Abstract
Background
Immune checkpoint inhibitors (ICIs) improve outcomes in microsatellite instability-high/deficient mismatch repair (MSI-H/dMMR) metastatic colorectal cancer (mCRC). Pembrolizumab is a first-line option (KEYNOTE-177), but new data from CheckMate 8HW support nivolumab–ipilimumab. In the absence of head-to-head trials, indirect comparisons using reconstructed individual patient data (IPD) provides an approach for inferring relative efficacy and safety.
Methods
Progression-free survival (PFS) curves from KEYNOTE-177 and CheckMate 8HW were reconstructed using IPDfromKM. Phase II curves assessed cross-phase consistency. Cox models estimated hazard ratios (HRs), and grade 3/4 immune-mediated adverse events (IMAEs) contextualized efficacy.
Results
Trial arms were comparable overall but varied in follow-up and treatment-line inclusion. Compared with pembrolizumab, nivolumab–ipilimumab showed longer estimated PFS [HR 0.64, 95% confidence interval (CI) 0.49-0.83]. Single-agent nivolumab and pembrolizumab demonstrated similar outcomes (HR 0.96, 95% CI 0.75-1.23), while reconstructed estimates for nivolumab–ipilimumab versus nivolumab were consistent with published CheckMate 8HW data (HR 0.66, 95% CI 0.53-0.82). Grade 3/4 IMAEs were more frequent with dual ICI versus nivolumab (odds ratio 2.2, P = 0.004) but not pembrolizumab (P = 0.25).
Conclusions
Indirect IPD analysis suggests nivolumab–ipilimumab may prolong PFS versus pembrolizumab in MSI-H/dMMR mCRC, although with increased toxicity. Survival estimates for single-agent nivolumab and pembrolizumab appeared comparable. These findings highlight the utility of IPD reconstruction for indirect comparisons where head-to-head evidence is unlikely.
Key words: colorectal cancer, immune checkpoint inhibitors, ipilimumab, microsatellite instability, nivolumab, pembrolizumab
Highlights
-
•
Dual ICIs may improve PFS versus single agents in MSI-H/dMMR mCRC.
-
•
Nivolumab and pembrolizumab demonstrated similar estimated PFS.
-
•
Dual ICIs had higher rates of grade 3/4 IMAEs versus single-agent nivolumab.
Introduction
Immune checkpoint inhibitors (ICIs) have improved outcomes in microsatellite instability-high/deficient mismatch repair (MSI-H/dMMR) metastatic colorectal cancer (mCRC). Pembrolizumab, based on results from the KEYNOTE-177 trial, became the standard of care (SOC) in this setting in many regions.1 However, recent findings from the phase III CheckMate 8HW trial now provide randomized data for nivolumab–ipilimumab or nivolumab as alternative options.2,3
Although CheckMate 8HW demonstrated the superiority of dual ICI over nivolumab alone, there remains a lack of evidence comparing this new regimen against pembrolizumab. In the absence of prospective comparisons, we attempted to reconstruct individual patient data (IPD) from public Kaplan–Meier (KM) curves of both trials to provide an indirect comparison of nivolumab–ipilimumab with pembrolizumab. Furthermore, we examined the comparability of single-agent nivolumab and pembrolizumab as a contextual analysis of programmed cell death protein 1 (PD-1) inhibitor outcomes in this setting. As a sensitivity and validation step, we evaluated the consistency of outcomes across phase II and III trials and validated the reconstructed IPD against published values to assess reliability.
