Immune Checkpoint Inhibitors in Metastatic Bladder and Other Solid Malignancies: How Long is Enough?

DURABLE RESPONSES AND PROLONGED SURVIVAL AFTER IMMUNE CHECKPOINT THERAPY

Programmed cell death protein-1 (PD-1) and programmed cell death ligand-1 (PD-L1) checkpoint blockade immunotherapy elicit antitumor responses that translate into durable tumor shrinkage and prolonged survival in multiple cancer types. Response rates vary by tumor type, line of treatment, and single or combination therapy. In 2014, pembrolizumab was the first monoclonal antibody targeting the PD-1/PD-L1 axis that was granted accelerated FDA approval for patients with advanced or metastatic melanoma, the tumor histology with the most robust long-term clinical data. In patients with treatment-naïve and pretreated advanced or metastatic melanoma, long-term follow-up (≥4.5 years) on the KEYNOTE-001 and KEYNOTE-006 trials demonstrated response rates of approximately 40% [1, 2]. The median time to response was approximately 3 months, and the median duration of response was unprecedented (not reached in KEYNOTE-001 and 53.5 months in KEYNOTE-006). In these studies, approximately 15% of patients had a complete response (CR), and this subset of patients did particularly well long-term. Over 90% had ongoing responses in KEYNOTE-001, and the 24-month progression-free survival (PFS) was 85.4% in patients who achieved CR in KEYNOTE-006. Notably, the estimated 5-year overall survival (OS) exceeds 33% in both studies on long-term follow-up, reflecting the “long tails of survival curves,” which is a hallmark of immunotherapy.

In advanced or metastatic urothelial carcinoma (UC), the long-term activity of PD-1/PD-L1 immune checkpoint inhibitors was best evaluated in KEYNOTE-045, a randomized open-label phase 3 trial comparing pembrolizumab and investigator’s choice chemotherapy (paclitaxel, docetaxel or vinflunine) in advanced or metastatic UC after progression on platinum-containing chemotherapy [3] At a median follow-up of 62.9 months, the objective response rate was higher for pembrolizumab (21.9% vs. 11.0%), and the duration of response was higher for pembrolizumab (29.7 months vs. 4.4 months). Accordingly, 5-year OS favored pembrolizumab (14.9% vs. 8.7%), with median OS not reached for the subset of patients who achieved CR or partial response (PR) (n = 59). Taken together, the early pembrolizumab trials in melanoma as well as more recent trials of immune checkpoint inhibitors targeting the PD-1/PD-L1 pathway in UC and other solid malignancies, have consistently demonstrated the potential for durable responses and prolonged survival in a subset of patients (Table 1).

DISCONTINUATION OF IMMUNE CHECKPOINT THERAPY BEFORE DISEASE PROGRESSION

There is growing clinical evidence that patients with various tumor histologies who stop immune checkpoint inhibitors for reasons other than disease progression may continue to derive long-term clinical benefits. Due to the maturity of follow-up, much of these data stem from post-hoc analyses of trials in advanced or metastatic melanoma, which evaluated outcomes of patients who stopped treatment for toxicity or after completing a defined period of treatment.

In KEYNOTE-001, patients with ipilimumab-naïve and ipilimumab-treated advanced or metastatic melanoma were treated with pembrolizumab (one of three dose regimens). Patients were allowed to discontinue pembrolizumab if they had received at least 6 months of treatment and at least 2 treatments after confirmed CR. 105 of 655 (15.8%) patients had confirmed CR as the best overall response. At a median follow-up of 43 months, 67 patients who achieved confirmed CR elected to stop treatment and proceeded to observation without further cancer-directed therapy. Among these 67 patients, the median time to overall response was 3 months, the median time to CR was 13 months, and the median time receiving pembrolizumab was 23 months. After a median off-treatment interval of 22 months, most of these patients (n = 61, 91%) maintained CR [4].

In the pivotal phase 3 KEYNOTE-006 study, which demonstrated the superiority of pembrolizumab over ipilimumab in advanced or metastatic melanoma, patients who were randomized to the pembrolizumab arms (10 mg/kg q2wk or 10 mg/kg q3wk) continued treatment for up to 2 years or until disease progression or intolerable toxicity [2]. A total of 556 patients received pembrolizumab, 103 (19%) of whom completed 2 years of pembrolizumab. Of these 103 patients, 21 (20%) achieved complete response, 69 (67%) partial response, and 13 (13%) stable disease. After a median follow-up of 34.2 months from completion of pembrolizumab, the 24-month PFS from pembrolizumab completion for the 103 patients was 78.4%. The 24-month and 36-month overall survival estimates were 95.9% and 93.8%, respectively. Responses were ongoing in 16 of 21 (76%) patients with CR, 53 of 69 (77%) patients with PR, and 7 of 13 (54%) with stable disease (SD) [2].

