The decision to perform pulmonary metastasectomy is usually now made by specialist sarcoma teams and is based on factors such as the interval since primary surgery, the number and rate of growth of metastases, and their response to chemotherapy. The surgical approach may be videothoracoscopic or by thoracotomy, and surgery may be through staged lateral thoractomies or bilateral through an anterior approach. The pulmonary resections are also ‘individualised’ depending on the location, size and number of metastases, with an implicit commitment to spare as much lung parenchyma as possible. It may be this degree of variability which makes data tabulation difficult. The authors of the 2011 EOI analysis acknowledged that “Amongst the limitations is the limited information on how the recurrences were treated. However, all patients were treated in experienced sarcoma centres and it is likely that all patients received the best available treatment for their recurrence. This includes, whenever possible, complete resection of a local recurrence and/or surgical treatment of all distant recurrences in case of resectable disease.”²

In 2006, the UK National Institute for Health and Clinical Excellence (NICE) which issues guidance to the National Health Service in England and Wales, published a manual for commissioners of cancer service on ‘Improving Outcomes for People with Sarcoma’.³ The manual is more about organisation than guidance on clinical practice and states that “The management of chest wall, intrathoracic sarcomas and pulmonary metastases requires a combination of skills available from a sarcoma multidisciplinary team (MDT) and a thoracic surgeon, often combined with plastic surgical reconstructive skills.” Included in this guidance is the advice on surveillance for the appearance of pulmonary metastases and states in that context “None of the 21 patients who presented between follow-up visits with symptomatic pulmonary metastases were considered candidates for potentially curative surgical resection of their metastases. Resection of pulmonary metastases was performed for 24 of the 36 patients whose asymptomatic recurrence was discovered by surveillance chest x-ray or staging CT scan”³ based on evidence reviewed.^{4 5} There is evident readiness to operate on asymptomatic pulmonary metastases in sarcoma patients but evidence for the practice, or guidance as to which patients are believed to benefit, cannot be inferred from this practice manual.

We retained 18 articles published since 1990 for inclusion in the systematic review: five report on first and subsequent pulmonary metastasectomy for bone sarcoma,^10–14 six on soft tissue sarcoma^15–20 and four on mixed sarcoma series.^21–24 The information in tables 1​1​–4 are extracted from these 15 studies which include data on the patients’ first pulmonary metastasectomy. Three of the 18 are confined to repeat pulmonary metastasectomy.^25–27 One of these²⁷ contained 14 patients rather than the specified minimum of 20 patients but is a further report providing outcomes for repeat metastasectomy in some of the patients reported from the same institution and was therefore included.¹³

With respect to research methodology, there were no randomised controlled trials. There was one comparison study in which patients who had undergone videothoracoscopic resection were matched with patients who had undergone a thoracotomy approach.²⁴ There were no protocol-based prospective studies. There was one retrospective cohort study of data obtained from a Cancer Centre's institutional Tumor Registry.¹¹ In other reports, cases were identified from databases which were held, as far as we could determine, at an institutional level^{10 12–17 20 23 26 27} or departmental level.^{18 19 21 24 25} It appears that many of the clinical data were retrieved by retrospective case note review. A statement in a report from the M.D. Anderson Cancer Centre (MDACC) is probably representative of the approach to data collection: “A prospective surgical database was used to identify metastasectomy patients and missing clinical data were supplemented in a retrospective manner.”²³

In most studies, the stated purpose of analysis was to report survival following first^{10–21 23 24} or repeated pulmonary metastasectomy.^{22 25–27} In 10 of the 18 reports, statistical analyses were performed to identify patient and tumour characteristics associated with improved survival following first pulmonary metastasectomy.^{11 13 14 16 18 19 21 23 25 26}

