Eligible studies directly comparing IMRT with either 2D-RT or 3D-RT in the radiotherapeutic management of HNSCC were identified through a systematic search of the medical literature using a validated search strategy. An electronic search of Medline via PubMed was conducted from January 1995 onwards till May 2017 with no language or publication status restrictions. The Cochrane Central Register of Controlled Trials (CENTRAL) and Database of Abstracts of Reviews of Effectiveness (DARE) were also searched electronically from inception till May 2017. Details of the search strategy are presented separately (S1 Appendix). Electronic search was further supplemented by hand-searching of review articles, cross references and conference proceedings.

For inclusion in the meta-analysis, trials had to be RCTs, include previously untreated patients with non-metastatic HNSCC, offer curative-intent treatment either in the definitive or post-operative adjuvant setting, and not be confounded by additional therapeutic differences between the two groups. Trials using brachytherapy boost in addition to external beam RT were also included, provided the boost was offered in both arms. Trials using chemotherapy (induction, concurrent, or adjuvant) were also considered eligible provided the chemotherapy regimen (drugs, dosage, scheduling) was identical in both arms. For trials with more mature data published or presented at a date later than the index publication, relevant data was also extracted from the update.

Two reviewers (TG and SK) independently extracted relevant data from individual studies with discrepancy, if any, being resolved by consensus. Xerostomia was the primary outcome of interest while LRC, OS, and QOL were secondary endpoints. Outcome data was extracted and pooled using the Cochrane methodology [22] for meta-analysis using the fixed-effects or random-effects model as appropriate and expressed as risk ratio (RR) or hazard ratio (HR) with respective 95% confidence intervals (CIs). Trials were subgrouped appropriately according to RT technique (2D-RT or 3D-RT) for toxicity outcomes (xerostomia) and site of primary (nasopharynx or laryngo-pharynx) for efficacy outcomes (LRC and OS). In addition, analysis for late xerostomia was done at different time-points (6-month, 1-year, 2-year and 3-year) for better interpretation and informed decision-making. The analysis, interpretation, and reporting of results also included a risk of bias assessment [22] for all included individual studies and grading [23] of the strength of recommendation. All analyses were done using Review Manager (RevMan) version 5.3 and GRADE profiler (GRADEpro) version 3.6.1 (The Nordic Cochrane Centre, Cochrane Collaboration, 2008).

Treatment characteristics and clinical outcomes of patients included in the 7 RCTs directly comparing IMRT with 2D-RT/3D-RT are summarized in Tables 1 and 2 respectively. Salivary gland toxicity (xerostomia) was the primary endpoint for 5 of the 7 included RCTs [15–18,20], and secondary endpoint [19,21] in two of them. Only a single study each used LRC [21] and OS [19] as primary endpoints. Assessment of salivary gland toxicity was variable and heterogeneous in terms of scoring criteria and time-points of assessment. One study [15] assessed salivary function by stimulated whole salivary flow-rates and defined severe toxicity as post-RT salivary flow <25% of pre-RT flow [27]. Another study [19] reported radiation toxicity including xerostomia by the Common Toxicity Criteria (CTC) version 3.0 [28]. Five studies [16–18,20,21] reported xerostomia using the Radiation Therapy Oncology Group (RTOG)/European Organization for Research and Treatment of Cancer (EORTC) morbidity criteria [29], while one study [17] primarily used the Late Effects on Normal Tissue/Subjective Objective Management Analytic (LENT SOMA) scale [30,31]in addition to the RTOG criteria. Five of 7 studies were rather small comprising of less than 100 patients (in both arms). Only two studies, one from China that included over 600 patients with nasopharyngeal cancers [19] and the other from France [21] led by Groupe d’Oncologie Radiotherapie Tete Et Cou (GORTEC 2004–01) with a planned accrual of over 300 patients with oro-hypopharyngeal cancers were adequately powered for disease-related outcomes. Unfortunately, GORTEC 2004–01 study [21] had to be terminated prematurely (after randomizing 188 patients) due to slow accrual and adoption of IMRT as standard treatment in France. Three studies [15,16,19] were restricted to patients with nasopharyngeal cancer while the remaining 4 studies [17,18,20,21] included patients with non-nasopharyngeal sites such as oropharynx, hypopharynx, larynx, and oral cavity. Four studies used 2D-RT as the control arm [15–17,19], while 3D-RT was the control arm in 3 studies [18,20,21]. QOL was assessed by the EORTC general Quality-of-Life Questionnaire (QLQ-C30) and specific Head Neck (HN35) module in 4 studies [15,17,18,20] and a customized xerostomia questionnaire (XQ) in one study [16]. One study each used Medical Outcomes Short Form (SF36) [15] and modified XQ [17] in addition to EORTC QOL questionnaires for additional information on patient-reported outcomes.

There was modest heterogeneity in the included trials allowing statistical pooling of results. The primary outcome of interest i.e. xerostomia was reported in all included studies. All 7 studies reported significant reduction in moderate to severe acute xerostomia (during or within 3-months of completion of RT). The use of IMRT was associated with a 36% relative risk reduction in grade 2 or worse acute xerostomia (RR = 0.64, 95%CI = 0.49–0.84; p = 0.001) compared to 2D/3D-RT (Fig 2). On subgroup analyses stratified by technique of irradiation (Fig 2), the magnitude of benefit with IMRT over 2D-RT (RR = 0.66, 95%CI = 0.47–0.95; p = 0.02) was comparable to its benefit over 3D-RT (RR = 0.62, 95%CI = 0.44–0.86; p = 0.004). More importantly, IMRT significantly reduced the risk of late grade 2 or worse xerostomia (RR = 0.44, 95%CI = 0.34–0.57; p = 0.00001) compared to non-IMRT techniques (Fig 3). Given the expected gradual recovery of xerostomia over time, data on late xerostomia was also extracted and pooled at different time-points as previously described. The risk of grade 2 or worse xerostomia was consistently and significantly reduced with IMRT at all defined time-points (Fig 3). This reduction is late grade 2 or worse xerostomia with IMRT remained statistically significant even on subgroup analyses comparing IMRT with either 2D-RT or 3D-RT.

