key: cord-0841912-wzkz2zu0
authors: Treasure, Tom; Farewell, Vernon; Macbeth, Fergus; Batchelor, Tim; Milošević, Mišel; King, Juliet; Zheng, Yan; Leonard, Pauline; Williams, Norman R.; Brew‐Graves, Chris; Fallowfield, Lesley
title: The Pulmonary Metastasectomy in Colorectal Cancer cohort study: Analysis of case selection, risk factors and survival in a prospective observational study of 512 patients
date: 2021-05-05
journal: Colorectal Dis
DOI: 10.1111/codi.15651
sha: 0d1d2d2e293b9c0b3576b139fa6017d4e950e62c
doc_id: 841912
cord_uid: wzkz2zu0

AIM: We wanted to examine survival in patients with resected colorectal cancer (CRC) whose lung metastases are or are not resected. METHODS: Teams participating in the study of Pulmonary Metastasectomy in Colorectal Cancer (PulMiCC) identified potential candidates for lung metastasectomy and invited their consent to join Stage 1. Baseline data related to CRC and fitness for surgery were collected. Eligible patients were invited to consent for randomization in the PulMiCC randomized controlled trial (Stage 2). Sites were provided with case report forms for non‐randomized patients to record adverse events and death at any time. They were all reviewed at 1 year. Baseline and survival data were analysed for the full cohort. RESULTS: Twenty‐five clinical sites recruited 512 patients from October 2010 to January 2017. Data collection closed in October 2020. Before analysis, 28 patients with non‐CRC lung lesions were excluded and three had withdrawn consent leaving 481. The date of death was known for 292 patients, 136 were alive in 2020 and 53 at earlier time points. Baseline factors and 5‐year survival were analysed in three strata: 128 non‐randomized patients did not have metastasectomy; 263 had elective metastasectomy; 90 were from the randomized trial. The proportions of solitary metastases for electively operated and non‐operated patients were 69% and 35%. Their respective 5‐year survivals were 47% and 22%. CONCLUSION: Survival without metastasectomy was greater than widely presumed. Difference in survival appeared to be largely related to selection. No inference can be drawn about the effect of metastasectomy on survival in this observational study.

The Pulmonary Metastasectomy in Colorectal Cancer (PulMiCC) study enrolled 512 patients who were being considered for lung metastasectomy from October 2010 to January 2017. Nested within this observational study was the PulMiCC randomized controlled trial (RCT) of 93 patients. The RCT was reported in Colorectal Disease in 2020 [1] . The widely believed large survival benefit from lung metastasectomy was not evident although a small late effect cannot be excluded. Patient-reported outcomes-quality of life and health utility-declined at a similar rate in the metastasectomy and the unoperated control arms [2, 3] .

The PulMiCC study was run in the context of a firm belief in the clinical effectiveness of lung metastasectomy. In response to a paper arguing the case for a trial [4] , the European Journal of Cardio-Thoracic Surgery published an Editorial in 2017 proclaiming 'Surgery for pulmonary metastases is a pillar of modern thoracic surgery' [5] . In support of lung metastasectomy for colorectal cancer (CRC), the authors cited two retrospective single-institution studies [6, 7] .

These provided a pooled 5-year survival of 60% in groups of 165 and 113 patients collected over 10 and 12 years. The Work Force of Evidence Based Surgery of the Society of Thoracic Surgeons based its recommendations for lung metastasectomy on the assumption that 5-year survival without resection is zero [8] . Neither statement was supported by control data. These publications in the leading thoracic surgical journals invite the conclusion that lung metastasectomy provides 60% survival benefit. This widely held belief and the climate of certainty resulted in multidisciplinary teams (MDTs) having difficulty randomizing patients into the PulMiCC trial ( Figure 1 ).

Although PulMiCC found no difference in survival, a small survival advantage due to resection of lung metastases that prove to be the only site of residual CRC cannot be discounted. But the trial was large enough to refute the improbable 0% [8] and the less extreme estimate of 5% more generally cited [9] . The 5-year survival in the control group was 30% (95% CI 15.3-45.7%), significantly above 5% (P < 0.001) [1] .

