Active and latent tuberculosis infections in patients treated with immune checkpoint inhibitors in a non-endemic tuberculosis area

Abstract

Background

Immune checkpoint inhibitors (ICIs) blocking inhibitory immune pathways (e.g., programmed cell death protein-1/-ligand1 [PD-1/PD-L1]) have revolutionized cancer therapy for numerous malignancies. There have been an increasing number of cases of active tuberculosis (TB) reported in association with ICI use, and recent data suggest alterations in immune responses in TB by ICI. The aim of this study was to characterize the frequency of latent tuberculosis infection (LTBI) and active TB in a large cohort of ICI-treated patients in a low TB incidence area.

Methods

We conducted a retrospective review of all ICI-treated patients tested for TB between January, 1997 and August, 2018. Data extracted included patient demographics, TB risk factors, latent/active TB diagnosis and treatment, tumor type, ICI used, immunosuppressive medications, and mortality related to TB.

Results

We identified 1844 ICI-treated patients, including 30 abnormal TB test results. Two patients were diagnosed with active TB, both prior to starting ICI therapy. One patient was treated for TB prior to starting ICI and the other patient was successfully treated concurrently. Seven patients were diagnosed with LTBI and none developed active TB. Twenty patients had indeterminate interferon gamma release assays (IGRA).

Conclusion

Despite recent reports of TB in patients taking ICI, we found no patients developing TB during ICI therapy in our large retrospective cohort of ICI-treated cancer patients in a non-endemic TB area. The high rate of indeterminate IGRA results suggests the need for prospective research with better diagnostics to quantify the actual risk of TB in this patient population.

Introduction

Immune checkpoint inhibitors (ICIs) are being increasingly used to treat various malignancies across all disease stages. One of the most clinically relevant inhibitory immune pathways is the programmed death-1 (PD-1) receptor/programmed death ligand-1 and -2 (PD-L1 and PDL-2) pathway. The PD-1/PD-L1 pathway is involved in maintenance of self-tolerance by downregulating normal immune responses [1, 2]. Clinically used ICIs are antibodies blocking either the PD-1/PD-L1 axis or cytotoxic T-lymphocyte associated protein-4 (CTLA-4) in order to activate antitumor immunity in numerous cancer types, including melanoma, non-small cell lung cancer, and lymphoma [3–6].

Immune-related adverse reactions or events (iAEs) occur when ICIs induce autoimmune responses and are well-known side effects of these medications [7, 8]. The management of ICI iAEs commonly requires systemic immunosuppression with steroids. Unlike traditional chemotherapeutic agents, ICIs do not cause immunosuppression, but rather immune activation making infectious complications theoretically less likely than during standard cytotoxic cancer therapies. Although various infectious complications have been reported, it remains unclear whether these infections are complications of ICI or concomitant immunosuppression used to manage iAE [9].

The role of the PD-1 pathway in host–pathogen interactions is not fully understood, and data have shown mixed effects of PD-1-pathway inhibition for various chronic infections [10, 11]. Mycobacterium tuberculosis (Mtb) represents a pathogen of significant clinical and epidemiologic importance in cancer patients worldwide. There have been a small, but growing number of reported cases of TB in ICI-treated cancer patients [12–26]. Furthermore, a recent review of reported adverse events to the FDA found 72 cases of TB following ICI use, most of which were in patients treated with nivolumab [27].

While these TB case reports are concerning, the development of active TB is somewhat counterintuitive in patients taking “immune-activating” medications. Possible mechanisms to explain the development of TB in ICI-treated cancer patients include direct disruption of cellular immune response against TB by concurrent medications (e.g., cytotoxic chemotherapy or steroids), cancer-induced immunosuppression, or a direct effect of ICIs [28, 29]. Alternatively, patients could develop an immune reconstitution-like reaction to subclinical TB infections when taking ICIs, similar to that developed in AIDS patients during the recovery of their CD4⁺ T-cell population [14–17, 30].

Within this context, our study aims to further characterize the frequency of latent tuberculosis infection (LTBI) and clinical characteristics of ICI-treated patients suspected to have active TB in a high-volume referral center for cancer patients in Minnesota, a low TB incidence area (3.1 new TB cases per 100,000 population in 2018) [31]. Herein, we report the results of the largest retrospective study to date to characterize the prevalence, clinical characteristics, and outcomes of ICI-treated cancer suspected of having LTBI or active TB at Mayo Clinic.

Materials and methods

We conducted a retrospective search of the electronic medical records to identify all patients who received ICIs and underwent TB testing (sputum, skin, blood, urine) at Mayo Clinic (Rochester, Minnesota) between January 1, 1997 and August 31, 2018. First, all patients who had an ICI mentioned in the medical record were identified. Using this patient cohort, we then identified all of those patients who had any TB testing performed during the study period. We included patients as early as 1997 to capture patients with abnormal TB testing years prior to ICI administration. Patients treated with ICI and who had abnormal TB testing, including those with indeterminate IGRA results, were included after chart review (Fig. 1).

