For several decades, lobectomy has been the standard treatment for peripheral small-sized NSCLC, supported by evidence from phase III trials that indicated its superiority over limited resection in terms of survival and local control [2]. Advances in imaging technology have enabled the detection of less invasive lung cancers [13], which has prompted a shift in practice toward sublobar resections for selected cases, with favorable outcomes [5, 14,15,16]. Recent trials that reported sublobar resections, including segmentectomy and wedge resection, which were not inferior to lobectomy, have further confirmed these paradigm shifts in surgical procedures [17]. A phase III trial comparing lobectomy and segmentectomy for patients with small peripheral NSCLC revealed that segmentectomy, when adequate surgical margins are achieved, was comparable to lobectomy in terms of OS and RFS [6]. Twenty-three patients from our institution were registered in this trial, contributing to the overall accrual; however, during the accrual period, 61 eligible patients were identified at our facility, with 38 not enrolled. Recognizing that registration bias could affect trial outcomes, we conducted the present study to investigate the characteristics and clinical outcomes of the eligible but nonregistered patients. This allowed us to identify patterns in surgical decision-making that emphasized how clinical judgment and patient-specific factors influence trial participation, aiming to provide a unique perspective on the challenges of implementing evidence from controlled trials into daily practice.

Among the 38 nonregistered patients, 6 with high-grade malignant tumors on image findings underwent lobectomy, and another 6 without sufficient margins on segmentectomy required lobectomy. Conversely, two patients underwent segmentectomy because of low malignancy potential inferred from imaging. Comorbidities, although within the trial eligibility criteria, were a significant factor that affected registration. This study did not enroll 11 patients who were eligible but had multiple minor complications, after consultation with the patients, as a limited procedure was considered preferable. Thus, for 25 of the 38 patients, the surgeon’s indications may have affected the patient’s decision, resulting in non-enrolment despite eligibility for the trial. Similar situations likely occur at other participating institutions.

The patients enrolled in the present study may have been eligible for enrollment in the JCOG0802/WJOG4607L trial, considering that in the trial, the 5-year OS was 94.3% and 91.1% and the 5-year RFS was 88.0% and 87.9% for the segmentectomy and lobectomy groups, respectively. However, the patients included in the present study had poorer oncological outcomes. These facts indicate that surgeons’ clinical judgments, including recommendations or opinions about tumors suspected to have high malignant potential, as well as patient-specific factors, may affect decisions about trial participation. This study reflects the experience of a single institution and does not provide definitive evidence; however, it suggests the potential presence of selection bias that may affect trial enrollment patterns.

In clinical trials, selection bias occurs while excluding patients with a poor prognosis or those with favorable outcomes and a clear treatment advantage. Such biases frequently arise from physicians’ clinical judgment and the prioritization of individual patient care over trial participation. According to Rothwell, the impact of this bias is minimal when the tested treatment has a clear advantage over the control. However, when the treatment’s superiority is uncertain, the bias may threaten the study’s external validity [18]. Registration bias may not be eliminated even in large-scale, highly reliable prospective randomized controlled trials if the superiority of the treatment being tested is uncertain. This bias has been reported in randomized controlled trials (RCTs) across oncology as well as other disease areas. Identifying the characteristics of nonregistered patients, as in the present study, represents one approach to evaluating the external validity of trial findings. Recognizing its existence in clinical practice allows for a more nuanced interpretation of RCT results. For example, when using the results of a given RCT to guide patient selection or surgical strategy, clinicians can consider the possibility that certain high- or low-risk patient groups may have been excluded. This provides important context for assessing whether reported survival or recurrence rates are applicable to their own patient population. Thus, awareness of registration bias can facilitate more individualized and appropriate treatment decisions in routine clinical practice.

The CSPOR-LC03 trial, conducted in relation to the JCOG0707 trial, indicates the presence of these biases. The JCOG0707 trial was a phase III randomized controlled trial that compared the efficacy of UFT and tegafur–gimeraci–oteracil (S-1) in patients with stage I NSCLC, who had undergone complete resection [19]. The CSPOR-LC03 trial focused on patients who were eligible for JCOG0707 but were not registered. The CSPOR-LC03 trial emphasizes that even in large-scale phase III trials, registration bias affects results significantly, as healthier patients may be registered and those with poor prognoses may be excluded based on subjective clinical judgment [20]. Although a direct comparison is not possible, our results resonate with the conceptual framework emphasized in the CSPOR-LC03 analysis. This perspective emphasizes the importance of accounting for registration patterns when evaluating the generalizability of randomized trial outcomes, particularly in real-world clinical practice.

The 38 patients who were eligible for the study but were not enrolled suggest registration bias; however, this does not imply the superiority of either lobectomy or segmentectomy, and further studies are needed to clarify the role of segmentectomy. Interestingly, four of the five cases of recurrence in our cohort occurred in the lobectomy group, suggesting that the choice of surgical procedure may not always impact prognosis significantly. This finding does not contradict the results of the JCOG0802/WJOG4607L trial. In fact, our study aims to complement the results of the JCOG0802/WJOG4607L trial by demonstrating the existence of registration bias and highlighting the effects of real-world clinical decision-making on trial enrollment patterns.

The limitation of this study was the small number of patients analyzed. The limited number of segmentectomies (n = 7) made it impossible to compare survival outcomes between surgical procedures. This study is hypothesis-generating and not designed to evaluate the efficacy of a specific surgical procedure. Collaboration with other institutions, ideally involving all facilities participating in phase III trials, and the use of multicenter or registry-based studies could provide a larger and more diverse dataset, enabling more robust comparisons and deeper insights into the association between registration bias and trial outcomes. Future studies addressing these biases may help refine trial methodologies and improve the applicability of their results to real-world settings.