The treatment of early-stage non–small cell lung cancer (NSCLC) continues to pose a major challenge, as reflected by disappointing 5-year disease-free survivals rates of only 69% in patients in a randomized trial who underwent surgery and mediastinal lymph node dissection. Significant surgical toxicity continues to be noted in unselected patient populations, and centralization of surgical resections to high-volume centers may not reduce 30-day mortality rates. It is interesting to note that 90-day mortality rates are approximately twice those of the more commonly cited 30-day figures.[4, 5] Analysis in nearly 1300 patients who underwent surgery for early-stage NSCLC revealed a risk of disease recurrence ranging from 6% to 10% per person-year during the first 4 years after surgery. Challenges facing long-term survivors include the finding that > 50% of all deaths in 5-year survivors of NSCLC are also related to lung cancer, with a persistent risk of death from lung cancer observed up to 18 years from the time of diagnosis. Recent data indicate a far higher incidence of second primary lung cancers in this patient group than was previously appreciated.[6, 8]
Although an anatomical resection is accepted as the standard of care in patients with early-stage NSCLC, a significant percentage of elderly patients do not undergo any local therapy.[5, 9] Changing demographics of lung cancer have resulted in the elderly now representing the fastest growing patient subgroup. Comorbidities are more common in the elderly, and comorbidity in elderly patients with lung cancer is associated with higher overall mortality. It is in the latter patient subgroups that stereotactic ablative radiotherapy (SABR, or SBRT) has had a major impact on population survival rates.
SABR is typically characterized by the use of 4-dimensional radiotherapy planning scans and the delivery of high doses of radiotherapy ranging from 7.5 Grays to 18 Grays, typically in 3 to 8 once-daily fractions delivered with a high degree of precision. Current evidence suggests that SABR can be considered to be the preferred treatment in patients who are unfit for surgery. Population-based studies revealed disappointing survival after conventionally fractionated radiotherapy delivered in ≥ 30 fractions. A study that was conducted after the introduction of SABR into the Dutch population revealed an improvement in the median survival of 9.3 months in patients aged ≥ 75 years who underwent radiotherapy. SABR has been used mainly for patients with peripheral lung tumors who are unfit to undergo surgery. However, retrospective studies in patients who underwent SABR despite being fit to undergo surgery have reported 5-year overall survival rates in the range of 51% to 70%.[15, 16] These reports, together with the low reported toxicity of SABR, have led to interest in trials comparing both treatments, but these prospective trials have either failed to accrue patients or are doing so at a pace that is unlikely to provide any answer in the near future. In such a situation, propensity score-matched pair analysis can be used to obtain 2 comparison groups with similar known prognostic factor characteristics. The propensity score is the conditional probability of assignment to a particular treatment based on several factors or covariates (eg, sex, age, cTNM stage, etc). The score, ranging from 0 to 1, assigns a probability of having received a particular condition (eg, SABR) to each patient, and adjustment for the scalar propensity score can remove bias due to all observed covariates.
In this issue, Varlotto et al described the results of such a retrospective matched-pair analysis, pooling databases from 4 institutions accrued between 1999 and 2008, with all eligible SABR cases being proven on pathology. First, prognostic factors influencing overall survival, any disease recurrence, and local tumor control were identified in the combined set of 137 patients treated with SBRT and 180 patients who underwent surgery. These factors were then used to perform a matched-pair analysis, with correction for potential biases in treatment assignment between the 2 groups using the propensity score as a covariate. Matched-pair analysis by pathology, patient age, size, and patient sex was performed for evaluating locoregional and disseminated control. For evaluating differences in overall survival, the analysis by Varlotto et al also controlled for the Charlson Comorbidity Index and aspirin use. After adjusting for propensity scores, overall survival was found to be correlated only with the Charlson Comorbidity Index. Similarly, the inclusion of propensity scores revealed that only tumor diameter, and not treatment modality, was correlated with locoregional control.
Some limitations of the study by Varlotto et al deserve mention. Details of the extent of lymph node dissection, including the number of lymph nodes examined, were not provided for in the surgical cohort. This may confound the analysis of locoregional control because compliance with guideline-specified surgical lymph node staging can be poor. For example, data from the National Cancer Institute's Surveillance, Epidemiology, and End Results program from 1998 through 2002 revealed that 62% of patients with pathologic N0 or N1 NSCLC had no mediastinal lymph nodes examined, although performing mediastinal lymph node staging was associated with a 7% reduction in all-cause mortality and a 11% reduction in lung cancer-specific mortality. Similarly, only 36% of the patients at 4 Dutch teaching hospitals had at least 3 mediastinal lymph nodes stations removed during surgery for lung cancer. An advantage for surgery over SABR is considered to be the ability to detect occult lymph node metastases and administer appropriate adjuvant postoperative chemotherapy. However, up to one-third of patients may not be able or willing to receive postoperative chemotherapy,[21, 22] thereby diminishing the perceived disadvantage of SABR.
The study by Varlotto et al found no differences in overall survival, total disease recurrence control, or locoregional control between SABR and surgical resection, findings that are supported in the recent literature. A Dutch propensity score-matched analysis comparing locoregional control after lobectomy using video-assisted thoracic surgery and SABR reported superior 1-year and 3-year locoregional control rates after SABR. The Dutch study did not observe any significant differences between both treatment groups with regard to rates of distant disease recurrence and overall survival. Another propensity score-matched study comparing patients treated with surgical resection (predominantly lobectomy) or SABR reported similar cause-specific survival rates and patterns of failure between the treatment groups. After both surgery and SABR, an early manifestation of distant metastases represents the predominant pattern of disease failure.[6, 13] The present finding that overall survival is correlated only with the Charlson Comorbidity Index is consistent with population-based data indicating that stage-specific 5-year survival rates in the Danish Lung Cancer Registry for patients with early-stage NSCLC and severe comorbidity were significantly lower than in patients without comorbid disease, with the survival impact of severe comorbidity found to be greatest during the first month after surgery, but persisting during long-term follow-up.
The findings of all these studies will add to clinical equipoise and should encourage enrollment in randomized clinical trials to better compare the effectiveness of surgery and SABR in patients with early-stage NSCLC. However, it should be recognized that patients enrolled in clinical trials comprise < 2% of all patients with cancer, and are not representative of the cancer population at large. Enrollment in trials of lung cancer also predicts for a lower hazard of death. Consequently, observational analyses of databases to compare outcomes between patients receiving different treatments will remain an appealing and useful tool in comparative effectiveness research due to their efficiency and generalizability to larger populations. Observational studies are likely to remain relevant for determining treatment choices in the frail and so-called “borderline” operable patients, who are most likely to benefit from SABR but are also the least likely to be included in clinical trials.