In a pooled post-hoc analysis of CheckMate 069 and CheckMate 067, double-blind phase 2 and phase 3 studies which demonstrated the superiority of ipilimumab plus nivolumab over ipilimumab monotherapy in treatment-naïve unresectable stage III or IV melanoma, 96 patients in the nivolumab plus ipilimumab arms discontinued treatment for treatment-related adverse events [5]. The median number of nivolumab and ipilimumab doses received in the patients who discontinued treatment was three. At a minimum follow-up of 18 months, these 96 patients had similar objective response rates (58.3% vs. 50.2%), PFS (median PFS 8.4 months vs. 10.8 months, p = 0.97), and OS (median not reached for both) compared to the overall population randomized to nivolumab plus ipilimumab [5].

Favorable long-term outcomes after discontinuation of immune checkpoint inhibitors in other advanced solid malignancies have also been reported; however, the proportion of patients with sustained durable responses after stopping treatment appears lower than that observed in advanced or metastatic melanoma trials. For example, in 5-year follow-up of KEYNOTE-024, a phase 3, randomized study of pembrolizumab versus platinum-based chemotherapy in treatment-naïve non-small cell lung cancer (NSCLC) with PD-L1 tumor proportion score of ≥50% and no sensitizing EGFR or ALK alterations, 39 of 154 (25.3%) patients randomized to pembrolizumab completed 2 years of treatment. Among these patients, the best response was CR in 4 (10.3%), PR in 28 (71.8%), SD in 6 (15.4%), and PD in 1 (2.6%). With a median follow-up was 5 years, 32 of these 39 (82.0%) patients were alive, and 18 patients (46.2%) were alive without PD or subsequent therapy. 36-month OS rate from the completion of 2 years of treatment was 81.4% [6].

Consistent with post-hoc analyses from clinical trials, real-world data evaluating the outcome of patients with advanced or metastatic melanoma who discontinued PD-1 inhibitors in the absence of progressive disease have also demonstrated that a substantial (≥70%) proportion of patients remained progression-free or were alive without the need for additional cancer-directed therapy after a median follow-up of ≥ 18 months[7, 8] Of note, patients who achieve CR appear to have the highest chance of durable response after stopping treatment, compared to PR then SD [8].

Comparable clinical data focused on UC is currently lacking; however, patients with UC are included in some large retrospective studies, including a cohort of 262 patients treated with PD-1/PD-L1 monotherapy across multiple cancer types in phase 1 clinical trials conducted in Gustave Roussy [9]. Twenty-five of the 262 (9%) patients in this cohort had UC. The maximum treatment duration was 12 months in durvalumab and atezolizumab trials, 24 months in pembrolizumab trials, and until disease progression in nivolumab trials. In this cohort, 39 (15%) of patients discontinued immunotherapy without the presence of disease progression (30 patients for prolonged response and 9 patients for toxicity). Fifteen of the 39 patients (38%) relapsed after stopping treatment, and the chance of relapse was lower when patients achieved deeper response at the time of treatment discontinuation: 12% for CR, 53% for PR, and 80% for SD. Of note, in this analysis, the median treatment duration prior to discontinuation was longer in patients who did not relapse. Treatment discontinuation within the first 12 months of PD-1/PD-L1 inhibitor treatment was associated with a higher risk of relapse than after 12 months (p = 0.002) [9].

The only study to date that evaluated the treatment duration of a PD1/PD-L1 inhibitor in a prospective randomized trial is CheckMate 153 [10]. In this study, patients with advanced or metastatic non-small cell lung cancer (NSCLC) who had received at least one prior systemic therapy and remained on treatment for 1 year were randomized to continue or stop nivolumab, regardless of response status at the time of randomization. When the subset of patients who had SD, PR, and CR at randomization was evaluated for efficacy (174 of 252 randomized) at a median follow-up of 13.5 months, PFS was found to be significantly longer for those who continued nivolumab (median PFS 24.7 months vs. 9.4 months, HR 0.56 [95% CI: 0.37–0.84]). OS also favored continuous treatment (median OS not reached vs. 28.8 months, HR 0.62 [95% CI: 0.42–0.92] in the IIT population). However, the primary endpoint was safety, and not all randomized patients were evaluated for efficacy, contributing to imbalances in baseline characteristics, including frequency of squamous histology. Therefore, further prospective studies with adequate statistical design are needed.

