Gastrointestinal Cancers: Timing Is Everything

In this edition of Oncology Scan, we have some transitions to announce. First, Stanley Liauw will be rotating off the editorial board. Dr Liauw has been an Associate Editor since 2013, during which time he has been an integral member of the editorial team, providing thoughtful reviews and carefully rendered, well-balanced decisions. We thank him for his service these past 4 years. Additionally, I will be stepping down as Senior Editor of the gastrointestinal section, and we are very pleased to announce that Salma Jabbour will assume the role as the new Senior Editor.

Dr Jabbour has served as an Associate Editor for nearly 2 years, and I have also very much appreciated her hard work and dedication in ensuring high-quality, timely, and thoroughly considered reviews and decisions. I am confident she will be successful in her new role as she takes over a group of outstanding Associate Editors. It has been a pleasure working with them, with Dr Anthony Zeitman, and the entire staff of the International Journal of Radiation Oncology, Biology, Physics (the Red Journal) to continually improve the journal’s quality through original research publications, special edition articles, these Oncology Scans, and newer article formats to engage the radiation oncology community. I would like to thank Anthony for this rewarding opportunity to be involved with the Red Journal and for his support, along with the rest of the Red Journal staff and editorial board.

We have selected 6 important articles for this edition of the Oncology Scan. The first is the recently published ESPAC-4 trial, showing that the combination of gemcitabine with capecitabine has improved survival compared with gemcitabine alone as adjuvant therapy for resected pancreatic cancer (1). Although the trial does not directly address a management decision in radiation oncology, the results are important nonetheless in moving the needle forward toward improving survival and systemic control, hopefully paving the way for an increased role for radiation therapy to meaningfully improve survival through better local-regional control. The next study is a secondary analysis of the landmark ACT II trial for anal cancer, looking at optimal timing of clinical assessment (2). The results should give clinicians guidance on when to assess for treatment response and when diagnostic and therapeutic interventions should and should not be considered. In keeping with the theme of timing, the next 2 articles are important prospective trials for rectal cancer. The Stockholm III trial, which previously reported interim toxicity results (3), was recently published showing outcome data (4). This trial tested short-course radiation therapy with 4 to 8 weeks’ delay before surgery compared with the traditional short course and surgery 1 week later and long-course radiation, showing no difference in local-regional control or survival with delay after short-course radiation. The second article reports results of the GRECCAR-6 trial, which examined 7 versus 11 weeks’ delay after chemoradiation and surgery for rectal cancer (5). These data help shed further light on the issue of optimal timing of pathologic response assessment, which is relevant to consider given the increasing interest in alternative treatment approaches including local excision and nonoperative management (NOM) for rectal cancer. Finally, the last 2 articles, which come from the same author group, use the National Cancer Database (NCDB) to describe the use of NOM in the United States and its associated outcomes (6, 7). We believe these articles are important for readers to be aware of because the authors use these data to caution against the use of NOM, though for reasons discussed below, these conclusions may be called into question.

Neoptolemos et al. Comparison of adjuvant gemcitabine and capecitabine with gemcitabine monotherapy in patients with resected pancreatic cancer (ESPAC-4): A multicentre, open-label, randomised, phase 3 trial. Lancet 2017. (1)

Summary:

In this randomized trial, 730 patients with an R0 or R1 resection for pancreatic cancer were randomized to adjuvant gemcitabine chemotherapy (1000 mg/m², weeks 1–3) for 6 monthly cycles, alone or in combination with capecitabine (1660 mg/m², days 1–21) for 6 monthly cycles. The primary endpoint was overall survival (OS). Patients were drawn from 92 hospitals across the United Kingdom, France, Germany, and Sweden.

With a median follow-up of 43.2 months, the trial showed improved OS (P=.032) with gemcitabine and capecitabine compared with gemcitabine alone (Table 1).

Table 1.