Methods
Phase III trials evaluating ICIs in MSI-H/dMMR mCRC were identified through a review of current clinical guidelines, major oncology conference proceedings, and published studies in this setting. The analysis focused on selected trials suitable for IPD reconstruction, and eligible studies were required to report KM survival curves with accompanying numbers at risk that were sufficient for reconstruction. KEYNOTE-177 and CheckMate 8HW were international, open-label phase III trials evaluating ICIs in MSI-H/dMMR mCRC. In KEYNOTE-177, patients with untreated MSI-H/dMMR mCRC were randomly assigned 1 : 1 to receive pembrolizumab or chemotherapy, with dual primary endpoints of progression-free survival (PFS) and overall survival. Patients in CheckMate 8HW were randomly assigned (2 : 2 : 1) to receive nivolumab–ipilimumab, single-agent nivolumab, or chemotherapy, with MSI status centrally confirmed. Dual primary endpoints included PFS comparing nivolumab–ipilimumab with chemotherapy (first-line) and nivolumab–ipilimumab with nivolumab (all-lines). Both studies permitted crossover from chemotherapy to ICI at progression. Because KEYNOTE-177 determined MSI status using local testing rather than central review, we restricted analyses to the locally tested MSI-H/dMMR subgroup in CheckMate 8HW to maintain methodological consistency. Additionally, although KEYNOTE-177 was restricted to treatment-naïve patients, CheckMate 8HW included both first-line and previously treated patients, representing a potential source of heterogeneity for the comparisons.
PFS curves from KEYNOTE-177, CheckMate 8HW, and relevant phase II trials (KEYNOTE-164, CheckMate 142) were digitized using WebPlotDigitizer.1, 2, 3, 4, 5, 6, 7 Time-to-event data were reconstructed using the IPDfromKM package in R.8 This method uses KM coordinates, numbers at risk, and total events to iteratively estimate individual-level data while preserving curve structure. Reconstructed KM curves were generated independently from digitized survival data using IPDfromKM and are presented to visually reflect the data. Formal risk-of-bias assessment was not carried out because the analysis was restricted to prospectively selected clinical trials and was intended as a comparative reconstruction of published survival data rather than a systematic evidence synthesis. Cox proportional hazards (PH) models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for all pairwise comparisons, excluding comparisons violating the PH assumption based on Schoenfeld residuals. As the available evidence formed an open treatment network without closed loops connecting all interventions, formal assessment of network consistency between direct and indirect evidence was not feasible and was therefore not carried out. Median PFS was calculated from the reconstructed survival times, and landmark survival probabilities were estimated at months 3, 6, 12, 18, 24, and 36. Validation was carried out by comparing reconstructed HRs and medians to published trial values. The frequency of grade 3/4 immune-mediated adverse events (IMAEs), defined according to the Common Terminology Criteria for Adverse Events (CTCAE) in the original trial publications, was determined by summing reported events by treatment arm. Pairwise two-sided Fisher’s exact tests were then conducted to compare these frequencies between treatment arms, with Benjamini–Hochberg adjustment applied for multiple comparisons. Analyses were conducted in R (version 4.4.2) using publicly available data; ethics approval was not required.
Results
Baseline characteristics of included trial arms are summarized in Table 1. Patient populations were broadly similar across trials in terms of age, Eastern Cooperative Oncology Group performance status, and tumor sidedness. KEYNOTE-177 had a 26-month longer median follow-up compared with CheckMate 8HW. Frequencies of grade 3/4 IMAEs and associated pairwise comparisons are summarized in Table 2. Nivolumab had the lowest frequency of reported grade 3/4 IMAEs (7.6%) followed by pembrolizumab (10.5%) and nivolumab–ipilimumab (15.6%). Nivolumab–ipilimumab had significantly higher odds of grade 3/4 IMAEs compared with single-agent nivolumab [odds ratio (OR) 2.22, 95% CI 1.34-3.76, adjusted P = 0.004], whereas no significant differences were observed in other pairwise comparisons.
Table 1.