IMMUNE CHECKPOINT RETREATMENT AFTER TREATMENT DISCONTINUATION

Clinical data on patients with disease response or stabilization on PD-1/PD-L1 immune checkpoint inhibitors, followed by an off-treatment period and subsequent retreatment for progressive disease, are limited in sample size, heterogeneous in methodology, and descriptive in nature. In the aforementioned KEYNOTE-006 trial, 13 patients with advanced or metastatic melanoma received a second course of pembrolizumab, 6 (46%) of whom had CR, 6 (46%) had PR, and 1 (8%) had SD as the best response to the initial course of pembrolizumab. With a median follow-up of 14.3 months after retreatment, 3 (23%) patients had CR (2 of whom had surgical CR before starting the second course of pembrolizumab), 4 (31%) had PR, 3 (23%) had SD, and 1 (8%) had PD as the best response to retreatment. Response to retreatment was unevaluable for 2 patients. Of the 6 patients with CR after the first course of pembrolizumab, 3 (50%) recaptured CR upon retreatment, two of whom had undergone surgical resection with no evidence of disease before pembrolizumab reinitiation. The remaining patients who had CR after the first course of pembrolizumab attained PR (n = 1) and SD (n = 1). One patient was unevaluable for the response after retreatment [2].

In the KEYNOTE-024 study, patients with advanced or metastatic NSCLC who had completed 35 cycles of pembrolizumab or achieved confirmed CR with at least 6 months of pembrolizumab and at least 2 additional cycles of treatment after CR were eligible to receive a second course of pembrolizumab upon progression [6]. Twelve patients were retreated with pembrolizumab, and their initial responses for the first course of pembrolizumab were CR (n = 1, 8%), PR (n = 8, 67%), and SD (n = 3, 25%). Upon retreatment, 4 (33%) attained objective responses, all of which were partial. Six (50%) of the retreated patients had SD, and 1(8%) had PD as the best retreatment response. One patient was unevaluable in this cohort.

In patients with advanced or metastatic UC, post-hoc analysis of the phase 3 KEYNOTE-045 (second-line post platinum) and phase 2 KEYNOTE-052 (first-line cisplatin-ineligible), trials evaluated pembrolizumab retreatment in patients who either stopped pembrolizumab after achieving CR before 2 years or completed 2 years of treatment with best response of CR, PR or SD. Patients had investigator-confirmed radiographic PD after treatment cessation, ECOG performance status of 0 or 1, and no intervening cancer-directed therapy before pembrolizumab retreatment [11]. The median time off-treatment was 7.7 months (IQR 3.6–16.5 months) in the KEYNOTE-045 cohort and 13.0 months (IQR 9.2–16.6 months) in the KEYNOTE-052 cohort. Of 11 patients retreated with pembrolizumab in KEYNOTE-045, all achieved CR (n = 3, 27%), PR (n = 2, 18%), or SD (n = 6, 55%) with a disease control rate (DCR) of 100%. Of 10 patients retreated in KEYNOTE-052, 1 (10%) patient achieved CR, 4 (40%) patients PR, and 4 (40%) SD, with DCR of 90%. However, only a minority of patients who had CR or PR with the first course of pembrolizumab recaptured CR or PR, respectively, during retreatment (4 of 10 in KEYNOTE-045 and 2 of 10 patients in KEYNOTE-052) [11].

Taken together, clinical data to date (Table 2) suggest that retreatment with PD-1/PD-L1 immune checkpoint antibodies after achieving clinical benefit and stopping treatment for reasons other than progression has the potential to restore antitumor activity; however, the depth of response may be lower than that of initial treatment. Of note, this contrasts with retreatment after progression with prior immune checkpoint inhibitor therapy. Recent data from the CONTACT-03 trial showed that the addition of atezolizumab to cabozantinib did not improve clinical outcomes and led to increased toxicity compared to cabozantinib alone in patients with locally advanced or metastatic renal cell carcinoma after progression on an immune checkpoint inhibitor [12]. Therefore, there is currently no prospective data to support continuing or rechallenging with a PD-1/PD-L1 inhibitor after progression on these immunotherapy agents.

