Concomitant medications and immune checkpoint inhibitor therapy for cancer: causation or association?

ABSTRACT

The majority of cancer patients assume concomitant medications for the treatment of cancer-related symptoms or co-morbidities. As immune checkpoint inhibitors expand in the treatment of a widening range of malignancies, drug–drug interactions have become an area of increasing interest due to the potential for some concomitant medications to exert immune-modulatory effects and influence outcomes from immunotherapy. Here, we review the evidence supporting this association across selected drug classes including antibiotics, proton pump inhibitors, metformin, and opioids.

Introduction

Immune checkpoint inhibitors (ICIs) have fast risen to the forefront in the therapeutic armamentarium against cancer. After initial evidence of activity in previously untreatable malignancies such as melanoma,¹ monoclonal antibodies inhibiting cytotoxic T-cell lymphocyte-associated antigen-4 (CTLA-4) and the programmed cell death-1 receptor/ligand (PD-1/PD-L1) interaction have swiftly expanded as a therapeutic option to other tumor sites including lung, bladder, renal, head and neck cancer and many others.² As the number of actionable drivers of anti-cancer immunity continues to expand, ICIs are set to play an increasingly relevant role in the management of cancer either as monotherapy or in combination with chemotherapy and targeted agents.³ While capable of inducing cancer immune rejection primarily through the reversal of T-cell exhaustion,⁴ ICIs are not universally effective and response rates are variable. Phenotypic characteristics of the tumor or the host response against malignancy have been shown to enrich for responses to immunotherapy. A number of traits including tumoral PD-L1 expression,⁵ number of non-synonymous somatic mutations,⁶ diversity in the host’s microbiome,⁷ and many others have been evaluated as predictive correlates of response to ICI in clinical trials and in routine practice. However, mechanisms of primary refractoriness and acquired resistance to immunotherapy are still incompletely understood, a factor that limits the use and clinical effectiveness of ICI in the clinic.⁸

Drug–drug interactions (DDI) represent a key area of interest in the context of systemic anti-cancer treatment. Traditionally, desired pharmacodynamic (PD) interactions between different classes of cytotoxic compounds have guided the development of polychemotherapy regimens leading to incremental oncological benefit by overcoming resistance to each chemotherapy class. On the other hand, unwanted DDIs at the pharmacokinetic (PK) level are potentially responsible for toxicity to cytotoxic and targeted anti-cancer therapeutics.⁹ As most monoclonal antibodies, currently available ICIs have a reproducible PK profile¹⁰ that is minimally influenced by concomitant therapies.¹¹

However, in view of the immune-mediated mechanism of action of ICI, considerable interest has been devoted to the investigation of how concomitant therapies with a potential immune-modulatory effect might interact with the PD of ICI, potentially modifying their efficacy or enhancing their toxicity in cancer patients.

In this commentary, we have evaluated five selected drug classes that are commonly administered alongside ICIs whose immune-modulatory effects have been postulated to interact with ICI and influence efficacy and/or toxicity outcomes, namely: antibiotics, corticosteroids, proton pump inhibitors, metformin and opioids.

Antibiotic therapy

The human body and the gut mucosal surfaces, in particular, are inhabited by over 100 trillion commensal bacteria which live in a symbiotic relationship with the host. Taxonomic features of the gut microbiota including diversity and enrichment of Faecalibacterium, Bifidobacterium and Bacteroides spp.^12,13 have been shown to associate with the probability of response to ICI.¹³

Broad-spectrum antibiotic therapy (ATB) can cause long-standing perturbation to the gut microbiota,¹⁴ and in view of the increasing relevance of this domain in cancer immune tolerogenesis ATB has been one of the most exhaustively explored class of concomitant medications in the context of ICI therapy.¹⁵

A vast number of retrospective as well as prospective studies has shown consistent and reproducible evidence of worse outcome in ATB-exposed patients across a range of malignancies and windows of exposure (Table 1). In general, the majority of studies suggest that patients treated with ATB have shorter overall (OS), progression-free survival (PFS) as well as reduced overall response rates (ORR) compared to ATB-unexposed patients.

Table 1.