Colorectal Disease has recently published a big data study of pulmonary metastasectomy for CRC [10] . Here we report data from the full PulMiCC cohort. The large majority of the patients were not randomized and so we cannot draw any valid inferences about survival attributable to lung metastasectomy. Nevertheless, the characteristics of the large sample of patients and their outcomes are available to inform the Association of Coloproctology of Great Britain and Ireland (ACPGBI) IMPACT initiative and the design of future research in this uncertain area [11, 12] .

F I G U R E 1 Concerned by the low rate of randomization into the RCT the Independent Data Monitoring Committee asked for an investigation. The three largest recruiting centres (Bristol, Liverpool and Sheffield) provided reasons for not randomizing in 155 patients. Among 78 patients for whom the MDT overrode the patients' wishes to be in a controlled trial, 77 (99%) were operated on. When patients made their own decision they were more evenly divided, demonstrating group equipoise. The results are given in the Sankey chart. Diagram created using SankeyMATIC (http://sanke ymatic.com/) | 1795 TREASURE ET Al.

The PulMiCC cohort study was planned around the PulMiCC RCT.

The study was administered by clinical trials staff based at the recruiting hospitals under the direction of the site's principal investigator (PI) who was either a thoracic surgeon or an oncologist. Patients who were potential candidates for lung metastasectomy were given written information and an explanatory DVD to take home. A healthcare professional training DVD was also available for clinicians to aid their discussions with patients.

Interested patients were invited to sign Stage 1 to be assessed for lung metastasectomy within PulMiCC and to be registered by the trials unit. They consented to collection of baseline information: sex, age, height and weight to derive body mass index, the interval since primary CRC resection, whether they had prior liver metastasectomy, the number of lung metastases, carcinoembryonic antigen assay (CEA) and tests of lung function. Results are summarized as hazard ratios (HRs) with confidence intervals (CIs). An HR greater than 1 indicates that larger values of the risk factor are associated with poorer survival and an HR less than 1 indicates that larger values are associated with higher survival. For categorical risk factors, an HR >1 indicates that poorer survival is associated with the presence of the risk factor (e.g., for male sex) or with a risk factor category compared to the reference category as for performance status using the Eastern Cooperative Oncology Group (ECOG) scores.

For descriptive purposes, survival curves were also estimated based on the fixed grouping of the patients into these three strata.

The times of origin for these three curves are different and are the dates of cohort entry. These classifications are retrospective and therefore analyses using these from the time of cohort entry are ad hoc because the classes are determined after cohort entry and, more importantly, by decisions made by the patient and MDT following cohort entry.

In all, 512 patients were recruited into Stage 1 of the PulMiCC trial from October 2010 to January 2017 from 25 clinical sites, mainly in England but also in China, Scotland, Serbia and Sicily (Table 1) . Of the 512 patients, 31 were excluded for reasons given in Table 2 Table 3 ). Data collected up to October 2020 are used in this analysis. The source, mix and eventual outcome of the three strata are shown in Figure 3 .

The baseline risk factors for the patients in the three strata are shown in Table 4 . The proportions of metastases in the three strata ( Figure 4) show the predominance of solitary metastases (69%) in the elective metastasectomy group (Table 5) . Table 6 presents the results from single-factor analyses of baseline factors derived from stratified Cox regression models. The table includes the number of deaths examined in each analysis as these vary depending on the number of patients for whom information is available on the various potential risk factors. The factors demonstrating evidence of a higher mortality risk were higher values of log(CEA), a shorter interval from CRC to cohort entry and higher ECOG classes with a suggestive effect for male sex. Multiple lung metastases and prior liver metastases also demonstrated limited potential for an increase in mortality so were also considered for multi-factor analyses. There was some evidence that the effect for log(CEA) was different in different strata. Estimated HRs and 95% the elective no metastasectomy, elective metastasectomy and randomized groups respectively with no effect being demonstrable in the elective metastasectomy group. Table 7 presents results from multi-factor stratified Cox regression models including variables demonstrating some potential for a relationship in single-factor analyses. The set of models presented is based on gradually decreasing numbers of deaths depending on the availability of risk factor information for the variables included with a final model with the four most significant variables. These were male sex, interval (shorter) from CRC to cohort entry, prior liver metastases and log(CEA). Again, there was evidence of differential effects for log(CEA) with estimated HRs and 95% CIs of 1.57 (1.20, 2.04), 1.14 (0.92, 1.43) and 1.52 (1.16, 1.98) in the elective no metastasectomy, elective metastasectomy and randomized groups respectively.

