Immune checkpoint inhibitor does not increase risks of poor outcomes in cancer patients with COVID-19 infection: A systematic review and meta-analysis


 Background The association between prior exposure to immune checkpoint inhibitor (ICI) and outcomes of cancer patients with coronavirus disease 2019 (COVID-19) infection has yet to be systematically evaluated. As the current evidence remains equivocal, this meta-analysis aims to investigate the effects of ICI treatment on COVID-19 prognosis, including mortality, severity, and hospitalization.Methods Eligible studies published up to 14 September 2020 were included and assessed for risk of bias using the Quality in Prognosis Studies tool. A random-effects meta-analysis was conducted to estimate the pooled effect size along with 95% confidence intervals (CIs). The quality of body evidence was evaluated using the modified Grading of Recommendations Assessment, Development, and Evaluation (GRADE) framework.Results Six studies involving a total of 1647 COVID-19-infected cancer patients were included in the systematic review. We discovered that prior ICI exposure was not associated with COVID-19 mortality (odds ratio [OR] 0.93 [95% CI: 0.37-2.36]; I2=30%), severity (OR 1.15 [95% CI: 0.40-3.28]; I2=0%), and hospitalization (OR 1.35 [95% CI: 0.64-2.87]; I2=51%). However, we discovered that prior exposure to chemoimmunotherapy predicted COVID-19 severity (OR 8.19 [95% CI: 2.67-25.08]; I2=0%), albeit with small sample size. GRADE assessments resulted in moderate-quality evidence for mortality, while the other outcomes yielded very low-to-low-quality evidenceConclusion Our findings indicated that ICI treatment should not be adjourned nor terminated during the current pandemic. Rather, COVID-19 vigilance should be increased, especially in patients receiving chemoimmunotherapy. Further studies with larger ICI cohorts are required to confirm our findingsTrial registration number: This project has been prospectively registered at PROSPERO (registration ID: CRD42020202142) on 4 August 2020.


Introduction
The current coronavirus disease 2019 (COVID-19) pandemic has brought upon a signi cant burden in global economy and health, resulting in millions of cases and nearly one million of death.
[1] Recent reports have suggested that cancer patients are more vulnerable to COVID-19-related death and complications [2][3][4]; thus, meticulous management to prevent further deterioration in such patients are essential. In light of this, the question remains whether to postpone or continue active cancer treatment, including immune checkpoint inhibitor (ICI), which exerts immunomodulatory functions [5]. To the best of our knowledge, the current evidence on the effect of prior ICI treatment on cancer patients infected with COVID-19 remains equivocal [6][7][8]. Therefore, this meta-analysis aims to explore the association between ICI and COVID-19 outcomes in cancer patients, thus providing the best available evidence to guide real-time treatment decisions in such cases.

Data extraction and risk of bias assessment
The following information were extracted from each included studies: (1) author and year of publication; (2) study characteristics, including recruitment period, study design, settings, location and sample size; (3) patient characteristics, including mean/median age, proportion of male patients, comorbidities, cancer types, adjuvant therapies, and characteristics of ICI (i.e. last exposure to treatment, drug class); and (4) outcomes.
The primary outcomes in this review were the risk of poor prognosis (i.e. mortality and severity) among COVID-19-infected cancer patients. Whenever possible, outcomes were further investigated per criterion according to the WHO interim guidance, viz. rate of hospitalization, intensive care unit (ICU) admission, invasive ventilation, acute respiratory distress syndrome (ARDS), and shock.[12] Data extraction were performed by one review author (GL) using a pre-speci ed form in MS Excel® for O ce 365 MSO ver. 2002 (Microsoft Corporation, Redmond, WA, 2018). A second investigator (RAB) checked the accuracy of these data extractions, and any disagreements resolved by the consensus between the authors.
Any reported effect size types (i.e. hazard ratio [HR], odds ratio [OR], relative risk [RR]) were incorporated in this study. When only binary data were provided, unadjusted odds ratio were calculated from the frequency of events and sample sizes. [13] In the case of studies reporting multiple adjustment models, we extracted the adjusted model incorporating the greatest number of covariates.
The included studies were further assessed for risk of bias by using the Quality in Prognosis Studies (QUIPS) tool [14], where the overall bias risk was judged to be low, moderate, and high risk. Risk of bias assessments were conducted by two independent reviewers, and any discrepancies were resolved by a third adjudicator in a blinded fashion. Details on the QUIPS checklist can be found on Supplementary  [16]. In the case of studies involving overlapping populations, analyses were prioritized to the largest-sized study. Outcomes were pooled as ORs, RRs, or HRs separately along with their 95% con dence intervals (CIs) by using the generic inverse variance methods. Both unadjusted and adjusted outcomes were extracted and synthesized in this study; however, adjusted estimates were prioritized for reporting and interpretation whenever available. Due to the likeliness of unexplained heterogeneity [9], a DerSimonian-Laird random-effects model was used [17]. Heterogeneity between studies was investigated with Cochran's Q test and I 2 statistics. According to I 2 values, heterogeneity was classi ed as negligible (0-25%), low (25-50%), moderate (50-75%), or high (>75%), while the signi cance level for Q statistics were set at 10%.
Subgroup analyses were performed only for outcomes with at least two studies in each subset. A priori, we de ned subgroups according to study design, location, sample size, and risk of bias; while additional subsets based on the presence of adjuvant therapy (i.e. ICI monotherapy vs ICI plus chemotherapy), cancer type (i.e. lung vs non-lung cancer), and comparator group (i.e. no treatment, chemotherapy, targeted therapy, radiotherapy, and surgery) were determined posteriori. On the other hand, sensitivity analysis was conducted by leave-one-out analysis and simultaneous exclusion of studies with high-bias risk. A P-value of <0.05 indicated statistical signi cance. When the number of studies was adequate (n10) [18], potential publication bias was investigated visually by contour-enhanced funnel plot [19] and quantitatively with Egger's [20] and Begg's test [21].
Lastly, the overall quality of evidence was assessed with the modi ed version of Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework for prognostic reviews [22], where the certainty of evidence was rated as high, moderate, low, or very low according to the judgements of these following domains: phase of investigation, study limitation, inconsistency, indirectness, imprecision, publication bias, moderate/large effect sizes, and exposure-response gradients.