The relationship between selected classes of concomitant medications and immune checkpoint inhibitors for cancer therapy

Antibiotics
Study	Study design(n)	Cancer type(n)	Immuno-therapy type	PPI window- pre ICI + during ICI	ATB patients (n)	Median OS (months) ATB+ vs ATB-	Median PFS (months) ATB+ vs ATB-	ORR (%) ATB+ vs ATB-
Pinato DJ, et al. (2019)16	Prospective multi-center (196)	NSCLC (119), Melanoma (38), others (39)	Anti PD-L1/PD-1	−30 d and during ICI	15%	2 vs 26 ↓ OS (p < .01)	-	Progression rate 81% vs 44% (p < .001)
Kim H, et al. (2019)17	Retrospective (234)	NSCLC (131), others (103)	Anti PD-1, Anti PD-L1, Anti CTLA-4	−60 d	46.2% (108)	5 vs 17 ↓OS (p < .001)	2 vs 4 ↓PFS (p < .001)	18 vs 26 ↓ORR (p = .038)
Chalabi M, et al. (2020)18	Retrospective (1512)	NSCLC	Anti PD-L1	−30 d and +30 d	22.3% (169)	8.5 vs 14.1 ↓OS (p = .001)
Zhao S, et al. (2019)19	Retrospective (109)	NSLC	Anti PD-1	−30 d and +30 d	18.3% (20)	6.07 vs 21.87 ↓OS (p = .0021	3.73 vs 9.63 ↓PFS (p < .001)	15 vs 22.6 ↓ORR (p = .662)
Ahmed J, et al. (2018)20	Retrospective (60)	NSCLC	Anti PD-1	−14 d and +14 d	28.3% (17)	24 vs 89 ↓OS (p = .003)	↓PFS (significant)	29.4 vs 62.8 ↓ORR (p = .024)
Derosa L, et al. (2018)21	Retrospective (360)	RCC (121) NSCLC (239)	Anti PD-1, Anti PD-L1, Anti-PD1, Anti CTLA-4	−30 d	RCC: 13% (16) NSCLC: 20% (48)	RCC: 17.3 vs 30.6 ↓OS (p = .03) NSCLC: 7.9 vs 24.6 ↓OS (p < .01)	RCC: 1.9 vs 7.4 ↓PFS (p < .01) NSCLC: 1.9 vs 3.8 ↓PFS (p = .03)	NSCLC: 13 vs 23 ↓ORR (P < .01)
Schett A, et al. (2020)22	Retrospective (218)	NSCLC	Anti PD-L1	−60 d and +30 d	15.1% (33)	1.8 vs 15.4 ↓OS (p < .01)	1.4 vs 5.5 ↓PFS (p < .01)	↑rate of disease progression and a lower number of CR/PR
Elkrief A, et al. (2018)23	Retrospective (74)	Melanoma	Anti PD-1, Anti CTLA4	−30 d	13.5% (10)	10.7 vs 18.3 ↓OS (p = .17)	2.4 vs 7.3 ↓PFS (p < .01)
Tinsley N, et al. (2019)24	Retrospective (291)	Melanoma (179) NSCLC (64) RCC (48)	Anti PD-1	−14 d and +42 d	32% (92)	10.4 vs 21.7 ↓OS (p = .002)	3.1 vs 6.3 ↓PFS (p = .003)
Corticosteroids
Study	Study Design (n)	Cancer Type	Immuno-therapy Type	Steroid prescription and window	Early steroid users (n)	Median OS (months) ES+ vs ES-	Median PFS (months) ES+ vs ES-	ORR (%) ES+ vs ES-
Arbor KC et al. (2018)25	Retrospective multicentre (640)	NSCLC	Anti PD-1/PD-L1	Within 30 d or on the day	14% (90)	Center 1: 5.4 vs 12.1 ↓OS (p < .001) Center 2: 3.3 vs 9.4 ↓OS (p < .001)	Center 1: 1.9 vs 2.6 ↓PFS (p < .001 Center 2: 1.7 vs 1.8 ↓PFS (p < .001)	Center 1: 6 vs 19 ↓ORR (p = .02) Center 2: 8 v 18 ↓ORR (p = .2)
Scott SC, et al. (2018)26	Retrospective (210)	NSCLC	Anti PD-1/PD-L1	>10 mg prednisolone	12% (25 patients)	4.3 vs 11 ↓OS (p = .006)
Ricciuti B, et al. (2019)27	Retrospective (650)	NSCLC	Anti PD-L1± CTLA-4	>10 mg prednisolone at the time of immunotherapy within 24 hours	14.3% (93 patients)	4.9 vs 11.2 ↓OS (p < .001)	2.0 vs 3.4 ↓PFS (p = .01)
Fuca G, et al. (2018)28	Retrospective (151)	NSCLC	Anti PD-1	>10 mg prednisolone at the time of immunotherapy within 1 month	23% (35)	4.86 vs 15 ↓OS (p < .001	1.98 vs 3.94 ↓PFS (p = .003)
Proton Pump Inhibitors
Study	Study Design (n)	Cancer Type	Immuno-therapy Type	PPI window	PPI patients (n)	Median OS (months) PPI+ vs PPI-	Median PFS (months) PPI+ vs PPI-	ORR (%) PPI+ vs PPI-
Chalabi M, et al. (2020)18	Retrospective (1512)	NSCLC	Anti PD-L1	Within 30 d prior to and 30 d after starting ICI	30.9% (234)	9.6 vs 14.5 ↓ OS (p < .01	1.9 vs 2.8 ↓PFS (p < .01)
Mukherjee S, et al (2018)29	Retrospective (158)	Melanoma, NSCLC	Anti PD-1/PD-L1	Concomitant	46.2%	4.9 vs 3.4 ↑PFS (p = .77)	↑PFS (p = .77)
Zhao S, et al. (2019)19	Retrospective (109)	NSLC	Anti PD-1	Within 30 d prior to and 30 d after starting ICI		11.9 vs 23.9 ↓OS (p = .754)	9.62 vs 6.23 ↑PFS (p = .343)
Homicsko K, et al. (2018)30	Retrospective (140)	Melanoma	Anti PD-1			↓OS	↓PFS	↓ORR
Metformin
Study	Study Design (n)	Cancer Type	Immunotherapy Type	Control (n)	Intervention (n)	Median OS (months) Met+ vs Met-	Median PFS (months) Met+ vs Met-	ORR (%) Met+ vs Met-
Afzal MZ, et al. (2018)31	Retrospective (55)	Melanoma	Anti PD-1/Anti CTLA-4	No metformin (33)	Metformin twice daily for at least 1 week during ICI use (22)	46.7 VS 28 ↑OS	19.8 VS 5 ↑PFS	68.2 VS 54.4 ↑ORR (p = .31)
Afzal MZ, et al. (2019)32	Retrospective (50)	NSLC	Anti PD-1	No metformin (29)	With metformin for at least 1 week (21)	11.5 VS 7.6 ↑OS (p = .5)	4.0 VS 3.0 ↑PFS (p = .6)	41.4 VS 30.7 ↑ORR (p = .4)