Trial and treatment arm characteristics for included phase III studies
| Characteristic | KEYNOTE-177 |
CheckMate 8HW |
|||
|---|---|---|---|---|---|
| Pembrolizumab n = 153 | Chemotherapy n = 154 | Nivolumab–ipilimumab n = 354 | Nivolumab n = 353 | Chemotherapy n = 101 | |
| Median follow-up | 73 months | 47 months | |||
| Immunotherapy dosing/chemotherapy regimens allowed | 200 mg every 3 weeks | FOLFOX, FOLFIRI with or without bevacizumab or cetuximab |
Nivolumab 240 mg, ipilimumab 1 mg/kg, every 3 weeks for four doses; then nivolumab 480 mg every 4 weeks | 240 mg every 2 weeks for six doses, then 480 mg every 4 weeks | FOLFOX, FOLFIRI with or without bevacizumab or cetuximab |
| Age, median (years) | 63 (24-93) | 63 (26-90) | 62 (52-70) | 63 (51-70) | 65 (26-87) |
| Sex | |||||
| Male | 71 (46%) | 82 (53%) | 162 (46%) | 190 (54%) | 45 (45%) |
| Female | 82 (54%) | 72 (47%) | 192 (54%) | 163 (46%) | 56 (55%) |
| Geographic region | |||||
| North America/Europe | 109 (71%) | 113 (73%) | 251 (71%) | 246 (70%) | 71 (70%) |
| Asia | 22 (14%) | 26 (17%) | 26 (7%) | 33 (9%) | 11 (11%) |
| Rest of world | 22 (14%) | 15 (10%) | 77 (22%) | 74 (21%) | 19 (19%) |
| ECOG performance status | |||||
| 0 | 75 (49%) | 83 (54%) | 192 (54%) | 183 (52%) | 52 (51%) |
| 1 | 78 (51%) | 71 (46%) | 162 (46%) | 170 (48%) | 49 (49%) |
| First-line metastatic treatmenta | 153 (100%) | 154 (100%) | 202 (57%) | 201 (57%) | 101 (100%) |
| Tumor sidedness | |||||
| Right | 102 (67%) | 107 (70%) | 241 (68%) | 240 (68%) | 68 (67%) |
| Left | 46 (30%) | 42 (27%) | 113 (32%) | 113 (32%) | 33 (33%) |
| Other/unknown | 5 (3%) | 5 (3%) | — | — | — |
| BRAF/KRAS/NRAS status | |||||
| Wild type | 43 (28%) | 38 (25%) | 83 (23%) | 103 (29%) | 23 (23%) |
| KRAS or NRAS mutant | 33 (22%) | 39 (25%) | 83 (23%) | 89 (25%) | 21 (21%) |
| BRAF mutant | 35 (23%) | 44 (29%) | 106 (30%) | 85 (24%) | 24 (24%) |
| BRAF + KRAS or NRAS | 0 (0%) | 2 (1%) | 9 (3%) | 2 (1%) | 2 (2%) |
| Unknown | 42 (28%) | 31 (20%) | 73 (21%) | 74 (21%) | 31 (31%) |
ECOG, Eastern Cooperative Oncology Group; FOLFIRI, fluorouracil, leucovorin, irinotecan.
CheckMate 8HW ‘first-line metastatic’ defined by 0 prior lines of metastatic therapy via Interactive Response Technology criteria.
Table 2.
Summary of grade 3/4 IMAEs in ICI regimens
| Treatment arm | Patients (n) | Grade 3/4 IMAE, n (%) |
|---|---|---|
| Nivolumab–ipilimumab | 354 | 55 (15.6%) |
| Nivolumab | 353 | 27 (7.6%) |
| Pembrolizumab | 153 | 16 (10.5%) |
| Pairwise comparisons | Reference arm | OR [95% CI], P-adjusted |
|---|---|---|
| Nivolumab–ipilimumab | Nivolumab | 2.22 [1.34-3.76], P = 0.004 |
| Pembrolizumab | 1.58 [0.86-3.07], P = 0.25 | |
| Nivolumab | Pembrolizumab | 0.71 [0.36-1.47], P = 0.30 |
CI, confidence interval; ICI, immune checkpoint inhibitor; IMAEs, immune-mediated adverse events; OR, odds ratio.
