A systematic review and meta-analysis of treatment-related toxicities of curative and palliative radiation therapy in non-small cell lung cancer

Treatment-related toxicity is an important component in non-small cell lung cancer (NSCLC) management decision-making. Our aim was to evaluate and compare the toxicity rates of curative and palliative radiotherapy with and without chemotherapy. This meta-analysis provides better quantitative estimates of the toxicities compared to individual trials. A systematic review of randomised trials with > 50 unresectable NSCLC patients, treated with curative or palliative conventional radiotherapy (RT) with or without chemotherapy. Data was extracted for oesophagitis, pneumonitis, cardiac events, pulmonary fibrosis, myelopathy and neutropenia by any grade, grade ≥ 3 and treatment-related deaths. Mantel–Haenszel fixed-effect method was used to obtain pooled risk ratio. Forty-nine trials with 8609 evaluable patients were included. There was significantly less grade ≥ 3 acute oesophagitis (6.4 vs 22.2%, p < 0.0001) and any grade oesophagitis (70.4 vs 79.0%, p = 0.04) for sequential CRT compared to concurrent CRT, with no difference in pneumonitis (grade ≥ 3 or any grade), neutropenia (grade ≥ 3), cardiac events (grade ≥ 3) or treatment-related deaths. Although the rate of toxicity increased with intensification of treatment with RT, the only significant difference between treatment regimens was the rate of oesophagitis between the use of concurrent and sequential CRT. This can aid clinicians in radiotherapy decision making for NSCLC.

Statistical analysis. The pooled risk of toxicities by any grade, grade ≥ 3, and treatment related deaths were expressed as the total number of cases for each toxicity outcome divided by the total number of patients treated with the same type of treatment. Treatment regimens were categorised into palliative RT alone, curative RT alone, sequential CRT and concurrent CRT. We performed indirect comparisons to estimate the risk ratio for the comparison between palliative versus curative RT and sequential versus concurrent CRT. The Mantel-Haenszel fixed-effect method was used to obtain the pooled risk ratio and corresponding confidence interval. We used the fixed-effect method for all comparisons for consistency. Statistical heterogeneity was assessed by calculating I 2 . Cochrane Review Manager version 5.3 (Cochrane Collaboration, Copenhagen, Denmark) was used for the analyses.
Quality assessment. The risk of bias for each trial was assessed using the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions 14 . These include random sequence generation, allocation concealment, incomplete outcome data, selective reporting and other biases (such as method of assessing symptoms).