A key challenge in interpreting these data is to disentangle the mechanistic role of ATB exposure from the potential confounding effect of patient’s fitness, co-morbid burden and underlying biology of cancer in dictating outcomes. Patients receiving ATB often have poorer performance status than those who do not,¹⁷ suggesting that ATB use might mask features of more aggressive disease that are likely to influence outcome independent of the underlying microbiome-modulating effect of ATBs. Patients with poorer baseline reserve may be more likely to acquire life-shortening infectious complications.

Timing of ATB therapy appears crucial with regards to their influence on outcomes from ICI therapy. In a previous study by our group,¹⁶ we showed that ATB therapy had a significantly detrimental impact on OS (2 vs 26 months, HR = 7.4; 95% CI: 4.2–12.9; p < .001) only if given within 30 d from ICI commencement, with no significant difference in OS seen in patients who received ATB concurrently with ICI. Importantly, prior ATB therapy was associated with an increased proportion of ICI-refractoriness evaluated radiologically, lending further credence to the hypothesis that ATB might exert a “priming” effect on cancer-specific immunity by disrupting the gut microbiome.³³ To further substantiate the hypothesis of a causative rather than associative link,³⁴ broad-spectrum ATB covering both Gram-positive and negative species have been associated with significantly lower response rates (25% vs 61%, HR = 2.34; 95% CI: 1.5–3.65; p = .02) and shorter PFS (HR = 1.8; 95% CI: 0.86–3.89; p = .012), with no effect seen in patients treated with, narrow-spectrum ATB.²⁰

