Renal Toxicities in Cancer Patients Receiving Immune-Checkpoint Inhibitors: A Meta-Analysis

Aim: We performed a meta-analysis of the available clinical trials of immune-checkpoint inhibitors to assess risk differences and relative risks of renal toxicity. Methods: 17 randomized phase III studies were selected, including 10,252 patients. Results: The administration of immune-checkpoint inhibitors resulted in an overall low-grade, high-grade and all-grade renal toxicity Risk Difference of: 0.746% (95% CI 0.629% to 1.15%, p < 0.001—random), 0.61% (95% CI, 0.292–0.929%, p < 0.001—fixed) and 1.2% (95% CI, 0.601–1.85%—random), respectively. The pooled Relative Risk of low-grade, high-grade and all-grade renal toxicity was: 2.185 (95% CI 1.515–3.152—fixed), 2.610 (95% CI, 1.409–4.833, p = 0.002—fixed) and 2.473 (95% CI, 1.782–3.431, p < 0.001—fixed), respectively. An increased risk of renal toxicity was evident in some subgroups more than others. Conclusion: Immune-checkpoint inhibitors are associated with an increased risk of renal toxicity.


Introduction
In the recent past, the introduction of the immune-checkpoint inhibitors (ICIs) within our anticancer armamentarium has changed, for the better; the natural history of a number of malignancies, greatly improving both prognosis, as well as the overall clinical outcome (including quality of life) of many cancer patients.
To date, we are exploiting for therapeutic purposes, just two immune checkpoints negatively regulating the immune response: the programmed death receptor 1/programmed death receptor ligand 1 (PD-1/PD-L1) axis and the cytotoxic T-lymphocyte antigen 4 (CTLA-4)/CD80 axis.
Indeed, presently available ICIs target these receptor/ligand axes, which are located on the surface of either tumour or immune cells, ultimately leading to a de-inhibition of the immune response which is ontologically devoted to protect our organism from any external harm. As the immune system (and not the tumour itself) is the final "target" of these agents, their toxicity profile is mainly due to its inappropriate activation, which can cross-react against antigens expressed on normal tissues and organs.

Study Selection
The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [11] was followed to identify eligible studies. Two authors (FM, VM) reviewed the citations from PubMed from January 2008 to January 2019 obtained by combining the following words: "Nivolumab" (Bristol-Myers Squibb Pharma EEIG, Dublin, Ireland), "Durvalumab" (AstraZeneca AB, Södertälje, Sweden), "Ipilimumab" (Bristol-Myers Squibb Pharma EEIG, Dublin, Ireland), "Tremelimumab" (AstraZeneca AB, Södertälje, Sweden), "Pembrolizumab" (Merck Sharp & Dohme B.V., Haarlem, Netherland), "Atezolizumab" (Roche Registration GmbH, Grenzach-Wyhlen, Germania) or "Avelumab" (Merck Europe B.V., Amsterdam, Netherland). Our search was limited to human randomized phase III trials published in English focalized on the efficacy and safety of ICIs in cancer patients, which reported data on renal toxicity. Renal toxicity was considered according to RIFLE or AKIN criteria, therefore a primary alteration in serum creatinine >0.3 mg/dl or an increase >150% over the first measurement. In the presence of multiple publications of the same trial, we selected the most recent, or else the most complete in terms of adverse events reporting. Meeting abstracts were not included due to the risk of biases in reporting adverse events.
After checking the safety reports of these studies, renal toxicity was defined as: "creatinine increase", "renal failure/injury", or "nephritis".

Data Extraction, Clinical Outcomes, and Quality Assessment
Three authors (F.M., V.M., A.R.) in an independent manner found and registered data on: authors, study population, experimental and control arms, number and severity of renal adverse events (low-and high-grade).
All trials used either version two, three, or four of the National Cancer Institute's Common Terminology Criteria for Adverse Events (CTCAE) to report the nature and the severity of the adverse events of interest. The quality of the selected studies was classified according to the Jadad 7-items system based on the randomization process, blinding procedure, and patients' withdrawal [12].
The primary outcomes of our meta-analysis were: risk differences and relative risks among all trials included, and between specific populations including: patients who received PD-1/PD-L1 or CTLA-4 inhibitors alone, patients who received ICIs in combination with other agents (either chemotherapy or targeted agents), and patients who received CTLA-4 inhibitors alone or in combination with other agents, either immunologic or not.

