Association between Tumor Size at the Time of Disease Progression and Survival Outcomes
Article information
Abstract
Purpose
This study evaluates the prognostic significance of tumor size at disease progression (PD) and depth of response (DOR) in cancer patients.
Materials and Methods
We performed post hoc analysis using data from six prospective clinical trials conducted by the Korean Cancer Study Group. Patients with tumor size at PD was categorized into ‘Mild PD’ and ‘Significant PD’ based on the cutoff values of relative change from baseline using maximally selected rank statistics. The overall survival (OS) and progression-free survival (PFS) were compared between PD and DOR categories.
Results
Among the 194 evaluable patients, 130 experienced PD. A 35.48% decrease from baseline in tumor size at PD was chosen for the cutoff between mild and significant PD for OS (mild PD: tumor size from the baseline ≤ −35.48%; significant PD > −35.48%). The mild PD had superior OS compared to the significant PD (25.8 vs. 12.8 months; Hazard ratio [HR] 0.47, 95% CI 0.266-0.843, p=0.009). When using an exploratory cutoff based on whether the tumor size was below vs. exceeded from the baseline (mild PD: tumor size from the baseline ≤ 0%; significant PD > 0%), OS remained significantly longer in the mild PD (17.1 vs. 11.8 months; HR 0.60, 95% CI 0.392-0.932, p=0.021). The greatest DOR was associated with the longest OS and PFS (p<0.001 for both).
Conclusion
Tumor size at PD and DOR were significant prognostic factors for progressive disease. Maintaining a sufficiently reduced tumor size even during PD was associated with better survival outcomes.
Introduction
The Response Evaluation Criteria for Solid Tumors (RECIST) 1.1 have been adopted as the principal standard for assessing the efficacy of therapeutic agents in the clinical setting [1,2]. The categorization of overall responses included complete response (CR), partial response (PR, defined as a 30% reduction in the sum from baseline), progressive disease (PD, marked by a 20% increase from the nadir), and stable disease (neither PR nor PD). These criteria are straightforward, reproducible, and objective, enabling oncologists to actively use them in clinical trials and daily clinical practice to make treatment decisions [3]. PD indicates no further clinical benefit with the current treatment and often serves as a criterion for discontinuing chemotherapy.
However, the simplicity of RECEST 1.1 can be limiting, as they may not fully capture the diverse progression patterns observed during the natural course of cancers. Even if the primary target lesion demonstrates PR, the appearance of a new lesion is classified as PD, regardless of the size of the target lesion. While PR uses the pretreatment baseline sum as a reference, PD is assessed by the increase from the smallest sum during treatment. Therefore, it is noteworthy that in cases where there is a significant reduction in tumor burden due to a prior excellent response, even a small absolute increase from this smallest sum can exceed 20% in the axial diameter, leading to a classification of PD, although the tumor size may still be smaller than the baseline.
Tumor size is widely recognized as a significant prognostic factor in various cancer types [4-11]. Although controversial, larger tumor sizes are reportedly associated with poorer prognoses, especially when performing definitive locoregional treatments such as surgery. Consequently, we hypothesized that, within the subset of patients experiencing PD during treatment, the prognosis may vary depending on the tumor size at the time of PD. Based on this hypothesis, we assessed survival outcomes among patients with PD, focusing on the extent of tumor size reduction relative to baseline at the time of PD. Additionally, we investigated the prognostic differences associated with maximal reduction in tumor size during the treatment period.
Materials and Methods
1. Patient population
For this exploratory retrospective analysis, data were pooled from six prospective clinical trials conducted by the Korean Cancer Study Group. To determine changes in tumor size, only participants with at least one measurable lesion were included in the analysis. Patients who only had baseline tumor sizes but were not followed-up were also excluded. Data collected from each participant included age, sex, cancer type, drugs administered as interventions, response assessment results, tumor size, progression status, overall survival (OS), and progression-free survival (PFS). Information on each clinical trial included in this study is summarized in Table 1.
