Time-to-progression ratio as a potential study endpoint in early-phase oncology trials: pooled analysis of phase II trials from the Australian MoST program Authors: Derrick H. W. Siu, Frank P. Y. Lin, David M. Thomas, Subotheni Thavaneswaran, David Espinoza, Sarah Chinchen, Mandy L. Ballinger, R. John Simes, Chee Khoon Lee Published: 2026 ABSTRACT ABSTRACTPurposeTime-to-progression ratio (TTPr) measures intra-patient treatment benefit by comparing disease control on study treatment with that of the immediately preceding therapy. Its prognostic validity in the immune checkpoint inhibitor (ICI) era remains uncertain.MethodsWe performed a pooled patient-level analysis of eight phase II trials. Associations between TTPr, TTP on prior therapy (TTP1), and on-study TTP (TTP2/ PFS), objective response rate (ORR), with overall survival (OS) were evaluated using Cox regression. Landmark analyses at 6-months were used to mitigate guarantee-time bias.ResultsAmong 246 evaluable participants, median TTP1 and TTP2 were 3.9 and 2.1 months, respectively, with poor correlation (r = 0.02). TTPr was not associated with OS (TTPr ≥1.3 vs <1.3, HR 0.85, 95% CI 0.57–1.26). In contrast, TTP2 and ORR were strongly prognostic (TTP2 ≥2.1 months vs <2.1 months: HR 0.51, 95% CI 0.35–0.76, p < 0.001); (response vs non-response: HR 0.19, 95% CI 0.07–0.52, p = 0.001) and remained independently associated with OS in multivariable analyses. Findings were consistent across ICI and non-ICI trials.ConclusionsTTPr is not a robust prognostic measure in platform trials. Although this raises concern regarding its validity as a surrogate endpoint for OS, formal evaluation requires randomized treatment comparisons. Introduction Drug development in oncology faces growing challenges as cancers are progressively stratified into biomarker-defined subgroups. For rare cancers or those characterized by uncommon biomarkers, conducting randomized controlled trials (RCTs) based on overall survival (OS) may be infeasible. In these contexts, non-comparative study designs may play a crucial role in evaluating drug activity and generating evidence to support regulatory approval [1]. Basket and platform trials are now widely adopted in oncology, these adaptive trial designs enable early identification of promising drug candidates in biomarker-matched populations, offering more efficient development pathways [2,3]. Such trials are also often powered based on surrogate endpoints, where objective response rate (ORR) and progression-free survival (PFS) are accepted as substantial evidence to support accelerated drug approval [4]. However, the correlation between these surrogate endpoints and OS is often weak [5]. Therefore, identifying surrogate endpoints that more reliably reflect preliminary efficacy is essential to inform go/ no-go decisions in drug development. Time to progression (TTP) ratio [6], also referred to as the progression-free survival ratio or growth modulation index, is recognized as a useful endpoint for assessing preliminary activity of novel cytotoxic chemotherapies and molecular targeted agents in early phase trials [7–15]. TTP ratio (TTPr) compares the TTP of current line of systemic therapy (TTP2) with that on the immediately preceding line of systemic therapy (TTP1). TTP generally shortens with each successive line of treatment, a TTPr threshold of ≥1.3 has been proposed to indicate promising drug activity, suggesting that the investigational therapy achieves a clinically meaningful extension of disease control relative to the prior treatment [6]. A key advantage of TTPr lies in its intra-patient comparison, where each patient serves as their own control. This approach mitigates inter-patient variability in tumor biology and prognostic factors, thereby enhancing the interpretability of efficacy signals in heterogeneous populations—a feature particularly valuable in basket and platform trials. TTPr was mostly validated in the era of cytotoxic chemotherapy, the appropriateness of using TTPr to assess the activity of new classes of therapies remains unknown despite its adoption by several precision oncology platform studies. Of interest, immune checkpoint inhibitors (ICIs) exhibit distinct response dynamics and progression patterns that may differ from those observed with chemotherapies or molecular targeted agents. The role of TTPr has not been adequately studied in this context. While formal evaluation of surrogate endpoint validity and predictive utility is ideally conducted using data from randomized