Survival outcomes for each arm of ESPAC-4

Outcome	Gemcitabine	Gemcitabine + capecitabine
Median OS (mo)	25.5	28
2 year OS (%)	52.1	53.8
5-year OS (%)	16.3	28.8
Median OS (R0 resection) (mo)	27.9	39.5
Median OS (R1 resection) (mo)	23	23.7

On multivariate analysis, treatment arm, resection margin, postoperative CA 19–9, tumor grade, lymph node positivity, and maximum tumor size were significantly correlated with OS. Median dose intensity was 93% in the gemcitabine-alone arm and 83% in the gemcitabine and capecitabine arm, and 65% of gemcitabine patients received all 6 cycles, compared with 54% of gemcitabine and capecitabine patients. The distant relapse rate was similar between the gemcitabine-alone (66%) and the gemcitabine and capecitabine arms (65%). The overall local relapse rate was 50%, higher in the gemcitabine alone arm (53%) than in the gemcitabine and capecitabine (46%). Grade 3 to 4 adverse events were significantly worse for gemcitabine and capecitabine with respect to diarrhea (5% vs 2%), neutropenia (38% vs 24%), and hand-foot syndrome (7% vs 0%). No difference in quality of life was noted between the 2 arms.

Comments:

For resected pancreatic cancer, the ESPAC-4 study establishes that gemcitabine combined with capecitabine is a new standard of care in adjuvant chemotherapy regimen. This combination provides a median survival benefit of 2.5 months over gemcitabine alone. This trial also demonstrates outstanding collaboration among the European community, having accrued 732 patients, but it is not entirely uniform in the patient characteristics and styles of management. For example, the study design did not specifically dictate the method of follow-up, and options for follow-up could include hematology, clinical chemistry, and use of a tumor marker. Additionally, there did not seem to be a clear plan at which time points to monitor patients or to routinely image patients as part of their surveillance. Likewise, it is unclear what methods were taken to classify patients into resectable versus borderline resectable status preoperatively, if at all.

Although patients underwent up-front surgery, 60% of patients in this study had R1 resections, which are known to portend a worse survival rate than R0 status (8). It is known that the prognosis of pancreatic cancer resection may be improved with higher-volume pancreatic cancer hospitals (9). In addition, in the United States the paradigm has generally shifted to favor preoperative therapy if the pancreatic cancer isdeemed “borderline resectable,” because this term implies that up-front surgery would result in R1 resections, owing to vasculature abutment. For patents with borderline resectable pancreatic cancer, preoperative therapy should be considered (10, 11). In a prospective study consisting of induction FOLFIRINOX and 50.4 Gy of chemoradiation, 68% of patients underwent surgery and 93% had negative margins. Median OS was 21.7 months (12).

In the ESPAC-4 study, subjects with R1 status experienced median survival rates of 23 and 23.7 months, compared with patients with R0 resections who experienced a median survival of 27.9 and 39.5 months for gemcitabine alone and gemcitabine plus capecitabine, respectively. These data suggest that intensifying chemotherapy is not beneficial in the setting of positive margins. It is possible that the high rate of R1 resections resulted in higher local recurrence rates, and in total, 50% of patients had local recurrence at a site of relapse. However, the article does not correlate pattern of relapse to margin status, and local recurrence may have been alone or with synchronous systemic relapse (liver 41%, other intraabdominal 23%, lung 11%, bone 3%). These high rates of local recurrence beg the question of whether postoperative radiation therapy has a role.

An enduring question in pancreatic cancer management, both in the resectable and borderline settings, is whether radiation therapy is of benefit. To date there seem to be no survival benefits from the incorporation of postoperative radiation therapy in addition to traditional chemotherapy in modern series (13). However, with incremental improvements in systemic therapy, the potential for local-regional control to meaningfully impact OS increases. Ongoing prospective studies such as Radiation Therapy Oncology Group protocol 0848 will elucidate the role of radiation therapy in the postoperative setting, and the Alliance trial A021501 will do so in the preoperative setting for borderline resectable pancreatic cancers.

Glynne-Jones et al. Best time to assess complete clinical response after chemoradiotherapy in squamous cell carcinoma of the anus (ACT II): A post-hoc analysis of randomised controlled phase 3 trial. Lancet Oncol 2017. (2)

Summary:

The ACT II trial randomized 940 patients with squamous cell carcinoma of the anal canal (SCCA) in a 2 × 2 design. In the first randomization, patients were randomized to the standard of care of concurrent chemotherapy with mitomycin (12 mg/m² × 1) and 5-fluorouracil (5-FU) (1000 mg/m² per day on days 1–4 and 29–32) versus cisplatin (1 dose of 60 mg/m² on days 1 and 29) and 5-FU with radiation (50.4 Gy at 1.8 Gy per fraction). The second randomization was to 2 additional cycles of maintenance cisplatin and 5-FU chemotherapy versus no maintenance chemotherapy. The initial results showed no difference in OS or progression-free survival (PFS) between concurrent mitomycin/5-FU/radiation and cisplatin/5-FU/radiation and no difference in OS and PFS for maintenance chemotherapy versus no maintenance chemotherapy (14). This secondary analysis was undertaken to compare clinical complete response (cCR) rates at various time points. Clinical assessment was done at 11 weeks, 18 weeks, and 26 weeks after beginning chemoradiation.