The population from which the patients having pulmonary metastases were drawn, is given in seven publications (table 1).^{10–12 14 15 17 23} As can be seen from the footnotes to the table, no two denominators are defined in the same way and none are comprehensive at a community level. Some authors give an upper age limit (not more than 55,¹² 40¹⁴ or 20 years¹¹) but read in context this appears to be to match the data set of operated patients rather than a prior policy. Some series include all sites of primary disease while others are limited to limbs^{10 12} or trunk and limbs.¹¹ They variously include all sarcoma patients,²³ or only those with soft tissue sarcoma (STS).¹⁷ The proportion of the denominator population recorded as developing pulmonary metastases ranges between 18% and 50% while the proportion of those with pulmonary metastases who have an operation to remove them varied from 5% to 88%. The report with the largest data set (MDACC²³) reported that only 1% of sarcoma patients have a pulmonary metastasectomy. We have not found it possible to determine how much of the variation in the recorded data is attributable to varying selection in clinical practice, the different biology of tumours according to histology, tumour site or variation among patients. A large amount of the variation appears to depend on how wide the net is cast in capturing the denominator.

Among these 18 studies of pulmonary metastasectomy for sarcoma the inclusion criteria are much as those proposed by Thomford²⁸ that the cancer at the primary site was eradicated, controlled or amenable to control;^{10 12 13 15 17–22 24 26 27} that the metastatic lung disease was amenable to complete resection;^{10 11 13 15 18–21 24 26 27} that there was no metastatic disease elsewhere;^{10 12 13 15 17–22 24 26 27} and that the patient was expected to withstand the loss of lung tissue necessary to give clearance.^{10–12 15 17–22 24 26 27} In individual instances authors specified that there should be no mediastinal or chest wall involvement;¹⁷ absence of pericardial or pleural effusions;¹² that the overall operative risk was acceptable;²⁶ or that there was no other available more effective treatment.¹⁹ In one report, increased size on chemotherapy was allowable, but not an increase in the number of metastases.²⁴ One study with five subgroups gave group-by-group criteria which, read in context, appeared to be defined after exploration of the available data to facilitate analysis and reporting.²³ Criteria for inclusion or exclusion in the metastasectomy cohort were not found in three studies.^{14 16 25}

Data are given in table 2 for the 15 studies which include data on the first pulmonary metastasectomy, for a total of 1168 patients. The average age of bone sarcoma patients was 17 years based on 377 patients in four studies with calculable data,^{10 12 13} excluding the reports limited to patients aged <21 years¹¹ or <40 years.¹⁴ For soft tissue sarcoma the average age was 46 based on five studies including 277 patients^16–20 excluding a study where median and range were given.¹⁵

Male sex was predominant in bone sarcoma (65% of 532 patients) but not in soft tissue sarcoma reports (50% of 277 patients). These differences in age and sex preclude meaningful amalgamation of outcome data following pulmonary metastasectomy for bone and soft tissue sarcoma.

The interval between resection of the primary and first pulmonary metastasectomy was provided in 9/15 reports and was highly variable as can be seen from table 2. There is a degree of consistency in the median interval of 1–2 years but half of the authors providing data, operated on synchronous metastases (5 of 10). Repeat metastasectomy was performed in 43% of patients based on 14/18 reports in which the data could be extracted (table 3).

Chemotherapy was frequently used but schedules were variable both within and between publications. Some authors stated that preoperative and/or postoperative chemotherapy was given routinely in all cases^{10 21 26 27} but more often the practice varied.^{11–14 16 18 21} One paper states “The only constant was that when the disease-free interval was <2 years with a single lung metastasis, no chemotherapy was added to surgery” and another that it was at the discretion of the oncologist.¹⁹ One group used chemotherapy preoperatively only when there were six or more metastases.¹⁷ It was also implicit in the text of several papers that response to chemotherapy was part of the clinical evidence used to help select patients for surgery; non-responding and progressing patients were less likely to be selected for pulmonary metastasectomy and this information is not necessarily explicit in the report. This statement in the report from the MDACC is representative of this approach: “Those who developed metastatic disease early with multiple pulmonary nodules were treated initially with chemotherapy to determine the pace of disease progression, if any, on treatment. Patients responding to chemotherapy, those with stable disease, and those with slow progression were referred for resection while those with rapidly progressive metastatic disease received alternative chemotherapy treatment.”²³