PHYSICAL AND FINANCIAL TOXICITY FROM IMMUNE CHECKPOINT THERAPY

Immune checkpoint therapy targeting the PD-1/PD-L1 axis is generally well tolerated; however, a subset of patients develops immune-related adverse events (irAEs) that can be serious and, in rare incidences, even fatal. The incidence of grade ≥ 3 irAEs in patients receiving anti-PD-1/PD-L1 antibodies, CTLA-4 antibodies, or a combination of the two are approximately 6%, 24%, and 55%, respectively. The mechanism and spectrum of irAEs are distinctly different from toxicities seen with traditional chemotherapy or targeted therapy. Clinical presentations are heterogeneous, and a broad range of organs can be affected. Fatal irAEs occur in 0.4–1.2% of patients [13]. Reported causes of treatment-related deaths include colitis, encephalitis, pneumonitis, myocardial infarction, and liver failure [14].

Immune checkpoint antibodies have long half-lives, and the immune system may remain activated for a long period of time after these antibodies are cleared. While there are patterns in the kinetics of the main classes of irAEs, there is a wide dispersion of onset times of irAEs. Patients who have tolerated immune checkpoint therapy well can develop new irAEs many months or even years after starting treatment [15]. Chronic irAEs affect up to 40% of patients and are most commonly endocrine or rheumatological in nature but can affect other organs [13]. A real-world dataset of 437 melanoma and lung cancer patients that evaluated the incidence of late-onset and long-lasting irAEs showed that the cumulative probability of irAE onset from treatment initiation at 6, 12, and 24 months were 42.8%, 51.0%, and 57.3%, respectively. The rate of ongoing toxicity from the time of first toxicity onset at 6, 12, and 24 months were 42.8%, 38.4%, and 35.7%, respectively [16].

Longer duration of immune checkpoint therapy may be associated with a higher incidence of irAEs. Consistent with this, the CheckMate 153 study that randomized patients with advanced or metastatic NSCLC to continuous versus 1-year fixed duration of nivolumab reported a higher incidence of treatment-related AEs (TRAEs) in the treatment continuation arm (48.0% vs. 26.4%). A higher proportion of patients also experienced grade 3-4 TRAEs in the treatment continuation arm (9.4% vs. 3.2%), and more patients randomized to continue treatment experienced TRAEs that led to treatment discontinuation (9.4% vs. 1.6%). One treatment-related death was reported in the continuous treatment arm, and no grade 5 AE was reported in the 1-year fixed duration arm [10]. Therefore, if treatment can be discontinued while maintaining durable clinical efficacy, patients may be able to avoid serious and late irAEs with subsequent detriment to their overall health.

In addition to side effects related to treatment, immunotherapy can have a substantial financial burden on individual patients and the healthcare system. Countries in North America and Europe that utilize national-level health coverage systems are under pressure to provide access of these expensive agents to patients, which translates into substantial societal costs. In the United States, immune checkpoint inhibitors cost approximately $150,000 annually for drugs alone [17]. This imposes a major cost on the healthcare system; ultimately, these costs are translated into higher healthcare premiums across the entire population. For individual patients, the magnitude of financial impact depends on how well patients are insured. Some patients have minimal co-insurance and co-pay requirements; however, many are increasingly members of plans where significant co-payments result in financial hardship. In addition, patients may experience objective financial burden and subjective financial distress because of cancer treatment, termed “financial toxicity” [18, 19]. Studies have also shown that financial burden is associated with decreased patient-perceived quality of life (QOL) assessed by multiple validated instruments [18, 20]. Therefore, if a shorter course of immune checkpoint inhibitors can be used to derive comparable long-term clinical outcomes, this also may have significant personal health and health economics impact on patients.

CLINICAL IMPLICATIONS AND FUTURE DIRECTIONS

Immune checkpoint inhibitors targeting PD-1/PD-L1 are now an established core pillar of cancer therapy for a wide spectrum of advanced or metastatic solid malignancies. Treatment is generally continued until disease progression, development of an unacceptable level of side effects, or in some instances, after completion of 2 years of therapy. The length of treatment is based on how these therapies were administered in prospective registrational clinical trials. Post-hoc analyses after long-term follow-up in some of these trials, as well as real-world data, have demonstrated potential for durable response and long-term survival after immune checkpoint therapy is discontinued. Potential mechanisms underlying this durable benefit include immune-mediated eradication of tumor cells, the development of a balance between tumor cells and immune cells leading to chronic deadlock, and the generation of memory T-cells inhibiting tumor recurrence [21, 22].

