T Ã…berg, Tom Treasure
Pulmonary metastasectomy is analysed in this paper according to the rules of evidence-based medicine (EBM). Our knowledge of metastasectomy lacks several crucial factors necessary for the tenets of EBM: survival without surgery in this special group of patients; knowledge of the denominator from which the cases are selected; and the biological nature of the individual tumour. If metastasectomy were introduced today as a new treatment, it would not be accepted. Our analysis provides an alternative interpretation of 5-year survival rates of 40% after metastasectomy as being a result of selection of the patients at the benign end of the continuous spectrum of malignancies. It is therefore a statistical illusion. Given the many variables and the long-time course in many patients with or without metastasectomy, the effect of surgery can only be resolved by a randomized control two-armed trial, where the outcomes between an unoperated group (the natural history) and an operated group (the natural history modified by treatment) are compared. Absence of knowledge is readily accepted by many patients if candidly and respectfully explained and so a randomized trial is possible, appropriate and acceptable to a sufficient number of patients. To shed light on whether there is truly a survival benefit from metastasectomy, a randomized trial has been started. Pulmonary Metastasectomy in Colorectal Cancer (PulMiCC) has so far recruited 86 randomized patients and is open internationally. It is funded for a further 5 years. Interested groups are invited to join the trial. EBM has provided a logical way of defining the ‘indication’ for treatment. An ‘indication’ as used in medical English is the base around which we decide to recommend treatment to patients. It can be said to be the centre of the doctor's professional duty: to identify as unequivocally as possible the indication for treatment. To decide on an indication is to influence the fate of the patient. The indication has therefore to be surrounded by stringent and quality assured rules. The ideals of EBM have been of great value to patients, but at the same time we must not forget that the evidence has to be individualized. Indications for treatment may be reached from various forms of evidence and encapsulated in clinical guidelines or, when we are less certain, consensus statements. Clinical guidelines [1] seek to make irrefutable recommendations, based on the latest and best available evidence, while consensus statements are based on opinions and are more tentative. Nevertheless, we cannot expect even the most secure evidence-based guidelines to be implemented in 100% of patients. The guidance has to be personalized according to the treating doctor's view of the best interests of the individual patient. The ideals of EBM have also been of great value to doctors who have to advise for or against treatments. This can only be done conscientiously when we are able to point to evidence. The ideals of EBM have been of value to society so that we are all spared the distress and the cost of needless treatments. The introduction of new drugs and the indications for their use are now subject to stringent rules which must be adhered to. Surgical operations should be just as carefully considered. Society should not be asked to spend its precious resources on ineffective treatments. The nation's nurses and hospitals are a shared resource; the doctors as the main drivers of costs in a health system must be responsible for their proper use. The phrase Personalized Medicine came into use, as if in opposition to EBM, but it is a false dichotomy. The tightest evidence-based guideline should be overridden for some patients. That is to say they are ‘personalised’. But, with respect to lung metastasectomy, there are no clinical guidelines. The European Society of Thoracic Surgeons (ESTS) Lung Metastasectomy Working Group (LMWG) considered producing guidelines but concluded ‘… the level of evidence to support current practice is too low to set firm recommendations to the members of ESTS. In the absence of a randomized controlled trial looking at the effectiveness of pulmonary metastasectomy on survival and quality of life, it is unlikely that the current practice will ever be influenced’ [2]. More recently, the Society of Thoracic Surgeons (STS) has searched and not found evidence of the kind on which ‘trustworthy’ clinical guidance can be based. The result is that there are no guidelines, in Europe or the USA because there is no trustworthy evidence base [1]. The leaders of ESTS' LMWG directed readers to the PulMiCC trial proposal [3]. That was more than 5 years ago. Since then, there have been many further publications but they do not introduce new information based on better evidence. They come to the same conclusions as were reached by the early follow-up studies in the 1970s and the registry data in the 1990s. In this article, we will go back to basics and examine the sources of evidence on which we might build the indication for metastasectomy. In the very earliest days of surgery for congenital heart disease, Maurice Campbell (1891–1973) made the important decision that the large number of patients they had seen at Guy's Hospital with congenital heart disease, untreatable up to that time, should be categorized in as much detail as possible. He asked that their clinical course be meticulously recorded. Without that record, he argued, future cardiologists and surgeons would never know how the newly introduced operations had altered the ‘natural history’. What is the equivalent ‘natural history’ of lung metastases? The 5-year survival of patients with lung metastases is commonly