Whether videoscopic or open surgery was used, and if open through what incision, and the surgical technique used to resect the metastases, are summarised in table 4 for 12/15 papers including data on first metastasectomy operations.^{10–13 16–21 24} The remaining 3 of the 15 studies were not explicit with respect to the surgical approach. In the more common surgery for carcinoma metastasised to the lung, lymphadenectomy has become an important consideration.²⁹ In these reports concerning sarcoma patients, hilar nodes were not routinely dissected¹⁹ or maybe dissected ‘when necessary’.¹² This avoidance of lymphatic resection appears to be linked to the lower rate of lymphatic spread in sarcoma compared with other thoracic malignancy in which further spread from the metastases to mediastinal lymph nodes is frequent.²⁹

There is a strong evident preference for open surgery (96% of patients had a thoracotomy of some form) with considerable emphasis placed by several authors on the importance of manual palpation of the lung^{10–12 16 17 20 22} which cannot be achieved through a purely videoscopic approach. One study specifically addressed the question for whether the less invasive thoracoscopic approach might be as effective²⁴ and it was concluded that it might be an option if there are no more than two metastases but this was not derived from data analysis presented in the publication. The general use of thoracotomy, often bilateral and repeated metastasectomy in 43% of patients overall represents a high treatment burden for patients (tables 3 and ​and44).

TCR data for sarcoma were studied to provide some context to the overall survival rates of patients with sarcoma (figure 1). The Registry has employed its own four-level staging system since 1960 and stages around 60% of all solid tumours. The classification system uses information in the patients’ notes to determine if the disease is local (stage 1), has extension beyond the organ of origin (stage 2), has regional lymph node involvement (stage 3) or has metastasised (stage 4). Survival data by stage for two complete decades 1985–1994 and 1995–2004 for both bone and soft tissue sarcoma are provided in figure 1. For patients entered as stage 4 bone sarcoma (metastatic disease at the time of registration) in those two decades 5-year survival of 20% and 25% are recorded for bone sarcoma and 13% and 15% for soft tissue sarcoma.

The Registry does not include full data on treatment but does provide data on the highest surgery code. These are presented in an abbreviated form in table 5. According to the selection criteria set out above, since the stated first criterion for pulmonary metastasectomy was that a radical operation had been successful at the primary cancer site, it is among the 8% of bone sarcoma patients and 21% of soft tissue sarcoma patients that pulmonary metastasectomy patients would be found. Summary data of sex distribution and the median age for patients in two completed decades (1985–1994 and 1995–2004) with bone and STS are in table 6.

Five-year survival data are plotted against publication date (figure 2) and the size of the series (figure 3) for 14 of these 15 studies where the data are given, to allow for visual inspection of time or case volume. Three-year and/or 5-year survival for the 15 studies including first (and subsequent) metastasectomy data are plotted in figure 4.

Five-year survival data are set out in table 7 sorted by tumour type from 14 of the 15 studies including first (and subsequent) metastasectomy data. Together these provide data on 1196 patients having metastasectomy from as early as 1976¹⁶ to as recently as 2008.¹⁴ Overall about a third of patients who have had pulmonary metastasectomy for bone sarcoma and about a quarter who have had pulmonary metastasectomy for STS survive beyond 5 years. TCR five-year survival data for two complete decades 1985–1994 and 1995–2004 are included in the table to provide a reference measure of survival in all sarcoma patients in the registry who were classified as stage 4, that is, sarcoma metastasised at presentation/registration. Direct comparison cannot be made but it is a reminder that an implicit assumption that the 5-year survivals of the patients in the pulmonary metastasectomy series would have approached zero would be incorrect.