trials, important insights can still be gained by examining the prognostic association between candidate endpoints and OS. In this pooled analysis of individual patient data (IPD) from single-arm phase II studies, we aimed to: (1) evaluate the prognostic association of TTPr on OS across a range of different thresholds; (2) dissect the individual associations of TTP1,TTP2, and ORR with OS and quantify the relative contribution of TTP1 to TTPr to isolate its independent prognostic effect; and (3) compare the performance of TTPr in ICI versus non-ICI studies, to assess its validity across therapeutic modalities. Methods Study design We conducted a pooled analysis of IPD from the Australian Molecular Screening and Therapeutics (MoST) precision oncology program (ACTRN12616000908437) [16]. MoST is a master protocol that screens adults with refractory solid cancers using comprehensive genomic profiling. Eligible participants were enrolled in one of the 19 single-arm phase II trials investigating novel therapies matched to genomic alterations. TTPr was one of the pre-specified endpoints and was collected consistently across these trials. For the current analysis, we included eight studies with mature OS data. These were (1) MoST 2 (ACTRN12616001019493) – durvalumab plus tremelimumab in rare or neglected cancers [17], (2) MoST 7 (ACTRN 12620000,918921) – tremelimumab in advanced rare or neglected cancers [18], (3) MoST10 (ACTRN12620000568910) – palbociclib and avelumab in tumors with gain of function mutation in CDK4 and CCND1-3, or CDKN2A deletion or loss of function mutations [19], (4) MoST 6–8 (ACTRN 12617001000392) – olaparib and durvalumab in cancers with defects in the homologous recombination repair pathway [20], (5) MoST 11 (ACTRN12620000984998) – tildrakizumab in advanced osteosarcomas or soft tissue sarcomas [21], (6) MoST 5 (ACTRN12618000357257) – eribulin in advanced CD31+ angiosarcoma or other selected CD31+ sarcomas [22], (7) MoST 4 (ACTRN12618000281291) – vismodegib in tumors harboring PTCH1 or SMO mutations [23], and (8) MoST 8 (ACTRN12619001265167) – trastuzumab emtansine in tumors harboring HER2 amplifications or mutations [24]. The MoST program was approved by the St Vincent’s Hospital Sydney Human Research Ethics Committee (reference, HREC/16/SVH/23). Data analysis TTP1 was obtained through central adjudication of medical records. TTP2, equivalent to PFS, was calculated from the date of study enrollment to the date of progression or death of any cause. TTPr was determined to be non-evaluable for the following reasons: (1) if there was no prior systemic therapy, (2) if the prior systemic therapy was an adjuvant or neoadjuvant therapy, (3) prior therapy was interpolated with a palliative radiotherapy or surgery to the target lesion, (4) prior therapy was stopped due to toxicity or withdrawal of consent, or (5) had incomplete medical record. OS was defined as the time from study enrollment to death from any cause, with participants alive at last follow-up censored at that date. We assessed the association between TTPr status (≥1.3 vs <1.3) and OS using Cox proportional hazards regression. This association was further explored across varying TTPr thresholds. Additionally, the relationship between TTP1 and OS was also examined: participants were initially categorized by the median TTP1 distribution, and Cox regression analysis was performed to evaluate for prognostic differences, followed by sensitivity analyses across alternative TTP1 cut-points. Landmark analyses were conducted at 6 months to evaluate the influence of TTPr, TTP1, and TTP2, up to the landmark time on subsequent OS. Landmark at 6 months was selected based on our prior work which demonstrated 6-months PFS rate was a valid surrogate for OS in phase II ICI studies [25]. Participants who died before or at 6 months were excluded as they were no longer assessable for OS at the landmark. For those with TTP2 of ≥6 months, TTP2 was truncated at 6 months, termed modified TTP26m (mTTP26m), and survival beyond the landmark was analyzed. Similarly, modified TTPr (mTTPr6m) was calculated using mTTP26m. Sensitivity analyses were performed using a 3-month landmark (mTTP23m). The landmark approach was used to mitigate guarantee-time bias associated with temporal overlap of survival time between TTP2/ PFS and OS. Partitioning survival time using landmark analysis ensure comparability among participants who survived beyond the landmark [26]. Given its established role as a marker of treatment activity, ORR was