The results showed that among the total population of the 940 enrolled patients, the cCR rate was 52% at 11 weeks, 71% at 18 weeks, and 78% at 26 weeks. In the subgroup of 691 patients who had assessments at all 3 time points, the cCR rate was 64% at 11 weeks, 81% at 18 weeks, and 85% at 26weeks. In addition, 72% of patients who did not have a cCR at 11 week sultimately did at 26 weeks, where as 93% of patients who had a cCR at 11 weeks maintained it at 26 weeks. There was no difference in attaining a cCR between the cisplatin arm and the mitomycin arm. Tumor size, nodal status, no involvement of neighboring organs (at 11 weeks), and sex (at 26 weeks) were significantly correlated with cCR. Finally, OS and PFS were significantly correlated with cCR (Table 2).

Table 2.

Correlation of overall and progression-free survival with achieving a clinical complete response

	5-y overall (n = 940)		5-y progression-free survival survival (n = 940)
Time point	% Survival	P	% Survival	P
11-wk		.0005		.0002
cCR	83		75
No cCR	72		63
18-wk		<.0001		<.0001
cCR	84		75
No cCR	59		53
26-wk		<.0001		<.0001
cCR	87		80
No cCR	46		33

Abbreviation: cCR = clinical complete response.

Comments:

This important secondary analysis from the ACT II trial addresses the timing of response assessment for patients with SCCA after definitive chemoradiotherapy. This population requires particularly thoughtful clinical monitoring, because early diagnostic procedures may contribute to unnecessary salvage abdominoperineal resection for patients who would otherwise achieve a cCR with continued surveillance. The decision of when to intervene with diagnostic or therapeutic procedures presents a common challenge, with only 52% of patients in the ACT II trial achieving a cCR at the initial 11-week evaluation. Importantly, evaluation time points were defined relative to the start of chemoradiotherapy, with the 11-week evaluation occurring approximately 6 weeks after completion of chemoradiotherapy. This early time point is suggested both by US and European consensus guidelines as the time of initial, cautious clinical assessment, rather than a final decision point for salvage treatment (15, 16). Current National Comprehensive Cancer Network (NCCN) guidelines recommend initial evaluation at 8 to 12 weeks after completion of chemoradiotherapy (14–18 weeks after the start of chemoradiotherapy), with close surveillance recommended for patients with nonprogressive but persistent disease, and consideration of biopsy at 6 months (16).

The ACT II study confirms that a subset of patients with SCCA will achieve a delayed but cCR to therapy, with 5-year PFS and OS equivalent to those patients achieving an early cCR. The authors interestingly suggest that such a delayed response effect may have contributed to the unexpectedly high success rate (50%) of salvage therapy (9 Gy) administered in the previously reported Radiation Therapy Oncology Group protocol 8704 (17). These findings are reassuring for patients who achieve a cCR at any time point up to 26 weeks after treatment initiation, but the ideal approach for patients who do not achieve a cCR at 26 weeks remains unknown. The 5-year OS and PFS in this group remain poor, although abdominoperineal resection may provide long-term survival in selected patients (18–20). Better prediction of patients who will ultimately have an incomplete response would provide opportunity for early salvage, noting that a benefit of early compared with delayed salvage in this poor risk group is unknown. Increased vigilance may be warranted at the 18-week evaluation, because the proportion of patients who will achieve a cCR between assessments 2 and 3 is substantially smaller (an increase from 81% to 85%) than the increment between assessments 1 and 3 (64%−85%). Clinical judgment in the context of multidisciplinary evaluation remains imperative, based both on individual patient risk and likely success of salvage treatment. For example, patients with high-risk tumors that are initially fixed on physical examination but become mobile and potentially resectable at the 6-week posttreatment time point may be considered for early biopsy under current NCCN guidelines. Overall, this secondary analysis confirms that favorable clinical outcomes are expected for patients achieving a cCR at any time point up to 26 weeks after treatment initiation and affirms the appropriateness of cautious surveillance for those with an incomplete response at the early 11-week assessment (6 weeks after treatment completion). Further study is needed to select for patients at an earlier time point who will not achieve a cCR by the 26-week evaluation and to develop more effective salvage treatments for this poor risk group.

