If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Department of Thoracic Oncology and CIC1425, Hôpital Bichat-Claude Bernard, Assistance Publique Hôpitaux de Paris, Université Paris-Diderot, Paris, FranceU830 INSERM “Genetics and Biology of Cancers, ART Group”, Curie Institute, Paris, France
Immune checkpoint inhibitors (ICIs) have dramatically improved patient outcomes in a variety of tumor types, but with variable efficacy. Recent research has suggested that antibiotic-induced disruption of the microbiota may impact ICI efficacy. We performed a systematic review and meta-analysis of studies that assessed the impact of antibiotic use on the survival of patients diagnosed with NSCLC and treated with ICI. We systematically searched Medline, the Cochrane Library, and major oncology conferences proceedings. Eligible studies mentioned hazard ratio or Kaplan–Meier curves for progression-free survival (PFS) or overall survival (OS) based on antibiotic exposure before or during ICI treatment. We identified 23 eligible studies. The impact of antibiotics was then evaluated in 2208 patients for PFS and 5560 for OS. For both PFS and OS meta-analyses, the between-study heterogeneity was high (Higgins and Thompson I2 of 69% and 80%, respectively). The pooled hazard ratio was 1.47 (95% confidence interval [CI]: 1.13–1.90) for PFS and 1.69 (95% CI: 1.25–2.29) for OS revealing a significantly reduced survival in patients with NSCLC exposed to antibiotics. The median OS was reduced on average by 6.7 months (95% CI: 5.1–8.4) in the patients exposed to antibiotics. The effect seems to depend on the time window of exposure with stronger effects reported when the patients took antibiotics [−60 days; +60 days] around ICI initiation. In patients with NSCLC, the findings of the meta-analysis indicate that antibiotic use before or during treatment with ICI leads to a median OS decreased by more than 6 months. Specifically, exposure shortly before or after ICI initiation seems to be particularly detrimental, whereas antibiotic use later during disease course does not seem to alter survival. Because PFS and OS were difficult to compare between studies owing to heterogeneity and the multiple confounding factors identified, further studies are needed to strengthen the understanding of this phenomenon.
The efficacy of conventional treatments for metastatic NSCLC (mNSCLC) remains modest. The overall survival (OS) does not exceed 50% at 1 year after standard platinum-based chemotherapies.
Recent advancements in cancer immunotherapies have revolutionized the treatment of advanced and mNSCLC by targeting immune checkpoints, such as programmed death-ligand 1 or its receptor, programmed cell death protein 1 (PD-1). By blocking the immune escape mechanism of the tumor, programmed death-ligand 1 or PD-1 inhibitors have been reported to have fewer side effects and superior efficacy than conventional toxic chemotherapy with median OS exceeding 20 months in the first-line setting in some studies and 1 year survival reaching nearly 70% in combination with platinum-based chemotherapy.
Relative efficacy of checkpoint inhibitors for advanced NSCLC according to programmed death-ligand-1 expression: a systematic review and network meta-analysis.
Pembrolizumab versus docetaxel for previously treated, PD-L1-positive, advanced non-small-cell lung cancer (KEYNOTE-010): a randomised controlled trial.
Atezolizumab versus docetaxel in patients with previously treated non-small-cell lung cancer (OAK): a phase 3, open-label, multicentre randomised controlled trial.
Atezolizumab versus docetaxel for patients with previously treated non-small-cell lung cancer (POPLAR): a multicentre, open-label, phase 2 randomised controlled trial.
Nevertheless, more than 50% of patients who had mNSCLC and were treated with ICI and chemotherapy progressed before 1 year, indicating that tumors of these patients were primarily or became secondarily resistant to anti–PD-1 monoclonal antibodies’ immune-enhancing effect. In addition, the efficacy of these treatments remains poorly predictable for any given patient. As a result, there is an unmet medical need to better understand the mechanisms involved, to develop biomarkers predictive of response to ICI, and to develop therapeutic strategies to potentiate ICI.