A combined reconstructed PFS curve with landmark probabilities and pairwise HRs is shown in Figure 1. In indirect comparison, nivolumab–ipilimumab demonstrated longer estimated PFS compared with pembrolizumab (HR 0.64, 95% CI 0.49-0.83, P < 0.001), whereas single-agent nivolumab was comparable to pembrolizumab (HR 0.96, 95% CI 0.75-1.23, P = 0.74). Although both chemotherapy arms demonstrated high early PFS at 3 months, single-agent ICIs exhibited more pronounced early drop-offs compared with the chemotherapy controls. Notably, nivolumab–ipilimumab maintained a more stable PFS profile with a 3-month PFS of 78.0%, which was comparable to the CheckMate 8HW chemotherapy cohort (82.7%). An exploratory analysis of restricted mean survival time at 6 months showed no significant difference between nivolumab–ipilimumab and chemotherapy (difference 0.174 months; 95% CI −0.224 to 0.573, P = 0.39), further suggesting the absence of an early disadvantage with the dual ICI regimen. Comparisons using first-line and centrally confirmed MSI populations for nivolumab–ipilimumab yielded consistent overall findings (Supplementary Figure S1, available at https://doi.org/10.1016/j.esmogo.2026.100356).
Figure 1.
Indirect comparison of PFS between ICI regimens and chemotherapy in MSI-H/dMMR mCRC. Kaplan–Meier PFS curves reconstructed from KEYNOTE-177 and CM8HW for nivolumab–ipilimumab, nivolumab, pembrolizumab, and respective chemotherapy arms. Associated tables show (left) landmark survival probabilities at prespecified time points (3-36 months) and (right) pairwise HRs with 95% CIs and P values. HRs are based on reconstructed IPD from published Kaplan–Meier curves. The locally tested MSI-H/dMMR populations from CM8HW were included to maintain consistency in testing approaches with KEYNOTE-177.
CIs, confidence intervals; CM8HW, CheckMate 8HW; HRs, hazard ratios; ICI, immune checkpoint inhibitor; IPD, individual patient data; mCRC, metastatic colorectal cancer; MSI-H/dMMR, microsatellite instability-high/deficient mismatch repair; PFS, Progression-free survival.
IPD validation results are summarized in Supplementary Table S1, available at https://doi.org/10.1016/j.esmogo.2026.100356. HRs and median PFS values derived from reconstructed IPD closely matched reported values in the original trials. For pembrolizumab versus chemotherapy, both reported and reconstructed HRs were identical at 0.60 (95% CI 0.45-0.79). Median PFS values were also consistent within 0.1 to 0.2 months across all arms.
Phase II and III trial comparisons are shown in Figure 2. Survival estimates for nivolumab–ipilimumab were consistent across phases, with survival probabilities within 6% at all landmark points. Greater variability was observed with single-agent pembrolizumab and nivolumab, potentially reflecting differences in dosing, eligibility criteria, or patient selection.
Figure 2.
Cross-phase consistency of dual immune checkpoint inhibition and single-agent immunotherapy in MSI-H/dMMR mCRC. (A) PFS curves for nivolumab–ipilimumab from CM8HW, CM142 first-line cohort, and CM142 previously treated cohort. (B) Pooled PFS curve from both CM142 cohorts compared with CM8HW nivolumab–ipilimumab. (C) PFS curves for nivolumab from CM8HW and CM142. (D) PFS curves for pembrolizumab from KEYNOTE-177 (KN177) and KEYNOTE-164 (KN164) (≥1 prior line and ≥2 prior lines).
CM142, CheckMate 142; CM8HW, CheckMate 8HW; Ipi, ipilimumab; mCRC, metastatic colorectal cancer; MSI-H/dMMR, microsatellite instability-high/deficient mismatch repair; Nivo, nivolumab; Pembro, pembrolizumab; PFS, Progression-free survival.
Discussion
In this analysis, we carried out an indirect comparison of ICI regimens in MSI-H/dMMR mCRC using IPD reconstructed from phase II and III trials. Clinically, pembrolizumab remains an established first-line SOC in MSI-H/dMMR mCRC, and the absence of prospective comparative data limits interpretation of emerging alternatives. Our findings provide context for the potential use of nivolumab or nivolumab–ipilimumab as alternative options, particularly because the comparison was made in similarly defined populations based on local MSI testing.