Search results and study characteristics
The initial search yielded 898 records, of which 406 were deduplicated and 470 were excluded after title and abstract screening. The remaining 22 studies were retrieved for full-text assessments, where 14 studies were excluded due to inappropriate design (i.e. case series/reports with <10 patients) and an additional two due to inappropriate or unidenti able settings (see Additional methods in Supplementary Materials for further details). Consequently, six studies cumulating a total of 1647 patients were included in this review-among which 919 (55.8%) were male, and hypertension were the most reported comorbidities (37.8%; Table 1). Lastly, we expanded our search by using a non-human skill-dependent search method based on PubMed's 'similar articles' algorithm, in addition to manually hand-searching reference lists of included studies. No new studies were identi ed from these expanded searches. Details on the literature search strategy are illustrated on Figure 1. Among the included studies, two each were conducted in America [23,24] and Europe [25,26], one in Asia [27], and one [28] in multiple regions (i.e. Asia, Europe, and USA). All studies were conducted retrospectively, and all but one [23] was multicentered. frequent type was lung cancer (407 [24.7%]), followed by gastrointestinal (200 [12.1%]) and breast tumors (199 [12.1%]). With regards to ICI type, most patients received anti-PD-1 (94 [5.7%]), followed by anti-PD-L1 (49 [3.0%]) and anti-CTLA-4 (46 [2.8%]; Table 1).
Risk of bias assessment resulted in low risk for three studies [24,26,28], moderate risk for two studies [23,27], and high risk for one study [25]. Most of the studies yielded unclear risks in study attrition and confounding domains (Supplementary Figure S1), which may be explained by the fact that all studies were retrospective. Furthermore, two studies reported that their studies were underpowered due to small sample sizes [24,25], while the other two acknowledged their relatively small sample size as a potential limitation [27,28]. Details on the risk of bias assessment for each signaling question can be found on Supplementary Figure S2.