All 7 RCTs presented Kaplan-Meier curves and/or reported HRs for local and regional disease control allowing data extraction and pooling for loco-regional failures. Overall, the use of IMRT was associated with a 24% relative reduction (HR = 0.76, 95%CI = 0.57–1.01) in the risk of loco-regional relapse compared to 2D/3D-RT (Fig 4), which was statistically of borderline significance (p = 0.06) as the upper bound of the 95%CI was just touching the line of unity. However, there were notable differences in LRC between IMRT and 2D/3D-RT on subgroup analyses stratified by site of primary tumor, suggesting that impact may be dependent upon primary site. For nasopharyngeal cancers, use of IMRT resulted in a very significant (48%) relative reduction in the risk of loco-regional failure (HR = 0.52, 95%CI = 0.34–0.80; p = 0.003) compared to non-IMRT techniques (Fig 4). Conversely, for cancers arising in the laryngo-pharynx, there was no significant difference in LRC between IMRT and 2D/3D-RT (HR = 1.06, 95%CI = 0.71–1.58; p = 0.78) (Fig 4).

Five of the 7 RCTs provided extractable data on survival. In general, the use of IMRT was associated with a 30% relative reduction in the risk of death (HR = 0.70, 95%CI = 0.57–0.88; p = 0.002). However, this benefit of IMRT was again mostly dependent on primary tumor site and clearly driven by the large Chinese trial in carcinoma nasopharynx (HR = 0.57, 95%CI = 0.42–0.78; p = 0.0005). For patients with laryngo-pharyngeal caners, the use of IMRT was associated with a non-significant reduction in the risk of death (HR = 0.85, 95%CI = 0.63–1.15; p = 0.29) compared to 2D/3D-RT (Fig 5).

Although included as an endpoint in 5 RCTs, comparison of health-related QOL outcomes between IMRT and 2D/3D-RT was reported in 4 studies including one study that analyzed longitudinal evolution of QOL over time in both arms combined. There was substantial deterioration in most QOL domains immediately following RT which gradually improved over time [26]. There was consistent, gradual, though partial recovery of salivary function over time, which was significantly higher with IMRT than 2D-RT/3D-RT [15–17,26]. Xerostomia-related QOL scores were also significantly better preserved with IMRT [15–18] although global QOL was not very different [15,17] between IMRT and conventional techniques (2D/3D-RT). Lack of extractable data from most studies precluded any statistical pooling of QOL outcomes. Sensitivity analysis for all relevant outcomes of interest demonstrated lack of influence of any single study on the overall magnitude and direction of effect, interpretation, and conclusions (S1 Fig). A relatively symmetric funnel plot ruled out the presence of any significant publication bias in the weighted pooled meta-analysis (S2 Fig).

The present meta-analysis is based only on RCTs directly comparing IMRT with either 2D-RT or 3D-RT in HNSCC identified from the medical literature using a validated search strategy. Although there was some heterogeneity across included studies, they were quite similar in terms of study design, methodology, analyses, and reporting. Appropriate subgroup analysis was done after stratifying on technique of irradiation (2D-RT or 3D-RT) and site of primary tumor (nasopharynx or laryngo-pharynx). Data on xerostomia was also pooled at later time-points to ascertain whether the significant benefit of reduction in the risk of moderate to severe acute xerostomia with IMRT persisted over time. The quality of included studies was judged to be moderately high with low to unclear risk of bias for xerostomia (primary outcome measure). However, the quality of studies for disease-related outcomes was downgraded due to high-risk of bias. Notwithstanding, disease-related outcomes were also compared providing useful information on the efficacy of IMRT for HNSCC in the curative-setting. No significant publication bias was detected for any of the outcome measures in this analysis. However, certain caveats and limitations remain. Due to lack of complete reporting in index RCTs, this meta-analysis did not attempt any comparison between IMRT and 2D/3D-RT for other significant acute toxicities of comprehensive head-neck irradiation (apart from xerostomia) such as mucositis, dermatitis, and dysphagia/odynophagia. Although xerostomia remains the most debilitating long-term toxicity of radio(chemo)therapy, chronic dysphagia can also negatively impact health-related QOL in survivors [4,6]. This meta-analysis could not assess the impact of IMRT on late dysphagia as it was not an endpoint in any of the included primary studies. An ongoing RCT that compares dysphagia aspiration related structures (DARS)-sparing and dysphagia-optimized IMRT with standard IMRT should provide definitive answers [36]. Furthermore, this meta-analysis, could not quantify QOL difference, if any, between the two techniques due to lack of easily and readily extractable or available data. It is widely accepted that IMRT is associated with significantly higher cost and complexity, compared to conventional techniques. However, none of the trials included any analyses of cost-effectiveness of IMRT, precluding any such estimation in the meta-analysis. The time-frame of included studies in the meta-analyses was variable with potential differences in quality of RT; advances in dose calculation and heterogeneity correction algorithms over time could have further improved disease control and reduced toxicity, irrespective of RT delivery technique. Finally, individual patient data was not available for pooling in this meta-analysis which was based primarily on summary data extracted from the published medical literature.