three months

2. Individual symptoms

a. Pain

There was no significant difference, either at 1 month or at 3 months, in the change in pain score, on a scale of 0 to 100%, for which a positive change represented an improvement over baseline.

Figure 2aPain at one month

Figure 2bPain at three months

b. Bloating

There was no significant difference either at 1 month or at 3 months, in the change in distension score, on a scale of 0 to 100%, for which a positive change represented an improvement over baseline.

Figure 3aDistension at one month

Figure 3bDistension at three months

c. Bowel Habit

There was no significant difference, either at 1 month or at 3 months, in the change in bowel score, on a scale of 0 to 100%, for which a positive change represented an improvement over baseline.

Figure 4aChange in bowel score at one month

Figure 4bChange in bowel score at three months

3. Quality of Life

There was no effect on quality of life at one month or at three months. The scale used was not stated.

Figure 5aChange in quality of life at one month

Figure 5bChange in quality of life at three months

4. Adverse effects

2/31 people withdrew from the active group and 4/27 from the placebo group because of nausea and vomiting. The confidence interval was too wide to determine if there was a difference between groups.

Figure 6Adverse effects

B. Combined capsule of celandin, aloe vera and psyllium versus placebo

2. Individual symptoms

One study with 32 participants reported differences in bowel habits (including frequency and consistency) and pain scores for the final two weeks of treatment compared with those in the 14 day pre-intervention run in period (Odes and Madar 1991). Eleven participants (34%) were identified as having IBS-C; the rest had simple constipation.

a. Pain

There was no significant difference in pain scores (number of episodes of pain per week) between groups.

Figure 7Number of episodes of pain per week

b. Bowel Habits

Compared to the placebo group, people in the intervention arm of the trial experienced a significant increase of 3.6 (95% CI 1.51, 5.69) in the mean number of bowel movements per week.

Figure 8Stool frequency

The consistency of the stools also improved. There was a statistically significant decrease in laxative use in the intervention group of −0.8 (95% CI −1.12, −0.47) on a scale of 1 to 3 and 16/19 people considered their bowel symptoms improved compared to 4/13 of the control group.

Figure 8Stool consistency

No adverse effects were reported and all participants reported that the capsules were easy and convenient to use.

Comment

Whilst this study shows a significant positive effect for bowel habit, we noted that this was a small study (35 patients) in a population of which only one-third had IBS. In addition, the intervention used a combination of aloe vera (30%), celandin and psyllium (soluble fibre). Therefore, it was not possible to attribute the effect to aloe vera alone, and this study was not included in the evidence statements.

Adverse effects

The use of aloe by mouth for laxative effects can cause abdominal cramps and diarrhoea. Adverse effects, reported in a small number of studies, include low blood sugar levels and electrolyte imbalance, particularly lowered potassium levels.

Drug interaction

Use of aloe with laxative drugs may increase the risk of dehydration, electrolyte imbalance, potassium depletion and changes in blood pH.

Oral preparations of aloe have been reported to lower potassium levels, which may impact on the effectiveness of drugs used to manage heart rhythm disturbances, heart disease and renal disease.

Oral preparations of aloe may lower blood sugar, so have the potential to interact with drugs used in the management of diabetes.

Aloe vera should not be used by individuals who may be at increased risk from the aforementioned adverse effects, particularly people with heart disease, kidney disease, diabetes and blood disorders.

Types of studies

For intervention studies, randomised trials (RCTs) and quasi-randomised studies, examining the use of dietary manipulation/exclusion for the treatment of IBS were preferred. Crossover trials with a washout period of less than 2 weeks were included but treated with caution. Double-blind placebo controlled studies are technically difficult and most elimination diet studies use a non-randomised open dietary elimination and re-challenge design. This has the potential to introduce bias due to the large placebo effect identified in IBS patients. For this review non-randomised studies were also permitted. Studies were restricted to the English language, but the date was not restricted.

Types of intervention

Interventions were to be included if they referred to an exclusion diet (excluding certain foods) or an elimination diet (only allowing certain foods):

• Lactose restricted diet
• Elimination diet based on foods with IgG4 titres >250 μg/l
• Elimination diet based on food challenge and re-challenge
• Elimination diet based on patient-reported intolerance and re-challenge
• Elimination diet using lamb, rice and pears
• Fasting therapy.

Types of comparisons

The following types of comparisons were to be included:

• True diet versus sham diet
• Elimination diet and food challenge with foods that had been identified as potential causes of intolerance
• Fasting therapy versus usual treatment.

Sensitivity analyses

The following sensitivity analyses may be considered:

• Setting (primary/secondary care)
• Blinding of patients.

Study Design

One study had a crossover design: Symons (1992) stated that the patients were randomised to interventions on two days, following a 12 hour fast, but the washout period was not clear.

Setting: all the studies included patients in secondary care, many of whom had not responded to previous treatment for IBS symptom management.

The majority of studies took place in the UK (Atkinson 2004; Bentley 1983; Hawthorn 1991; Hunter 1985; Jones 1982; McKee 1987; Nanda 1989; Parker 1995; Smith 1985; Zar 2005; Zwetchkenbaum and Burakoff 1988). One took place in Japan (Kanazawa and Fukudo 2006), two in Europe (Böhmer and Tuynman 1996; Petitpierre 1985), two in the US (Drisko 2006; Zwetchkenbaum and Burakoff 1988) and one in Australia (Symons 1992).

Population

All studies included patients with a diagnosis of IBS, although the definition varied between studies. Four studies used Rome criteria: Rome I (Hawthorne 1991); Rome II (Atkinson 2004, mean duration of IBS over 10 years; Drisko 2006; Zar 2006); Zar (2006) also predefined the IBS type. Two studies used the Manning Criteria (Kanazawa and Fukudo 2006; Symons 1992). Two studies used a definition described by the author (Böhmer and Tuynman 1996; Parker 1995). Seven studies simply said the patients ‘had IBS’ (Jones 1982; Bentley 1983; McKee 1987; Petitpierre 1985; Smith 1985; Nanda 1989; Zwetchkenbaum and Burakoff 1988). None of the studies stated that any patients had IBS as result of gastrointestinal infection.

Atkinson (2004) did not state whether patients had bloating and/or pain but described the symptom severity to be severe. Another study described the duration and frequency of symptom episodes (Nanda 1989), and one study reported duration of symptoms and the percentage of patients with pain, bloating and urgency (Hawthorne 1991). The remaining studies did not state the severity of symptoms. The age range of patients was 18 to 80 years with the average mean age being approximately 28 to 44 years. None of the studies particularly identified elderly patients. All studies had more women than men.

Interventions

I. RANDOMISED TRIALS

A. True diet versus sham diet

1. Global symptoms

a. Number of patients with improvement in global symptoms

Atkinson (2004) randomised 150 patients to true and sham exclusion diets. They reported the number of patients with improvement in global symptoms. There was no significant difference between true and sham diets; however, the confidence interval was fairly wide.

Figure 1

b. Global improvement of symptoms score

Atkinson (2004) reported the final global symptom score on a scale from 0 to 500 (where lower scores are better). There was no significant difference between true and sham diets, although the true diet was favoured.

Figure 2

2. Quality of life

Atkinson (2004) reported a significant improvement in quality of life change-from-baseline scores for the true diet compared with sham diet, but the confidence interval was fairly wide. The mean difference was 38.0 (95%CI 2.36, 73.64), for a sham diet change from baseline of 50 points. The scale was 0 to 500.

Figure 3

3. Mental health

There was a small significant difference between the sham and true diet groups in the HAD anxiety scores (scale 0 to 21), but no significant difference in the depression scores.

Figure 4HAD anxiety

Figure 5HAD depression

Symptoms on reintroduction of excluded foods at end of trial period

Of the 131 patients who gave data at the end of the trial period in Atkinson (2004), 93 (41 in the true diet group and 52 in the sham diet group) agreed to attempt reintroduction of eliminated foods. The mean IBS symptom score significantly increased (i.e. worsened) more in the true diet group (83.3 points) than in the sham diet group (31 points, p=0.003; standard deviations not given).

The change in global symptom score also showed that significantly more patients in the true diet group worsened on reintroduction of foods to which they had IgG antibodies (i.e. those that had been excluded during the diet): 41.5% of these patients worsened on reintroduction of these foods, versus 25% worsening in the sham diet group on reintroduction of similar foods (to which they had not been shown to have antibodies), p=0.047. We noted though that the self-selecting group taking part in this section of the trial may not have been representative of the randomised groups.

B. Fructose-sorbitol solution dose 1 versus fructose-sorbitol dose 2

Fructose and sorbitol, when ingested together, are thought to provoke symptoms of IBS. Sorbitol is found naturally in fruits, particularly peaches, pears, and plums. It is also added to soft drinks and diet products. One study (Symons 1992) compared two concentrations of sorbitol in a mixed solution of fructose and sorbitol. Concentrations compared were 17.5 and 20 g/litre sorbitol (the fructose concentration was kept constant). We noted that the duration of the study was very short – two days in each phase.

II. NON-RANDOMISED STUDIES

All thirteen studies identified food intolerance with varying degrees of response to exclusion diets (Bentley 1983, Böhmer and Tuynman 1996, Drisco 2006, Hawthorn 1991, Hunter 1985, Jones 1982, Kanazawa and Fukudo 2006, McKee 1987, Nanda 1989, Parker 1995, Petitpierre 1985, Zar 2005, Zwetchkenbaum and Burakoff 1988), illustrated in Table 1.

Table 1

Non-randomised studies; exclusion diets and results.

The non-randomised studies have been grouped according to the type of diet, and the following general conclusions can be drawn:

• There is some evidence to suggest that a simple diet of lamb, pears and rice or one meat and one fruit may improve symptoms
• A low allergenic diet does not appear to give remission in IBS symptoms
• Food exclusion diets based on IgG antibody testing appear to be effective in improving symptoms
• A starvation diet significantly decreased symptoms, but this was in a group of hospitalised patients, and is not applicable to primary care
• A lactose restricted diet gave a significant decrease in symptom score for lactose intolerant patients, but not for the lactose tolerant group.

It should be remembered that this evidence is from non-randomised studies, so its overall quality is reduced.