Nivolumab is a fully human monoclonal antibody targeting the programmed death receptor-1 (PD-1) and blocking the binding of PD-1, expressed on T cells, with its ligands PD-L1 and PD-L2, present on antigen-presenting cells and cancer cells [1]. Nivolumab therapy induces the disruption of PD-1/PD-L1 signaling, thus restoring the ability of T cells to selectively recognize and kill cancer cells [2].

Regarding metastatic renal cell carcinoma (mRCC), in a randomized phase III trial (CheckMate 025), nivolumab compared with everolimus proved able to confer a 5.4-month improvement of median OS, with a more favorable safety profile [3, 4]. The magnitude of the clinical benefit was so relevant that on November 2015 the Food and Drug Administration (FDA) and on April 2016 the European Medical Association (EMA) approved nivolumab for mRCC patients who had received a prior line of treatment with anti-angiogenic agents (Administration USFD, FDA Expands Use of Immunotherapeutic to Kidney Cancer) [5, 6]. Afterwards, the Italian Renal Cell Cancer Early Access Program (EAP) started in July 2015 based on such clinical data, during the evaluation of nivolumab by EMA and the negotiations with the Italian Ministry of Health [7].

Immune-checkpoint inhibitors (ICIs) induce a peculiar spectrum of toxicities, different from the one determined by conventional chemotherapy, caused by an enhanced activity of the immune system and by systemic inflammation: the so-called “immune-related adverse events” (irAEs) [8]. These adverse events (AEs) display a wide variety of manifestations concerning grade of toxicity, generally mild although severe cases may occur, as well as number and type of organs involved. IrAEs could be dermatological (rash, vitiligo and pruritus), gastrointestinal (diarrhea, colitis, hepatitis, increase of amylase and lipase), endocrine (thyroiditis and hypophysitis), pulmonary (pneumonitis), renal (nephritis) and systemic (fever and fatigue) [9].

In retrospective studies, non-conclusive evidence has been collected about a possible association of the occurrence of irAEs with durable responses and survival benefit from ICIs, both for anti-CTLA4 and anti-PD1, first in advanced melanoma [10–15] and then in other tumor settings, particularly non-small cell lung cancer (NSCLC) [16–18]. Recently, a prospective clinical trial demonstrated that, in NSCLC patients treated with nivolumab, the early occurrence of irAEs (specifically rash and pyrexia, but not diarrhea) correlated with an enhanced tumor response (37% vs 17%) and a longer progression-free survival (PFS) (6.4 vs 1.5 months). The authors reported a stronger predictive value for the detection of toxicities at two weeks from the treatment start, compared to the six-week assessment [19].

Given the controversy over the potential association of irAEs with favorable clinical outcomes and the discordant conclusions obtained in different cancer populations, here we report the results of a secondary analysis of the Italian EAP for nivolumab in mRCC patients, with the specific aim to assess the correlation of AEs, specifically irAEs, with patients’ outcome in this large real-life mRCC population.

A total of 389 patients were enrolled between July 2015 and April 2016 and treated with at least 1 dose of nivolumab, thus they represented the study population for this analysis. Baseline patients’ characteristics are reported in Table 1. The median follow-up was 11.9 months (range 1–24.7 months) and patients were administered a median of 13 doses of nivolumab (range 1–49 doses).Tab1

Baseline patients characteristics

Specific details regarding drAEs reported in the Italian EAP and in the Checkmate 025 trial are shown in Table 2. Any grade and grade 3–4 drAEs occurred in 124 (32%) and 27 (7%) of patients in Italian EAP, respectively. No treatment-related deaths were recorded. Median time to appearance of drAEs was 1.4 months (range 0–11.4 months) and they were generally manageable with treatment as per protocol-specific guidelines.Tab2

Rates of drug-related adverse events reported in the CheckMate 025 trial and in the Italian Early Access Program of nivolumab in mRCC

DrAEs represented the reason for treatment discontinuation in 22 cases (7.9%), of which 10 (45%) were considered irAEs, including: grade 4 hyperglicemia (n = 1), grade 3 diarrhea (n = 1), grade 3 pneumonitis (n = 1), grade 3 bronchiolitis obliterans organising pneumonia (BOOP), grade 3 fatigue (n = 1), grade 3 skin toxicity (n = 1), grade 3 tremor (n = 1), grade 2 eyelid ptosis (n = 2), grade 2 liver toxicity (n = 1), grade 2 hypothyroidism (n = 1).