included to adjust for differences in efficacy across the investigational agents evaluated in this analysis. Univariable and multivariable landmark analyses were used to assess associations of TTP1, TTP2, ORR with OS. Finally, we compared the performance of TTPr and TTP2 between ICI and non-ICI subgroups. All statistical analyses were performed using R software (version 4.4.0). Results Participant characteristics Out of 343 participants, 246 (72%) with evaluable data for TTPr were analyzed. Ninety-seven participants were excluded due to non-evaluable TTP1, key reasons included (i) no prior systemic therapy (n = 40, 41%), (ii) prior systemic therapy stopped due to completion of treatment (n = 30, 31%), and (iii) prior systemic therapy stopped due to adverse events (n = 9%) (Figure 1). The evaluable cohort consisted of 131 females (53%). The median age was 59 years (range 18–85). The most common tumor types were sarcoma (n = 87; 35%), upper gastrointestinal (n = 42; 17%), colorectal (n = 25; 10%), and gynecological cancers (n = 19; 8%). A total of 177 (72%) participants received ICI containing treatments (Table 1). Figure 1.Consort diagram. TTP1, prior line of systemic therapy.A consort diagram showing participant eligibility and inclusion for analysis.Table 1.Baseline demographics of participants with evaluable time-to-progression ratio.VariableN = 2461Age at study registration59 (45, 68)Sex F131 (53%)Tumor type Colorectal25 (10%) Genitourinary7 (3%) Glioma16 (7%) Gynecological19 (8%) Neuroendocrine3 (1%) Other27 (11%) Sarcoma87 (35%) Thoracic15 (6%) Unknown primary5 (2%) Upper gastrointestinal42 (17%)Previous lines of therapy 197 (39%) 285 (35%) 3+64 (26%)Performance status 0115 (47%) 1121 (49%) 210 (4%)Study treatment involves ICI No69 (28%) Yes177 (72%)Previous exposure to ICI No230 (93%) Yes16 (7%)1Median (IQR); n (%). Consort diagram. TTP1, prior line of systemic therapy. Table 1.Baseline demographics of participants with evaluable time-to-progression ratio.VariableN = 2461Age at study registration59 (45, 68)Sex F131 (53%)Tumor type Colorectal25 (10%) Genitourinary7 (3%) Glioma16 (7%) Gynecological19 (8%) Neuroendocrine3 (1%) Other27 (11%) Sarcoma87 (35%) Thoracic15 (6%) Unknown primary5 (2%) Upper gastrointestinal42 (17%)Previous lines of therapy 197 (39%) 285 (35%) 3+64 (26%)Performance status 0115 (47%) 1121 (49%) 210 (4%)Study treatment involves ICI No69 (28%) Yes177 (72%)Previous exposure to ICI No230 (93%) Yes16 (7%)1Median (IQR); n (%). Baseline demographics of participants with evaluable time-to-progression ratio. 1Median (IQR); n (%). The median follow-up was 24.4 months (95% CI 18.6–28.8). Median OS was 10.0 months (95% CI 4.4–18.3). Median TTP1 was 3.9 months (range 0.2–33; IQR 2.1–7), a total of 78 of 246 participants (32%) had TTP1 >6 months. Median TTP2 was 2.1 months (range 0.3–37; IQR 1.7–5.5), a total of 56 of 246 participants (23%) had TTP2 >6 months, and median mTTP26m was 2.1 (IQR 1.7–5.5). The correlation between TTP1 and TTP2 was poor (spearman R = 0.02). Median TTPr was 0.70 (range 0.05–13; IQR 0.30–1.9) and median mTTPr6m was 0.66 (IQR 0.30–1.4). Median TTP1 was significantly longer in the TTPr <1.3 group compared to that in the TTPr ≥1.3 group (5.0 months vs 2.1 months, p < 0.01). 6-month landmark survival analysis according to modified TTPr status At the 6-month landmark, 161 (65%) participants were alive and included in the analysis. Survival beyond this landmark was not significantly different in the mTTPr6m ≥1.3 group vs the mTTPr6m <1.3 group (HR 0.85, 95% CI 0.57–1.26, p = 0.42). HR for survival did not vary significantly across a range of mTTPr6m thresholds (Figure 2). Figure 2.Overall survival beyond 6-month landmark according to modified time-to-progression-ratio status.A composite figure with two parts. A) A Kaplan–Meier line graph titled “6-month landmark survival” comparing two strata: modTTPr less than 1.3 and modTTPr greater than or equal to 1.3. The x-axis is labeled “Time in months” (months), with labeled ticks at 6, 12, 18, 24, 30 and 36. The y-axis is labeled “Survival probability” (unitless), ranging from 0.00 to 1.00 with labeled ticks at 0.00, 0.25, 0.50, 0.75 and 1.00. A dashed horizontal reference line is drawn at survival probability 0.50. Both survival curves start near 1.00 around 6 months and decline in stepwise drops over time. The modTTPr greater than or equal to 1.3 curve stays above the modTTPr less than 1.3 curve from about 6 to roughly 19 to 20 months, then the curves converge and run close together near 0.25 