We can reasonably deduce that

Five-year survival from 14 of the 15 studies reporting first (and subsequent) pulmonary metastasectomy operations. They are grouped by sarcoma type and then by mid-year of the series to aid visual inspection for time trends. TCR 5-year survival data for stage 4 patients are provided for two complete decades of data overlapping the reported series. These TCR patients all had metastases at presentation but not necessarily lung or lung only.

Data are not available in the publications concerning the fate of patients beyond 5 years and there are no narrative accounts of the clinical course of these patients. However, a number of the authors include, in their narrative, a statement of belief in cure for patients who have recurred in the lung or that their surgery has curative intent^{12 14 16 17 19 24 25} and the phrase ‘potentially curative resection’ is included in NICE guidance.³ Illustrative statements from recent publications are these from 2009 and 2010

Given the lack of effective systemic therapies, PM remains the only potentially curative treatment for STS lung metastases as long as all known disease can be completely resected with negative margins.¹⁶

We demonstrate that after repeated metastasectomies, a subset of patients can be cured.¹⁴

Other authors explicitly state the improbability of cure attributable to pulmonary metastasectomy. Antunes writes “The 5-year survival may reach 50%, although true cure is extremely rare, the majority of patients eventually dying of the disease.”¹⁰ And Sardenberg et al state “It should be emphasised that surgery does not change the biology of the tumor or the metastatic process, and a definitive cure for most patients represents the combination of host histology, tumor spread, response to systemic therapy, and surgical resection, which together render the patient free of disease.”¹⁹

Several reports include multivariate analysis to seek factors that might determine a greater or lesser survival rate. The interval between diagnosis or resection of the primary cancer and the metastasectomy surgery is the commonest factor reported as being significant^{11 18 19 21 23 25 26} survival usually being better if the interval was 12 months or longer. Fewer metastases was also associated with better survival^{13 14 18 19 23 25 26 30} most commonly at a number of about three or fewer. Female sex^{14 16} was also favourable. Patients in whom there was substantial necrosis following chemotherapy survived longer.^{10 11 14 22}

Efficacy, effectiveness and cost effectiveness are different measures of the attributes of a treatment. Pulmonary metastasectomy has been shown to be efficacious in that complete macroscopic clearance has been achieved; in appropriately selected cases, R0 resection of all known pulmonary metastases can be consistently accomplished. Whether pulmonary metastasectomy is effective in prolonging life requires proof that survival has been extended, by metastasectomy, beyond that which would have occurred without pulmonary metastasectomy. Cost-effectiveness requires, in addition to survival, measures of health gained, measures of health lost due to death and complications, and for these to be costed in comparison with any alternatives, including no treatment. This third measure, estimation of cost-effectiveness is outside the scope of this review and depends on first establishing effectiveness.

Evaluation of the effectiveness in preventing or postponing death by pulmonary metastasectomy is the common objective in these clinical reports. The existing practice is believed to be effective based on repeated experience worldwide for over 40 years. In 1971, thoracic surgeons at Memorial Sloane-Kettering reported on 22 patients with treated osteogenic sarcoma in whom they performed lung resections.³² The meticulous case by case communication of that experience merits revisiting (figure 5).

In the discussion that followed Beattie modestly states:

We reported these data with some reluctance, since they really constitute a progress report on a clinical research project underway at Memorial Hospital. We used osteogenic sarcomas because they are such serious tumors. There are occasional spontaneous regressions and good results; but you saw in the figures Dr Martini showed that with amputation we have had a 17% five-year cure rate. Of the 83% of patients who died, 5% lived 3 years. Very occasionally a patient would go on longer before dying.³²

Twenty years later³³ Beattie reported further

Twenty-year follow-up reveals that at least four of the six survivors at 10 years survived more than 19 years; one was lost to follow-up. The patient who died of metastatic osteogenic sarcoma more than 19 years after her first thoracotomy had a total of nine thoracotomies.