assumed to approach zero. Is this a realistic estimate? It was contradicted by Åberg's finding reported in 1980 of a 5-year survival rate of 25% for unoperated patients, similar to that of contemporary patients who had lung metastasectomy [4]. The data were few (3/12) and the confident limits are wide (6–57%) but they do not include zero [5]. A realistic estimate of the natural history of the disease is an absolute requirement. For present day patients with lung metastasis, we still do not have that essential information. There are registry data that provide the natural history of patients with cancer usually divided by stage. In Tables 1 and 2, we show cancer registry data for survival of all patients with colorectal cancer and sarcoma in the worst category in the registry, designated Stage 4. This includes only patients who already had metastases at the time of diagnosis. Even with metastases at presentation (and these are the minority in clinical series) there is a non-trivial number of 5-year survivors. It is of the order of 10% for colorectal cancer in the cancer registry. Therefore, the number for whom survival can be attributed to lung metastasectomy is already well short of the 40% that is generally inferred from follow-up studies. Survival rates of 40% after lung metastasectomy are compatible with selection of the most favourable patients with no beneficial effect from metastasectomy [7, 8]. For many diseases, the cancer registry survival data have improved decade by decade and hence the difference attributable to metastasectomy will have narrowed. We hope that improved survival is a reflection of better treatments and effective operations, but there are other statistical factors which we will now consider. Colorectal cancer Five-year survival data of patients with metastases at registration. Thames Cancer Registry cases registered as having metastases from colorectal cancer at the time of diagnosis in three decades from 1980 to 2009. It is unlikely that many of these patients underwent lung metastasectomy. For example the most active group in the Thames region at the time reported 29 patients having lung metastasectomy for carcinoma of any type in an 8-year period (<4 per annum) [6]. These 5-year survival figures provide an order of magnitude for the survival of registered patients with metastases of nearly 10%. If the most favourable quartile were identified as lung metastasectomy candidates but not operated on (10/25 = 0.4), ∼40% survival would be observed [7, 8]. The colorectal cancer patients have usually not had metastases at the time of the primary resection and have already survived longer than 2–3 years from the time of diagnosis before lung metastasectomy. The authors acknowledge Henrich Møller for providing the data for inclusion in these tables. Colorectal cancer Five-year survival data of patients with metastases at registration. Thames Cancer Registry cases registered as having metastases from colorectal cancer at the time of diagnosis in three decades from 1980 to 2009. It is unlikely that many of these patients underwent lung metastasectomy. For example the most active group in the Thames region at the time reported 29 patients having lung metastasectomy for carcinoma of any type in an 8-year period (<4 per annum) [6]. These 5-year survival figures provide an order of magnitude for the survival of registered patients with metastases of nearly 10%. If the most favourable quartile were identified as lung metastasectomy candidates but not operated on (10/25 = 0.4), ∼40% survival would be observed [7, 8]. The colorectal cancer patients have usually not had metastases at the time of the primary resection and have already survived longer than 2–3 years from the time of diagnosis before lung metastasectomy. The authors acknowledge Henrich Møller for providing the data for inclusion in these tables. Sarcoma Five-year survival data of patients with metastases at registration. Thames Cancer Registry cases registered as having metastases from bone or soft tissue sarcoma at the time of diagnosis in two decades from 1985 to 2004. The group designated to manage sarcoma in the Thames region at the time reported 43 patients having lung metastasectomy for carcinoma of any type in an 8-year period (∼5% per annum) [6], so these 5-year survival figures cannot be attributed to lung metastasectomy. Long survival after lung metastasectomy may reflect selection of patients with a favourable natural history of survival [9]. The authors acknowledge Henrich Møller for providing the data for inclusion in these tables. Sarcoma Five-year survival data of patients with metastases at registration. Thames Cancer Registry cases registered as having metastases from bone or soft tissue sarcoma at the time of diagnosis in two decades from 1985 to 2004. The group designated to manage sarcoma in the Thames region at the time reported 43 patients having lung metastasectomy for carcinoma of any type in an 8-year period (∼5% per annum) [6], so these 5-year survival figures cannot be attributed to lung metastasectomy. Long survival after lung metastasectomy may reflect selection of patients with a favourable natural history of survival [9]. The authors acknowledge Henrich Møller for providing the data for inclusion in these tables. Detection of metastases much earlier can now be achieved by more available health care and more sensitive diagnostic tests. If the diagnosis is made 1 year earlier, the recorded survival of the patient will be 1 year longer. This is called lead time bias and it creates an illusion of improving outcomes simply because the stop watch was started sooner. With the introduction of new methods of detection, we have been able to