Beyond factors classically considered for predicting ICI efficacy, such as the expression of ligands of immune checkpoints, tumor mutational burden, or tumor infiltration by immune cells, the intestinal microbiota has recently emerged as a potential predictor or modulator of response to ICI.
Strikingly, fecal material transferred from patients treated with ICI into germ-free mice modulated the response of mice tumors to anti–PD-1 treatment, echoing the effect witnessed in humans.
Patients who have stage IV NSCLC are particularly at risk for receiving antibiotics during the course of their disease or during the months preceding cancer diagnosis. First, they are mainly former or current smokers. Smoking is known to favor lung infection, because tobacco smoke impairs local epithelial immunity and cilia-induced mucus clearance.
Second, the median age of patients with NSCLC at diagnosis is frequently between 65 and 70 years, possibly supporting higher frailty to infections. Third, tobacco smoking-induced chronic obstructive disease of the small airways leads, by itself, to frequent bronchial infections with cough and chronic expectoration requiring iterative courses of oral antibiotics.
Considering the hypothesized interplay between microbiota and ICI efficacy, studies examined the impact of antibiotic treatments on the survival of patients with cancer, especially those with NSCLC. Most of these studies were performed retrospectively on a limited number of patients. Because of their conflicting results, there is currently no consensus on any significant impact or absence of impact of antibiotics with regard to the survival of patients treated with ICI. Nevertheless, now that a reasonable number of studies on this question have been released, we performed a systematic review and meta-analysis to provide answers to the following research question:
“Among patients diagnosed with NSCLC and treated with ICI, does the use of antibiotics before, during, or after ICI treatment initiation, rather than the absence of use, impact the progression-free survival or overall survival of patients?”
Methods
Search Strategy and Inclusion Criteria
Our systematic review protocol was submitted to the International Prospective Register of Systematic Reviews (PROSPERO CRD42019145675), and we followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines.
We performed a systematic literature search using Medline (through PubMed), the Cochrane Library, and major oncology conferences proceedings for studies reporting the impact of antibiotics on the efficacy of ICI and the survival of patients with cancer treated with ICI. Major oncology conferences included the American Society of Clinical Oncology meetings, the European Society for Medical Oncology meetings, the European Lung Cancer Congress, the World Conference on Lung Cancer, the International Association for the Study of Lung Cancer meetings, the American Association for Cancer Research meetings, and the Society for Immunotherapy of Cancer meetings. We did not apply any filter on the year of publication or language (although the queries were submitted in English). The queries were built using broadly defined medical terms (Supplementary Table 1).
Two investigators (LL and JC) independently searched the databases and conferences proceedings, screened titles and abstracts of the search results, and assessed the full-text articles, posters, and abstracts for eligibility. Any disagreement was resolved with the help of a third author (PAB). The references of the selected articles were also reviewed to detect other studies of interest.
Studies evaluating the impact of antibiotics on types of cancer other than NSCLC or with data aggregating different cancer types were excluded. Studies having the following data were included:
•
Data on the hazard ratio (HR) for OS or progression-free survival (PFS) based on the patients’ use of antibiotics before, during, or after treatment with ICI
or
•
Published Kaplan–Meier curves for OS or PFS based on the patients’ consumption of antibiotics before, during, or after a treatment with ICI.
We did not restrict the systematic review to specific ICI or cancer treatments. In the case of studies having potentially overlapping patient populations, only the largest and most up-to-date study was included.
Data Extraction
From each of the eligible studies, the following data were collected: author’s name, publication or presentation year, country, type of publication (e.g., publication, poster, and abstract), patients and cancer characteristics (e.g., number of patients, frequency of exposure to antibiotics, histology, cancer stage, Eastern Cooperative Oncology Group performance status [ECOG PS]), anticancer treatment characteristics (e.g., line of treatment, ICI received, and treatment scheme), antibiotic treatment (e.g., time window of exposure to antibiotics around ICI treatment initiation, reason for antibiotic use, duration of use, route of administration, and type of antibiotics), median OS and PFS, and HR for PFS and OS. These data were reported in a standardized data extraction spreadsheet for analysis.
