Comment on “Evaluating pathological complete response as a surrogate endpoint for long-term survival in patients with non-small cell lung cancer: a systematic review and meta-analysis”

Dear Editor, We read with great interest the recent article by Li et al[1], “Evaluating pathological complete response as a surrogate endpoint for long-term survival in patients with non-small cell lung cancer: a systematic review and meta-analysis.” The authors synthesize data from 14 randomized clinical trials that enrolled 4025 patients to examine whether pathological complete response (pCR) can serve as a surrogate for long-term outcomes, specifically event-free survival (EFS) and overall survival (OS) in resectable non-small cell lung cancer (NSCLC). They report correlations between higher pCR rates and improved survival and propose that pCR could accelerate drug development for early-stage NSCLC. We acknowledge the value of this comprehensive meta-analysis, but we have several methodological and interpretive concerns. First, we contend that pCR is an indispensable postoperative assessment metric, yet it is insufficient by itself as a primary endpoint for long-term efficacy trials in resectable NSCLC. The article cites the success of pCR as a surrogate endpoint in breast cancer, but breast and lung cancers exhibit fundamentally different biological behaviors. Breast cancer typically arises from mammary ductal or lobular epithelial cells, whereas lung cancer most often originates from bronchial mucosal or alveolar epithelium. These etiologic distinctions translate into different clinical management: breast cancer therapy is stratified by markers such as ER, PR, HER2, and the Ki67 index, which directly inform treatment selection[2]. Lung cancer is broadly classified into small cell lung cancer and non-small cell lung cancer, and requiring markedly different treatment approaches[3]. A second issue is that studies do not use a uniform definition of pCR. Variations include whether residual in situ carcinoma is permitted, whether lymph node status must be N0, and whether pathological sampling follows the IASLC standardized procedure. These methodological differences produce variability in reported pCR rates, yet the authors did not fully control for them or perform sensitivity analyses. In clinical practice, the extent and standardization of pathological sampling markedly affect pCR rates, yet the study treated this key factor too crudely and thus calls into question the stability of pCR as a surrogate endpoint. Third, despite substantial differences in study protocols, the article concludes rather uniformly that “pCR can be used as a surrogate endpoint.” The pooled trials encompassed markedly different treatment strategies – pure chemotherapy, immunotherapy combined with chemotherapy, immunotherapy monotherapy – along with varying stages, cycle numbers, and inclusion or exclusion of postoperative adjuvant therapy. Although the authors performed subgroup analyses, many subgroups comprised only three to five studies and yielded confidence intervals that nearly spanned 0 to 1; these results do not justify claims of a “strong correlation.” Consequently, the paper’s overall judgment appears optimistic and insufficiently rigorous. In recent years, numerous new regimens have emerged. Thus, relying solely on pCR to substitute for long-term survival (OS) assessment is inadequate. For example, the large phase III trials of immunotherapy combined with chemotherapy included in the article[4,5]. Most studies had short follow-up periods, and overall survival (OS) data were largely immature, leaving hazard ratio estimates unstable. Fourth, the article’s explanation of the surrogate-threshold effect (STE), a key indicator of surrogate validity, is inadequate. An STE below 1 generally indicates a strong association and high surrogate reliability. As STE increases above 1, the association weakens and may be ineffective. However, the STE values reported by the authors are extremely high. For example, their estimates imply that the pCR rate would need to be more than tenfold higher than that of the control group to predict a meaningful OS improvement, a scenario that is virtually never seen in clinical practice. Such high STEs indicate that pCR cannot reliably predict long-term survival, yet the authors did not explore this implication in depth and instead retained a relatively positive conclusion. A meta-analysis[6] reported that, in clinical trials of neoadjuvant immune checkpoint blockade for resectable non-small cell lung cancer, pCR and MPR correlated significantly with 2-year EFS but not with OS. In summary, although the authors provide a useful synthesis of current evidence, the methodological limitations described above substantially weaken the case for pCR as a reliable surrogate for long-term survival in resectable NSCLC. To address these shortcomings, I propose developing a composite scoring system for lung cancer that integrates multiple prognostic and predictive factors. The score could include MPR, along with PD-L1 expression[7], immune microenvironment status[8], changes in tumor marker levels, ctDNA[9], pathological risk features, and alterations in driver genes[10]. Subsequent validation should assess the correlation between this risk score and long-term outcomes, with the goal of establishing an early, comprehensive model for prognosis prediction.