In our real-world population-based study, we observed improved survival associated with second-line ipilimumab compared with second-line chemotherapy or targeted therapy (median OS 6.9 vs 4.9 months; Unadjusted HR = 0.65; Adjusted HR = 0.62). The estimated treatment effect was robust under different methods to adjust for the effect of subsequent treatment (HR range = 0.63 – 0.67). Moreover, despite more older patients receiving second-line ipilimumab (mean age: 61.7 years) compared to chemotherapy or targeted treatment (mean age: 55.2 years) in the real-world, the observed relative treatment effect was still consistent with those reported in the pivotal trial (HR = 0.66; 95% CI: 0.51 – 0.87)2. Additionally, we observed approximately half of the second-line ipilimumab patients completed the planned four doses. This was slightly lower than the 64.2% of patients who received all doses in the trial2.
In this study, the hazard ratio for overall survival (Unadjusted HR = 0.65; Adjusted HR = 0.62) was similar to the pivotal trial (HR = 0.66). The survival curve also plateaued around two to three years, similar to the pivotal trial. In contrast, the relative estimate of survival benefit, median OS, was shorter in the real-world compared to the pivotal trial for both the ipilimumab patients and the control patients. In particular, the real-world median OS was 2.9 months shorter (Real-world: 7.2 months; RCT: 10.1 months) in the ipilimumab group and 1.5 months shorter in the control group (Real-world: 4.9 months; RCT: 6.4 months), compared to clinical trial outcomes. As such, the magnitude of benefit for OS observed in the trial was approximately 3.7 months whereas the real-world effectiveness was 2.3 months. In our real-world setting, less ipilimumab patients were alive at 2-years (Real-world: 21.1%; RCT: 25%) and 3-years (Real-world: 14.3%; RCT: 25%)4. Similarly, less control patients were alive at 2-years (Real-world: 7.1%; RCT: 17%) and 3-years (Real-world: 4.7%; RCT: 10%)4. This difference may be attributable older patients receiving second-line ipilimumab in the real-world as compared to the trial (mean age: 61.7 vs 56.8 years). Moreover, 10% of treated and 14% of controls in the trial had central nervous system (CNS) metastases at baseline and received previous treatments for it. In our real-world study, a little over 20% of patients in each group had radiation to the brain prior to index systemic treatment, which may have affected absolute survival estimates.
In the population-based study conducted by Polkowska et al, which also compared second-line ipilimumab to second-line chemotherapy, the OS and HR of second-line ipilimumab was similar to our study and the trial. While the median OS were similar between the controls in both studies, patients receiving second-line ipilimumab in our study (median OS: 7.2; 95% CI: 5.3 – 8.7) had a slightly higher median OS than those patients in Poland (median OS: 5.9; 95% CI: 5.6 – 8.4), though the confidence intervals were overlapping between the two studies. Additionally, difference in median OS may also be affected by the difference in access to treatments. The authors mentioned that immunotherapies are only available for second and subsequent lines of treatment. While first-line immunotherapies were not publicly funded in Ontario at the time of our cohorts, around 5% of patients in each group had received non-ipilimumab immunotherapies from clinical trials or private payers. Furthermore, the third-line treatment availability may also differ between the two studies. In our study, controls were more likely to receive third-line ipilimumab while treated were more likely to receive third-line nivolumab or pembrolizumab. Despite the differences, both observed median OS for second-line ipilimumab was within the range observed in some of the single-arm studies, which ranged between 6.4 to 8.8 months10–14.
Our study has a number of notable strengths. We had an extensive collection of linked data, providing detailed information on patient characteristics and treatments for an entire population. Thus, we were able to adjust for many more potential confounders using propensity score methods in contrast to other population-based studies. These variables include rurality, socioeconomic status, previous resection, comorbidity status, and time from diagnosis to initiating second-line treatment. Additionally, we also used different sensitivity analyses to explore the effect of subsequent third-line treatment given that real-world data are non-randomized and that subsequent therapies can be a potential source of confounding. The survival benefit of second-line ipilimumab persisted after these adjustments.
Our study has several limitations. First, inherent to observational studies, our estimates may have been affected by residual confounding from unbalanced or unmeasured variables. Our analysis did not adjust for performance status, though we were able to control for differences in comorbidity, which could affect functional status. Additionally, given this is a pre-funding vs post-funding comparison, it is unlikely that the distribution of performance status of the population would change significantly. Data on lactate dehydrogenase (LDH) was also not available, though notably in the pivotal trial by Hodi et al, there was no significant difference in ipilimumab treatment effect based on LDH2, and it is unlikely that there would be imbalances given the pre/post design with nearly all patients receiving ipilimumab after funding. Second, in contrast to the randomized trial, the comparator in our study consists of historical controls, since once second-line ipilimumab became available, a very small number of patients received chemotherapy or targeted therapy. While historical controls avoid the small sample size of controls or confounding by indication after ipilimumab funding, historical comparators might be confounded by secular trends such as changes in clinical practice (e.g. radiation or resection practices for metastases, availability of other systemic therapies (see Appendix for public funding timeline)) that might bias in favor of ipilimumab. Lastly, after second-line ipilimumab was funded, some patients may have truncated their first-line treatment to access ipilimumab for second-line treatment as soon as possible, as there would have had been no option for first-line immunotherapy. This funding change may have resulted in differences in the extent of disease at the start of second line treatment between treatment groups. Drysdale et al. shown that 40% of patients who received first-line ipilimumab for MM had received prior chemotherapies for less than 60 days during a similar study period in Ontario15. Time between diagnosis to second-line treatment was adjusted for via the propensity score method to address potential bias, though residual imbalance remained in our study.
In addition to validating the efficacy observed in clinical trials, our study also informs policy decisions in Ontario. While ipilimumab was the first immunotherapy to enter the treatment landscape, other immunotherapies, PD1 inhibitors pembrolizumab and nivolumab, have become available in both first-line and second-line setting in Ontario, along with nivolumab in combination with ipilumumab16–19. Despite the availabilities of these therapies, the effectiveness of second-line ipilimumab is still of importance for patients who may not be able to tolerate aggressive immunotherapy or progress on combination targeted treatments. Potential future area of investigation can examine the effectiveness of second-line ipilimumab after first-line PD-1 inhibitors and the comparative effectiveness between ipilimumab monotherapy and combination therapies.