Statistical Methods
The number of patients receiving ICIs, as well as the number of renal adverse toxicity patients in both treatment and control arms were extracted from all selected studies. Risk Differences (RD) were evaluated as different incidences of renal toxicities between experimental and control arms. Furthermore, we evaluated Relative Risk (RR) as the ratio between incidence of renal toxicity in experimental and control arms; 95% Confidence Intervals (CIs) were subsequently calculated, as proposed by Altmann et al. [13]. Studies with no Immune-mediated adverse events in the treatment or control arms were corrected according to Yates [14].
Cochran's Q statistic was employed to test heterogeneity between studies. The Iˆ2 statistic (indicating the percentage of variance in a meta-analysis) was chosen for quantification of inconsistency. Both the inverse variance fixed-effects model (weighted with inverse variance), and the random effect model were adopted to estimate RD and RR among trials included. In case of non-statistically significant heterogeneity, we considered the result obtained by fixed-effects model, while in case of statistically significant heterogeneity results provide by the adoption of the random effect model was considered.
Data collection was obtained using Microsoft Excel (version 16.62, Microsoft, Redmond, Washington, DC, USA), while data analysis was performed with the MedCalc software (MedCalc Software, version 20.113, Ostend, Belgium).

Search Results
Three authors independently identified a total of 6083 potentially relevant articles on PubMed, EMBASE, Cochrane Library, focused on the efficacy and safety of ICIs in cancer patients. Of these, 5706 were primarily excluded for at least one of the following reasons: review articles, case reports, systematic review, meta-analysis, editorials, letters, or commentaries. Among the 377 remaining studies, only 27 were potentially relevant phase III trials. Of them, 10 were independently excluded by three authors for at least one of the following reasons: presence of ICIs in both experimental and control arm, or inadequate data on the toxicities of interest ( Figure 1).

Statistical Methods
The number of patients receiving ICIs, as well as the number of renal adverse toxicity patients in both treatment and control arms were extracted from all selected studies. Risk Differences (RD) were evaluated as different incidences of renal toxicities between experimental and control arms. Furthermore, we evaluated Relative Risk (RR) as the ratio between incidence of renal toxicity in experimental and control arms; 95% Confidence Intervals (CIs) were subsequently calculated, as proposed by Altmann et al. [13]. Studies with no Immune-mediated adverse events in the treatment or control arms were corrected according to Yates [14].
Cochran's Q statistic was employed to test heterogeneity between studies. The I^2 statistic (indicating the percentage of variance in a meta-analysis) was chosen for quantification of inconsistency. Both the inverse variance fixed-effects model (weighted with inverse variance), and the random effect model were adopted to estimate RD and RR among trials included. In case of non-statistically significant heterogeneity, we considered the result obtained by fixedeffects model, while in case of statistically significant heterogeneity results provide by the adoption of the random effect model was considered.
Data collection was obtained using Microsoft Excel (version 16.62, Microsoft, Redmond, Washington, DC, USA), while data analysis was performed with the MedCalc software (MedCalc Software, version 20.113, Ostend, Belgium).

Search Results
Three authors independently identified a total of 6083 potentially relevant articles on PubMed, EMBASE, Cochrane Library, focused on the efficacy and safety of ICIs in cancer patients. Of these, 5706 were primarily excluded for at least one of the following reasons: review articles, case reports, systematic review, meta-analysis, editorials, letters, or commentaries. Among the 377 remaining studies, only 27 were potentially relevant phase III trials. Of them, 10 were independently excluded by three authors for at least one of the following reasons: presence of ICIs in both experimental and control arm, or inadequate data on the toxicities of interest ( Figure 1). At the end of the process, 17 randomized phase III studies were selected and considered independently by three authors [15−31]. In Table 1, we have summarized the main characteristics of the included trials. At the end of the process, 17 randomized phase III studies were selected and considered independently by three authors [15][16][17][18][19][20][21][22][23][24][25][26][27][28][29][30][31]. In Table 1, we have summarized the main characteristics of the included trials. Overall, 10,253 patients have been examined in our analysis; 5721 received ICIs alone (n = 3450) or in combination (n = 2114) with other agents, 4490 received a PD-1 or PD-L1 inhibitor, while 781 received a CTLA4 inhibitor. All grade renal toxicities have been found in 134 patients enrolled in the experimental arms of the selected trials, and 44 in the control arm. High-grade renal toxicity has been observed in 37 patients enrolled in experimental arms and nine in control patients. The study in which the highest renal toxicity rate has been observed was Keynote-045 [22] in which overall renal toxicity occurred in 38 patients enrolled in experimental arm and 22 patients enrolled in comparator arm.

