Metastasectomy for colorectal cancer: are there clothes on the emperor?

The resection of liver and lung metastases in colorectal cancer became adopted into practice by the end of the 1990s.^1,2 Belief in the clinical effectiveness of liver metastasectomy is strongly held to the extent that a group of liver surgeons in the Netherlands state ‘On ethical grounds, a true randomised trial comparing both treatment strategies [that is surgical resection or not] in patients with resectable colorectal liver metastases has not been, and will not be performed’. ³ This view is shared by their colleagues internationally. However, we know from a recent meta-analysis of a succession of randomised controlled trials that when asymptomatic metastases are detected sooner, so that more people can be operated on, there has been no discernible survival benefit (Figure 1). ⁴ We also know from a recent meta-analysis that systemic treatments for metastatic colorectal cancer have been rigorously tested in randomised controlled trials but that an effect on survival attributable to metastasectomy remains unproven and open to doubt. ⁵ A trial of pulmonary metastasectomy in colorectal cancer ⁶ is currently being analysed. I will argue in this commentary for an equivalent trial to be undertaken for liver metastasectomy.

Figure 1.

Meta-analysis multicentre RCTs published in 2006, 2014 and 2016, which included 82% of the patients in the published systematic review.⁴ Hazard ratios are shown with 95% confidence intervals. It suggests if anything an excess of harm from detecting and treating metastases.

How metastasectomy came into practice

From the 1960s, lung metastasectomy was included in the treatment of osteogenic sarcoma which is characteristically a disease of the young; there may be only a few metastases confined to the lung.^7,8 Lung metastases from colon cancer are far more common, and thoracic surgeons accumulated substantial personal case series clinical. From the 1970s through the 1990s, follow-up studies appeared in increasing numbers.^9,10 A simplistic explanation for how lung metastases became a surgical target is that they are easily detected radiographically, showing up white against the black background of the lung. But lung metastases from colorectal cancer, detected and selected for surgery, are asymptomatic and generally would have remained so; colorectal lung metastases do not usually contribute to the process of dying.

Liver metastases contribute to the terminal clinical course towards death and unlike lung parenchyma, the liver regenerates allowing repeated operations, and so a more plausible mechanistic case can be made for liver metastasectomy. The relatively higher rates of lung versus liver metastases in rectal rather than colon cancer support the anatomically logical concept that the liver’s capillary bed filters cancer cells from the portal venous blood, halting systemic dissemination for a time and allowing a second chance of surgical cure.

The recurrence of apparently cured colon cancer, without symptoms until it was inoperable, led to a concept of ‘second-look’ laparotomy in the 1950s. ¹¹ That meant reopening the abdomen at intervals, in asymptomatic people, to take a look. The introduction of monitoring with the tumour marker carcinoembryonic antigen in the 1970s provided a means of identifying recurrence by a blood test. People could then be offered a second-look laparotomy by elevation of the tumour marker. If liver metastases were found, resection offered a means of rescue. Personal case series and registry data were used to establish the practice. At a time when survival with liver metastases was perceived to be less than a year or two, five-year survival following metastasectomy reached 30%. A trial was tentatively proposed from the Mayo Clinic in 1992; power calculations suggested that as few as 36 randomised patients might prove the survival benefit of liver resection. ¹ Randomisation to not have metastasectomy was considered unethical and no trials were done. Liver metastasectomy became established practice on observational data and randomised trials were stated to be ‘not only obsolete but unethical’. ¹

Does metastasectomy meet the criteria for acceptance without randomised controlled trial evidence?

Under circumstances when a condition is known to not spontaneously remit, and there is a large mechanistically explicable effect, with a clear time relationship from the intervention, a repeated observation is sufficient evidence for effectiveness. Relieving an obstruction from an airway ¹² and application of Thomas’s splint to a fractured femur ¹³ are two nice examples of when we can separate ‘signal from noise’. Attributing survival to the effect of piecemeal removal of blood-borne metastases does not meet these criteria. There is wide variation in tumour biology; multiple treatments may have been used, and the impossibility of excluding persistence of disease at other sites makes the link between the metastasectomy and survival inconclusive. Even if the surgeon feels sure and the grateful patient believes it to be so, being alive after five or 10 years (and by the way 10-year survivors after metastasectomy are few) can rarely be reliably ascribed to the effect of operation. It seems to me to be an instance when we cannot separate the signal from noise. ¹²

Alternative explanations for the apparent effectiveness of metastasectomy

In surgical follow-up studies, survival is implicitly or often explicitly attributed to the metastasectomy, but for whatever reason, some people with metastatic colorectal cancer (and it is more than 5%) remain alive at five years, with their metastases. ¹⁴ Favourable features for survival are metastases limited to one or two organs, a longer interval before they are diagnosed, no elevation of carcinoembryonic antigen, and, once detected, measurably slower rates of growth. These are all general prognostic features and indicators of the relative indolence of the cancer itself. In repeated follow-up studies, and in meta-analysis of them, we see the very same features used to select people for surgery: 60% of resected lung metastases are solitary, rarely more than three are resected, 50% have not appeared until three years after primary resection, and 60% have non-elevated carcinoembryonic antigen. In practice, about 3% of patients have a liver or lung metastasectomy. ¹⁵ The apparent survival effect might well be due to selection for surgery of likely survivors.