Outcomes
The summary of pooled unadjusted and adjusted effects can be found on Table 1, while the certainty of evidence according to GRADE assessment can be seen on Supplementary Table S3. GRADE assessment of the effects of prior ICI treatment on COVID-19 mortality resulted in moderate-quality evidence, while the remaining outcomes yielded very low-to-low-quality evidence. Publication bias assessments were not performed as no outcomes yielded more than 10 studies [29]. Furthermore, sensitivity analyses were only eligible for unadjusted and adjusted effects on mortality outcome.
We discovered that prior exposure to ICI was not associated with COVID-19 mortality (OR 0.94 [95% CI: 0.53-1.66]; Figure 3A), which was further ascertained by our ndings from the adjusted effects (OR 0.93 [95% CI: 0.37-2.36]; Figure 4A). Although both models yielded low and nonsigni cant heterogeneity (unadjusted estimates, I 2 =29%; adjusted estimates, I 2 =30%; both with p>0.10), we discovered that smaller-sized studies and studies with moderate-to-high risk of bias tend to yield excessively wide CIs. Nonetheless, we proved that these studies did not contribute much to the pooled effects as our estimates remained relatively robust following sensitivity analysis (Supplementary Figure S3A). Considering this, we judged the certainty of evidence for both qualitative and quantitative assessments to be moderate. Furthermore, subgroup analyses based on cancer type, presence of adjuvant therapy, and comparator group also revealed similar trends.
Similar to mortality, we also demonstrated a non-signi cant association between prior ICI treatment with severity and hospitalization ( Figure 3B-C). However, we were unable to establish a rm evidence as the observed low-to-moderate heterogeneity (severity, I 2 =28%; hospitalization, I 2 =51%) remained unexplained due to inabilities to perform subgroup and sensitivity analyses (see footnote of Table 1). Furthermore, study-speci c ndings showed equivocal trends, where Robilotti et al [23] revealed that prior ICI exposure was an independent risk factor of COVID-19 severity and hospitalization (Supplementary Table S4), while the other studies [24,25,27,28] stated otherwise. Meta-analysis of adjusted estimates on hospitalization outcome was not possible due to overlapping populations [23,24], while those of severity outcome revealed that ICI may increase risks of COVID-19 severity in cancer patients (OR 3.17 [95% CI: 1.44-6.94] in Figure 4B). Nonetheless, it is important to note that the observed effect was primarily driven by a single study [23], hence indicating that the independent model was unreliable. Considering this, we judged the quality of evidence on COVID-19 severity to be low for qualitative and very low for quantitative assessments, while those of COVID-19 hospitalization resulted in low-quality evidence for both assessments.
Subset analyses based on cancer type and presence of adjuvant therapy for hospitalization outcome revealed similar trends to those of mortality outcome. However, we discovered that concomitant use of ICI and chemotherapy was associated with COVID-19 severity (OR 8.19 [2.67-25.08]; I 2 =0%, p=0.441). Furthermore, we also found that the risk of severity was higher in patients with prior ICI treatment than patients without active cancer treatment (OR 2.39 [95% CI: 1.24-4.62], I 2 =0%), and that the risk of hospitalization was higher in patients receiving ICI than those receiving targeted therapy (OR 2.60 [95% CI: 1.35-5.01], I 2 =0%). Nonetheless, it is important to note that the observed effects may be primarily driven by a single study as most of the included studies yielded small sample sizes and wide CIs.