from about 22 to 36 months. Approximate visible values: at 12 months, modTTPr less than 1.3 is about 0.60 and modTTPr greater than or equal to 1.3 is about 0.70; at 18 months, about 0.35 versus about 0.45; at 24 months, both near about 0.25; at 30 months, both near about 0.22 to 0.25; at 36 months, modTTPr greater than or equal to 1.3 is near about 0.25 while modTTPr less than 1.3 is near about 0.10. A “Number at risk: n (percent)” table is shown below with time points 6, 12, 18, 24, 30, 36 months. For modTTPr less than 1.3: 107 (100), 58 (54), 28 (26), 17 (16), 6 (6), 0 (0). For modTTPr greater than or equal to 1.3: 54 (100), 36 (67), 21 (39), 6 (11), 2 (4), 1 (2). B) A forest-plot style summary table with columns “Modified TTPr threshold”, “n less than threshold”, “n greater than or equal to threshold” and “HR”. Six rows list thresholds 0.7, 1.0, 1.3, 1.5, 1.8 and 2.0. The corresponding counts and hazard ratios are: threshold 0.7: 69 and 92, HR 0.74; threshold 1.0: 85 and 76, HR 0.95; threshold 1.3: 107 and 54, HR 0.85; threshold 1.5: 113 and 48, HR 0.89; threshold 1.8: 120 and 41, HR 0.95; threshold 2.0: 128 and 33, HR 1.05. To the right, each row shows a square marker with a horizontal confidence interval line. The x-axis under the forest plot is labeled with numeric ticks 0.50, 0.71, 1.0 and 2.0 (unitless). The square markers align near their listed hazard ratio values, with the last row’s marker slightly above 1.0 and the others below 1.0.Two plots of 6-month landmark survival and hazard ratio across modified time-to-progression-ratio thresholds. Overall survival beyond 6-month landmark according to modified time-to-progression-ratio status. As a sensitivity analysis, we conducted Cox regression analysis on OS at the 6-month and 3-month landmarks modeling mTTPr as a continuous variable, mTTPr6m (HR 0.96; p = 0.50) and mTTPr3m (HR 0.97; p = 0.67) were not associated with improved OS (Supplementary Table S1a and S1b, respectively). 6-month landmark survival according to modified TTP2 Based on the median distribution of mTTP2, OS beyond the 6-month landmark was significantly longer in participants with mTTP26m ≥ 2.1 months compared to the mTTP26m < 2.1 months group (HR 0.51, 95% CI 0.35–0.76, p = 0.001). There was a consistent trend toward improved OS with increasing duration of mTTP26m (Figure 3). Figure 3.6-month landmark overall survival according to modified time-to-progression-2 status.The line graph displays survival probabilities over time, with two curves for mTTP2 less than median and mTTP2 greater than or equal to median, showing higher survival for the latter. The x-axis represents time in months (6, 12, 18, 24, 30, 36) and the y-axis indicates survival probability from 0.00 to 1.00. A dashed line at 0.50 and vertical guides at 12 and 18 months are present. Below, a table lists 'Number at risk: n (percent)' for both groups at each time point: 6 months (52, 109), 12 months (21, 73), 18 months (9, 40), 24 months (6, 17), 30 months (1, 7), 36 months (0, 1). The forest plot compares TTP2 thresholds with columns for n less than and greater than or equal to threshold and HR. Rows include: median (2.1 months): 52, 109, 0.51; 4 months: 91, 70, 0.35; 6 months: 105, 56, 0.29. Each row features a square marker with a confidence interval line. The forest plot x-axis ranges from 0.18 to 2.5.A Kaplan-Meier plot and forest plot depicting survival rates by modified time-to-progression-2 status. 6-month landmark overall survival according to modified time-to-progression-2 status. 3-month landmark survival according to modified TTP2 As a sensitivity analysis, we conducted 3-month landmark survival analysis using mTTP23m. 203 participants were alive and included in the analysis. OS beyond the 3-month landmark was also significantly longer in participants with mTTP23m ≥ 2.1 months compared to those with mTTP23m < 2.1 months (HR 0.42, 95% CI 0.31–0.59, p < 0.001). Supplementary Figure 1 further illustrates the HRs across varying mTTP23m thresholds. 6-month landmark survival according to TTP1 status Based on the median distribution of TTP1, OS beyond the 6-month landmark was not significantly different in participants with mTTP1 ≥ 3.9 months compared to those with mTTP1 < 3.9 months group (HR 0.77, 95% CI 0.53–1.11, p = 0.17). Figure 4 shows the HRs across varying TTP1 thresholds. Figure 4.6-month landmark overall survival according to time-to-progression-1 status.A. A Kaplan–Meier style line graph titled “6 months landmark survival” comparing two strata: “TTP1 < median (3.9 months)” and “TTP1 ≥ median (3.9 months)”. The x-axis is labeled “Time in months” with ticks at 6, 9, 12, 15, 18, 21, 24, 27, 30, 33, 36. The y-axis is labeled “Survival probability” from 0.00 to 1.00 with ticks at 0.00, 0.25, 0.50, 0.75, 1.00. Both curves start near 1.00 around 6 months and decline in steps. A dashed horizontal line marks 0.50 and dashed vertical lines mark about 15 and 16 months. Around 15 months, both curves are near 0.50. From about 18 to 36 months, the “TTP1 ≥ median (3.9 months)” curve stays above the “TTP1 < median (3.9 months)” curve, ending near about 0.15 to 0.20 versus about 0.05 to 0.10. Below, a “Number at risk: n (%)” table lists values at 6, 9, 12, 15, 18, 21, 24, 27, 30, 33, 36 months. For “TTP1 < median (3.9 months)”: 75 (100), 58 (77), 43 (57), 33 (44), 24 (32), 14 (19), 8 (11), 6 (8), 3 (4), 1 (1), 1 (1). For “TTP1 ≥ median (3.9 months)”: 86 (100), 72 (84), 51 (59), 38 (44), 25 (29), 19 (22), 15 (17), 8 (9), 5 (6), 3 (3), 0 (0). B. A forest-plot style chart with columns labeled “TTP1 threshold”, “n < threshold”, “n ≥ threshold” and “HR”. Four rows are shown: “2 months” with n < threshold 31, n ≥ threshold 130, HR 0.55; “median (3.9 months)” with 75, 86, HR 0.77; “6 months” with 104, 57, HR 0.83; “8 months” with 123, 38, HR 0.93. Each row has a square marker with a horizontal confidence interval line. The x-axis beneath the forest plot shows tick labels 0.35, 0.50, 0.71, 1.0, 2.5, with a vertical reference line at 1.0.Two plots show 6-month landmark survival by time-to-progression-1 groups and hazard ratios by threshold. 6-month landmark overall survival according to time-to-progression-1 status. 3-month landmark survival according to TTP1 status As a sensitivity analysis, we assessed OS beyond the 3-month landmark. OS was not significantly longer in participants with TTP1 ≥3.9 months compared to those with TTP1 <3.9 months (HR 0.74, 95% CI 0.54–1.01, p = 0.06). Supplementary Figure 2 further illustrates the HRs across different TTP1 thresholds. Cox regression analysis on OS In bivariate analysis adjusting mTTP26m for TTP1 (Table 2), TTP2 remained significantly associated with OS (HR 0.48, 95% CI 0.32–0.71, p < 0.001), whereas the association between TTP1 and OS was of borderline significance (HR 0.68, 95% CI 0.47–0.99, p = 0.05). In multivariate analysis adjusting mTTP26m for TTP1 and ORR (Table 2), TTP2 remained strongly associated with OS (HR 0.56, 95% CI 0.37–0.83; p = 0.004), the association between TTP1 and OS remained similar (HR 0.65, 95% CI 0.45–0.95, p = 0.03), ORR was strongly associated with OS (HR 0.21, 95% CI 0.08–0.59, p = 0.003), however the number of participants who achieved a response was small (16/159,10%).Table 2.6-month landmark Cox regression analysis, time-to-progression 1, time-to-progression 2, and objective response as dichotomized variables.VariablenUnivariate HR (95% CI)p-valueBivariate HR (95% CI)p-valueMultivariate HR(95% CI)p-valueTTP1 <3.9 months751.00 1.00 1.00 TTP1 ≥3.9 months860.77 (0.53–1.16)0.170.68 (0.47–0.99)0.050.65 (0.45–0.95)0.03mTTP26m <2.1 months521.00 1.00 1.00 mTTP26m ≥2.1 months1090.51 (0.35–0.76)<0.0010.48 (0.32–0.71)<0.0010.56 (0.37–0.83)0.004OR (SD/ PD)1431.00 ––1.00 OR (PR/ CR)160.19 (0.07–0.52)0.001––0.21 (0.08–0.59)0.003Abbreviations: CI – confidence interval; CR – complete response; HR – hazard ratio; OR – objective response; PD – progressive disease; PR – partial response; SD – stable disease; TTP – time to progression. Table 2.6-month landmark Cox regression analysis, time-to-progression 1, time-to-progression 2, and objective response as dichotomized variables.VariablenUnivariate HR (95% CI)p-valueBivariate HR (95% CI)p-valueMultivariate HR(95% CI)p-valueTTP1 <3.9 months751.00 1.00 1.00 TTP1 ≥3.9 months860.77 (0.53–1.16)0.170.68 (0.47–0.99)0.050.65 (0.45–0.95)0.03mTTP26m <2.1 months521.00 1.00 1.00 mTTP26m ≥2.1 months1090.51 (0.35–0.76)<0.0010.48 (0.32–0.71)<0.0010.56 (0.37–0.83)0.004OR (SD/ PD)1431.00 ––1.00 OR (PR/ CR)160.19 (0.07–0.52)0.001––0.21 (0.08–0.59)0.003Abbreviations: CI – confidence interval; CR – complete response; HR – hazard ratio; OR – objective response; PD – progressive disease; PR – partial response; SD – stable disease; TTP – time to progression. 