While there have been no randomised controlled trials of metastasectomy, there have been repeated trials of increased intensity of monitoring for metastases after primary resection; the diagnosis is advanced by one to two years, but there is no beneficial effect on survival. ⁴ A systematic review of controlled trials reports survival gains in metastatic colorectal cancer survival over 20 years. Written into the text of the meta-analysis are the telling words ‘some may argue’ and ‘others will disagree’ about the contribution made by metastasectomy. ⁵ Which way round is it? Is it that metastasectomy results in longer survival, or is that more and longer surviving patients provide the opportunity for more metastasectomy operations? Taken together, these two meta-analyses indicate room for doubt about the contribution made by metastasectomy. If trials for systemic treatments are lauded, why should surgery be exempt from the need for high-grade evidence?

Precedents for stopping surgery found not to benefit patients

Clinical practice changes constantly and not only because of ‘progress’ made by innovation and refinement of practice. There is a catalogue of retreats and reversals: things that were done and are no longer done because it was decided that they do not give benefit. ¹⁶ An example in breast cancer surgery is the move away from radical mastectomy. The practice held sway for 80 years in the form of progressively more radical surgery. Surgeons sought earlier diagnosis and performed more extensive surgery. However, it was observed that women still were dying of cancer, while others who had lesser operations lived a long time and died of something else. Meanwhile, they had been spared the disabling and disfiguring consequences of radical surgery. Radical mastectomy came under scrutiny in the 1960s and its use declined in the 1970s. ¹⁷ It should be noted that it was the doubt and decline in practice of radical mastectomy that set the scene for controlled trials, not the other way round.

Can we now think the unthinkable? It was explained repeatedly in the papers included in the historical analysis of the adoption of liver metastasectomy why randomised trials were not needed to see the benefit. ¹ The self-justification of those pioneer liver surgeons may or may not stand up to retrospective critical analysis but they have succeeded in ensuring that trials were not done. ‘Reversals’ such as that for radical mastectomy are remarkably common, ¹⁶ but doubt has to precede a randomised trial. The two recent meta-analyses approach the question from quite different perspectives and both raise doubt.^4,5 For how much longer can policies of detection and resection of metastases remain outside the now customary standards of evidence? Metastasectomy for colorectal cancer may now be a candidate for reversal.

How liver metastasectomy trials might be designed and carried out

The role of lung metastasectomy has been tested in the Pulmonary Metastasectomy in Colorectal Cancer trial, which provides a model which could adapted to test the effect of liver metastasectomy on survival and quality of life (see https://clinicaltrials.gov/show/NCT01106261). The design of the trial first established whether there is uncertainty within the cancer team concerning the merits of metastasectomy for an individual patient under consideration. In current clinical practice, most people with lung metastases from colorectal cancer are considered unsuitable for metastasectomy; a carefully selected minority are offered this surgery. It follows that if it is ‘no’ for many and ‘yes’ for some, there must be a transition through a zone of uncertainty. ¹⁸ Many factors on a multifactorial spectrum go into the balance pans for or against: the person’s own preference, the number of metastases, their location in the lung, when in the course of the disease they were diagnosed, the tumour marker elevation assay, the likelihood of cancer elsewhere, comorbidity and the impact resection would have on their pulmonary reserve. The weight of each factor is based on judgements rather than science and it is often what the doctors themselves call ‘gut feeling’ that tips the balance.

Potential candidates for lung metastasectomy were invited to join the study and over 500 are in the cohort and more than 90 consented to be randomised in the nested nested controlled trial. The non-randomised patients already form one of the largest prospective cohorts. The data are exceptional in being recorded on an intention to treat basis, including those considered for, but not having, metastasectomy. Also included are those in whom the objective of complete resection of all metastases was not achieved and those in whom the nodule proved to not be a colorectal metastasis. Existing reports characteristically comprise people who have had a metastasectomy and have had post operative confirmative histology, omitting those in whom metastasectomy was intended but not successfully completed.

When a multidisciplinary cancer team’s discussion is protracted, the discussion reveals indecision, or there is dissent, then uncertainty is signalled. In the argot of randomised trials, this may represent group equipoise. The patient should then be informed that there is uncertainty and be invited to have the decision made by random assignment. Randomisation is undeniably difficult but it is not impossible. ¹⁹ People who are assigned to surgery or ablation are then introduced to the surgeon, radiologist or oncologist according to the technical approach decided on clinical grounds. For those not randomised to an intervention, monitoring is continued. They and their doctors decide if and when treatment should be introduced. If it is true as stated that liver metastasectomy for colorectal metastases ‘is one of the greatest advances in surgical practice of recent times’, ²⁰ the benefit should be seen with relatively small numbers and the trial could be halted. If on the other hand the effect is more marginal, the wider medical world and the public should have the evidence available to them.