Regarding irAEs, any grade irAES occurred in 76 (20%) patients, of which 40 (10%) grade 1, 27 (7%) grade 2, 9 (2%) grade 3 and 1 (< 1%) grade 4 AEs, respectively. Considering the five pre-specified categories, 30 irAEs (8%) were cutaneous, 17 (4%) endocrine, 7 (2%) hepatic, 19 (5%) gastro-intestinal and 4 (1%) pulmonary. Further details are illustrated in Table 3.Tab3

Rates of irAEs in the Italian Early Access Program of nivolumab in mRCC

At a median follow-up of 12 months, the median PFS was 4.5 months (95% CI 3.7–6.2 months) and the 1-year OS rate was 63% in the overall study population. In patients with and without drAEs, median OS was 22.5 months (95% CI not yet evaluable) versus 16.4 months (95% CI 12.1–20.7 months), 1-year OS was 69.0% versus 59.7% and 2-year OS was 46.6% versus 43.6%, respectively, p = 0.01 (Additional file 1: Figure S1A). In patients who discontinued treatment for the occurrence of drAEs, 1-year OS was 65%. A similar significant result was obtained considering selectively the population with the occurrence of irAEs (n = 76) versus no reported irAEs, where median OS was not reached versus 16.8 months (95% CI 13.0–20.6 months), 1-year OS was 75.4% versus 59.8% and 2-year OS was 66.9% versus 36.8%, respectively, p = 0.002 (Fig. 1a).Fig1Fig. 1

Survival analysis for immune-related adverse events. Kaplan-Meier curves for overall survival in patients stratified for the occurrence of immune-related adverse events (a) and with landmark analysis at 6 weeks (b)

We performed this analysis in order to assess the safety and efficacy of nivolumab in mRCC patients in the real-world setting in Italy, in an Early Access Program, and the potential association between irAEs and survival outcome.

Nivolumab proved to be safe and well-tolerated in the routine clinical practice. The most common drAEs occurred were fatigue, rash, diarrhea, thyroid disfunctions, both hypothyroidism and hyperthyroidism, and pyrexia, consistently with previous literature evidences regarding nivolumab treatment, including a recent pooled analysis on the safety profile of nivolumab in patients with advanced melanoma [15].

Interestingly, in our analysis we found a lower rate of drAEs in the Italian EAP, both any grade and grade ≥ 3, as compared to those of the nivolumab arm of the clinical trial Checkmate 025 [3, 4, 7]. These findings highlight the importance of real-world data since they are able to give key messages about the efficacy and safety results of new anti-cancer agents in the routine clinical practice, thus helping the development and the management of drugs.

We showed that patients reporting treatment-related toxicity had a significantly longer OS. The most remarkable finding is that the selection of AEs potentially determined by the nivolumab-related immune system activation, the irAEs, leads to observe a stronger association with an improved survival, confirmed in a multivariable model.

Several studies have reported a potential association between the occurrence of irAEs during immunotherapy and treatment efficacy. First, in the setting of advanced melanoma, the development of irAEs in patients treated with CTLA4 inhibitors proved to be significantly correlated with higher tumor response rate and probability of survival [10–13, 20]. Similarly, during treatment with nivolumab, patients experiencing any grade of irAEs displayed a significant overall survival (OS) improvement [14], whereas in a recent study any grade irAEs were associated with an increased overall response rate (ORR) but not with longer progression free survival (PFS) [15]. Afterwards, similar evidences were collected in non-small cell lung cancer (NSCLC) patients treated with nivolumab, where those experiencing irAEs had a higher ORR and an increased PFS, thus identifying the occurrence of irAEs as an independent predictor of treatment response [18]. Nevertheless, other evidences did not support the correlation between immune-related toxicity and benefit from immunotherapy. Specifically, in two cohorts of patients affected by metastatic melanoma treated with ipilimumab, the incidence of irAEs and the use of systemic corticosteroids to treat immune-related toxicity were not found to be associated with ORR, OS and time to treatment failure [21, 22].

Our findings support the association between irAEs and survival benefit from nivolumab. However, one of the possible bias of this analysis is the influence of treatment duration, since its increase leads to a prolonged nivolumab exposure and therefore to a potentially higher likelihood of AEs occurrence. To this purpose, we performed a landmark analysis at the median time of appearance of AEs, 6 weeks, and the association of irAEs occurrence with an improved OS was confirmed. Additionally, early versus late onset irAEs, defined as occurring before or after the cutoff of 6 weeks from the treatment start, showed a non-significant trend towards a poorer OS, however further studies should collect more solid evidence about this topic.