Erlandsson et al. Optimal fractionation of preoperative radiotherapy and timing to surgery for rectal cancer (Stockholm III): A multicentre, randomised, non-blinded, phase 3, non-inferiority trial. Lancet Oncol 2017. (4)

Summary:

The Stockholm III trial was a randomized, phase 3 noninferiority trial to address the optimal fractionation regimen and timing of preoperative radiation therapy for patients with any stage of resectable, non-metastatic rectal cancer within 15 cm from the anal verge. Initially patients were randomized 1:1:1 to 3 treatment arms: (1) short-course radiation therapy (25 Gy in 5 fractions) with surgery within 1 week (SRT); (2) short-course radiation therapy with surgery delayed 4 to 8 weeks later (SRT-delay); or (3) long-course radiation therapy (50 Gy in 25 fractions) without concurrent chemotherapy, followed by surgery 4 to 8 weeks later (LRT-delay). All patients underwent total mesorectal excision (TME). The primary endpoint was time to local recurrence, and secondary endpoints were OS, postoperative mortality, postoperative complications, late complications, acute radiation toxicity, sphincter preservation rate, and quality of life (to be reported later). The primary comparison was between SRT (reference group) and SRT-delay or LRT-delay.

The trial was a noninferiority trial, with treatment considered noninferior if the hazard ratio did not exceed 1.7 of a 1-sided 90% confidence interval based on an initial assumption of a 15% cumulative incidence of local recurrence. The trial was designed to have 80% power by enrolling 840 patients. Because local recurrence rates were found to be significantly lower than anticipated early during enrollment, the trial was amended and repowered such that noninferiority would be shown if the hazard ratio did not exceed 2.5 or 3.6. Owing to multiple reasons, including hospital preference for SRT or SRT-delay, the trial was also amended such that hospitals were allowed the option to enroll in a 2-arm randomization between SRT or SRT-delay instead of the 3-arm randomization. Patients in the 3-arm randomization were analyzed separately from the 2-arm randomization. In addition, patients in the SRT and SRT-delay groups were pooled and analyzed together from both randomization groups. In total, 840 patients from 1998 to 2013 were enrolled: 385 in the 3-arm randomization and 455 in the 2-arm randomization between the 2 short-course arms. In total there were 357 SRT, 355 SRT-delay, and 128 LRT-delay patients enrolled. Adjuvant chemotherapy became standard after 2007, which was given to 15% of all patients (no difference among the patient groups). Minimum and median follow-up was 2.0 and 5.2 years, respectively.

No differences were seen in cumulative incidence of local recurrence, cumulative incidence of distant metastases, or OS among groups in the 3-arm randomization or between the 2 SRT arms after pooling from the 3- and 2-arm randomizations. SRT-delay was found to be noninferior to SRT for all oncological outcomes, including local recurrence. The cumulative incidence of local recurrence (as any event) was low at 2%, 3%, and 5% in the SRT, SRT-delay, and LRT-delay patients, respectively, in the 3-arm randomization (nonsignificant). Although there was no statistically significant difference in acute radiation toxicity requiring hospitalization among the 3-arm randomization groups, more patients in the SRT-delay group were hospitalized owing to radiation toxicity compared with SRT patients in the pooled SRT analysis (7% vs <1%). However, complications were less frequent in the SRT-delay patients compared with SRT patients in the pooled SRT analysis (any postoperative complication 41% vs 53%, any surgical complication 28% vs 36%, respectively). Late complications were similar among groups in either the 3-arm randomization or pooled SRT analysis: most commonly bowel obstruction, 11%; and pelvic abscesses, 7%.

Comments:

The Stockholm III trial is unique in that it is the only published phase 3 trial that has attempted to address the question of surgery timing after SRT. Other published and ongoing phase 3 trials that have studied different timings of surgery after RT have been in the context of comparing SRT with LRT with concurrent chemotherapy (21–24). Although LRT-delay patients were included in this trial, the reported oncologic outcomes and complication results were focused on comparing the SRT and SRT-delay patients. This trial provides the first level 1 evidence that delaying surgery to 4 to 8 weeks after SRT, similar to what is done in LRT treatment, is superior to immediate surgery after SRT, owing to significantly lower postoperative complications. This benefit may come at the price of increased severe radiation toxicity when delaying surgery after SRT but does not seem to be any more frequent than what was seen in the LRT-delay patients in this study.