Interestingly, the duration of ATB appears equally important in influencing patients’ prognosis. Patients receiving multiple ATB courses or a single-prolonged course lasting >7 d had the worst overall PFS (median PFS, 2.8 months; HR, 2.625; p =.026) and OS (median OS, 6.3 months; HR, 1.904; p = .009) compared to ATB-unexposed controls (median PFS, 6.3 months; median OS, 21.7 months). Single, shorter courses of ATB had no significant effect on survival.²⁴

While the current body of evidence uniformly points toward a detrimental effect of ATB in influencing oncological outcomes from ICI, there is virtually no mechanistic evidence to delineate the immune-biologic foundations of such a profound shift in responsiveness and survival following ATB exposure. In absence of more solid evidence, clinicians should promote antimicrobial stewardship in ICI recipients and ensure that ATB prescribing is guided by best evidence, clinical judgment and precautionary principles to avoid potentially detrimental effect on response and survival from immunotherapy.¹⁹ Prospective research addressing the causal roots of this association is key to promote the development of approaches that may favorably manipulate diversity and composition of gut bacteria to synergize with ICIs; for instance, pro-biotics, pre-biotics and fecal microbiota transplantation.⁷

Corticosteroid therapy

ICIs are known to cause a wide range of immune-related adverse events (irAEs) affecting skin, endocrine, pulmonary and gastrointestinal systems through nonspecific dysregulation of self-tolerance leading to immune pathology.³⁵ Immunosuppressive treatment with corticosteroids is the first-line therapy for the majority of irAEs,³⁶ a point that has led to growing concern over the potentially detrimental effect of corticosteroids on outcomes from ICI. In addition to irAEs, indications for corticosteroid use including preexisting co-morbidities such as auto-immune conditions, as well as for alleviation of cancer-related symptoms such as anorexia/cachexia, pain or treatment of central nervous system metastases to reduce peritumoral edema.²⁷ These conditions themselves are often associated with a poorer prognosis, and thus can make it difficult to meaningfully evaluate the effects of steroid use in combination with ICIs due to potentially independent factors contributing to reduced prognosis in patients taking corticosteroids.

Interpreting this association is made even more difficult by the fact that patients assuming ≥10 mg of prednisolone equivalent daily are generally excluded from clinical trials,³⁷ and that the occurrence of irAE per se seems to predict for improved efficacy of ICIs, adding a further layer of complexity to the study of preexisting versus concomitant corticosteroid therapy.^37,38 While it is accepted that corticosteroids act through a wide variety of mechanisms ranging from activation of glucocorticoid response elements (GREs) resulting in inhibition of IL-1 and IL-6 transcription,^39,40 and impairment of the CD28 costimulatory pathway leading to diminished T-cell function,⁴¹ it is not yet fully known the extent to which these mechanisms might affect cancer immune rejection in the context of ICI therapy.

Unsurprisingly, clinical studies in this area have yielded contradicting results (Table 1).

In one of the initial studies, Arbor et al.²⁵ looked at 640 patients with non-small cell lung cancer (NSCLC) and found that baseline use of steroids, defined as ≥10 mg prednisone equivalent, was associated with a reduced overall response rate (6% vs 19%; p = .02), median PFS (1.9 vs 2.6 months; HR = 1.7, p = .001) and OS (5.4 vs 12.1 months; HR = 2.1; p < .001) in patients treated with PD-(L)1 blocking agents.

However, a follow-on study by Ricciuti et al.²⁷ focusing on the indication of corticosteroid therapy found no significant difference in OS between patients receiving <10 mg vs. ≥10 mg prednisone equivalent for cancer-unrelated indications (10.7 v 11.2 months, respectively, p = .77). Confirmation of worse outcome in patients receiving corticosteroids for palliative indications suggests a predominant role of adverse disease features in dictating poorer survival outcomes, as opposed to immune-modulatory effects of corticosteroids.