Quality of Studies
The Jadad scoring system was employed to assess the quality of studies and was investigated by three authors independently. Follow-up time was adequate for all studies. Jadad scores are listed for each trial in Table 1; the mean scores were 4 (range, 3 to 5) underlining the fair quality of the selected studies.
Overall Risk Difference between experimental and comparator arm was statistically significant in low grade, high-grade and all-grade analysis (Table 2). Also, RR analysis demonstrated a significant higher risk to develop renal toxicity for patients who received an ICI (Table 2). Heterogeneity among trials included in analysis was statistically significant in low and all grade RD analysis (I2: 62%, 52.3%, respectively) while no statistically significant heterogeneity has been observed in G3-5 analysis (I2: 0 %) ( Table 3). Kidney toxicity at every grade for each category of ICIs (PD-1/P-L1 and CTLA4) are shown in the Supplementary Materials.

Renal Toxicity in Patients Receiving Anti CTLA-4 Agents
Only two trials were considered for this outcome [20,31], for a total of 781 patients. In this subpopulation RD analysis showed a statistically significant risk difference in lowgrade analysis while no statistically significant risk difference was found in high-grade and all-grade analyses. RR analysis revealed a statistically significant RR only in all-grade toxicity ( Table 2). Heterogeneity was significant only in all-grade RR analysis (Table 3). Low grade RD was 0.74% (95% CI, 0.15-1.34 p = 0.014-fixed). Pooled RR in all-grade toxicity was: 2.137 (95% CI 1.087 to 4.199 p = 0.028-random)