Fig. 1.

Study Enrollment flowchart

Data extracted included patient demographics, country of origin, TB test results, TB risk factors (smoking status, HIV status, healthcare worker), BCG status, treatment of latent or active tuberculosis, tumor type, ICI used and duration of use, iAEs, presence/absence of immunosuppression, and mortality related to TB. We estimated the theoretical risk of TB reactivation for patients with positive tuberculin skin test (TST) and/or interferon gamma release assay (IGRA) results by using McGill University’s “The Online TST/IGRA Interpreter” multifactorial prediction modeling tool [32]. Patients with discordant IGRA test results (positive test followed by negative test) were excluded unless there was clinical information supporting a diagnosis of LTBI or TB. The study was approved by the Mayo Clinic Institutional Review Board (# 18-007227).

Results

Our electronic medical records search yielded 7474 potential cases, including 1844 patients who physically received ICI therapy at Mayo Clinic, Rochester. The latter cohort primarily included local patients from South Eastern Minnesota, while the remaining patients reflect Mayo Clinic’s referral population. We identified 30 patients treated with ICI and who had abnormal TB testing (Fig. 1). The average age at the time of TB testing was 58 years (range: 21–81 years). The majority were White/Caucasian (n = 27) men (n = 23) and either current or former smokers (Table 1).

Table 1.

Summary of patient demographics and clinical characteristics

	Number (%)	TB or LTBI (%)	Indeterminate IGRA (%)
Number of patients	30	8	20
Gender
Female	7 (23.3)	2 (25)	5 (25)
Male	23 (76.7)	6 (75)	15 (75)
Race/ethnicity
White/Caucasian	27 (90)	6 (75)	20 (100)
Hispanic/Latino	1 (3.3)	1 (12.5)	0
Asian	1 (3.3)	1 (12.5)	0
Middle Eastern	1 (3.3)	0	0
High risk birth country	1 (3.3)	1 (12.5)	0
BCG vaccination	3 (10)	3 (37.5)	0
Tobacco use
Current	6 (20)	2 (25)	8 (40)
Former	13 (43.3.)	4 (50)	8 (40)
Never	11 (36.7)	2 (25)	4 (20)
HIV status
Positive	0	0	0
Negative	17 (57.6)	5 (62.5)	11 (55)
Unavailable	13 (43.3)	3 (37.5)	9 (45)
Healthcare worker
Yes	1 (3.3)	1 (12.5)	0
No	30 (26.7)	7 (87.5)	20 (100)
Average age (at time of TB testing) (years)	58.2 (range: 21–81)	54.3	59.7

Lung cancer, melanoma, and lymphoma comprised the majority of tumors treated (n = 11, 7, and 7, respectively), and pembrolizumab was the most commonly prescribed ICI (n = 16) (Table 2).

Table 2.

Summary of the frequency of different immune checkpoint inhibitors, treated malignancies, and immune-related adverse side effects

	Number (%)
Immune checkpoint inhibitors used	33
Atezolizumab	3 (9.1)
Ipilimumab	4 (12.1)
Ipilimumab + Nivolumab	3 (9.1)
Nivolumab	7 (21.2)
Pembrolizumab	16 (48.5)
Tumors treated	30
Lung	11 (36.7)
Melanoma	7 (23.3)
Lymphoma	7 (23.3)
Renal cell carcinoma	2 (6.7)
Fallopian tube carcinoma	1 (3.3)
Adrenocortical carcinoma	1 (3.3)
Prostate	1 (3.3)
Immune-related adverse events (iAE)	9
Pneumonitis	3 (33.3)
Colitis	2 (22.2)
Hepatitis	1 (11.1)
Thyroiditis	1 (11.1)
Hypophysitis	1 (11.1)
Inflammatory arthritis	1 (11.1)

Twenty-nine patients underwent IGRA testing, 27 of whom had IGRA testing alone and 2 patients had both TST and IGRA testing. IGRA test results showed 8 positive tests and 1 negative test (had a positive TST). Twenty patients had indeterminate IGRA results due to poor mitogen response (Table 3). Fourteen of the patients with indeterminate IGRA testing had undergone immunosuppressive therapy 3 months prior to, during, or after TB testing.

Table 3.

Summary of diagnostic testing results for latent tuberculosis infection

TB test performed	Number (%)
Tuberculin skin test	3
Positive	2 (66.7)	Average size 18 mm (17, 19)
Negative	1 (33.3)
IGRA	29
Positive	8 (27.6)
Indeterminate	20 (69)	IGRA testing relative to ICI (11 prior to, 1 during, 8 after initiation ICI therapy)
Negative	1 (3.4)

The 2 patients with both TST and IGRA testing had discordant results: one patient with a positive TST and negative IGRA, and the other with negative TST and positive IGRA. Both tests were completed within 10 days in both cases. The patient with a negative IGRA test followed by a positive TST test had unknown BCG vaccination status, but had emigrated from Iraq through Jordan. He did not have a prior diagnosis of LTBI or active TB and was not diagnosed with LTBI (Table 4).