When results from both univariate and multivariate analyses were reported, results from the multivariate analysis were preferred. When HR for OS or PFS was not available, it was estimated from the Kaplan–Meier curves using the approach described by Tierney et al.,
The impact of antibiotic use on the efficacy of ICI measured in terms of PFS and OS for patients with NSCLC was analyzed. HR values were used to compare the PFS and OS of patients based on exposure to antibiotic treatments either before, during, or after the start of the ICI treatment compared with no antibiotic use. The weighted average of median PFS and OS reported for patients exposed and not exposed to antibiotics was also computed using the weight attributed in the meta-analysis.
When studies included different time windows of antibiotic exposure, we collected the results for each time window. When pooling studies, we only included one unique time window for each, selected as the longest time window available with data in the study.
We also performed a subgroup analysis to visualize the impact of different time windows of antibiotic exposure on ICI efficacy. Because time windows of exposure vary strongly among studies, we selected cutoff points (1) corresponding to frequently reported periods to keep a high number of studies and (2) discriminating studies investigating antibiotic use strictly before ICI treatment initiation, strictly after ICI treatment initiation, studies with antibiotics before and after, and lastly, studies monitoring the use of antibiotics during the entire ICI treatment (sometimes up to 1 year after ICI treatment initiation). These different studies were split into the following subgroups:
1.
Group A [−90 days; 0]: exposure to antibiotics in the 90 days before ICI treatment initiation;
2.
Group B [−60 days; 60 days]: exposure to antibiotics between 60 days before and 60 days after ICI treatment initiation;
3.
Group C [0; 60 days]: exposure to antibiotics within 60 days after ICI treatment initiation;
4.
Group D [−90 days; ∞]: exposure to antibiotics within 90 days before ICI treatment initiation and during the entire ICI treatment.
In this analysis, the different time windows of antibiotic exposure reported in the same study could be considered for different groups, but each study was represented only once per subgroup.
An analysis of publication bias was conducted using a funnel plot analysis. An influence analysis was also conducted using a leave-one-out approach.
Statistical Analysis
We calculated the pooled HR for PFS and OS using random-effects model. This type of model was chosen to best accommodate the high heterogeneity we expected from the included studies.
The random-effects models were calculated using the Knapp–Hartung method and the Sidik–Jonkman estimator
The systematic search yielded 1449 results: 537 publications from peer-reviewed journals, 760 posters or abstracts, 149 clinical trials, and three Cochrane Reviews. After a first screening, we retained 60 studies (16 publications, 25 posters, and 19 abstracts), evaluating the impact of antibiotics on the survival of patients with cancer treated with ICI with dates of publication ranging from 2017 to 2019, and reporting data from European, North American, South American, and Asian patients. A total of 21 studies were then excluded because they reported results on cancers other than NSCLC or contained results aggregating data relative to different cancers. A total of 16 studies were further excluded either because of redundancies to other studies or because the HR for PFS and OS was neither reported nor assessable. A screening tree of the selection process is displayed in Figure 1.
Figure 1Flow diagram of included and excluded studies. AACR, American Association for Cancer Research; ASCO, American Society of Clinical Oncology; ELCC, European Lung Cancer Congress; ESMO, European Society for Medical Oncology; HRs, hazard ratios; IASCL, International Association for the Study of Lung Cancer; OS, overall survival; PFS, progression-free survival; SITC, Society for Immunotherapy of Cancer; WCLC, World Conference on Lung Cancer.