Prospective clinical trials with thoughtful patient selection and adequate statistical design are needed to answer the clinically important question of optimal duration of PD-1/PD-L1 immune checkpoint therapy in advanced or metastatic solid cancers. The answer to this question is unlikely to be one-size-fits-all and will likely depend on (1) tumor histology, (2) genetics (e.g., mismatch repair and circulating tumor DNA status), (3) treatment regimen (monotherapy vs. combination with anti-CTLA-4 antibody vs. combination with chemotherapy), and (4) disease context (CR/PR/SD and duration of response). These factors should be proactively considered during the design of clinical trials and incorporated into eligibility criteria and stratification factors. Studies should also prospectively capture the incidence of irAEs as well as health-related QOL and economic impact. A likely challenge that these studies that de-intensify treatment will face is patient enrollment, particularly if randomized. A phase 3, randomized non-inferiority trial which compared the continuous treatment to stopping PD-1/PD-L1 immune checkpoint therapy at approximately 1 year in advanced or metastatic UC was stopped for low accrual (Alliance A031901). Table 3 summarizes ongoing prospective trials evaluating the potential to stop anti-PD-1/PD-L1 immune checkpoint inhibitors, predominantly in advanced and metastatic melanoma. Results of these studies should pave the way for and guide future studies in other tumor histologies.

In addition to the optimal duration of treatment, the dose and frequency of PD-1/PD-L1 immune checkpoint inhibitors also need to be optimized. In phase 1 clinical trials, immune checkpoint inhibitors have consistently demonstrated full target engagement and comparable efficacy at lower doses than the approved dosages of these agents. Pembrolizumab showed full target engagement at dosages of ≥ 1 mg/kg every 3 weeks, and response rates were comparable at dosages between 2 mg/kg and 10 mg/kg [23, 24]. Nivolumab showed comparable response rates, survival, and target engagement at dosages between 0.1 mg/kg to 10 mg/kg every 2 weeks. Further, target occupancy lasted for at least 2 months at dosage of 0.3 mg/kg, which is 10 times lower than the recommended phase 2 dose of nivolumab 3 mg/kg [25]. Prospective randomized trials are ongoing evaluating longer cycle length or lower dosages of PD-1/PD-L1 inhibitors, including REFINE-Lung (NCT05085028) and NVALT-30 Dedication (EudraCT 2020-000493-15) in NSCLC. Therefore, the optimization of dosing (i.e. reduced unit dose), schedule (i.e. longer cycle length), and treatment duration (i.e. shorter total duration) have the potential to reduce costs and toxicity related to treatment exposure. These studies should be supported with enthusiasm by the medical community.

The development and validation of biomarkers that detect minimal residual disease (MRD) may be a powerful tool to help select the best patients to stop immune checkpoint treatment early, specifically those who have non-detectable MRD in addition to radiographic CR. Such technologies could include circulating tumor DNA (ctDNA) and cell-free methylated DNA patterns [26, 27] and may also be utilized for longitudinal disease monitoring to detect early disease progression. For example, a phase 2 study evaluating patients with advanced or metastatic solid tumors treated with pembrolizumab showed that baseline ctDNA level, as well as changes in ctDNA level with treatment, correlates with PFS and OS after immunotherapy [26]. The ongoing Alliance A032103 study will evaluate ctDNA prospectively in patients with muscle-invasive bladder cancer who have adverse pathologic features after radical cystectomy. Patients with detectable ctDNA are randomized to standard-of-care adjuvant nivolumab versus nivolumab plus relatlimab (treatment intensification). Conversely, patients with undetectable ctDNA are randomized to standard-of-care adjuvant nivolumab versus surveillance with the potential to receive nivolumab if ctDNA becomes detectable (treatment de-intensification). Similar biomarker-based, patient-adapted study designs could be utilized in the advanced and metastatic setting to inform optimal treatment duration.

Until more clinical data emerges, patients with advanced or metastatic malignancies treated with PD-1/PD-L1 immunotherapy should continue treatment per standard practice. In the absence of unacceptable toxicity, treatment should be continued for 2 years or until disease progression. Patients who are interested in stopping treatment earlier should be encouraged to participate in clinical trials evaluating treatment de-intensification, whenever possible. Patients should also be counseled that based on clinical data to date, response rate after retreatment may be lower than that with the first course of treatment. If an MRD assay is commercially available for the patient’s tumor type, this could be utilized to guide personalized management decision.