The LRT-delay data seem to be more of an afterthought because of the substantially lower number of patients compared with the SRT and SRT-delay patients, likely owing to treatment bias for some form of SRT within the participating Swedish centers. Although the primary comparison in the study was not between SRT-delay and LRT-delay, it is noteworthy that LRT-delay did not have any advantages in any of the oncologic outcomes. Because concurrent chemotherapy was not used in the LRT-delay patients, the applicability of these data is diminished, given that the current clinical standard is concurrent 5-FU–based chemotherapy with LRT. It is not clear whether SRT-delay is equivalent to LRT-delay with concurrent chemotherapy in terms of local control, particularly for more locally advanced tumors that, for instance, threaten the circumferential resection margins and require optimal tumor down-staging.

What is clear is that level 1 data are mounting in favor of using SRT-delay as a standard alternative to LRT treatment. Randomized trials from Poland and the Trans-Tasman Radiation Oncology Group Trial that compared SRT with LRT-delay with concurrent chemotherapy have consistently shown lower severe acute toxicities with SRT and no differences in severe late toxicities, local control, disease-free survival, and OS (21, 25). Critics of these trials point out concerns regarding higher positive circumferential radial margins (21) and possible higher local failures in subgroups of patients with distal tumors with SRT (25). These concerns may potentially be addressed by resequencing and moving systemic chemotherapy up front in the preoperative setting. As the authors have pointed out, systemic chemotherapy could be interdigitated with SRT-delay before surgery such that patients are receiving full-dose systemic chemotherapy to treat micrometastatic disease at an earlier time point. This treatment approach may provide a win-win scenario (patients get treated with systemic therapy earlier and the overall duration of radiation treatment is shortened), and it is currently being studied in the phase 3 setting, with promising results (23, 24, 26).

Although LRT is still more widely practiced in the United States and other countries, SRT may have a brighter future as a worldwide standard treatment option given the greater flexibility it provides for integrating with systemic chemotherapy. Further details in future publications on the quality of life and late toxicity results from this trial will provide additional insight on the SRT-delay approach.

Lefevre et al. Effect of interval (7 or 11 weeks) between neoadjuvant radiochemotherapy and surgery on complete pathologic response in rectal cancer: A multicenter, randomized, controlled trial (GRECCAR-6). J Clin Oncol 2016. (5)

Summary:

In this randomized trial from France, 265 patients with locally advanced rectal cancer (T3, T4, or N+) of the mid- or lower rectum were treated with standard of care long-course chemoradiation with 45 to 50 Gy with concurrent intravenous 5-FU or capecitabine and then randomized to a 7-week delay or an 11-week delay after the end of chemoradiation until TME. The primary endpoint was pathologic complete response (pCR) rate, and the secondary endpoints were postoperative morbidity, sphincter preservation rate, OS, and disease-free survival. The trial was powered to detect an increase of pCR rate from 12% with 7-week delay to 26% with 11-week delay. Investigators allowed a ±5-day window from the assigned date to complete the TME.

In total, 265 patients were enrolled, with 253 patients having undergone TME with pathologic specimens available for response assessment and postoperative morbidity data. Overall, 80% of patients underwent surgery within the allowed time window for the assigned group; 6% of patients had surgery earlier, and 14% had surgery later. No difference was seen between the 7-week and 11-week arms for distant metastases during the waiting period (2% in both arms) or the rate of sphincter preservation (90% vs 89%, P=.726). No difference in pCR rate was seen between the 7-week group and the 11-week group by intention to treat (15% vs 17%, P=.598) or by per-protocol analysis (17% vs 16%, P=.78). Although laparoscopic TME was the most common approach used (83%), a nonsignificantly higher proportion of patients in the 11-week group required conversion from a laparoscopic to an open TME compared with the 7-week group (15% vs 10%, P=.26), and 7 of 8 cases of conversion due to pelvic dissection difficulties occurred in the 11-week group. Operative time was 15 minutes longer in the 11-week group (P=.35). The completeness of the TME was lower for the 11-week group compared with the 7-week group (78% vs 90%, P=.0156). Overall postoperative morbidity (45% vs 32%, P=.04) and medical complications (33% vs 19%, P=.01) were worse in patients in the 11-week group compared with the 7-week group, and hospital stay was 2 days longer for the 11-week group, although this difference was not statistically significant.