With a time origin of cohort entry, Figure 5 displays Kaplan-Meier estimated survival curves and 95% CIs of three groups of patients retrospectively classified in the strata: non-randomized and no metastasectomy, elective metastasectomy, and randomized. There are usually more missing data in the elective non-metastasectomy group because, if metastasectomy is decided against for any reason, other investigations may serve no purpose. CEA is an exception. It was required for minimization but the value was not always in the record. It is a paradoxical marker because its elevation on screening may prompt a search for metastases but its elevation is a negative predictor for survival after metastasectomy and may contribute to a decision against operating.

of solitary and multiple metastases is shown in Figure 7 . The number of lung metastases is a risk factor of particular relevance for metastasectomy patients and is considered further in the next section.

All the curves in Figure 5 show a relatively slow rate of decline initially. This is a hallmark of many cancer survival studies-whether randomized or observational-in contrast to cancer registry data presentations when there is characteristically an early sharp fall which becomes progressively flatter. For randomized patients, the plateau may be because patients have to have a good probability of short-term survival (1-2 years) often defined by entry criteria. This is also seen among patients in whom the decision is individualized. All patients in the cohort were being considered for lung metastasectomy and therefore had favourable features. This provides for a further bias in addition to the selection of patients on prognostic features. The median interval F I G U R E 4 Three numbers of lung metastases in the three strata based on available data as in Figure 5B ,C. This effect carries over to the initial period after the metastasectomy as seen by the initial plateaus in Figure 6 after deaths on the day of operation. We suggest that this is a form of guaranteetime bias [13] . It can occur when survival is timed from enrolment, and is compared across groups defined by a classifying event occurring sometime during the subsequent passage of time. In this case, patients who show other sites of disease, or other adverse features, fall outside the standard criteria for metastasectomy [14] . But metastasectomy remains an option for patients whose progression trajectory indicates the likelihood of longer survival. It is carried out if survival appears to be 'guaranteed' for a reasonable length of time. In assessment for liver metastasectomy there is a period of assessment before the final decision is made. It is a conscious policy referred to by liver surgeons as the 'test of time' [15, 16] . The relatively low hazard ratio of 1.19 for multiple vs. single metastases in Table 6 is not directly comparable with results derived only from metastasectomy patients. In the meta-analysis of 24 reports including 2589 patients already referred to, the hazard ratio for multiple vs. solitary metastases, for patients having a metastasectomy, was 2.04 [9] . The better survival of patients with a solitary metastasis has been found repeatedly in case series, in the International Registry of Lung Metastases [18] and in systematic reviews [19, 20] . Figure 4 shows that among patients selected for the elective metastasectomy stratum the solitary metastasis rate was nearly double that in the unoperated patients (69% vs. 35%). A large difference was seen in survival after 2 years among patients with a difference in this risk factor.

The analysis of the randomized stratum in PulMICC produced an HR of 2.15 which is comparable with that in the meta-analysis, while there was no evidence of an effect in the elective metastasectomy patients. A likely explanation is that in more recent practice, aware of the hazard of multiple metastases, the MDT will only recommend metastasectomy if the balance of other risk factors is favourable. In randomized patients, minimization prevents trading off risk factors.