Discussion
We carried out a meta-analysis of 17 randomized clinical trials to evaluate the incidence of renal toxicity among patients receiving ICIs.
In our analysis, we demonstrated that the administration of ICIs is associated to an increased risk of renal toxicity of low-and high-grade. Moreover, this correlation was confirmed also in a specific subpopulation of patients including patients who received ICIs alone and in patients receiving ICIs in combination with other agents. Regarding studies exploring CTLA-4 inhibitors, no statistical significant differences in terms of RD (for highand all-grade toxicity) and RR (for low-and high-grade toxicities) were found. However, the relatively limited number of trials analysed could not have the adequate power to detect a difference between incidences of these side events.
The estimated incidence of ICIs associated AKI (ICIs-AKI) ranges from 1.4% to 4.9%, and the most common renal lesion is represented by tubulointerstitial nephritis (TIN). Nevertheless, several other immune-mediated pathologies have been associated, including various glomerulonephritides [32][33][34]. Though it is difficult to describe a ICIs induced nephropathy because, to our knowledge, none of these medications showed a direct effect on any kidney structure neither biopsy proven studies.
The study conducted by Cortazar et al. shows that there is a variable and often prolonged delay between ICIs initiation and the development of AKI, that the frequent rate of extrarenal immune-related adverse events (irAEs) occurring concomitantly or immediately preceding the AKI, that patients who have concomitant extrarenal irAEs have a lower likelihood of kidney recover, that patients who were rechallenged with ICIs therapy did not develop recurrence of ICIs-AKI and, as expected, that those who fail to achieve kidney recovery after ICIs-AKI show increased mortality [35]. In a large retrospective study conducted by Seethapathy et al., the authors underline that irAEs may be a novel risk factor for adverse kidney outcomes, since those patients resulted at increased risk of sustained eGFR loss. Those findings suggests that adverse kidney outcomes after ICIs may have a different mechanism than typical TIN, however, has to be considered that these patients are not exempted from other causes of AKI such as volume contraction, ischemic tubular injury or obstructive nephropathy [36]. In a recent metanalysis conducted by Magee DE et al., the authors describes the adverse event profile for immunotherapy agents compared with chemotherapy in solid organ tumors. The authors found an incidence of AKI in 1.31% of the studied patients that is lower than ours. Nevertheless the authors choose only the studies that compared treatment with ICI with standard chemotherapeutic regimens; therefore it would be possible that other studies considering ICIs alone or in combination with other agents influenced our results [37]. Therefore, is of utmost importance to perform long term trials to analyze acute and chronic renal toxicity of ICIs.
Our analysis was limited by several factors, including the different setting, disease and modality of administration (combination or alone) of ICIs. It is thus not surprising that higher renal toxicity rates have been found in the Keynote-045 study, which compared pembrolizumab to investigator-choice chemotherapy in patients progressed to standard first line therapy for advanced or metastatic urothelial carcinoma [22]. It is indeed probable that the previous administration of platinum compounds (as well as the specific disease site) could partially justify this finding.
Furthermore, our meta-analysis per definition cannot capture the possible role of a preexisting subclinical renal damage, due to co-morbidities (e.g., hypertension and diabetes) or previous therapeutic interventions (e.g., the above mentioned renal damage from previous platinum-based therapies or from previous nephrectomy). The estimation of the exact incidence of renal toxicity among patients receiving ICIs is of particular interest, since these agents are often administered in patients with pre-existing renal failure of some degree.
Here, two other key issues emerge. First of all, very limited data are available about the administration of ICIs in patients with moderate or severe renal failure, or even in patients with end-stage renal disease on dialysis. In urothelial cancer, ICIs have been administered in patients unfit to receive standard platinum-based chemotherapy, thus allowing the enrolment of patients with a certain degree of impaired renal function, ultimately demonstrating that the immunological treatment did not raise the risk of renal failure [38]. Unfortunately, the creatinine or creatinine clearance limits imposed by the vast majority of randomized controlled phase III trials do not allow for the confirmation of the activity and, more importantly, safety of ICIs in patients with more advanced renal impairment. From a pharmacological viewpoint, this makes little sense, if any, considering that ICIs are monoclonal antibodies which are cleared by the reticulo-endothelial system, and not by the kidneys [38]. Indeed, retrospective case series of cancer patients on dialysis treated with these agents are accumulating, suggesting the safety of this therapeutic approach [39][40][41][42].
Another point is that, in trials conducted in the era of modern immunotherapy, renal toxicity is almost always reported as "creatinine increase" that is extremely incomplete.
Not to take into account that "You do not see what you're not looking for", meaning that immune-related renal toxicity is, in our opinion, often under-diagnosed. It is indeed common to observe, during the first months of treatment with ICIs, an initial rise in creatinine levels, which subsequently stabilizes, which could indeed be due to an autoimmune response against the kidneys.

Conclusions
As suggested by our study, ICIs are associated with an increased risk of renal toxicity. Although usually of a low-grade, these adverse events are, in our opinion, too often underdiagnosed. Thus, the risk of renal toxicity should be always considered before treatment starts, and the cooperation of nephrologist and oncologist has to be encouraged in the presence of a deterioration in renal function. For sure, the use of ICIs should not be discouraged in patients with pre-existing renal impairment, even when on dialysis.

Summary Points
In recent years, immune-checkpoint inhibitors revolutionized the treatment scenario and outcomes of cancer patients. Few studies have exanimated the effective incidence of renal toxicity mediated by immune-checkpoint inhibitors.
We selected clinical trials evaluating safety and efficacy of immune-checkpoint inhibitors in patients with solid tumors. We identified 17 randomized phase III studies, including 10,252 patients. Of these, 5721 received ICIs alone (n = 3450) or in combination (n = 2114) with other agents, 4490 received a PD-1 or PD-L1 inhibitor, while 781 received a CTLA4 inhibitor.
We performed a meta-analysis aimed at evaluating the Risk Difference (as different incidences of renal toxicities between experimental and control arms) and Relative Risk (as the ratio between incidence of renal toxicity in experimental and control arms).
Immune-checkpoint inhibitors are associated with an increased risk of renal toxicity of low-and high-grade.
An increased risk of renal toxicity was evident in specific subpopulation of patients, such as patients that received immune-checkpoint inhibitors alone or those receiving immune-checkpoint inhibitors in combination with other agents.

Funding:
The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript.