Electronic Search

A medical librarian searched MEDLINE^®, EMBASE^®, and the Cochrane Controlled Trials Registry for English-language articles. The overall search was performed for a period dating 12 months before the final literature search for CER No. 20 (September 28, 2009) through April 2013. For SBRT, the literature was searched for the period January 1, 1990, through April 2013. The entire search was updated May 1, 2014.

In both searches together, we identified 7,130 unique titles and screened 284 in full text. Of the latter, 15 (N=1,781) met the inclusion criteria, including one RCT (Gupta 2012, N=60).²⁵ In the updated search, we identified a second citation to an RCT (Rathod 2013).²⁶ Because the latter was the same study and included the same patients as the Gupta RCT, it was not double-counted in the total number of patients; however, it reported additional, different outcomes that were reviewed herein and so is counted in that context. Thus, in 14 studies that reported unique data, including the RCT of Rathod,²⁶ 3DCRT and IMRT were compared.^25,27–39 One study compared 3DCRT and SBRT³⁴; none compared IMRT and SBRT. As in CER No. 20, no evidence was identified on PBT. The flow of articles through the screening and study selection process is shown in the PRISMA diagram (Figure 2).

Figure 2

PRISMA diagram for disposition of literature search results. ^a Twelve studies addressed both Key Questions 1 and 2. ^b Overlapping patient population refers to the studies in which the same patients were included in more than one study. In this case, (more...)

We note that although CER No. 20 was published in final form in 2010, we had obtained the final data for PARSPORT,⁴⁰ a key RCT, from the investigators at the time we updated the CER No. 20 literature search. Because the PARSPORT findings appeared in CER No. 20, they were not included in this report.

Gray Literature (Publication Bias)

The study selection criteria for this update stipulate exclusion of abstracts or other non-peer-reviewed or non-full-length studies. Therefore we did not include any information based on comprehensive searches of meeting abstracts. We examined the bibliographies of all papers screened in full text to identify peer-reviewed articles the electronic search may have missed.

We accessed the Web site ClinicalTrials.gov to identify ongoing phase 3 RCTs that would meet the criteria for inclusion based on our protocol. We identified two phase 3 RCTs of conformal RT in head and neck cancer that are recruiting patients. The first trial (National Clinical Trial [NCT] 01893307) is designed to compare IMRT and PBT in the treatment of oropharyngeal cancer. The primary outcome is the incidence of any late-onset (> 90 days) grade 3 toxicity during the 2 years after completion of RT. The second RCT (NCT01216800) is designed to compare the effects of IMRT and 3DCRT on auditory function (hearing) when used as adjuvant therapy in patients who have undergone surgical resection of parotid tumors. A MEDLINE search of the NCT number(s) and title(s) did not yield any published results; it is unknown whether any data have been reported. Examination of a Scientific Information Packet from one manufacturer of RT equipment did not yield published evidence to add to this update.

Description of All Included Studies

We identified 15 reports that met the inclusion criteria for this CER update. All are generally described in this section; other details and results specific to a particular Key Question are considered in the relevant subsections to follow.

Study Limitations

According to the USPSTF criteria for assessing the risk of bias of individual studies, the Gupta RCT was rated “fair,” whereas the 13 nonrandomized studies were rated “poor.” The rationale for the ratings is provided in Table B-1.

We assigned a “fair” USPSTF rating to the Gupta RCT, primarily because the study was not double-blinded, particularly its outcomes assessments. Furthermore, the investigators did not clarify the meaning of their reported “modified intention-to-treat” method. Although this point is for practical purposes moot, because they reported a 97 percent followup rate in each of two study arms, the USPSTF method stipulates assignment of this rating. Gupta reported aggregated survival results in patients with tumors in different sites. However, the distribution of tumor sites and characteristics between arms was similar. Overall, the two study arms were statistically similar and comparable. We note also the report of Rathod and colleagues, a RCT that included the same patients of Gupta but reported additional outcomes. ²⁶ We include evidence from this report but do not double count it in the PRISMA diagram or summations of included evidence. Its risk of bias is identical to that of the Gupta study using the USPSTF system..

The 13 nonrandomized studies were retrospective database analyses, one of which used a historical comparator group. All of the included nonrandomized studies reported results in aggregate, mixing outcomes achieved in heterogeneous groups who may not have received the same treatment(s). Overall, we rated these studies as poor according to the USPSTF criteria.

Overview

Tables 4 and 5 depict key comparative acute (< 90 days post-treatment) and late (> 90 days post-treatment) toxicity outcomes reported by each relevant study; a blank cell in any table means that the toxicity was not reported in that study. Acute and late toxicity outcomes were not collected consistently across studies. Only eight (53%) studies reported acute toxicities.^{25,27,29,31,33,36–38} Nine (60%) studies reported late toxicities.^{25,27–29,31,33,34,36,39} Only two studies reported QOL evidence according to RT modality, including the RCT by Rathod.²⁶

Table 4

Summary of key reported acute (<90 days post-treatment) comparative toxicity outcomes.

Table 5

Summary of key reported late (>90 days post-treatment) comparative toxicity outcomes.

Because toxicities were inconsistently reported, we focused this update, as we did CER No. 20, on those toxicities prominently associated with RT in the head and neck: dysphagia, salivary gland function, and xerostomia. We also only consider toxicities of grade 2 or greater according to accepted criteria, such as those of the Radiation Therapy Oncology Group or the NCI Common Terminology Criteria for Adverse Events. Grades greater than 2 are those that have direct impact on patient outcomes and can adversely affect treatment delivery.

Patients in all studies, except that of Dirix (2010),³¹ received chemotherapy as part of treatment; those treated by Dirix received postoperative RT. In general, investigators did not adjust results to account for chemotherapy-associated toxicities independent of RT-associated toxicities, which complicates interpretation of toxicity evidence for many adverse events (e.g., mucositis). This is somewhat ameliorated by our focus on studies in which chemotherapy regimens are similar between study arms, thus potentially isolating the effect of the RT modality on such outcomes.

Key Points

• The results of primary interest for this Key Question comprise comparative acute (< 90 days) and late (> 90 days) radiation-associated dysphagia, salivary gland dysfunction, xerostomia, and QOL.
• New comparative evidence assessed in this update strengthens the conclusion from CER No. 20 that the risk of grade 2 or higher late xerostomia is significantly lower in patients treated with IMRT than with 3DCRT.
• Although we identified evidence on other key toxicities (e.g., mucositis, dysphagia, skin toxicities, osteoradionecrosis of the jaw) and QOL, the reported rates compared between modalities were inconsistent. Thus, evidence on adverse events other than late xerostomia remains insufficient to alter conclusions of CER No. 20.
• Post-treatment toxicities were reported inconsistently across studies, precluding comparisons within the body of evidence. We are uncertain whether the limited evidence on RT-associated toxicities overall reflects their absence or that the investigators did not systematically collect or chose not to report them.
• The best quality evidence comprises one small (N=60), fair quality RCT (Gupta, 2012) in which 3DCRT and IMRT were compared in a regimen of CCRT to treat patients with cancer of the hypopharynx, larynx, and oropharynx. Key findings of this study relevant to Key Question 1 pertained to late xerostomia and salivary gland dysfunction.
• One nonrandomized, poor quality study of 3DCRT versus SBRT did not report on primary outcomes for Key Question 1.
• One fair quality study and one poor quality nonrandomized study reported QOL outcomes related to treatment with 3DCRT or IMRT

Qualitative Synthesis

In Table 6, we aggregate new comparative evidence related to Key Question 1 on toxicities actually reported in studies according to the intervention comparison, treatment regimen, and timeframe (acute vs. long-term). We identified no evidence from patients stratified according to tumor site(s), so we did not include tumor information in this table. Although we collected evidence on lesser NCI Common Terminology Criteria for Adverse Events or Radiation Therapy Oncology Group grades, as shown in Tables B-4 and B-5, here we present grade 2 or higher toxicities, which are likely to adversely impact patient management, hospitalization, and survival outcomes. The last two columns of Table 6 show reported proportions for each toxicity and any statistically significant results by study if so achieved.

Table 6

Key Question 1: Qualitative evidence synthesis for key reported comparative grade 2 or higher toxicity outcomes and QOL.

Strength of Evidence for Key Question 1

To evaluate the SOE, we used an approach specifically developed by the AHRQ EPC program and referenced in the “Methods Guide.”²⁰ This approach is based on a system initially described by the GRADE Working Group. It explicitly addresses four required domains: risk of bias, directness, consistency, and precision, as outlined in the Methods section.

Table 7 shows the SOE for new evidence on the comparative effects on QOL and toxicities of 3DCRT, IMRT, SBRT, and PBT in the treatment of head and neck cancer patients.

Table 7

SOE for Key Question 1: Adverse effects and QOL.

