Resection of asymptomatic primary tumour in unresectable stage IV colorectal cancer: time to move on from propensity matched scores to randomized controlled trials.

Colorectal cancer (CRC) remains a major global disease burden with over 1.2 million new cases each year – about half of those who get the disease will die within 5 years of diagnosis1. At the time of diagnosis, some 20% already present with stage IV disease, of which only a minority (15‐20%) are amenable for attempt at curative resection (usually for liver metastasis), either as an upfront combined resection or as a staged approach with or without combinations of neoadjuvant systemic treatment 2. Despite several improvements in systemic therapy and modern surgical strategies for attempts at cure, the majority of patients with stage IV disease are unresectable and only amenable to palliative strategies. For patients who have a symptomatic primary (e.g. obstruction, perforation or bleeding) a surgical resection or stoma may be warranted if other attempts at bypassing the problem or as a bridge to later surgery is not possible (e.g. endoscopic stenting)3. In contrast, for those patients with stage IV disease having an asymptomatic primary tumour (either in colon or rectum) the role and potential benefit (if any) – and, notably, also the potential harm – of resection of the primary tumour has been much more controversial. Thus, the debate continues4, 5, 6, 7.

Contestants against primary resection would argue that palliation is not possible if the patient does not have any symptoms to “palliate” and the limiting factor for survival is control of the metastatic disease and so systemic therapy should be prioritized. Proponents for resection of the primary tumour argue that reducing the tumour load reduces the disease burden and even disease progress, makes systemic therapy more effective and pre‐emptively manages potential complications from occurring. Both sides may be right – and wrong.

Indeed, a more aggressive approach to unresectable stage IV disease was seen in the past, with every 3 in 4 patients having the primary tumour resected, with a drop to just over every one in two being resected in the latter time period8, 9. Of notice, as the resection rates have dropped, the survival rates doubled from 8.6% in 1988 to 17.8% in 2009 (P<0.001)9. However, using the same datasets, other investigators have come to different conclusions regarding resections of asymptomatic primary tumours in otherwise unresectable stage IV colorectal cancer10. Notably, these cohort studies sampled over longer time periods are biased towards the multiple factors that changed with the time and development in diagnostics, management and available systemic treatment that have not been controlled for in the comparative analyses.

In this issue of the Journal, a nationwide cohort study from the Netherlands investigated survival after primary tumour resection in unresectable stage IV CRC11. Using a propensity score matched approach, they found a survival benefit for those who underwent primary tumour resection (n=2746) compared to systemic chemotherapy (n=3345). When matched by propensity scores in a 1:1 fashion, resection was beneficial for survival when combined with systemic therapy both before and after resection, yet resection upfront with subsequent chemotherapy proved to be best. The authors conclude that this treatment should be entertained more often as an option, even for those with no symptoms from the primary tumour.

While the findings may truly be so that primary tumour resection provides for a survival benefit, several points need to be considered before jumping to conclusions. Notably, propensity scores can be used in several ways12, of which propensity score matching is one. This involves a matched modelling of assembled pairs of two interventions (in this case resection or no resection) to selected baseline characteristics in an attempt to reduce bias. It works similar to a randomized trial, except it is not randomized and it is not a trial. Propensity score matching is still based on matching of the available, chosen data for matching. In the current study, a total of 1737 (28,5%) of the patients were not ‘matchable’ (one third resected primary, two thirds non‐resected) and thus excluded. As the authors matched pairs based on year of diagnosis, age, tumour location, morphology and number of organs with metastases (1 or >1), several factors that may influence survival were not matched for. Notably, presence of comorbidity (e.g. ASA score, Charlson‐Deyo score, presence of any organ function deficits etc) or functional status (e.g. ECOG status level) were not available and thus not controlled for. Also, TNM stage was not matchable due to the nature of those non‐resected (lacking a pTNM status). Thus, it may still be so that those who received surgery up front were deemed fitter and more likely to tolerate a surgical procedure. Or, alternatively, had less extensive disease overall.