Table 4.

Summary of treatment and outcomes of patients with latent tuberculosis infection and tuberculosis

	Risk of developing active TB (annual/cumulative)*	ICI therapy/anti-TB treatment	Outcome
Latent tuberculosis infection
Subject 6	1.23% per year/1.23%	Pembrolizumab/none	Died secondary to malignancy 2 years later without development of TB
Subject 8	0.1% per year/5.9%	Nivolumab/isoniazid (INH) × 14 months	Subsequently treated to nivolumab for 5 years without development of TB Follow-up duration: 20 years
Subject 12	1.99% per year/30%	Atezolizumab/INH × 6 months	Subsequently treated with ICI for 3 months Died 1 year after LTBI diagnosis secondary to malignancy
Subject 13	1.24% per year/19.8%	Nivolumab/None	Not treated due to age and co-morbidities Died secondary to malignancy 6 months later
Subject 14	0.73% per year/23.93%	Atezolizumab/INH × 1 month	Follow-up IGRAs,  including T-SPOT.TB test, were negative and initial test was questioned as false positive Died 4 years later secondary to malignancy
Subject 15	1.47% per year/4.41%	Nivolumab/INH × 9 months	Nivolumab for 3 months No recurrence of TB after 1 year of follow-up
Subject 26	1.95% per year/62.4%	Pembrolizumab/none	Pembrolizumab for > 1 year after positive IGRA No TB reactivation after 1 year of follow-up
Active tuberculosis
Subject 18		Nivolumab + Ipilimumab/INH, pyrazinamide, ethambutol, moxifloxacin (then INH and rifampin, later rifabutin) × 12 months	Treated with nivolumab for 5 months Died 3 years later secondary to malignancy No recurrence of TB
Subject 25		Pembrolizumab/INH, rifampin for at least 5 months	TB managed at outside facility Tolerated pembrolizumab for 6 months several years after treatment No progression or recurrence of TB at 3 years of follow-up

*Estimates are based on McGill University’s “The Online TST/IGRA Interpreter” multivariate predictive modeling tool. The cumulative risk estimate of TB reactivation is defined as up to the age of 80 years old

No patient developed active TB during or after receiving ICIs. Two patients treated with ICI during the study period were diagnosed with active TB prior to starting ICIs. One patient was diagnosed with TB just prior to starting ICI therapy (nivolumab and ipilimumab) and was successfully treated with concurrent TB therapy. The other patient was also diagnosed with active TB prior to ICI therapy and TB treatment was started at their home institution in Southeast Asia based on imaging and clinical suspicion of TB, but in the absence of microbiologic confirmation of active TB. The patient was treated with 3 anti-TB drugs for 2 months followed by rifampin and isoniazid. Additional details regarding his treatment course were unavailable. He subsequently returned to our institution and received pembrolizumab without documented recurrence of TB (Table 4).

Seven patients were diagnosed with LTBI and 5 started LTBI therapy, however only 3 patients completed the treatment. One patient was lost to follow-up, one patient was not treated for unclear reasons, and another patient discontinued therapy after 1 month due to liver enzyme abnormalities (Table 4).

Discussion

In this large retrospective study, we investigated the frequency, clinical presentation, and outcomes of LTBI and active TB in patients treated with ICIs at a high-volume referral center. A small number of patients were diagnosed with active TB or LTBI in an area with a low-incidence of TB. All patients with LTBI and active TB were diagnosed prior to ICI treatment. We did not identify any patients with LTBI and TB reactivation or active TB infection during or after ICI therapy in our patient cohort. Nevertheless, the high frequency of indeterminate IGRA results due to low-mitogen control responses in cancer patients precludes the exact determination of the number of LTBI patients in our study. However, and most importantly, there were no instances of active TB infection after receiving ICIs, no deaths related to TB, and all TB patients were able to successfully complete anti-TB therapy while on ICI drugs.

The high rate of indeterminate IGRA results is by itself an interesting finding for this population of cancer patients. Out of 397 IGRA tests, 42 were indeterminate (10.6%) (20 of whom met criteria for study inclusion). Studies in patients with TB and HIV co-infection have shown the lower the CD4 counts, the higher the rate of indeterminate results in IGRAs [33]. Our study’s high rate of indeterminate results seems even higher than the rates reported in immunosuppressed patients with HIV and on biologic medications [33, 34]. The reason for the unexpected IGRA results remains unclear and our data suggest that better diagnostic testing for LTBI is needed for cancer patients.