An implicit assumption is embedded in many subsequent reports that none of these patients would have lived beyond 5 years without metastasectomy. There are other reasons, apart from having had pulmonary metastasectomy, for patients with pulmonary metastatic disease to be alive at and beyond 5 years. These patients are carefully selected and although it is not possible to put a reliable figure on it, they are a minority of all patients with the disease (table 1).² It is appropriate to select patients for surgery and yet comments such as “survival of (surgically treated) patients was significantly better than of patients ineligible for metastasectomy, p<0.00001”¹⁴ is not an appropriate or meaningful statistical comparison: the difference is evident but how much is due to the selection and how much to the surgery, cannot be determined.

The identifiable factors for the selection of suitable patients include fewer metastases and a longer interval between the diagnosis and treatment of the primary and the resection of pulmonary metastases.^{34 35} Others which appear in some analyses are tumour doubling time and the size of the nodule(s) which under surveillance is a proxy for rate of growth. These are prognostic features for survival under any circumstance. We also know from the TCR data that there are 20–25% of patients with osteosarcoma and about 15% with soft tissues sarcoma, with metastases at the time of registration, who are alive beyond 5 years (tables 6 and ​and7).7). The narrative accounts record that there are some natural long-term survivors: “one patient not operated on remains alive 18 years after not having surgery.”¹⁴ “Long-term survivors appear to belong to a subset of patients with indolent, lung-only disease.”³⁶ These natural survivors, who are likely to have slower progression, and fewer metastases, are likely to be disproportionately frequent among patients selected for metastasectomy³⁷ for the selection process is not random as illustrated in figure 6. More than 30 years ago Aberg first proposed that selection might be the major factor determining survival after pulmonary metastasectomy³⁸ and returned to this question in 1997.³⁹ It is of note that none of the authors cite Aberg. It is known that ‘citation distortions create unfounded authority’⁴⁰ which has been confirmed in the context of metastasectomy by a citation network analysis.⁴¹

In fact, the effect of selection may easily be underestimated. Multivariable analysis cannot detect more than a limited number of factors in these relatively small series. All of these factors may exhibit covariance and they are all indices of the relative aggressiveness of a cancer. Failure for one or more of these factors to reach significance in any particular Cox model does not refute the overall finding. Furthermore, if the knowledge of previous finding leads to the exclusion of some patients, the range of that variable is reduced and it is less likely to be found in subsequent analyses.⁴²

In this group of young and therefore physically resilient patients there seems to be a strong desire to never say no, and to push the boundaries for selection for metastasectomy. It is well established that patients with more metastases are unlikely to live long after metastasectomy. Nevertheless, there is no apparent upper limit in many of the reports nor was there for 85% of surgeons in the ESTS survey.⁶ Among the reports of pulmonary metastasectomy in the present systematic review some surgeons report very high numbers of metastases resected. For example “The authors have removed as many as 80–100 nodules during a single thoracotomy”²² and in the COSS report “The highest number of pulmonary nodules surgically removed was 250”’¹ are out of line with the observational evidence that above a count of relatively few metastases, outcomes are too poor to justify this surgery. Maybe these high numbers reflect the observation that the preoperative count of nodules underestimates the true extent of the disease “It is interesting that despite the presence of only three or four nodules on many CT scans, up to 50 or more nodules were found and removed in a number of these patients.”²² The issue of the relative reliability of modern imaging versus surgeons’ palpation of the lung has been considered elsewhere with varying conclusions.^{43 44}

The extent of necrosis caused by chemotherapy evident in the excised nodule was found to be a favourable feature for survival.¹⁰ The clinical response or failure to respond to chemotherapy was given as factor in selection in several series. Both of these observations are evidence that chemotherapy is having an effective in these patients and has had a demonstrable effect on their disease. Why then should a survival difference be attributed to the surgery?