Results
Eligible studies. We identified 49 eligible trials  with a total of 10,388 patients, of which 8609 were evaluable for toxicity (Fig. 1). The overall trial characteristics are shown in Table 1. There was variability in the reporting of symptoms, with various versions of 5 different toxicity grading criteria used in 39 of the included trials. 8 trials included stage IV patients accounting for 1835 patients. 5 of these were palliative trials and the remaining 3 trials only had a small proportion (23 patients) of stage IV disease. Treatment characteristics of included trials are summarised in Table 2 demonstrating the heterogeneity with respect to the study design, toxicity scoring criteria, treatment arms, RT dose fractionation and chemotherapy regimen. There was a wide range of RT dose fractionation used, from 10 Gy in 1 fraction for palliative RT, up to 74 Gy in 37 fractions in concurrent CRT. Most chemotherapy regimens were platinum-based. 2 studies assessed elderly patients 17,18,60 . Treatment-related death. The overall rate of TRD was low on indirect comparisons, highest in concurrent CRT (3.1%), followed by sequential CRT (2.3%), curative radiation alone (2.4%) and palliative radiation (0%). In the 6 trials 20,22,27,33,51,53 that directly compared concurrent with sequential CRT, TRD from concurrent was higher than sequential CRT but the difference was not statistically significant (5.1% vs 2.7%, p = 0.05) (Fig. 2). In the one trial 30  Oesophagitis. Grade ≥ 3 oesophagitis from concurrent CRT was statistically significantly higher than sequential CRT in 9 trials (22.2% vs 6.4%, p < 0.0001) (Fig. 3A). Any grade oesophagitis from concurrent CRT was also statistically significantly higher than sequential CRT in 3 trials (79.0% vs 70.4%, p = 0.04) (Fig. 3B). 2 trials 48,59 compared any grade oesophagitis between curative RT and palliative RT, this was higher in curative but the difference was not statistically significant (35.4% vs 26.6%, p = 0.06) (Fig. 3C). Trials were not sufficient for meta-analysis in other comparison groups in assessing grade ≥ 3 or any grade oesophagitis.
Pneumonitis. In the 7 trials that directly compared concurrent with sequential CRT, Grade ≥ 3 pneumonitis from concurrent was higher than sequential CRT, but not statistically significant (11.1% vs 8.7%, p = 0.26) (Fig. 4A). 2 trials 38,53 directly compared any grade pneumonitis, demonstrating the rate from concurrent CRT was not statistically significantly higher than sequential CRT (10.0% vs 5.7%, p = 0.16) (Fig. 4B). Trials were not sufficient for meta-analysis in other comparison groups in assessing grade ≥ 3 or any grade pneumonitis.
Neutropenia. Neutropenia was reported in different time intervals following chemotherapy or not specified. Selective reporting of febrile neutropenia was also identified. 3 trials directly compared Grade ≥ 3 neutropenia between concurrent and sequential CRT. Rates from concurrent was higher than sequential CRT, but not statistically significant (58.2% vs 55.4%, p = 0.56) (Fig. 4C). Trials were not sufficient for meta-analysis in assessing any grade neutropenia.  www.nature.com/scientificreports/ www.nature.com/scientificreports/  www.nature.com/scientificreports/ Cardiac adverse events. 4 trials directly compared grade ≥ 3 cardiac events between concurrent and sequential CRT. The rates from concurrent was higher than sequential CRT, but not statistically significant (4.3% www.nature.com/scientificreports/ vs 1.8%, p = 0.08) (Fig. 4D). Trials were not sufficient for meta-analysis in other comparison groups for grade ≥ 3 or any grade cardiac events.
Pulmonary fibrosis and myelopathy. Pulmonary fibrosis and radiation myelopathy were poorly reported in the studies. Only 7 trials reported pulmonary fibrosis and 9 trials reported myelopathy across all treatment groups; meta-analysis to compare between groups was not feasible. The rate of pulmonary fibrosis (any grade) was higher in the palliative RT and curative RT arms than the sequential CRT and concurrent CRT. This finding is strongly influenced by a single study by Nestle et al. which reported 100% rate of pulmonary fibrosis based on imaging rather than clinical symptoms.
Toxicity stratified by stage. Trials with only stage III NSCLC comparing sequential versus concurrent CRT were analysed (see Fig. 5). The rate of grade ≥ 3 oesophagitis was statistically higher for concurrent CRT (16.6% vs 7.4%, p < 0.0001), whilst the difference in rates of treatment related death, grade ≥ 3 pneumonitis and grade ≥ 3 cardiac events were not statistically significant. Trials were not sufficient for analysis stratified by stage for stages I, II or IV disease.  Tables 3 and 4). Any grade toxicities were also lower with palliative compared with curative radiation alone. On pooled comparisons, any grade oesophagitis (RR 0.40; CI 0.32-0.50) and pneumonitis (RR 0.04; CI 0.02-0.09) were significantly less with palliative compared with curative RT alone (see Tables 3 and 4).
The range of reported grade ≥ 3 oesophagitis was 0 to 41.4% for concurrent CRT and 0 to 20.4% for sequential CRT, whilst any grade oesophagitis ranged between 46.4% and 100% for concurrent CRT and 36.4% to 100% in sequential CRT.

Risk of bias.
Reporting bias were identified with incomplete data and selective reporting of toxicities in most studies, resulting in an overall high risk of bias. Funnel plots were generated to visually assess for publication bias. Symmetrical funnel plots were obtained for comparison groups (> 5 studies) between sequential and concurrent CRT in grade 3 oesophagitis, pneumonitis and treatment-related deaths.