However, erosion of the effect of a risk factor as the practice becomes more selective has been observed in the use of risk factors in case selection [21] . A simple analogy might help explain. Looking at performance data of a large pool of high school basketball players an analyst noted that youths ≥2 m tall were higher scorers. They were preferentially picked for the county team. When the analysis was run for elite teams, height was no longer such a strong discriminator.

They were nearly all very tall. Blackstone and Lauer have called this effect 'work up bias' [22] . The coach, however, continued to select for the squad one or two players ≥1.7 m tall who could almost unerringly shoot and score from 15 to 20 m away. As does the coach, an MDT looks at all the factors. If a patient has two to three metastases TA B L E 7 Results from multi-factor analyses using a stratified Cox regression model that have remained much the same over a year of observation, they know that patient is likely to have a longer survival.

The major limitation of the PulMiCC cohort study is that it is not a randomized comparison and cannot provide evidence about survival benefit attributable to metastasectomy. The best currently available data are in the PulMiCC RCT report [1] .

We can confidently state that the 5-year survival of people with lung metastases from CRC is not zero. Among 481 patients in the cohort, 169 patients did not have a metastasectomy and 37 of them were 5-year survivors (22%, 95% CI 16%-29%). It is possible but very unlikely that a few of these survivors did not in fact have malignant lung lesions. We specifically sought information on long survivors, however treated, and the PIs did not report any case in whom the diagnosis of CRC had been wrong.

In most cases after lung metastasectomy CRC recurs sooner or later, but it is quite feasible that in some cases the lung metastases are the only residual site of disease and metastasectomy is curative.

We hear anecdotal accounts [23] but documented proof of diseasefree survival at a long interval after lung metastasectomy are yet to be seen. Reported long-term survivors have usually also had systemic treatments. The majority of patients in this cohort with and without metastasectomy had systemic treatments. An analysis of those treatments is the subject of a further report.

The big data study in the English NHS already referred to showed that lung metastasectomy is highly selective, being used on only 2.3% of patients with resected CRC in 2013 [10] . A study in the Korean National Health Insurance Database of 2573 CRC lung metastasectomy patients found a similar rate of 2.5% [24] . The lung metastasectomy operations referred to as 'a pillar of modern thoracic surgery' were performed at a rate of about one a month in the two series cited in support [5] -more a flying buttress than a pillar. This is a highly selective practice. discounted PulMiCC as too small [25] . Instead, they chose to use a non-randomized follow-up study in which 48 had metastasectomy.

PulMiCC's metastasectomy arm fell short of their unspecified cutoff with only 46 patients but there was a control group. Also recommended for consideration was peritoneal surgery. They cited an RCT in support. It had the same surgery in both arms but the authors concluded 'that high-quality surgery is of value' which cannot be derived from the abstract cited. The position with respect to liver resection was summed up by surgeons at Memorial Sloan Kettering:

'We took as our point of departure the assumption that there will never be an RCT to answer the question of if liver resection has a role in the management of CRLM [colorectal liver metastases], or even to quantify its exact benefits' [26] It seems questionable that the practitioners of a particular form of surgery should seek to bar the way to its evaluation. We are aware that the IMPACT initiative of ACPGBI includes active detection of metastases for resection but meta-analysis of the many trials of more vs. less intensive screening protocols have found no overall survival benefit [27, 28] .

Thoracic surgeons have addressed the question of the clinical effectiveness of lung metastasectomy by participating in the PulMiCC studies, the RCT and this cohort, but resolution of this matter is unlikely to come from specialist thoracic surgeons engaged in a very small part of the overall treatment of advanced CRC. ACPGBI have made a commitment to improving the care of patients with advanced CRC [11] . That should include reducing and avoiding the use of ineffective treatments if only to make room in the budget for adopting new ones. It is noteworthy that, in the process of prioritizing patients for cancer treatments during the COVID-19 pandemic, metastasectomy was deemed low priority.