6-month landmark Cox regression analysis, time-to-progression 1, time-to-progression 2, and objective response as dichotomized variables. Abbreviations: CI – confidence interval; CR – complete response; HR – hazard ratio; OR – objective response; PD – progressive disease; PR – partial response; SD – stable disease; TTP – time to progression. When TTP1 and TTP2 were analyzed as continuous variables in a bivariate model (Table 3), TTP2 again remained significantly associated with OS (HR 0.77, 95% CI 0.70–0.85, p < 0.001), whereas TTP1 showed no significant association (HR 0.98; p = 0.29). In multivariate analysis assessing TTP1 and TTP2 as continuous variable and ORR as categorical variables, TTP2 remained significantly associated with OS (HR 0.81, 95% CI 0.73–0.90, p < 0.001), whereas TTP1 showed no significant association (HR 0.98; p = 0.27). ORR also demonstrated strong association with OS (HR 0.29, 95% CI 0.10–0.81, p = 0.02).Table 3.6-month landmark Cox regression analysis, time-to-progression 1 and time-to-progression 2 as continuous variable, objective response as categorical variable.VariablenUnivariate HR (95% CI)p-valueBivariate HR (95% CI)p-valueMultivariate HR (95% CI)p-valueTTP11610.99 (0.96–1.03)0.600.98 (0.95–1.02)0.290.98 (0.95–1.02)0.27mTTP26m1610.77 (0.70–0.86)<0.0010.77 (0.70–0.85)<0.0010.81 (0.73–0.90)<0.001OR (PD/ SD)1431.00 ––1.00 OR (PR/ CR)160.19 (0.07–0.52)0.001––0.29 (0.10–0.81)0.02Abbreviations: CI – confidence interval; CR – complete response; HR – hazard ratio; ORR – objective response; PD – progressive disease; PR – partial response; SD – stable disease; TTP – time to progression. Table 3.6-month landmark Cox regression analysis, time-to-progression 1 and time-to-progression 2 as continuous variable, objective response as categorical variable.VariablenUnivariate HR (95% CI)p-valueBivariate HR (95% CI)p-valueMultivariate HR (95% CI)p-valueTTP11610.99 (0.96–1.03)0.600.98 (0.95–1.02)0.290.98 (0.95–1.02)0.27mTTP26m1610.77 (0.70–0.86)<0.0010.77 (0.70–0.85)<0.0010.81 (0.73–0.90)<0.001OR (PD/ SD)1431.00 ––1.00 OR (PR/ CR)160.19 (0.07–0.52)0.001––0.29 (0.10–0.81)0.02Abbreviations: CI – confidence interval; CR – complete response; HR – hazard ratio; ORR – objective response; PD – progressive disease; PR – partial response; SD – stable disease; TTP – time to progression. 6-month landmark Cox regression analysis, time-to-progression 1 and time-to-progression 2 as continuous variable, objective response as categorical variable. Abbreviations: CI – confidence interval; CR – complete response; HR – hazard ratio; ORR – objective response; PD – progressive disease; PR – partial response; SD – stable disease; TTP – time to progression. In a sensitivity analysis at the 3-month landmark, bivariate analysis showed that both mTTP23m (HR 0.39, 95% CI 0.28–0.52, p < 0.001) and TTP1 (HR 0.63, 95% CI 0.46–0.87, p = 0.005) were significantly associated with improved OS (Supplementary Table S2a). In multivariate analysis, mTTP23m (HR 0.46, 95% CI 0.33–0.64, p < 0.001), TTP1 (HR 0.61, 95% CI 0.44–0.84, p = 0.003), and ORR (HR 0.20, 95% CI 0.07–0.54, p = 0.002) were significantly associated with OS. When TTP1 and TTP2 were analyzed as continuous variables in a bivariate model (Supplementary Table S2b), both remained significantly associated with OS; however, the effect size for TTP1 was modest (HR 0.97; p = 0.04). In multivariate analysis, mTTP23 m (HR 0.54, 95% CI 0.42–0.70; p < 0.001) and ORR (HR 0.20, 95% CI 0.07–0.56, p = 0.002) were strongly associated with OS, whereas the effect size for TTP1 and OS remained modest (HR 0.96, 95% CI 0.93–0.99; p = 0.02). Immunotherapy vs non-immunotherapy studies There was no significant difference in 6-month landmark OS between participants who received ICI vs non-ICI treatments (HR 1.06, 95% CI 0.70–1.62, p = 0.78) (Supplementary Figure 3). HR for OS also did not vary significantly across different TTPr thresholds regardless of whether the participant received ICI or not (Supplementary Figure S4a, b, respectively). 6-month landmark OS according to mTTP26m in ICI vs non-ICI trials OS beyond the 6-month landmark was significantly longer in participants with mTTP26m ≥2.1 months compared to the mTTP26m <2.1 months group (HR 0.54, 95% CI 0.34–0.85, p = 0.008) for ICI trials but borderline significant for non-ICI trials (HR 0.48, 95% CI 0.23–1.00, p = 0.05). Figure 5 illustrates the HRs across varying mTTP26m thresholds. Figure 5.6-month landmark overall survival according to modified time-to-progression-2 status in immunotherapy (5a) and non-immunotherapy (5b) trials.Image A depicts the ICI subgroup with a table and forest plot showing modified time to progression thresholds, sample sizes and hazard ratios. The x-axis ranges from 0.18 to 2.5 with specific ticks. Three rows detail: median (2.1 months) with 34 below and 79 above threshold, hazard ratio 0.54; 4 months with 62 below and 51 above, hazard ratio 0.35; 6 months with 73 below and 40 above, hazard ratio 0.28. Squares indicate hazard ratios with confidence intervals. Image B shows the non-ICI subgroup with similar forest plot parameters. Rows detail: median (2.1 months) with 18 below and 30 above threshold, hazard ratio 0.48; 4 months with 29 below and 19 above, hazard ratio 0.33; 6 months with 32 below and 16 above, hazard ratio 0.29. Each row includes a square at the hazard ratio with a confidence interval line.Forest plots show hazard ratio changes with time to progression 2, lower values at higher thresholds. 