Summary of the pooled clinical trials
2. Categorization of patient group based on the tumor size at the time of PD
Maximally selected rank statistics for OS were used to investigate the optimal cutoff value for the relative change (%) in tumor size (sum of target lesions) that would most significantly differentiate the two comparison groups [14]. Additionally, in a clinically pragmatic and statistically conservative approach, a subsequent exploratory analysis was conducted according to whether the sum of the measured lesions at PD remained the same or less than at baseline, or exceeded the baseline instead of the statistically determined cutoff value.
3. Assessment
Response evaluations for all solid tumors, except lymphoma, were conducted based on RECIST 1.1 criteria. Lymphoma was evaluated according to the revised Cheson 2007 criteria [12]. We assessed progression dynamics by categorizing PD into two distinct classifications: ‘Mild PD’ and ‘Significant PD’ based on the relative change in the sum of the measured lesions in patients with PD. Mild and significant PD were defined based on the extent of change in the sum of measurable lesions at the point of PD compared to the baseline measurement before treatment. If the reduction was below the given statistically determined cutoff value described above, it was classified as mild PD; if it exceeded the cutoff value, it was classified as significant PD. Thus, both terms indicated PD, but they were distinguished based on the extent of reduction from baseline. OS was defined as the time from the start of treatment to the date of death, or if the participant was still alive, to the last date of follow-up in the clinical trial. PFS was defined as the time from the start of treatment to the date of confirmed progression or death, or to the last follow-up if no progression or death occurred, whichever occurred first. The depth of response (DOR) was defined as the maximum tumor shrinkage compared to the baseline target lesion, represented as a percentage, regardless of any specific time during the treatment [13].
4. Association between survival outcomes and DOR
For statistical comparison, DOR was categorized into quartiles as follows: quartile 1 (Q1), a decrease of more than 0% and less than or equal to 25% compared to the baseline (−25% ≤ Q1 < 0%), Q2: a decrease of more than 25% and less than or equal to 50% (−50% ≤ Q2 < −25%), Q3: a decrease of more than 50% and less than or equal to 75% (−75% ≤ Q3 < −50%), Q4: a decrease of more than 75% and up to 100% (−100% ≤ Q4 < −75%) including those where the sum of target lesions showed CR. Individuals with no reduction in tumor size (tumor size (tumor size ≥ 0%) were classified as having ‘No shrinkage’. PFS and OS were also compared based on the median value of each participant’s DOR among the total population, regardless of PD.
5. Statistical analysis
The patient characteristics were analyzed and summarized using descriptive methods. Survival outcomes were assessed using Kaplan-Meier curves. The log-rank test was performed to determine whether there was a significant difference in the survival curves of each group. Hazard ratios (HR) were calculated using a Cox regression hazard model for both unadjusted and adjusted analyses. Univariable results were adjusted for clinical variables, including age, sex, and study type. The ‘survival,’ ‘survminer’ and ‘maxstat’ packages were used for calculating the cutoff point distinguishing between mild PD and significant PD in the R program (R Core Team, 2023). All other statistical analyses were performed using IBM SPSS Statistics for Windows ver. 20 (IBM Corp.).
Results
1. Patient characteristics
A total of 318 participants were included in the pooled clinical trial. After excluding patients with no measurable target lesions and those lacking follow-up data for tumor size measurements after baseline, the total number of evaluable patients was 194. Among them, 130 patients were confirmed to have PD. Patient demographics are summarized in Table 2.