Finally, 23 studies were included for systematic review (Table 1), that is, nine publications, eight posters, and six abstracts, of which 21 were performed retrospectively. Overall, 2208 patients diagnosed with NSCLC were included in the meta-analysis for PFS and 5560 patients in the meta-analysis for OS.
Table 1Characteristics of the Studies Included in the Meta-Analysis
Barrón F, Arrieta OG, Cardona A, et al. Relevance of antibiotic use on clinical activity of immune checkpoint inhibitors in Hispanic patients with advanced NSCLC (CLICAP-ABs). Poster presented at: 2019 World Conference on Lung Cancer; September 7–10, 2019; Barcelona, Spain.
Chalabi M, Cardona A, Nagarkar D, et al. Effects of antibiotics and proton pump inhibitors in NSCLC patients treated with atezolizumab and docetaxel: pooled analysis of the OAK and POPLAR. Poster presented at: 2018 ESMO Immuno-Oncology Congress; December 13–16, 2018; Geneva, Switzerland.
Negative association of antibiotics on clinical activity of immune checkpoint inhibitors in patients with advanced renal cell and non-small-cell lung cancer.
Fidler MJ, Hogue C, Kuzel T, et al. Impact of antibiotic usage on survival during checkpoint inhibitor treatment of non-small cell lung cancer (NSCLC). Poster presented at: 2019 World Conference on Lung Cancer; September 7–10, 2019; Barcelona, Spain.
Baseline absolute lymphocyte count and ECOG performance score are associated with survival in advanced non-small cell lung cancer undergoing PD-1/PD-L1 blockade.
Khan U, Pena C, Brouwer J, et al. Impact of antibiotic use on response to treatment with immune checkpoint inhibitors. Poster presented at :2019 Gastrointestinal Cancers Symposium; January 17-19, 2019; San Francisco, California.
Kulkarni A, Kumar M, Fellows Pease D, Wang Y, DeFor TE, Patel M. Impact of antibiotics and proton pump inhibitors on clinical outcomes of immune check point blockers in advanced non-small cell lung cancers and metastatic renal cell cancer. Poster presented at: 2019 ASCO Annual Meeting; May 31–June 4, 2019; Chicago, IL.
Metges J, Michaud E, Gabbas M, et al. Impact of anti-infectious and glucocorticoids on immunotherapy : nivolumab and pembrolizumab follow-up for lung cancer in a French study. Poster presented at: ESMO 2018 Congress; October 19–23, 2018; Munich, Germany.
Mielgo Rubio X, Chara L, Sotelo-Lezama M, et al. Antibiotic use and PD-1 inhibitors: shorter survival in lung cancer, especially when given intravenously. Type of infection also matters. Poster presented at: 2018 World Conference on Lung Cancer; September 23–26, 2018; Toronto, Canada
Mielgo Rubio X, Aguado C, Sereno M, et al. Early antibiotic use affects the efficacy of first line immunotherapy in lung cancer patients but route of administration seems to be decisive. Poster presented at: 2019 World Conference on Lung Cancer; September 7–10, 2019; Barcelona, Spain
Riudavets M, Mosquera J, Garcia R, et al. Impact of corticosteroids and antibiotics on efficacy of immune-checkpoint inhibitors in patients with advanced non-small cell lung cancer. Poster presented at: 2019 World Conference on Lung Cancer; September 7–10, 2019; Barcelona, Spain
Rounis K, Papadaki C, Makrakis D, et al. Correlation of various clinical, imaging and laboratory parameters with outcome in patients with metastatic non-small cell lung cancer (NSCLC) treated with immune checkpoint inhibitors (ICIs): results from a prospective, observational, single institution study. Poster presented at: ELCC 2019- European Lung Cancer Congress; April 10–13, 2019; Geneva, Switzerland.
Schett A, Rothschild S, Mauti L, et al. Prognostic impact of the use of antibiotics in patients with advanced non-small cell lung cancer (NSCLC) receiving PD-(L)1 targeting monoclonal antibodies. Poster presented at: ELCC 2019- European Lung Cancer Congress April 10–13, 2019; Geneva, Switzerland.