stage patients more precisely. Some patients diagnosed as Stage I lung cancer are now classified as Stage II because we can now see previously undetectable lymphatic involvement. This is called stage migration. Similarly, better detection of mediastinal node involvement with lung metastases, and the exclusion of these patients, produces a group for metastasectomy with a better natural history. Any difference in observed survival would be due to better selection, not to the metastasectomy. With computed tomography (CT) we were able to detect lung metastases. With the addition of positron emission tomography to CT, some patients with lung metastases are now seen to have unsuspected sites of cancer elsewhere in the body. This results in more exclusions, which has the effect of narrowing down the selection of patients in whom metastasectomy is ‘indicated’. The survival results will inevitably appear to be better. Higher 5-year survival in this more highly selected group may be a consequence of selection rather than metastasectomy. There is a distinction made in EBM between ‘efficacy’ and ‘effectiveness’ [10]. Unfortunately, the two words are more or less interchangeable in every day English but in the language of EBM, efficacy is used for interventions that can be seen ‘to work’. So if a surgeon removes a solitary lung metastasis and the pathologist reports cancer-free margins, the operation has efficacy; the metastasis has been removed. Whether the operation is ‘effective’ in achieving the desired clinical outcome, which is to improve survival, the pathologist cannot say. Even after 5 years, patients may die of their same cancer and the metastasectomy can then be seen to have been ineffective as a means of cure [11]. There have been many clinical situations where repeated observation and experience were sufficient proof for a treatment to become established [11, 12]. Surgeons do not hesitate to relieve tension pneumothorax or to retrieve an inhaled object obstructing the trachea. Rightly no one asks them for randomized controlled trial (RCT) evidence. Cataract and hip surgery, and the relief of mitral and aortic stenosis, all entered practice without RCTs. In these examples, the surgeon deals with a single, clearly evident cause of the patient's problem, which is then promptly relieved by a mechanistically rational intervention. If these principles are used to test the effectiveness of lung metastasectomy, it fails the test. There are multiple factors involved and the time scale is measured in years, so the patient's survival at any time point up to and beyond 5 years cannot unequivocally be attributed to that surgical action. It is worth noting that, for treatments that can be seen to be effective by observation alone [11], the more severe the problem, the greater is the benefit to the patient. It is for that reason we are prepared to observe a small pneumothorax or a mild degree of aortic stenosis: it is for the severe cases that we know we should intervene. This is in contrast to lung metastasectomy. Patients selected for metastasectomy are the least severely affected among a heterogeneous they have metastases, longer to their and more EBM not to the patient's survival at any time point to the lung metastasectomy, without control We must that lung metastases are the cause of Patients in whom the only of cancer is the lung metastasis have a for survival, at least in the short They are not to die time so survival data are clearly in this There is an that we must to a lung metastasis because it in may to the mediastinal and from there as we will this is not the on which the practice of metastasectomy is the results are to be better if metastasectomy is a period of It is for a surgeon to to a tumour. were seen as a The in it is to all The of the indications for metastasectomy was that there should be a solitary or very few the should have been and that the patient should have the pulmonary to and the of lung to the These were and were the of set years It was also that there should have been an of time the primary This a further period of observation after the metastasis evident to that there was no or cancer the many follow-up this was found to be on years the primary surgery in cases of sarcoma and years for colorectal cancer This that only a few of all patients with lung metastases the and paper from is an example that was as the practice established The 5-year survival rate was 40% but the data were from patients operated on between and with a very wide of cancer That is an of than patients a this paper in for it to have been and more with it being the of for surgery, have several In the of lung metastasectomy, there are two by this is the in the degree of only one patient had a metastasectomy in a period If they found patients for metastasectomy that how large was the denominator from which the patients were A is the absence of any control These to Åberg's of patients who were but not have lung metastasectomy. He found that some of them survived 5 years without metastasectomy [4]. The in this is the Registry of Lung in The analysis of patients, meticulously and clearly that the favourable factors for survival were an the primary resection of greater than years and a solitary The authors their words They these which is These are for cancer survival of treatment rather than being of a beneficial effect of metastasectomy The authors were also in their for further In their the registry would of surgery and other to be by randomized the registry as the that the in the registry of the few patients who from operation after being to have it would to the value of the That would have provided the of the unoperated survival for patients who