As clinical evidence continues to evolve, particularly in oncological indications other than NSCLC, the provision of immunotherapy in patients receiving chronic corticosteroid therapy should be the focus of individualized decision-making, taking into account the primary indication for steroid therapy. While corticosteroids do not seem to influence outcome from ICI if prescribed in response to irAEs,⁴² initiation of immunotherapy in immunosuppressed patients yields the potential for underlying immune pathologies to exacerbate, leading to potentially life-threatening complications.⁴³

Proton pump inhibitors

Proton pump inhibitors (PPI) are selective inhibitors of the H⁺/K⁺ ATPase widely used prophylactically and for the treatment of gastrointestinal conditions like peptic ulcers or gastroesophageal reflux disease (GORD).

Their potential to act at the cancer/immunity interface is multifaceted and so is the potential direction of the association with outcomes. It is thought they can cause immune-suppression by way of their ability to reduce the expression of adhesion molecules by inflammatory cells or alter the secretion of pro-inflammatory cytokines amongst other mechanisms.⁴⁴

Alongside direct immune-modulatory consequences thought to negatively influence outcomes to ICIs,⁴⁵ PPI therapy exerts important consequences on the gut microbiome⁴⁶ due to its direct changes to gastric pH and delayed gastric emptying.⁴⁷ PPI use has been shown to significantly decrease the diversity of the gut microbiota and induce both positive and negative selection of specific bacterial species within the gut.⁴⁸

The effects of PPI, however, may not be exclusively detrimental: pre-clinical evidence has highlighted the anti-tumor potential of this drug class as a result of their capacity to neutralize the acidic tumor microenvironment, a concept that lacks full validation in patients.⁴⁹

Retrospective analyses on two randomized control trials of atezolizumab in NSCLC (POPLAR and OAK) assessed the impact of ATB and PPI had on survival outcomes for patients who were randomly assigned between ICI and chemotherapy. For the ICI patients, OS survival was significantly shorter in the PPI receivers (9.6 versus 14.5 months, HR 1.45, 95% CI 1.20–1.75, p = .0001) as was PFS (1.9 versus 2.8 months, HR 1.30, 95% CI 1.10–1.53, p = .001). Interestingly, PPI use was associated with a greater risk of progression or death in the immunotherapy but not in the chemotherapy control arms, corroborating a specific immune-modulatory action of PPI in the context of ICI.¹⁸

Similar results were observed in patients with melanoma treated with PD-1-targeted therapies, where PPI use was associated with reduced PFS, OS and response rates. Analysis of pre-treatment samples showed PPI treated patients to have increased neutrophil and lymphocyte counts as well as higher NCAM1 and CSF3-R levels, suggesting a prognostically adverse pro-inflammatory status in these patients which is postulated to link PPI use with reduced efficacy of ICI therapy.³⁰

Much more limited evidence exists in other cancer types: a smaller retrospective study of 95 patients of mixed histologies (24% head and neck, 18% gastric) showed tumor-specific differences in response but not in PFS and OS.⁴⁵

Robust recommendations for PPI use cannot be inferred given the retrospective nature of currently available evidence. Heterogeneity in dosing or PPI type prescribed has often been disregarded.²⁹ Some studies were underpowered to look at differences in outcomes across different cancer types and many studies did not look at the effects of other influencing factors, for instance, co-prescription of corticosteroids or antibiotics. The association between PPI use and the risk of nephrotoxicity from ICI therapy is of interest and warrants further evaluation in prospective studies.⁵⁰