The use of stereotactic radiotherapy to treat 'oligometastases' has been commissioned by NHS England on the basis of very weak evidence, and this may well supersede the use of surgical metastasectomy [29] .

Colorectal MDTs are in the best position to implement trials of treatments for advanced CRC. Commenting on yet another round of the homoeopathy vs. allopathy debate The Lancet recognized that allopathy might have RCT evidence but homoeopathy has a following and while 'doctors need to be bold and honest with their patients about homoeopathy's lack of benefit' they should also be honest 'with themselves about the failings of modern medicine to address patients' needs for personalized care'. The IMPACT initiative seeks to personalize care [11] but all systemic treatments have been introduced on the basis of controlled trial evidence. To paraphrase

The Lancet, doctors need to be bold and honest with themselves about the failure to even seek proof of the effectiveness of local treatments. This is a declared research priority of the ACPGBI [12] .

None of the authors has a conflict of interest with respect to any of the content of this submission.

Committee London-Hampstead 10/H0720/5. Approval for follow-up as an audit of practice 11 Feb 2019 HAMPSTEAD, NRESCommittee.

London-(HEALTH RESEARCH AUTHORITY).

Pulmonary Metastasectomy in Colorectal Cancer: updated analysis of 93 randomized patients-control survival is much better than previously assumed

Pulmonary Metastasectomy versus Continued Active Monitoring in Colorectal Cancer (PulMiCC): a multicentre randomised clinical trial

Pulmonary Metastasectomy in Colorectal Cancer: health utility scores by EQ-5D-3L in a randomised controlled trial show no benefit from lung metastasectomy

Analysis of pulmonary metastasis as an indication for operation: an evidence-based approach

Surgery for pulmonary metastases: quo vadis?

Factors influencing survival after complete resection of pulmonary metastases from colorectal cancer

Risk factors for lymph node metastases and prognosticators of survival in patients undergoing pulmonary metastasectomy for colorectal cancer

Expert consensus document on pulmonary metastasectomy

Risk factors for survival after lung metastasectomy in colorectal cancer patients: a systematic review and meta-analysis

National variation in pulmonary metastasectomy for colorectal cancer

Improving Management of Patients with Advanced Colorectal Tumours: Association of Coloproctology of Great Britain and Ireland

Use of a modified Delphi approach to develop research priorities for the Association of Coloproctology of Great Britain and Ireland

Challenges of guaranteetime bias

The surgical treatment of metastatic tumours in the lung

Surgical management and outcomes of colorectal cancer liver metastases

If a picture is worth a thousand words, take a good look at the picture: survival after liver metastasectomy for colorectal cancer

Effect of public reporting of surgeons' outcomes on patient selection, "gaming," and mortality in colorectal cancer surgery in England: population based cohort study

Long-term results of lung metastasectomy: prognostic analyses based on 5206 cases

Pulmonary metastasectomy in colorectal cancer: a systematic review and quantitative synthesis

Reported outcome factors for pulmonary resection in metastatic colorectal cancer

The use of scoring systems in selecting patients for lung resection: work-up bias comes full-circle

Caveat emptor: the treachery of work-up bias

Pulmonary Metastasectomy in Colorectal Cancer: has the randomized controlled trial brought enough reliable evidence to convince believers in metastasectomy to reconsider their oncological practice?

Pulmonary metastasectomy in colorectal cancer: a population-based retrospective cohort study using the Korean National Health Insurance Database

Management of colorectal cancer metastases to the liver, lung or peritoneum suitable for curative intent: summary of NICE guidance

Questioning why more patients with colorectal liver metastases are not referred for metastasectomy

Meta-analysis of colorectal cancer follow-up after potentially curative resection

Follow-up strategies for patients treated for non-metastatic colorectal cancer

Stereotactic ablative body radiotherapy in patients with oligometastatic cancers: a prospective, registry-based, single-arm, observational, evaluation study

Data available on request from the authors.

https://orcid.org/0000-0001-9358-7610