Comments:

In rectal cancer, retrospective studies have suggested higher rates of pathologic response with longer delays of 6 to 8 weeks or greater between chemoradiation and TME (27–29). Sloothak et al (30) showed in a Dutch population study that <10 to >11 weeks from the end of chemoradiation was associated with the highest pCR rate (18%) compared with <10 weeks and >11 weeks. Prospective, randomized data have been limited, and the impact on surgical complications has not been carefully considered. The Lyon 90–01 study was a randomized, prospective trial that found a higher pCR or near pCR rate for a 6 to 8 week interval (26%) compared with a 2-week interval (10%), without a significant difference in perioperative morbidity (31). The GRECCAR-6 study analyzed the effect of waiting an additional month after chemoradiation to TME on pCR and surgical morbidity. It found that pCR rates were similar, whereas surgical morbidity was higher with the longer delay. The increase in surgical morbidity was found when evaluating patients by the Dindo classification for surgical morbidity and medical complications (driven primarily by urinary symptoms) within 90 days of surgery. Technical aspects of surgery also seemed to be adversely affected, although not all of these differences were statistically significant.

The GRECCAR-6 study further contributes to our understanding of how to sequence multimodal treatment in locally advanced rectal cancer. When TME follows traditional chemoradiation, this study suggests that the right balance between tumor response and surgical complication may be appropriately addressed by the 5- to 10-week time frame proposed by the NCCN guidelines (32). Either a shorter delay (such as 1 week [4]) or a longer delay (>10 weeks) may adversely affect this balance. The optimal timing of surgery, however, is subject to change with the shifting paradigm of multimodal treatment. Adjuvant chemotherapy, which has traditionally been administered postoperatively, may have a role in the preoperative setting and influence both clinical response and surgical outcomes. A nonrandomized prospective trial by Garcia-Aguilar et al (33) demonstrated that pCR after chemoradiation was significantly higher after gradual increases in the time interval to TME up to 20 weeks, provided that FOLFOX chemotherapy was given during that delay. In this experience, there was no increase in complications or adverse effect on the surgeon’s rating of the technical difficulty of resection, despite a greater appreciation of pelvic fibrosis. Similarly, a randomized trial by Bujko et al (26) comparing short-course RT and 3 cycles of FOLFOX against long-course chemoradiation showed no increase in surgical complications, suggesting that a longer delay from the end of short-course RT to TME may not be harmful in the setting of ongoing preoperative therapy. The question of how to optimally time surgery is one that will continually be evaluated, especially as we modify initial therapy and test whether we can safely forgo either radiation (NCT01515787) or surgery (NCT02008656), depending on response. For now, however, when offering patients standard doses of 5-FU–based long-course chemoradiation for locally advanced rectal cancer, we have level 1 evidence to support a 7-week, rather than an 11-week, interval.

Ellis et al. Long-term survival after chemoradiotherapy without surgery for rectal adenocarcinoma: A word of caution. JAMA Oncol 2017. (7)

Summary:

In these 2 studies, Ellis and colleagues use the NCDB to examine the patterns of care in the United States with respect to the use of NOM among rectal cancer patients, examine factors that correlate with its use, and report on survival outcomes for patients treated with NOM. In the patterns of care study, the NOM cohort included patients with rectal cancer in the NCDB from 1998 to 2010 treated with chemoradiation for which surgery was “not part of the first course of treatment.” Patients were excluded if their treatment was classified as palliative, they had stage IV disease or prior cancer history, they refused surgery, surgery was recommended but not performed, and surgery was contraindicated because of risk factors. In the outcomes study, which only included patients from 2004 to 2008, the NOM cohort included only stage II and III rectal cancer patients treated with chemoradiation. Patients who died within 1 year of diagnosis were excluded to account for immortal timebias.

