Examining the use of pre-operative radiation therapy in addition to perioperative chemotherapy: National Cancer Data Base vs. Surveillance, Epidemiology, and End Results.

Abstract

We appreciate the interest and efforts taken to conduct a Surveillance, Epidemiology, and End Results (SEER) analysis, the thought-provoking results, and comments from Xiao et al in regard to our recent report using the National Cancer Data Base (NCDB) to examine combined modality strategies in patients with gastric and gastroesophageal (GE) junction adenocarcinomas.1 Although both large, US-based data resources have proven useful in cancer outcomes research, several key differences between SEER and the NCDB in terms of methodology, reporting standards, and available variables merit an in-depth discussion. Owing to its population-based nature, the granularity of treatment-related variables in SEER is particularly relevant here. Chemotherapy is coded as “yes” or “no/unknown,” and radiotherapy is coded as either “beam,” “other,” or “no/unknown.” SEER does not capture any information regarding the sequencing of treatments with regard to surgery (given only before surgery vs only after vs both), the number of chemotherapy agents, the anatomic region receiving radiation, or the radiation dose. All these parameters were used in our inclusion/exclusion criteria to define the patient population of interest, and were vital toward answering our research question. Based on the above, patients included in the validation study described by Xiao et al may have been compared with each other in a way such that it does not directly address our original query (eg, comparing patients receiving neoadjuvant chemoradiation [CRT] vs those treated with perioperative chemotherapy [POC]). Other combinations also are possible due to missing information, such as postoperative CRT versus neoadjuvant chemotherapy. A percentage of patients likely received either preoperative chemotherapy or CRT. In this light, their results for the T3 to T4 subset would be compatible with results noted in the long-term follow-up of the PreOperative therapy in Esophagogastric adenocarcinoma Trial (POET) study and a similar meta-analysis.2, 3 Moreover, because SEER unfortunately does not differentiate between no treatment and unknown treatment, it leaves the possibility that a percentage of patients who were categorized as not receiving radiation (such as the POC group) actually received radiation, thereby further complicating interpretation. These issues regarding loss of granularity also have implications for propensity score matching. For the above reasons, we believed that the NCDB was better suited to address our query, and have concerns that the SEER analysis described by Xiao et al is being represented as a validation study. Xiao et al identified 1270 patients, which potentially may have fit our study's inclusion criteria. However, based on the total number of patient records, the NCDB is nearly 4 times greater in size than SEER. Due to differences in sample size, and other reasons as described above, we would be particularly interested in reviewing the demographic, disease, and treatment characteristics of their study cohort, as well as details regarding the inclusion/exclusion criteria used and the methodology used for propensity score matching. Xiao et al referenced 3 studies4-6 that they argued were inconsistent with the findings of our study.1 In the 3 retrospective studies, the study investigators compared treatment outcomes between patients treated with neoadjuvant CRT compared with POC. No patients reportedly received adjuvant chemotherapy after neoadjuvant CRT. However, our study compared outcomes in patients receiving POC with or without preoperative radiotherapy.1 Thus, our study cohort that included radiation as a modality, also received adjuvant chemotherapy. This is relevant because the value of adjuvant chemotherapy after neoadjuvant CRT for GE junction cancers cannot be discounted. As reported by Mokdad et al, the addition of adjuvant chemotherapy conferred a substantial overall survival benefit (median overall survival of 40 months vs 34 months: hazard ratio, 0.79; stratified log-rank P < .001).7 These 3 studies analyzed only patients with cancers of the lower esophagus and GE junction, whereas approximately 31% of the patients in our study had gastric adenocarcinoma primary tumors.1 Based on the above, we would argue that these studies do not conflict with our findings. Last, both analyses suffer from limitations inherent to all retrospective and large database studies. Results from studies in the prospective, randomized setting, which seek to address the question of optimal initial strategy for both gastric and GE junction adenocarcinomas, are maturing and greatly awaited. No specific funding was disclosed. The authors made no disclosures.