Patient’s characteristics
2. Clinical significance of tumor size at PD
Maximally selected rank statistics revealed that a 35.48% decrease in tumor size (sum of target lesions) at the time of PD served as a significant demarcation between mild and significant PD in relation to OS (S1 Fig.). When applying a cutoff of −35.48% (mild PD: tumor size from the baseline ≤ −35.48%; significant PD > −35.48%), 30 patients were classified as mild PD and 100 as significant PD (Table 2). Patients with mild PD showed a statistically superior OS compared to those with significant PD (25.8 vs. 12.8 months; HR, 0.47; 95% confidential interval [CI], 0.266 to 0.843; p=0.009) (Fig. 1A). In a similar manner, to apply it clinically in a more practical and intuitive manner while adopting a conservative approach, the tumor size during PD was divided into patients whose tumors grew to the size at baseline or smaller, and those whose tumors grew larger than the baseline (mild PD: tumor size from the baseline ≤ 0%, n=73; significant PD > 0%, n=57) (S2 Table). In these comparisons, the OS was significantly longer in the mild PD (17.1 vs. 11.8 months; HR, 0.60; 95% CI, 0.392 to 0.932; p=0.021) (Fig. 1B). When adjusted for sex, age, and study type, patients with mild PD consistently showed a lower mortality rate than those with significant PD (HR, 0.45; 95% CI, 0.276 to 0.741; p=0.002) (Table 3).
Overall survival based on pre-specified cutoff values of tumor size (sum of target lesion) at the time of disease progression: (A) mild disease progression (PD): tumor size from the baseline ≤ −35.48%; significant PD > −35.48%, (B) mild PD: tumor size from the baseline ≤ 0%; significant PD > 0%. CI, confidence interval; HR, hazard ratio.
Multivariable analysis on the overall survival among patients with PD
3. Association between DOR and survival
The DOR waterfall plot from pooled trials among the total population is shown in S3 Fig. The OS was significantly different according to the DOR quartiles (p < 0.001) (Fig. 2A). The most substantial reduction in tumor size observed in Q4 corresponded to the best median OS (not reached). Conversely, the least reduction in tumor size, noted in Q1, showed the shortest median survival (12.7 months). This was comparable to the no shrinkage group (15.2 months, p=0.740). The PFS showed the same pattern. Q4 exhibited the longest median PFS (46.7 months), which decreased toward Q1 (p < 0.001) (Fig. 2B). The no shrinkage group had the shortest median PFS (2.8 months). Similar statistical patterns were observed in the binary comparison based on the median DOR value (−26.9%) for each individual. The median OS was not reached for the below-median group vs. 12.8 months for the above-median group (HR, 0.41; 95% CI, 0.270 to 0.627; p < 0.001). The median PFS was 15.2 months for the below-median group vs. 5.3 months for the above-median group (HR, 0.30; 95% CI, 0.208 to 0.423; p < 0.001) (S4 Fig.).
Comparison of survival based on depth of response among all participants, with patients divided into quartiles. (A) Overall survival. (B) Progression-free survival.
Discussion
Our findings indicate that survival may vary depending on the degree of increase in tumor size upon reaching PD, favoring a relatively small tumor size at PD. Survival analysis according to DOR further supported the association between a decrease in tumor size and favorable prognosis. Of note, the finding that OS remained significantly superior in the mild PD group, even when baseline tumor size was used as an alternative cutoff (i.e., 0% reduction) instead of the 35.48% reduction, underscores the clinical importance of tumor size in PD. Interestingly, a 35.48% reduction from baseline is similar to the RECIST 1.1 definition of a PR, which requires a 30% reduction. This implies that patients undergoing PD may still have a relatively favorable prognosis if the sum of their target lesions remains at or below the initial PR levels. An additional exploratory analysis using baseline tumor size as an alternative reference point was conducted because the −35.48% threshold was deemed impractical for routine application in clinical practice. This approach included individuals with larger tumors in the mild PD category compared to when a −35.48% cutoff was used, further validating the clinical significance of maintaining or reducing tumor size relative to the baseline. This implies that maintaining the tumor size at or below its initial baseline measurement at the start of treatment could potentially be associated with a more favorable prognosis, even in cases of progressive disease.