There is a firm belief that if sarcoma recurs in the lungs, and the patient is still within the criteria for resection, further resections should be performed. “The authors concluded that patients persistently free of the primary osteosarcoma who developed recurrent resectable metastatic disease of the lung should be considered for reoperation a second, third, or fourth time, as these patients had similar DFI curves after five years.”²⁶ This belief is supported by what might be inappropriate data interpretation. Consider these statements for example: “Prognostic factors for increased survival included 3 or greater redo pulmonary operations’;²³ “patients with complete resection for recurrent pulmonary metastasis show a significantly better prognosis after repeat pulmonary metastasectomy”²⁷ and “repeat metastasectomy for recurrent pulmonary metastasis also provided a favorable overall survival (p<0.041).”²⁰ To undergo a second metastasectomy a patient has to have survived, and been without evidence of disease for a reasonable length of time, to meet the criteria for each subsequent operation. The problem is exemplified by Sardenberg and colleagues. Survival was measured from the first thoracotomy for pulmonary metastasectomy (confirmed with the first author) and was 15, 45 and 48 months, in 35 patients having only one surgical episode, 24 who had 3 and 13 who had three. The authors provide a statistical analysis (p=0.077) of the association between more thoracotomies and longer survival. They neglect the fact in their interpretation that survival after surgery, and for a reasonable period of time, was a requirement to move to the next analytical group.¹⁹ This way of presenting the data maximises survivor bias: ongoing survival is an entry criterion to having a further metastasectomy operation.

We found no data, or even a narrative account, concerning the effect of surgery on symptoms. The patients who are regarded as candidates for metastasectomy are generally detected on surveillance and presentation with symptoms probably distinguishes patients as being not suitable for metastasectomy, either because it represents extrapulmonary disease or the pulmonary disease is too advanced. The evidence cited in NICE guidance³ is as follows: “Detection on the basis of symptoms occurred in 21 patients. Fifteen of these patients presented between scheduled visits. Seven patients were symptomatic primarily on the basis of their metastatic pulmonary disease. These patients had diffuse metastatic disease in all cases, with documented synchronous recurrence outside the lung, and none was resectable.”⁵ What is implicit in many of the reports reviewed is that ultimately they call a halt to repeated thoractomies because the patient respiratory function will not withstand further surgery and loss of lung tissue.

Two publications^{45 46} which appeared in 2011 after completion of our review and data analysis, and following submission, have been brought to our attention by reviewers. They come from prominent North American institutions and appeared in a specialist thoracic surgical journal, The Annals of Thoracic Surgery. The report from Brigham and Women's Hospital, Boston, Massachusetts states in its title “Repeated and aggressive pulmonary resections for leiomyoma metastases extends survival.” The report is of 82 patients with a variety of bone and soft tissues sarcomas between 1989 and 2004. Repeat metastasectomy was performed in 28/82 with some patients having three, four or five thoracic operations. Operated patients with leiomyosarcoma (31/82) had a median survival of 70 months compared with 24 months for other sarcoma subtypes. No control data for survival among comparable patients, not operated upon, are provided. The text confirms that the surgery was repeated and aggressive but that survival was extended as a result cannot be inferred from the data presented for the reasons given already.

The other report, from the Massachusetts General Hospital, is of 97 patients with 13 sarcoma subtypes operated on for pulmonary metastases between 2002 and 2008.⁴⁶ They report that of 69% of 29 patients who had multiple operations were alive at 5 years compared with 41% of 60 patients who had a single operation and find the difference to be statistically significant and the bottom line of the conclusions reads “Repeated pulmonary metastasectomy in select patients may improve survival despite recurrent disease.” But patients have first to be survivors to be candidates for surgery, a point the authors themselves make in their discussion “Patients in whom disease rapidly recurred after surgery (either as a local recurrence or disseminated disease) were probably selected out from repeated surgical resection.”⁴⁶ Patients who thus meet the criteria for a second operation must necessarily live longer making the argument circular and the statistical test spurious.

Neither of these papers provides evidence on symptomatic benefit for these patients.