Metformin

Metformin is the drug of choice for patients with type 2 diabetes and is a complex regulator of multiple signaling pathways which result in increased insulin sensitivity and reduce liver gluconeogenesis.³¹ Its role as an immune-modulatory agent capable of reprogramming the metabolism of the tumor microenvironment has been of great interest. Metformin inhibits complex 1 of the mitochondrial electron transport chain, an essential metabolic source for cancer proliferation.⁵¹ Via activation of the LKB1/AMPK pathway, metformin inhibits mTOR signaling and down-regulates pro-inflammatory pathways, making the tumor microenvironment less tolerogenic via downregulation of PD-L1,⁵² and over-expression of major histocompatibility complex I.⁵³ Metformin can also reduce immune exhaustion by increasing IL-10 production to enhance CD8 + T-cell function,⁵² and concurrently inhibit CD39/CD73 on myeloid-derived suppressor cells.³² Other favorable properties include shifts in the gut microbiota composition through alteration of microbial folate and serine/methionine metabolism,⁵² and reduction of intra-tumoral hypoxia.⁵⁴

Despite the multi-faceted anti-tumor properties of metformin, clinical studies have been disappointing in confirming a beneficial effect. In a retrospective study of 50 NSCLC patients metformin use correlated with numerically longer median OS (11.5 vs 7.6 months, p = .5), PFS (4 vs 3 months, p = .6) values and higher response rates (41.1 vs 30.7%, p = .4) in patients administered both metformin and an anti-PD-1 inhibitor in combination vs those only receiving ICIs, although these did not meet statistical significance.³² A larger study of 224 patients confirmed the lack of association.⁵⁵ The ongoing open-label phase Ib trial by Kubo et al.⁵⁶ is evaluating metformin concomitantly with nivolumab as a strategy to overcome immunotherapy-refractoriness.

Opioids

Opioid use can influence many physiological functions because of its varied receptor locations; its primary analgesic effect is achieved by targeting μ receptors within the central nervous system. Another target is on the receptors found in the enteric nervous system which is what coordinates action within the gut. This influence on gastrointestinal function has many effects, one such example that is widely observed being opioid-induced constipation and nausea.⁵⁷

Opioids may induce significant changes to the gut microbiota,⁵⁸ an observation supported by clinical and pre-clinical evidence.^59–61 Opioid use can also impair the gut barrier function and induce gut bacterial translocation, triggering downstream inflammatory responses.⁶²

Evidence around opioid use is limited, with only one study evaluating the effects of opioids on the clinical outcome of ICIs, albeit as a secondary aim. It was found that its use was associated with a significantly lower PFS and OS (median PFS 4.5 vs. 8.1 months, HR 1.79, 95% CI 1.12–2.85, p = .010; median OS 8.6 vs. 26.3 months, HR 3.08, 95% CI 1.71–5.55, p < .001) compared to those patients who did not use opioids.⁶³

While interesting, the immune-mediated contribution to the adverse prognosis of ICI recipients who are concurrently on opioids must be weighed against the prognostically adverse value of a symptomatic oncological diagnosis, which might have triggered opiates use in the first place and confounded the association.

Conclusion

Concomitant medication use is common in cancer patients and the postulated immune-modulatory effects of certain drugs have led to an increased interest in DDIs in the context of ICI therapy. While biologic plausibility behind an immune-mediated influence on outcomes from ICI can be inferred for the majority of drug classes described here, the strength and robustness of the association vary across DDI and tumor type considered. Even in the case of antibiotics, where evidence of an adverse interaction is stronger, the debate between causal versus associative relationship has not been fully resolved. As the knowledge around DDI in the context of ICI therapy expands, clinicians should use this information to carefully review the appropriateness of prescribing on a case by case basis to ensure optimal efficacy of ICI is maintained without detriment to the treatment of cancer-related symptoms or co-morbidities.

Funding Statement

DJP is supported by grant funding from the Wellcome Trust Strategic Fund [PS3416].

Disclosure of potential conflict of interest

DJP received lecture fees from ViiV Healthcare, Bayer Healthcare and travel expenses from BMS and Bayer Healthcare; consulting fees for Mina Therapeutics, EISAI, Roche, Astra Zeneca; received research funding (to institution) from MSD, BMS. There are no other personal or financial conflicts of interest to disclose.

Authors’ contributions

Study concept and design: DJP

Acquisition of data: NH, MN

Analysis and interpretation of data: NH, MN, DJP

Drafting of the manuscript: NH, MN, DJP

Critical revision of the manuscript for important intellectual content: All the authors.

Statistical analysis: N/A

Obtained funding: DJP

Study supervision: DJP