Thompson J, Szabo A, Arce-Lara C, Menon S. Microbiome & immunotherapy: antibiotic use is associated with inferior survival for lung cancer patients receiving PD-1 inhibitors. Poster presented at: 2017 IASLC 18th World Congress; October 15–18, 2017; Yokohama, Japan.
Tien Phuc D, Madnukeshwar Hedge A, Cherry CR, et al. Antibiotic use and overall survival in lung cancer patients receiving nivolumab. Poster presented at: 2018 ASCO Annual Meeting; June 1–5, 2018; Chicago, IL.
ABX, antibiotics; ASCO, American Society of Clinical Oncology; CI, confidence interval; ELCC, European Lung Cancer Congress; ESMO, European Society for Medical Oncology; HR, hazard ratio; mOS, median overall survival; mPFS, median progression-free survival; OS, overall survival; PFS, progression-free survival; SITC, Society for Immunotherapy of Cancer; WCLC, World Conference on Lung Cancer.
a Start and end dates are given relatively to the initiation of ICI treatment.
The included studies were largely heterogeneous in terms of the reported patients’ characteristics (Supplementary Table 2). Sex, age, tumor histology, stage, and ECOG PS score were not consistently reported. Among the 5745 patients included in the 23 studies selected, there was a slight majority of men (58.3% in the 19 studies reporting this metric; minimum: 40%, maximum: 82%). The median age ranged from 62 years to 75 years (14 studies reported these data). Most patients (71.2%) had a nonsquamous histology versus 28.3% who had a squamous NSCLC (based on 16 studies). A large majority of patients had an ECOG PS score between 0 and 1: only 7.4% of patients had a score of at least 2 (based on nine studies). These patients’ characteristics were consistent with the clinical features and epidemiology of those with advanced NSCLC retrieved from clinical practice,
except for ECOG PS (because a substantial percentage of patients with mNSCLC have a ECOG PS of 2 or 3 at diagnosis).
Although most patients were in their second line of treatment or more, the studies generally aggregated patients treated with ICI along different lines. This is consistent with the history of the approval of ICI drugs. Almost all approved ICI in varying proportions were found: pembrolizumab, nivolumab, atezolizumab, avelumab, durvalumab, tremelimumab, ipilimumab, combinations, and combinations of ICI with chemotherapies or targeted therapies (detailed data in Supplementary Table 3).
The prevalence of antibiotic use ranged from 3% to 80% based on the studies and time window of exposure considered. Reason and duration of treatment and route of administration of antibiotics were unevenly reported. Only eight studies reported the duration of antibiotic treatment (Supplementary Table 3) but without any analysis of the impact of this feature on clinical outcomes. Almost all classes of antibiotics were seen in varying proportions. Beta-lactams and fluoroquinolones were among the most reported molecules, and respiratory tract infections were the most common indications reported for antibiotics. Given the high variability of each study and the nonavailability of individual data, it was not possible to analyze the effects of treatment duration or of specific antibiotic classes or molecules on clinical outcomes.
Effects of Antibiotics on PFS
The HR for PFS was given in 11 studies, and it was assessed from the Kaplan–Meier curves from six further studies. Observed HR ranged from 0.5 to 3.45. All studies were small or medium in size. The heterogeneity factor was high (I2 of 69%), which was expected, given the differing time windows of antibiotic exposures and variability in cancer stages and treatment lines.
The random-effects model yielded a pooled HR of 1.47 with 95% CI of 1.13–1.90 suggesting a significant detrimental effect of antibiotic use around ICI treatment initiation on the PFS of patients (Fig. 2). Using the same weighing of studies, the median PFS was reduced by 1.2 months (95% CI: 0.8–1.7] in patients exposed to antibiotics.