Discussion
In lung cancer clinical decision making, the consideration of toxicity is essential. As expected, patients receiving palliative RT had lowest toxicity, followed by curative RT alone, sequential and highest with concurrent chemoradiation. The benefit of this review is to provide better estimates of each toxicity effect compared to individual trials.
Acute oesophagitis is one of the main morbidities from lung irradiation. The large differences between individual trials makes it difficult for clinicians to estimate the toxicity in the process of informed consent. The grade ≥ 3 rate with curative RT without chemotherapy is low (0.5%). However, only 4 trials were included, 2 of which included stage I patients only 17,19,25,49 . Although this toxicity is significantly higher with concurrent CRT, it should not be used alone as a factor to preclude concurrent treatment. Oesophagitis can be managed with nutritional support and admission and rarely leads to late stenosis. In addition, IMRT have reduced the incidence of this 65,66 . This difference in oesophagitis rates should be considered as oesophagitis may impact on survival 67 .
Pneumonitis occurs sub-acutely and is the main toxicity of concern as it can result in death. Although the risk of any grade pneumonitis is high for all curative radiotherapy, the risk of Grade 3 + pneumonitis is < 10%. In addition, we found no significant difference between concurrent versus sequential CRT. This suggests that decisions regarding the sequencing of treatment should not be based on the anticipated risk of pneumonitis. However, the increasing use of adjuvant or palliative immunotherapy when combined with prior radiotherapy may potentially increase future pneumonitis risk.
Cardiac toxicity encompasses a range of disorders. Nearly all studies reported cardiac toxicity as a general outcome "cardiac" rather than specifying individual events. The pathophysiology and dose resulting in an event is likely to differ. The risk of grade ≥ 3 toxicity has been correlated with pre-existing cardiac disease and mean heart dose 68 . In breast cancer, Darby et al. found the rates of major coronary events increased linearly with the mean heart dose by 7.4% per Gray 69 . Moreover, data from RTOG 0617 showed heart dose is an independent factor for overall survival 70 . However, a systematic review which includes 3 studies from RTOG 0617 found that heart dose-volume parameters were not consistently associated with survival or cardiac toxicity 71 . Although reduction of heart dose is ideal, any de-escalation of therapy should be carefully weighed against the resulting inferior cure rates [71][72][73][74] .
Toxicities for the elderly population are not well established due to the lack of and under-representation in randomised trials. The EORTC and SIOG groups recommended chemotherapy to be considered only in selected fit elderly patients, as the added toxicity may outweigh survival benefit 75 . In this systematic review, there were only two included studies that specified elderly toxicity rates, reflecting the need for randomised trials in this group to aid determine best suitable treatment. www.nature.com/scientificreports/ www.nature.com/scientificreports/ Only one trial included adjuvant immunotherapy and no palliative immunotherapy was used. The studies included treatment with various radiation technique, dose fractionation, including escalated therapy (radiation dose 24,61 or systemic therapy). Due to changes in radiation technique, older studies (prior to 2000) were not included in this review. Advanced radiation technique such as 3-dimentional compared with 2-dimensional palliative RT to improve conformality can reduce toxicities 76 . Secondary analysis from the RTOG 0617 also confirms that IMRT was associated with lower rates of severe pneumonitis and cardiac doses in locally advanced NSCLC 77 . We did not review toxicities relating to SBRT or the impact of non-chemotherapy systematic therapy.
There are several limitations which are inherent to systematic reviews of randomised trials 78 . Whilst the selected good performance status patients may limit the generalisability, the rates reported in this review may be higher due to escalation of treatment in the experimental arms. On the other hand, real-world patients may also have pre-existing comorbidities and other patient factors which could increase toxicity. This review was unable to analyse toxicity rates based on dose-volume parameters due to insufficient data published in the trials included. Moreover, the pooled rates reported are averages of the toxicity from treatment in different stages. This likely results in an overestimation of risk for those with stage I compared to III disease 79 . The incidence of toxicities reported are crude estimates between the number of patients with toxicities and the total number of patients treated. Actuarial estimates provides a more accurate determination toxicities prevalence 80 . The findings from this review should be interpreted with some caution.
We included randomised trials as protocols with prospective data generally provides better-quality toxicity data. In concordance with Sivendran et al. on adverse event reporting in cancer clinical trial publications, we  www.nature.com/scientificreports/ identified selectivity and heterogeneity with reporting toxicities in trials 81 . The quality of studies examined ranged from low to high, contributed by reporting bias. There was variation in timing of reported toxicities, different toxicity grading criteria used and limited studies on quality of life. This highlights the need for trials to report reliable toxicity data, ideally under standardised criteria and in conjunction with the Consolidated Standards of Reporting Trials (CONSORT) recommendations 82 .
To the best of our knowledge this is the only available review of toxicity data in recent trials that compares and provides estimates of palliative radiotherapy, curative radiotherapy, sequential and concurrent chemoradiotherapy regimens. The only statistically significant difference between treatment regimens was the rate of oesophagitis with concurrent versus sequential CRT. This information is clinically useful and should be considered by clinicians and patients when weighing up the established survival benefits with the toxicity of the different treatment options. www.nature.com/scientificreports/