are similar to having metastasectomy. That is the ‘natural history’ as There is important registry in which data were to practice as as possible. surgeons data on patients of all lung metastasectomy operations in their in a The point was patients who had a metastasectomy. The to the outcomes was as a as was the survival of patients metastasectomy. this of we can that than of patients with colorectal lung metastases have metastasectomy. This is with other in which this information can be If we this level of selection knowledge of for survival, it less that follow-up studies include 40% 5-year [7, 8]. These data are from a of the of time of in a lung cancer the for providing the data to this The are on the from the to the by the time in days were and the was We have the It can be seen that to have much longer than and lung The to the would have many years to a due to the of cancer and would have been unlikely to cause the of the patient. The patient would have with the before it any are seen for lung metastases and the same might be time as a means of patients for metastasectomy was in patients operated on from to In patients, the time was found to from to more than days by these patients, had metastasectomy and Patients with of less than days years whether operated on or Patients were and the metastases after They operated on with rates of do that is of course to any to them as as possible before they The was that a of patients with who had a period of and who had metastasectomy all survived for 5 years Patients with the metastases their a that has been used before in the of lung metastasectomy These various lead to that the of survival of patients after 5 years to the metastasectomy operation is a illusion with an of If we just on the data a of follow-up studies of nearly patients that survival was if there was more than one metastasis and an of less than years. The is the same as that reached in the follow-up studies in the 1970s with by the in and with a more in Nevertheless, the surgeon authors concluded ‘… it to surgery for patients with two to is the point of the same analysis and an ever more statistical result and then it with the words to We an alternative It to not only to to operations without a realistic of benefit to patients, but it is to a that is so far at with the evidence. not the patient with metastases the in this and other metastasectomy has been to not provide What is now is that and are to be effective based on the for surgical metastasectomy. They are being as a less way of the same They are being introduced without on no more than surgical without evidence from control data There are and a result of better selection, less surgery or treatment, or for a of these 5-year survival rates of after lung metastasectomy colorectal cancer as an are now around the we become able to the more for we might see even better 5-year survival rates to That would further the in the effectiveness of metastasectomy, but it would be a statistical There is new evidence that a to the practice of metastasectomy. In the of colorectal the indication for lung metastasectomy, there is a that patients who have had primary resection should be to detect metastases as early as possible so that they can be There has been a of at detection of metastases with the to of the diagnosis with current The finding is that more not lead to improved This has been found in RCTs. It one might that earlier detection for treatment would improve That has not been the the tenets of EBM, the effectiveness of metastasectomy is one that a randomized trial to an An has the that it that the and factors are similar in It is in value only by of several RCTs. A treatment with as much as lung metastasectomy would not now be introduced into practice without RCTs. An is A of patients with lung metastases of whom in are to be 5-year based on Thames Cancer Registry data The patients are according to metastases and longer between primary surgery and metastasectomy These are They are also with longer survival after metastasectomy and are used for If of these patients are among the in the that would result in the observed 40% survival by selection Patients with three or more metastases and an of 1 year and of the are In the PulMiCC trial, may any patient whom they may but the patients between the possible and are the where there is most to be The carefully selected patient with a lung metastasis, with favourable can expect a survival, not but years, the The metastases that can be seen and are very unlikely to be the cause of the patient's If metastases are that might treatment to relieve the but in they are and to or distress the end of These should be shared with the patient. In some surgeons say they the patient hope is not medical and and by other than the many patients would be to be spared an operation that will not benefit The time of the year or two of their can be in better than surgery and Surgeons are to point that there is very low with metastasectomy. is not a of in introduced highly to of drugs to patients with of was based on Metastasectomy in low of and but surgery should only be done for a treatments because they do is not and interventions have a they up and time, and they a of we know that the of with a metastasis may be for some patients. information the nature of the disease and the of with or without a metastasis, the patient may still on an The indication then their We are not or and should be a without of the patient's is not to lung metastasectomy but to base the indications on rational the patient's and an interpretation of the evidence. information I What do Our current is based on of all the evidence that we are and if this evidence we would of