The patterns of care study showed an increase in the use of NOM from 2.4% in 1998 to 5% in 2010. Patients who were black, uninsured or enrolled in Medicaid, and treated at a low-volume center were more likely to be treated with NOM. Within high-volume centers, the same trend of increased use in black and uninsured/Medicaid patients was seen. In the outcomes study, patients treated with NOM had worse OS compared with patients treated with surgery, even after adjusting for race, insurance status, and other clinical factors found to be associated with OS. The authors raise concern about the use of NOM and call for more comparative effectiveness studies.

Comments:

The primary standard of care for stage II/III rectal cancer includes neoadjuvant chemoradiation followed by TME and adjuvant chemotherapy (34). A fraction of these patients will achieve a cCR after radiation, and a smaller fraction will be found to have a pCR at the time of surgery. Achieving a CR is associated with an excellent prognosis in this patient population, which has raised the question of whether an NOM strategy might be feasible in patients with a cCR. Researchers have sought to address this issue of NOM in rectal cancer after a cCR, with numerous studies including both retrospective and prospective study designs (35). Despite promising results, we lack randomized data comparing NOM with the standard TME. Therefore, TME following chemoRT remains the standard of care in rectal cancer.

The authors of these 2 studies evaluate the patterns of care and survival associated with NOM for rectal cancer using data from the NCDB. Before addressing their results and conclusions, one must consider the inherent limitations associated with the NCDB and their analytic approach. The first issue to consider relates to the authors’ definition of NOM, which depended critically on a variable within the NCDB that assesses the “reason for no surgery.” Determining the “reason for no surgery” relies on a cancer registrar retrospectively evaluating clinical notes with the goal of determining the reason for why surgery was not performed. As one could imagine, even under the best of circumstances, determining the rationale for no treatment would be challenging to accurately assess, exposing the study to the possibility of bias. In their study design, when considering this “reason for surgery” variable, the Ellis study excluded a substantial number of nonoperative patients who did not receive surgery for reasons other than the notation that surgery was “not part of the first course of treatment.” Furthermore, if one looks at the total number of stage I-III rectal cancer patients receiving no surgery, the results tell a different story (Fig. 1).

Fig. 1.

This figure shows 94,029 patients with stage I-III rectal adenocarcinoma diagnosed between 2004 and 2013 within the National Cancer Database. Black dots indicate the annual number of patients who did not receive surgery (local excision or proctectomy) divided by the total number of rectal cancer patients in that year. The dotted line shows the trend over time of nonoperative management in rectal cancer. Image courtesy of James D. Murphy.

In contrast to the Ellis study, our findings presented here looking at a more generalized NCDB study cohort do not show an increase in the NOM of rectal cancer over time. These contradictory results stem from subtle but important differences in patient selection, which emphasizes the importance of clearly defined study methodology to minimize bias in database studies.

Another important limitation in these analyses comes from the lack of treatment-related details within the NCDB. In particular, the authors cannot ascertain response after chemoradiation within the NCDB. Although the NOM studies mentioned above include only patients with a cCR, the NOM cohort in the Ellis study includes patients for whom surgery was “not part of the first course of treatment.” One could hypothesize numerous reasons why surgery was “not part of the first course of treatment,” including patients with rapidly progressive disease, poor tolerance to treatment, declining performance status, comorbidity that arises after diagnosis, and other reasons that portend a worse prognosis compared with patients with a cCR. Studies using NCDB data cannot control for these factors, and these patient selection issues have the potential to dramatically bias outcomes in favor of patients treated with surgery. Beyond the question of selection bias, information on the surveillance regimen for these patients is not taken into account, which is an absolutely crucial component in all NOM strategies (36) and perhaps the single most critical flaw of this analysis. A treatment approach that does not include surgery is not the same as an NOM treatment, because the latter is reserved strictly for complete clinical responders and includes rigorous close surveillance to address recurrences early so as not to compromise long-term disease control. Other important factors excluded from this analysis are details about surgery, radiation, and chemotherapy, and cause-specific survival (as non-cancer survival could differ between study cohorts and influence results [37]).

The question of comparative effectiveness between NOM and standard of care trimodality therapy represents an important research priority in the field of rectal cancer. The NCDB and other cancer registries contain large numbers of patients in an easily accessible format, which provides a strong attraction to use these data for comparative effectiveness research. However, large patient numbers do not protect against the biases noted above, and overall tumor registries such as the NCDB are poorly suited to address these comparative effectiveness questions. Continued efforts with carefully designed multi-institutional prospective studies are needed to define the role of NOM in the treatment of rectal cancer.