6-month landmark overall survival according to modified time-to-progression-2 status in immunotherapy (5a) and non-immunotherapy (5b) trials. Discussion In this IPD analysis of 246 participants from eight phase II MoST studies, OS was not meaningfully associated with TTPr across a broad range of thresholds. Similarly, TTP1 was not consistently associated with OS. By contrast, our landmark analyses showed that longer on-study disease control during the first 6 months, as measured by mTTP26m, was associated with improved subsequent OS. These findings indicate that proportional delay in progression captured by TTPr does not appear to be prognostic, whereas absolute disease control achieved on the current investigational therapy, is a clinically relevant determinant of OS. Prior studies using chemotherapy or targeted therapy reported a positive association between TTPr and OS [7–9,11]. In contrast, our analysis did not demonstrate TTPr to be prognostic regardless of treatment types. This discrepancy is likely attributable to differences in (i) analytical methodology, (ii) diverse cancer types and treatments with wide variations in TTP1, and (iii) the rigor in data collection for TTP1. In all prior studies, TTPr had been evaluated in relation to OS without appropriate temporal separation between the assessment of TTPr and the survival outcome. As TTPr is mathematically dominated by TTP2, any association between on-study disease control and OS may be inflated due to the shared survival time rather than treatment-related benefit, resulting in a misleading prognostic signal. In contrast, our study employed landmark survival analyses by anchoring progression status at predefined timepoints and explicitly separating the intervals used to define TTPr from subsequent OS. Landmark analysis mitigates guarantee-time bias which arises when a time-dependent variable—such as TTP2—can only be observed in patients who survive long enough to experience the defining event. In such settings, patients with longer TTPr must, by definition, survive to reach progression, artificially inducing a correlation between TTPr and OS due to shared survival time rather than true prognostic effect. Although landmark approaches necessarily exclude patients who die prior to the landmark, this restriction is intrinsic to the methodological objective of the present analysis. Specifically, because our aim was to characterize disease-response trajectories on prior therapy (TTP1) and on-study treatment (mTTP2) that inform subsequent OS, meaningful interpretation requires restriction to patients who survive long enough for disease control to be observed and quantified. Accordingly, the resulting estimates represent conditional survival effects among patients alive at the landmark, rather than unconditional treatment effects from study initiation. TTPr is sensitive to extreme values of either TTP1 or TTP2, which can be influenced by tumor biology or relative treatment effects. TTPr may appear high when TTP1 is very short due to aggressive disease biology or in the setting of an ineffective prior line of treatment, even if absolute disease control on study is modest (e.g., TTP1 1.1 months, TTP2 3.5 months, yielding TTPr 3.3). Conversely, TTPr may be low when TTP1 and TTP2 are of similar duration, despite favorable tumor biology and prolonged disease control with the investigational therapy (e.g., TTP1 16.1 months and TTP2 11.2 months, yielding TTPr 0.70). These examples illustrate that while TTPr may capture relative improvement over prior therapy, it does not necessarily reflect absolute treatment benefit or correlate consistently with survival. Additionally, our data demonstrated that TTP1 was significantly shorter in the TTPr ≥1.3 cohort, and vice versa, and TTP1 correlated poorly with TTP2. Our findings were consistent with other studies where there is no definitive correlation