Further, one needs to scrutinize the choice of endpoint in the current setting of non‐curable disease. When there are no symptoms to alleviate, one may question if overall survival is the best outcome measure to address. The added months of survival may truly be valuable for the individual patient, but one should consider at which risk and at what price this comes. For one, the current study had a 30‐day mortality rate in the resected patients at 9%, which was similar for the systemic therapy group. These numbers are staggering, and far beyond those seen in any controlled trial of systemic therapy. However, they may actually reflect the truth in a real‐life setting and, also, points to the risk of adversary effects of “just doing something” in a palliative setting. Given that all emergency resections (known for a very high perioperative mortality) were excluded, the rate of short‐term deaths in what should be considered ‘elective’ surgery is considerable. For patients aged over 75 years, the 30‐day mortality rate was even higher at 15%. The message is that either ‘palliative’ approach comes at a high risk of short‐term death, knowing that these patients were likely deemed better performers than the ‘best supportive care’ candidates that were not included in the analyses. So, with a high risk of death (one in ten) for a relatively short survival benefit (of a few months) in otherwise asymptomatic patients, one may question if overall survival is the correct endpoint to consider. Further, doing landmark analysis of survivors beyond a 6 months period (during which time a considerable number of patients die) further skews the data towards favourable reporting of survivor bias (you have to be alive in order to live longer). Further, perioperative risk for anastomotic leaks is higher in patients with metastatic disease13, and this again strongly influences survival14.

Arguably, quality of life may be far more important in this patient group, whose longevity is limited but for whom symptom control is paramount. Data on QoL are very limited but may improve with surgical resection, as demonstrated in a very small (n=24), uncontrolled, cohort study15. No other studies have reported QoL data in this setting thus far16, which only testifies to the glaring lack of research in this field17.

Several systematic reviews have been conducted16, 18, 19. Based on limited data from few and heterogeneous studies with considerable reported bias, the median survival is reported at 15.2 months (range 10‐30.7 months) in the resection group and 11.4 months (range 3‐22 months) in the non‐resection group, with a notable overlap in the reported ranges for both. This likely reflects the poor selection of appropriate candidates and a huge case‐mix in disease burden, functional status and additional comorbidity. Also, a number of previously reported propensity matched studies have shown a variation in patient selection and outcomes10, 20, 21, 22, 23, 24, 25, but presents a reserved but favourable view on primary tumour resection for selected patients. As evidenced from one recent metaanalysis of heterogeneous studies, the patients that may benefit most from primary tumour resection are those who are young, fit, have a good performance status, and have minimal systemic disease19. However, the propensity matched cohort and their associated findings have reached a limit in terms of information for decision‐making. It is now time to move on to randomized trials.

Indeed, a lack of randomized controlled trials (RCTs) in this area is what keeps the debate alive. Thus, several ongoing randomized trials (the German‐Austrian SYNCHRONOUS26 trial, the Dutch‐Danish CAIRO427 trial, the Korean multicenter trial28, and the Chinese trial29) and their expected results are much needed (Table 1). In addition to survival outcomes, at least three of the RCTs have stated QoL as secondary outcomes, and as such these measures will be of additional value for decision‐making for both patients and caregivers alike.

Table 1

Overview of ongoing randomized controlled trials for resection of primary tumour in unresectable stage IV colon or rectal cancer.

Trial/name	Origin	Trialc	Inclusion age	Primary outcome	sample (n)	Est. Completed, date
SYNCRONOUS	Germany, Austria	ISRCTN30964555	≥18 yearsa	OS, 3 yrs	800	July 2019
CAIRO4	Netherlands, Denmark	NCT01606098	≥18 years	OS, 5 yrs	360	August 2020
CCRe‐IV	Spain	NCT02015923	≥18 years	OS, 2 yrs	336	November 2016
Korean multicenter	Korea	NCT01978249	20‐90 yrsb	OS, 2 yrs	480	April 2018
China multicenter	China	NCT02149784	18‐75 years	OS, 3 yrs	480	July 2019

for colon cancer primary only; rectal cancers are excluded

colon and upper rectum cancers

from either registry ISRCTN (www.isrctn.com) or NCT (www.Clinicaltrials.gov).

OS denotes Overall survival