In our indirect comparison, nivolumab–ipilimumab was associated with longer estimated PFS than pembrolizumab, with consistent benefit across landmark time points. Furthermore, nivolumab–ipilimumab did not demonstrate the early drop-off in survival commonly seen with single-agent ICIs, suggesting improved control of early disease progression and a more stable chemotherapy-sparing strategy. This efficacy advantage was observed despite a more heavily pretreated population in CheckMate 8HW, where 42% of patients had received prior systemic therapy, compared with the exclusively first-line setting of KEYNOTE-177. This observation is consistent with the suggested potential benefit of PD-1/cytotoxic T-lymphocyte associated protein 4 (CTLA-4) blockade in MSI-H/dMMR tumors and the use of dual ICI, as its benefit persisted even in a less favorable treatment context.
The relative efficacy of reconstructed dual ICI compared with single-agent nivolumab was consistent with the primary findings of CheckMate 8HW and therefore interpreted as confirmatory.3 However, dual ICI was also associated with significantly higher odds of grade 3/4 IMAEs compared with single-agent nivolumab (OR 2.2, P = 0.004), although no statistically significant difference in toxicity was observed versus pembrolizumab (P = 0.25). These safety considerations reinforce the importance of careful patient selection when weighing potential benefits and risks. Single-agent nivolumab showed PFS comparable to pembrolizumab, serving as a contextual reference supporting consistency of PD-1 inhibitor outcomes across trials. However, differences in trial inclusion criteria, particularly line of therapy and prior treatments, should be considered when interpreting these findings.
Cross-trial differences in chemotherapy control outcomes further illustrate the impact of patient selection. The chemotherapy arm in KEYNOTE-177 carried out better than in CheckMate 8HW (HR 0.75, 95% CI 0.54-1.03), despite similar eligibility criteria, suggesting that differences in clinical and molecular characteristics of recruited patients influence results.
Consistency of dual ICI outcomes across trial phases supports reliability, whereas single-agent outcomes were more variable. However, differences in trial design (including patient population and immunotherapy dosing) likely contributed to this discrepancy. CheckMate 142 used a lower ipilimumab dose as part of its reduced-dose strategy, which may explain some of the variation relative to CheckMate 8HW. Although differences in trial design and dosing are relevant, this pattern also raises the hypothesis that evolving immune regulatory pathways may influence treatment responsiveness over the disease course. As MSI-H/dMMR tumors evolve and progress through prior chemotherapy, the tumor microenvironment may develop redundant mechanisms of immunosuppression that limit the efficacy of PD-1 blockade alone. From a biological standpoint, CTLA-4 inhibition facilitates T-cell priming and the expansion of new T-cell clones in the lymph nodes, whereas PD-1 blockade primarily restores function in exhausted effector T cells within the tumor.9,10 By targeting both the priming and effector phases of the immune response, dual ICI therapy may provide a more robust clinical benefit that remains consistent regardless of a patient’s prior treatment history or the state of local T-cell exhaustion. These interpretations remain speculative and are intended to provide biological context rather than mechanistic inference, as the present analysis cannot directly evaluate changes in the tumor microenvironment over time.
Reconstructed survival curves yielded HRs and medians highly consistent with reported trial results, with HR differences ≤0.02 and median PFS within 0.2 months. This internal validation supports the feasibility of our reconstruction approach and its value in comparative effectiveness research when individual-level trial data are unavailable.
Limitations include potential inaccuracies in KM digitization and assumptions in IPD reconstruction. This approach cannot account for unmeasured confounders or differences in baseline characteristics. Furthermore, indirect comparisons are inherently subject to bias due to differences in trial design, follow-up, and patient populations (cross-trial heterogeneity). These findings should therefore be considered exploratory rather than definitive evidence of comparative efficacy. Nonetheless, internal validation and consistency across sensitivity analyses support the robustness of our findings.
In summary, our exploratory analysis of reconstructed IPD from phase III trials indirectly suggests that nivolumab–ipilimumab may offer improved PFS over pembrolizumab in MSI-H/dMMR mCRC, but with greater toxicity. Single-agent nivolumab and pembrolizumab appear similar in efficacy. IPD reconstruction provides a practical approach to estimate relative treatment effects in settings where direct comparisons are unlikely to emerge, supporting its utility for evidence generation and comparative analysis.