in the duration of disease control between successive lines of therapy [8,12,27], this observation suggests that disease trajectory is dynamic, and challenges the assumption that TTPr adequately accounts for the natural history of cancer to enable robust intra-patient comparison. In our analysis, TTP2 consistently demonstrated significant association with improved OS, irrespective of landmark or modeling methods. Similarly, ORR was also significantly associated with improved OS, however the small sample size limited the robustness of its interpretation. In contrast, the interpretation of TTP1 was more nuanced. TTP1 was not consistently associated with improved OS. In scenarios where both TTP1 and TTP2 were associated with improved OS, the effect size of TTP1 was consistently smaller than that of TTP2 across multiple analyses. Taken together, these data suggest that TTP1 is a less robust prognostic marker than TTP2, which consistently showed a significant association with OS, with larger effect sizes and greater reproducibility across landmark times and modeling approaches. TTP2 reflects more contemporaneous disease behavior and treatment effect, and is therefore likely to exert a greater influence on subsequent survival than the more temporally distant TTP1. This finding was also consistent across both ICI and non-ICI trials, and aligns with our prior work demonstrating that 6-month PFS is a valid surrogate for 12-month OS in ICI studies [25]. Identifying endpoints suitable for earlier read-out in single-arm studies remains challenging. Although our findings demonstrated that TTP2/PFS correlated more strongly with OS than TTP1, these endpoints are less informative in single-arm settings, where progression rates vary substantially across heterogeneous patient populations and tumor types. Further, our study was unable to definitively address the predictive performance of TTPr. Future research that aims to further determine the prognostic and predictive properties of TTPr should consider comparing the performance of TTPr against other measures of treatment activity (e.g., PFS). These are ideally performed using randomized trial data. Further, as the use of real-world data has gained significant renewed interest, high quality external synthetic control data, together with thoughtful statistical methodology on indirect comparison, may provide valuable insights on comparative effectiveness analyses and endpoint evaluation. Limitations This study has several limitations. First, our analysis was conducted using single-arm studies and thus our analyses are limited to evaluation of the prognostic, but not the surrogate, properties of these endpoints. Second, the sample size of fewer than 250 participants was modest. Third, subgroup analyses restricted to a single histology or treatment type were limited by small patient numbers. Histology is known to influence survival even when treated with the same biomarker-directed therapy [28,29], a larger sample size would allow greater insight into the influence of histology on survival. Fourth, TTP1 data were collected outside clinical trial settings and retrospectively obtained prior to study enrollment, with no standardization of imaging frequency or measurement methodology. The potential inaccuracies and variability introduced may hinder TTP1’s true capability to capture disease natural history. Fifth, the upper limit of TTP2 is fixed at the respective landmarks (i.e., at 3 or 6 months), this inherently restricts the range of possible TTPr values, especially for patients with short TTP1. Finally, this analysis was subject to cohort selection bias. Clinical trial participants were more likely to have favorable prognostic features compared to the real-world population, therefore limiting the generalizability of our results. Despite these limitations, our exploratory study provides important insights into the prognostic value of TTPr and related endpoints. Conclusions In conclusion, the findings from our study indicate that TTPr is not prognostic once temporal bias is addressed. Its potential predictive utility in comparative treatment settings warrants further investigation in randomized trials where relative treatment effects can be evaluated. Although the lack of prognostic association raises concern regarding its validity as a surrogate endpoint, formal surrogate endpoint validation would require analyses within randomized trial settings. Supplementary Material Supplemental Material Supplemental Material STROBEchecklist_ttpr.docx