There may be multiple reasons for the increasing number of anecdotal reports of patients developing TB while on ICI. First, these agents are being increasingly used in areas with high prevalence of TB and in conjunction with other immunosuppressive medications, including cytotoxic chemotherapy agents. Along these lines, corticosteroids are also commonly used for the management of iAE in ICI-treated patients, which can independently increase the risk for developing TB reactivation in LTBI and progression to primary TB in this already immunocompromised population of cancer patients.

In addition, there are probably other immunologic factors contributing to the development of TB in these patients which requires further investigation. Studies in animal model and ex-vivo studies suggest that both a deficient and excessive immune response to Mtb infection is detrimental to disease control and outcomes [35–42]. Moreover, recent data suggest that ICI can result in tuberculosis reactivation via dysregulation of TNF-α, IL-18 and IFN-γ in granulomas and increased bacterial loads in the subset of granulomas with increased inflammatory changes and caspase 1 activation [28, 29]. In addition, alterations of the intestinal microbiome, specifically an abundance of Ruminococcaceae and other microbiota, have been associated with an increased risk of developing TB in a macaque model treated with ICI, which was started early in the infection process [29]. Consequently, the timing of ICI therapy in relationship to TB infection may be critically important in determining the effectiveness of the host immune response against Mtb infection in these cancer patients.

Taken as a whole, this data support the concept that the immune system has a spectrum of activity that changes over time and must balance pathogen control and immune-mediated pathology to optimally protect the host against Mtb infections [43]. Thus, the dynamic immune response to Mtb is complex and blockade (or at least partial blockade) of the PD-1 and cytotoxic T-lymphocyte associated protein-4 (CTLA-4) pathways pharmacologically may have multiple concurrent effects on human response to Mtb infections with variable infection outcomes [44].

However, our data suggest that TB reactivation represents a rare phenomenon in patients in a low TB incidence area and that the accumulating reports may represent unusual events that are occurring due to multiple factors that might not be necessarily associated directly to ICI effect.

To support this interpretation of our study findings, a recent report from South Korea found an incidence of TB to be 0.3% in all cancer patients taking ICI, whereas a study done in Japan demonstrated an even higher incidence of 1.7% in patients with lung cancer [45, 46]. While South Korea and Japan are high and medium-incidence countries for TB, respectively, a study done in low-incidence country (Germany) showed none of the 70 patients with melanoma and metastatic Merkel cell carcinoma treated with ICIs developed TB, despite 9 patients with positive IGRAs in that cohort [47].

Our study has several limitations, including the retrospective nature of data collection and the small number of TB tested patients. Furthermore, the actual number of patients with LTBI in our institution taking ICI cannot be accurately determined since LTBI screening is not routinely performed in this patient population. There was also a high proportion of indeterminate IGRA results, which also precluded us from having a more accurate estimation of LTBI in our study cohort. Additionally, our study had no comparison group and probably included mostly patients at low-risk for TB exposures.

In conclusion, the risk for TB reactivation and acute TB infection appears to be low in large referral centers located in a low TB incidence area. However, future studies looking at large numbers of untreated vs. treated LTBI patients receiving ICI and followed for a prolonged period of time would be needed to estimate the effect of ICI immunotherapies in TB reactivation. Although screening for LTBI in patients starting TNF-alpha inhibitors is recommended, the role of routine screening for LTBI in patients taking ICI in low TB incidence settings remains unclear. Furthermore, improved diagnostics would be necessary to accurately diagnose LTBI in this patient population with a high frequency of low-mitogen indeterminate IGRA results.

Abbreviations

TB

Active tuberculosis

iAE

Immune-related adverse event

ICI

Immune checkpoint inhibitor

IGRA

Interferon gamma release assay

LTBI

Latent tuberculosis infection

Mtb

Mycobacterium tuberculosis

PD-1

Programmed cell death protein-1

PD-L1

Programmed death ligand 1

TST

Tuberculin skin test

Funding

The authors did not receive support from any organization for the submitted work.

Declarations

Conflict of interest

GRS has no relevant financial or non-financial interests to disclose. TP participated in Advisory Boards for AstraZeneca and Novocure, outside the scope of the submitted work. All fees were paid to Mayo Clinic. PE participated in a short-term advisory scientific board for DiaSorin Molecular, which was outside the scope of the submitted work. Honorarium was paid to Mayo Clinic. PE has also filed two patent applications related to immunodiagnostic laboratory methodologies for latent tuberculosis infection for intellectual property protection purposes. To date, there has been no income or royalties associated with those filed patent applications. Otherwise, PE does not have other conflicts of interest COI to disclose. This paper’s contents are solely the responsibility of the authors and do not necessarily represent the official views of the National Institutes of Health or Mayo Clinic. No other financial or material support for this work was provided to the authors.