Figure 2Forest plot of hazard ratios for progression-free survival of patients diagnosed with NSCLC and exposed to antibiotics versus not exposed to antibiotics around immune checkpoint inhibitor treatment initiation. ABX, antibiotics; CI, confidence interval; HR, hazard ratio.
HR for OS was given in 18 studies, and it was assessed from the Kaplan–Meier curves from three further studies. Observed HR ranged from 0.5 to 9.3. All studies were small or medium in size except one having nearly 2000 patients. The heterogeneity factor, I2, was 80%, which corresponds to a very high level of heterogeneity.
The random-effects model yielded a pooled HR of 1.69, with 95% CI of 1.25–2.29, suggesting a significant detrimental effect of antibiotic use around ICI treatment initiation on the OS of patients (Fig. 3).
Figure 3Forest plot of hazard ratios for overall survival of patients diagnosed with NSCLC and exposed to antibiotics versus not exposed to antibiotics around immune checkpoint inhibitor treatment initiation. ABX, antibiotics; CI, confidence interval; HR, hazard ratio.
Using the same weighing of studies, the median OS was reduced by 6.7 months (95% CI: 5.1–8.4) in patients exposed to antibiotics.
Variability of Time Windows of Exposure to Antibiotics
All studies exhibited slightly varying time windows of exposure to antibiotics, ranging from 90 days before to 365 days after ICI treatment initiation. As hypothesized in several of the included studies, the time window of exposure to antibiotics seemed to have been a critical factor explaining the different outcomes. Supplementary Figures 1 to 4 present the HR for PFS and OS based on the duration of the time window of exposure to antibiotics. Studies having a very long duration of exposure correspond to studies in which antibiotic use was collected during the entire ICI treatment.
To further explore the hypothesis formulated in multiple studies that the time window of exposure was important, we regrouped study data into four subgroups based on the time windows of antibiotic exposure. The pooled HR for PFS and OS in each subgroup are displayed in Figures 4 and 5. It can be observed that there was no significant impact of antibiotics on efficacy when the time window of exposure included ICI treatment duration (group D). When only considering exposure in the 90 days before ICI treatment initiation (group A), we noticed a higher HR. Nevertheless, it was not significantly different from one. The detrimental effect of antibiotics on ICI efficacy was significant when considering the time windows of [−60 days; +60 days] (group B) and [0; +60 days] (group C) around ICI treatment initiation. From this meta-analysis, we hypothesize that it is particularly critical to use antibiotics during a limited time window shortly before and after ICI treatment initiation, whereas the antibiotic detrimental effects vanish at later times.
Figure 4Forest plot of hazard ratios for progression-free survival of patients diagnosed with NSCLC and exposed to antibiotics versus not exposed to antibiotics, according to the time window of antibiotic exposure. Group A: antibiotic exposure in the following timeframe [−90 days; 0] relative to immune checkpoint inhibitor treatment initiation; group B: [−60 days; 60 days]; group C: [0; 60 days]; group D: [−90 days; ∞]. ABX, antibiotics; CI, confidence interval; HR, hazard ratio.
Figure 5Forest plot of hazard ratios for overall survival of patients diagnosed with NSCLC and exposed to antibiotics versus not exposed to antibiotics, according to the time window of antibiotic exposure. Group A: antibiotic exposure in the following timeframe [−90 days; 0] relative to immune checkpoint inhibitor treatment initiation; group B: [−60 days; 60 days]; group C: [0; 60 days]; group D: [−90 days; ∞]. ABX, antibiotics; CI, confidence interval; HR, hazard ratio.
The included studies had Newcastle–Ottawa scale scores ranging from 4 to 8 (Supplementary Table 4). Funnel plots for PFS and OS are available in Supplementary Figures 5 and 6. We observed a lack of publication in the lower-left corner, implying that studies having a high SE, which demonstrates a positive impact of antibiotics on ICI efficacy, might not have been published.