The reason why the mild PD group was associated with a better prognosis remains an open question, but several hypotheses can be proposed. First, a smaller tumor size itself may be an independent favorable prognostic factor, as indicated by multiple studies showing that a low tumor burden or smaller tumor size correlates with better outcomes [4-11]. Second, when PD occurs, patients with smaller tumors may take longer to reach a lethal tumor burden of approximately 1012 cells [15]. However, since subsequent treatment after PD can vary greatly among patients, this assumption appears to oversimplify complex tumor kinetics. Third, it can be assumed that in the mild PD group, the rate of tumor growth was relatively slower than that in the significant PD group. Considering that the treatment can reduce the cell proliferation rate as a result of the cytostatic effect [16], it is possible that the anticancer effects of the current treatment could be maintained even in cases of PD. This suggests that maintaining the same treatment strategy may be beneficial without switching to the next-line treatment immediately after PD. Fourth, the tumors may still remain small when PD occurs, simply because they have significantly reduced in size from baseline owing to an excellent response to treatment. In this study, a higher DOR was associated with better prognosis, supporting this finding. The DOR is known to be strongly associated with prognosis not only in cytotoxic chemotherapy but also in immunotherapy and targeted therapy for various cancer types [13]. Our study also revealed that DOR is a reproducible prognostic factor regardless of the type of anticancer treatment (cytotoxic, immunotherapy, or targeted therapy), which is consistent with previous reports.
An important consideration is that our results should not be misinterpreted as suggesting that treatments that reduce tumor size, such as debulking surgery, can improve prognosis in metastatic or unresectable disease settings. This is because our study focused on observing changes in tumor size as outcomes of anticancer therapy responses rather than examining the impacts of deliberate size reduction interventions at any given time.
This study has several limitations. First, this study amalgamated data from six independent clinical trials for analysis, which introduced substantial heterogeneity owing to variations in patient clinical profiles, tumor types, and therapeutic interventions. One study in lymphoma patients used the Cheson criteria instead of RECIST 1.1. Although the Cheson criteria differ from RECIST, they are also fundamentally based on target lesion size, which, in our view, justified including this study. Moreover, because all data were derived from participants in prospective clinical trials, factors such as performance status and comorbidities that could significantly affect prognosis were homogeneous and consistently favorable. Furthermore, adjustments were made for clinical factors, including the study type, to maintain data reliability. Second, as previously described, differences in survival can be influenced by various factors, such as intrinsic tumor biology and patterns of subsequent treatment, in addition to tumor size at the time of PD. Therefore, these aspects should be carefully considered when interpreting the data from this study. Finally, our analysis was based solely on changes in the sum of the target lesions in patients with PD. Hence, we also included patients who developed PD unrelated to changes in the target lesion sum, such as the development of new lesions or progression solely to non-target lesions. Our findings suggest that these patients might still have a better prognosis if they are classified as having mild PD based on their target lesions. As PD does not always indicate treatment failure, our findings provide insight beyond the use of PD in systemic therapy.
In conclusion, our study found a significant correlation between the changes in the size of target lesions during anticancer treatment and the prognosis of PD. If the reduction was maintained at or below approximately 35% compared to baseline, or even at or below the same baseline size (mild PD), better survival was observed compared to those exceeding this dimension. These findings are supported by the observation that a greater DOR is associated with a better prognosis. Thus, even in patients with PD, a favorable prognosis can be expected if the tumor size is substantially reduced.
Electronic Supplementary Material
Supplementary materials are available at Cancer Research and Treatment website (https://www.e-crt.org).
Notes
Author Contributions
Conceived and designed the analysis: Maeng CH, Kim BJ, Keam B.
Collected the data: Maeng CH, Kim BJ, Keam B.
Contributed data or analysis tools: Ahn MJ, Choi IS, Zang DY, Kim BH, Kwon M, Heo DS, Keam B.
Performed the analysis: Maeng CH, Kim BH, Kwon M.
Wrote the paper: Maeng CH, Keam B.
Conflict of Interest
Conflict of interest relevant to this article was not reported.
Funding
This study was supported in part by the Korean Cancer Study Group (KCSG). This study was supported by the National R&D Program for Cancer Control through the National Cancer Center (NCC) funded by the Ministry of Health and Welfare, Republic of Korea (HA22C0012).
Acknowledgements
We thank Trial Informatics Co., Ltd. for constructing the KCSG Clinical Trial Data Warehouse.