Acknowledgments
Funding
This research was supported by the Canadian Institutes of Health Research (CIHR) Canada Graduate Scholarships-Master’s Program and the BC Cancer Foundation and BC Cancer Rising Stars Award [grant number F25-02941].
Disclosure
JML has received consulting fees from Amgen, Eisai, Ipsen, Novartis, Taiho, AstraZeneca, Sanofi, and Merck and has received research funding from Personalis and Ipsen. His research is supported by a Michael Smith Health Professional Investigator Award. SG has received consulting fees from Amgen, Eisai, Ipsen, Novartis, Taiho, AstraZeneca, Merck, and BMS Canada. JPSV received consulting fees/honoraria from Incyte, Pfizer, Astellas, and Ipsen. All other authors have declared no conflicts of interest.
Supplementary data
References
- 1.André T., Shiu K.K., Kim T.W., et al. Pembrolizumab versus chemotherapy in microsatellite instability-high or mismatch repair-deficient metastatic colorectal cancer: 5-year follow-up from the randomized phase III KEYNOTE-177 study. Ann Oncol. 2025;36(3):277–284. doi: 10.1016/j.annonc.2024.11.012. [DOI] [PubMed] [Google Scholar]
- 2.André T., Elez E., Van Cutsem E., et al. Nivolumab plus ipilimumab in microsatellite-instability-high metastatic colorectal cancer. N Engl J Med. 2024;391(21):2014–2026. doi: 10.1056/NEJMoa2402141. [DOI] [PubMed] [Google Scholar]
- 3.André T., Elez E., Lenz H.J., et al. Nivolumab plus ipilimumab versus nivolumab in microsatellite instability-high metastatic colorectal cancer (CheckMate 8HW): a randomised, open-label, phase 3 trial. Lancet. 2025;405(10476):383–395. doi: 10.1016/S0140-6736(24)02848-4. [DOI] [PubMed] [Google Scholar]
- 4.Le D.T., Kim T.W., Van Cutsem E., et al. Phase II open-label study of pembrolizumab in treatment-refractory, microsatellite instability-high/mismatch repair-deficient metastatic colorectal cancer: KEYNOTE-164. J Clin Oncol. 2020;38(1):11–19. doi: 10.1200/JCO.19.02107. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.André T., Lonardi S., Wong K.Y.M., et al. Nivolumab plus low-dose ipilimumab in previously treated patients with microsatellite instability-high/mismatch repair-deficient metastatic colorectal cancer: 4-year follow-up from CheckMate 142. Ann Oncol. 2022;33(10):1052–1060. doi: 10.1016/j.annonc.2022.06.008. [DOI] [PubMed] [Google Scholar]
- 6.Overman M.J., McDermott R., Leach J.L., et al. Nivolumab in patients with metastatic DNA mismatch repair-deficient or microsatellite instability-high colorectal cancer (CheckMate 142): an open-label, multicentre, phase 2 study. Lancet Oncol. 2017;18(9):1182–1191. doi: 10.1016/S1470-2045(17)30422-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Rohatgi A. WebPlotDigitizer (version 4.5) https://apps.automeris.io/wpd4/ Available at.
- 8.Liu N., Zhou Y., Lee J.J. IPDfromKM: reconstruct individual patient data from published Kaplan–Meier survival curves. BMC Med Res Methodol. 2021;21(1):111. doi: 10.1186/s12874-021-01308-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Wei S.C., Levine J.H., Cogdill A.P., et al. Distinct cellular mechanisms underlie anti-CTLA-4 and anti-PD-1 checkpoint blockade. Cell. 2017;170(6):1120–1133.e17. doi: 10.1016/j.cell.2017.07.024. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Postow M.A., Callahan M.K., Wolchok J.D. Immune checkpoint blockade in cancer therapy. J Clin Oncol. 2015;33(17):1974–1982. doi: 10.1200/JCO.2014.59.4358. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.