The evidence we identified for this update supports an SOE rating of “moderate” for the comparison of 3DCRT and IMRT in a regimen of CCRT, showing a benefit of IMRT in significantly reducing the incidence of late grade 2 or higher xerostomia. Two other studies showed a statistically significant reduction in the incidence of late grade 2 or higher xerostomia in two other treatment regimens (RT with or without CCT, postoperative RT). New evidence on any other RT-associated toxicity is insufficient to form conclusions on a benefit or harm of 3DCRT compared with IMRT.

Overview

In this section we summarize evidence on comparative oncologic outcomes for different RT types. As noted in the Methods section, we did not seek evidence from other study designs (e.g., single-arm observational) that may report additional outcomes not captured in the comparative studies.

Table 8 depicts key oncologic outcomes reported by each relevant study; a blank cell in each table means the outcome was not reported in that study. Not all outcomes were collected in each study. Outcomes of primary interest are overall survival, local control (no evidence of primary tumor), or locoregional control (no evidence of primary tumor or regional metastatic spread) among patients treated with IMRT compared with 3DCRT. Other oncologic outcomes were inconsistently reported across the body of studies, as shown in Table 8.

Table 8

Summary of key reported comparative oncologic outcomes.

Key Points

• The results of primary interest for this Key Question comprise overall survival, local control, and locoregional control.
• As we found in CER No. 20, comparative evidence assessed in this update was insufficient to draw relative conclusions on any oncologic outcomes.
• The key oncologic outcomes were not reported universally across studies, so we could not make comparisons across a larger body of evidence.
• The best quality evidence comprises one small (N=60), fair quality RCT (Gupta, 2012) in which 3DCRT and IMRT were compared in a regimen of CCRT to treat patients with cancer of the hypopharynx, larynx and oropharynx.
• Two additional nonrandomized, poor quality studies reported on the key oncologic outcomes with 3DCRT and IMRT in a regimen of CCRT among patients with cancer of the hypopharynx, larynx, oral cavity, nasopharynx, and oropharynx.
• One study of 3DCRT versus SBRT reported overall survival and local control in a regimen of RT with or without CCRT among patients with nasopharyngeal cancer. However, 22 percent of unidentified patients in the 3DCRT arm received concurrent high-dose rate brachytherapy so the oncologic outcomes are not included in this synthesis.

Qualitative Synthesis

In Table 9, we have aggregated new evidence related to Key Question 2 on comparative oncologic outcomes actually reported in studies according to the intervention comparison, treatment regimen, and timeframe. We identified no evidence from patients stratified according to tumor site(s), so we have not included tumor information in this table. Further, we did not identify any evidence on differences in oncologic outcomes related to the HPV status of patient tumors. The last two columns of Table 9 show reported proportions for each outcome and statistically significant results if attained.

Table 9

Key Question 2: Qualitative evidence synthesis for key reported comparative oncologic outcomes.

In general, evidence on tumor control and survival outcomes is sparse. Table 9 shows that statistically significant differences were inconsistently reported for overall survival, local control, or locoregional control among studies of 3DCRT versus IMRT in any regimen compiled there. All abstracted data are shown in detail in Table B-6.

Strength of Evidence for Key Question 2

To evaluate the SOE, we used an approach specifically developed by the AHRQ EPC program and referenced in the Methods Guide.²⁰ This approach is based on a system initially described by the GRADE Working Group. It explicitly addresses four required domains: risk of bias, directness, consistency, and precision, as outlined in the Methods section.

Table 10 shows the SOE for the comparative effects of 3DCRT, IMRT, SBRT, and PBT on oncologic outcomes in the treatment of head and neck cancer patients. The criteria we used to arrive at the SOE ratings are outlined in the Methods section of the update. Details on how the SOE ratings were determined are summarized in Table 10.

Table 10

SOE for Key Question 2: Oncologic outcomes.

We determined that new evidence, including one “fair” quality RCT (Gupta, 2012),²⁵ is insufficient to support a conclusion on the relative effect of IMRT and 3DCRT on overall survival or locoregional control rates in a regimen of CCRT. New evidence is insufficient to form conclusions on the effect of any other RT modality comparison for any oncologic outcome in any other regimen we identified in this update.

Key Points

• In CER No. 20, no comparative studies addressed these issues. In this update, we did not identify any new evidence that specifically addressed Key Question 3.
• Therefore insufficient evidence exists to form conclusions about the comparative effects or SOE on 3DCRT, IMRT, SBRT, or PBT based on specific patient and tumor characteristics.

Key Points

• In CER No. 20, no comparative studies addressed these issues. In this update, we did not identify any new evidence that specifically addressed Key Question 4.
• Therefore insufficient evidence exists to form conclusions about the comparative effects or SOE on 3DCRT, IMRT, SBRT, or PBT based on specific patient and tumor characteristics.