Interestingly, two recent studies obtained different results according to the intensity of the toxicity. Recently Judd et al, in a retrospective series of non-melanoma patients treated with anti-PD-1 agents obtained a trend towards an improved ORR in patients experiencing any irAE, but this association was statistically significant for low grade irAEs [23]. It confirmed the previous data by Weber et al, showing that in the setting of melanoma patients who received nivolumab monotherapy, those with any grade irAEs had a significantly better ORR, while for G3–4 irAEs this correlation was not significant [15]. This could reflect the tumor biolological aggressiveness, or, on the other side, it may be explained by the negative effects of high grade AEs, potentially dangerous or able to impair the adequate assessment of tumor response. In our results, G3–4 drAEs were significantly associated with a lower 1-year OS rate as compared to G1–2 drAEs, while G3–4 versus G1–2 irAEs showed a non-significant trend towards a shorter survival outcome.

The reasons underlying this potential association between irAEs occurrence and patients’ outcome are still to be determined, although some hypotheses have been advanced. IrAEs are caused by the unbalancing of the immune system induced by immune checkpoint blockade, possibly generated by the cross-reactivity between tumor neoantigens and normal tissue antigens [24]. These shared antigens could be involved in this process, although this seems not to be a convincing explanation in light of the low incidence rate of multiple irAEs in the same patient. Other possible mechanisms were postulated concerning anti-CTLA4 antibodies: on the one side the role of immunosuppressive regulatory T cells (T reg) is questioned, on the other ipilimumab could induce a non-specific increase in endogenous T-cell response mediated by dendritic cells or paracrine cytokine stimulation. Another interesting theory advocates a potential causality of this association irAEs-treatment efficacy due to the interaction between immunotherapy and polymorphisms in the genes involved in the response to ICIs [25]. Finally, considering anti-PD-1/PDL-1 inhibitors, they are deeply involved in the regulation of humoral immune response, since they proved able to modulate B cells both directly and via T-cell mediation. The altered production of auto-antibodies could in turn mediate the development of irAEs, thus possibly explaining the association between immune-related toxicity and treatment response [26]. A possible evidence in support of this hypothesis is the finding that the presence of pre-treatment anti-thyroid antibodies is an independent predictor of response to nivolumab in NSCLC patients [18].

Another key point is the definition of irAEs, since no well-established criteria have been identified yet and literature evidences are not clear upon this topic. We decided to select five distinct categories of irAEs (cutaneous, endocrine, hepatic, gastro-intestinal and pulmonary), according to the best known pathogenetic correlation between toxicity and immunotherapy. However, further studies focusing specifically on the mechanisms of action of ICIs could better clarify which AEs should be considered immune-related. The highest level of evidence collected in other disease settings regarding the association of irAEs with response to immunotherapy is about cutaneous toxicity, especially vitiligo, followed by endocrine alterations. Specifically, in melanoma patients, cutaneous irAEs were shown to be associated with clinical benefit and better outcome from anti-PD-1 treatment, with a major role of vitiligo [27–29] and in NSCLC dermatological and endocrine irAEs, specifically antibody-mediated thyroid disfunctions, resulted to be positively correlated with tumor responses or improved survival to anti-PD-1 ICIs [16, 17]. Although limited by the small numbers, we observed that cutaneous, endocrine and gastro-intestinal irAEs were significantly associated with improved OS, similarly to unselected irAEs, while hepatic and pulmonary irAEs conditioned a poor OS outcome. Nevertheless, no definitive conclusions could be derived from these results, considering the limited number of patients experiencing any specific class of irAEs.

At our knowledge, this is the first study investigating the potential association of irAEs with survival in the setting of mRCC. Our results highlight the importance to recognize treatment-related toxicities, specifically those potentially underlying immune-related mechanisms, since these display a significant association with an improved patients’ outcome from immunotherapy with anti-PD-1 ICIs. The limitations of this study are inherent to the specific nature of EAP, basically the high heterogeneity of the patients’ population and its retrospective nature, that could have impaired the power of the statistical analysis and limited to drive robust conclusions upon the results achieved. Moreover, as previously discussed, a precise definition and categorization of irAEs is still lacking, and the classification that we performed is not validated or standardized.

In conclusion, the appearance of irAEs strongly correlates with a survival benefit in a real-life population of mRCC patients treated with nivolumab. Further studies could confirm the potential role of the incidence of irAEs as a predictor of response to ICIs as well as explore the underlying mechanisms leading to the development of immune-related toxicity. As a consequence, criteria should be identified to establish whether a precise timing of occurrence of irAEs or specific organ-related toxicities could better predict the treatment outcome, with the aim to better personalize the treatment management in this disease setting.