Discussion
This meta-analysis showed that prior exposure to ICI treatment was not associated with poor prognosis in COVID-19-infected cancer patients. We demonstrated that there was a moderate-quality evidence that ICI did not increase the risks of COVID-19 mortality, while the certainty of evidence for other outcomes yielded very low-to-low quality, which was primarily due to the observed equivocal trends where, unlike other studies, Robilotti et al [23] reported an independent association between ICI and COVID-19 severity and hospitalization-which may partly be explained by the fact that the included patients had worse conditions than general cancer populations, thus potentially overestimating the trends. [23]. Although we observed that ICI may independently increase the risk of COVID-19 severity, the interpretation of our ndings were limited as study-speci c estimates were imprecise and the model was primarily driven by a single study [23]. Furthermore, the observed risk may be elucidated by the relatively prevalent hematological cancer patients, which was a known risk factor of ICI-related immune adverse events [30]. This was further ascertained by Robilotti et al, stating that the risk may potentially be accentuated in patients with metastatic and hematologic cancers. [23] In addition, we also discovered that concomitant use of ICI and chemotherapy was associated with COVID-19 severity. Although further studies are required to con rm these trends, the effects may potentially be elucidated by the fact that patients receiving chemoimmunotherapy are at higher risks of immune-mediated adverse events [30,31], thus predisposing these patients to COVID-19 complications.
The dilemma to whether continue, postpone, or even terminate active cancer treatment, including ICI, persisted during the current COVID-19 pandemic. While physicians are expected to prioritize patients' safety, it is also important to ensure that patients receive timely treatment. Several reports and guidelines have regarded ICI as unsafe during the pandemic, and have advised the postponement of such treatments due to safety concerns [32][33][34]. These are based on two hypothetical adverse interactions between ICI and COVID-19 infection. First, recent reports have suggested that COVID-19 infection may mask ICI-related pneumonitis symptoms, thus potentially delaying essential treatments. [7,35] Although this might be detrimental considering that ICI-related pneumonitis accounts for about one-third of treatment-related deaths, their incidence is relatively rare. Furthermore, the risks of ICI pneumonitis tend to be augmented in early ICI recipients and super-responders [36], suggesting that a prompt and accurate risk strati cation, in addition to increased COVID-19 vigilance, may be able to mitigate this issue.
In addition, early hypotheses postulated that ICI may worsen COVID-19 outcomes due to potential immune hyperactivation [36][37][38], where they may upregulate pro-in ammatory cytokines [36,37] and over-activate CD8 T-cells [38]-resulting in the dysregulation and exhaustion of T-cells [36,39]. This hypothesis was supported by the fact that severe COVID-19 cases were associated with lymphopenia and immune hyperactivity [6,37]thus suggesting that ICI may synergistically exacerbate cytokine storm in COVID-19 infection [38]. Nevertheless, a recent report by Di Cosimo et al stated that the occurrence of cytokine storm in COVID-19 patients was more likely to be driven by direct viral damage rather than immunemediated in ammation [40]. Moreover, recent studies have argued the potential role of ICI on the prevention and management of COVID-19 infection. ICI has exhibited immunity protection against several infectious agents [37], while also restoring cellular-mediated immunocompetence resulting in increased viral control [6,41]. In addition, ICI may enhance immune response to viral antigens without triggering adverse immune reactions [40], thus further suggesting the potential therapeutic utility of ICI.
Altogether, these ndings indicated that ICI treatment should not be unnecessarily deferred during the current pandemic; but rather, COVID-19 vigilance in ICI-treated cancer patients, especially those receiving chemoimmunotherapy, should be increased. This is especially important to ensure prompt diagnosis and treatment of COVID-19 infection, thus preventing adverse outcomes in such patients. The decision to continue or suspend ICI treatment should be based on case-by-case approaches [36,42], where treatment adjustments may be performed to mitigate risks of COVID-19 infection by reducing patients' contacts to medical system, rather than due to ICI-related safety concerns. Furthermore, alternative decisions to speci c cohorts may be adopted, e.g. early treatment discontinuation in patients with complete or prolonged response [8], adjustments of treatment interval or modality [7,43], or adjournments of ICI treatment in high risk patients (e.g. elderly, patients with history of immune-related adverse events and/or comorbidities) [43].
This study has several limitations. Although our ndings rejected the early hypotheses stating that ICI may cause deleterious effects to COVID-19 cancer patients, study paucity and small-sized cohorts limited the interpretation of our results. Furthermore, some models (i.e. severity and hospitalization) were also limited by the predominant moderate-to-high-bias risk, which was further worsened by the observed heterogeneity in hospitalization outcome, thus resulting in the judgement of very low-to-low-quality evidence. Nonetheless, we demonstrated a relatively consistent and robust ndings that ICI was not associated with COVID-19 mortality. Moreover, although all included studies were conducted retrospectively, the studies involved diverse populations, thus ascertaining the generalizability of our ndings. Despite this, it is worth noting that none of the included studies directly compared the risks between different ICI classes, implying that further studies with larger ICI cohorts are required to con rm our ndings and to explore the observed effects. In addition, we also advise future studies to speci cally investigate the association between prior ICI exposure and COVID-19 severity in hematological cancer patients and in patients receiving chemoimmunotherapy, as the current evidence was still equivocal. Lastly, although our eligibility criteria may introduce language bias, we did not discover any non-English article during the literature search process, thereby suggesting that any potential bias was insigni cant.
To the best of our knowledge, this is the rst meta-analysis conducted to evaluate the association between prior ICI exposure and COVID-19 outcomes. Although our ndings were limited due to study scarcity and small-sized cohorts, we were able to establish a moderate-quality evidence on the non-signi cant relationship between ICI and COVID-19 mortality. We hope that our ndings may encourage physicians to increase COVID-19 vigilance and perform risk-bene t assessments on each ICI-treated cancer patients, rather than indifferently deferring ICI treatment throughout the pandemic, which may cause signi cant harms to cancer patients in the long run.

Conclusion
In the end, this meta-analysis adds to the growing body of evidence suggesting that ICI was not associated with poor COVID-19 outcomes in cancer patients. Our ndings indicated that the adjournment of ICI treatments during the pandemic is unwarranted, but rather COVID-19 vigilance should be performed more rigorously, especially in patients receiving chemoimmunotherapy. Further studies with larger cohorts and higher quality of evidence are required to con rm our ndings on COVID-19 severity and hospitalization. This project received no speci c grant from any funding agency in the public, commercial, or not-for-pro t sectors.

Con ict of interest
The authors declare no con ict of interest  [23], of which Luo et al [24] was prioritized due to larger sample size. CI, con dence interval; ICI, immune checkpoint inhibitor; OR, odds ratio