To assess the variability and consistency of our results, we performed an influence analysis of both OS and PFS with results found in Supplementary Figures 7 and 8. Although some studies influenced much of the overall heterogeneity of the meta-analysis, the overall effects remained largely significant, even when the most powerful analyses were removed.
Discussion
This study reveals that the use of antibiotics around ICI treatment initiation was correlated with a lower OS and PFS of patients. Although some studies presented conflicting results, the magnitude in the survival decrease seems clinically relevant and well attested by numerous studies. During the analysis, we favored HR obtained from multivariate analyses to offset as many confounding factors as possible, including comorbidities, age, sex, concomitant drugs, or other medical variables, in hopes of defining antibiotic use as an independent risk factor.
Our systematic review and meta-analysis can certainly not find causality between antibiotic use and ICI efficacy nor can it elucidate the underlying mechanism of action involved. It can only suggest and contribute to the thought-provoking hypothesis that there may be a link between antibiotics' exposure and ICI efficacy. The principal interest of our study was to place into perspective all findings that we could identify as published in the past 2 years. To our knowledge, this study is among the first systematic reviews of literature and meta-analyses assessing the impact of antibiotics on the survival of patients with NSCLC treated with ICI. It is also the largest curated list of literature on the topic, having regrouped publications in peer-reviewed journals, posters, and abstracts collected from conferences proceedings. The inclusion of abstracts of works not published elsewhere was designed to mitigate (albeit not perfectly) publication bias.
Apart from the growing body of literature on microbiota, all mechanisms of interactions between the microbiota and metabolic functions (e.g., the immune system) are not well known. Although it is well-established that all these functions can be altered when antibiotics are used, the extent to which the antibiotic offense on the microbiota translates into medically meaningful phenomenon is all but clear. Nevertheless, an impact of the antibiotics is supported by elegant experimental data. In tumor-bearing mice, broad-spectrum antibiotics compromised the antitumor effect of anti–PD-1 monoclonal antibodies. In contrast, fecal microbiota transplantation from patients with cancer who responded to anti–PD-1 monoclonal antibodies to antibiotic-treated mice restored the antitumor effects of the PD-1 blockade.
If antibiotics really affect the efficacy of ICI, we can hypothesize that the microbiota might be the missing link.
Many other phenomena may be hidden behind the correlation established herein, and multiple confounders within studies may have influenced the reported outcomes. Patients who received antibiotics certainly suffered from bacterial infections that could also have affected the immune system and ICI efficacy with no role of microbiota. Patients needing antibiotics were mainly treated for respiratory infections, such as pneumonia, which can be serious and even fatal. This could also have distorted the results, because there was no way to dissect whether death was due to comorbidity or cancer without individual patient data and cause of death. Similarly, without individual data, it was not possible to restrict the analysis to the subset of patients who received antibiotics for prophylaxis to assess their effects in the absence of an underlying complicated infection. Patients under antibiotics might also have been weaker, although this was not reflected in the ECOG PS scores, which imperfectly reflects comorbidities. They also may have had a higher tumor burden and a faster tumor progression because of intrinsic genetics or cell biology background. There may have also been a mixture of contributions. For example, liver metastases are often associated with worse prognosis in patients receiving ICI. This is also true for different metastasis sites (not systemically reported) to the disease burden. The role of corticosteroids, known to affect immune response, also remains unresolved, and the studies included in the meta-analysis offered contrasting views on their impact. We also noted previously that patients in different line settings of therapy were often aggregated into the included studies. Nevertheless, it is reassuring to see that the PFS endpoint, which considers not only death but also tumor progression, was also affected by the antibiotic treatment. Furthermore, when it was documented, the overall response rate (assessed based on the Response Evaluation Criteria in Solid Tumors criteria, mainly by investigators themselves rather than centrally and independently assessed) was also lower in patients treated with antibiotics around ICI treatment initiation (data not given).
The main limitation of this work was the strong heterogeneity of included studies, although we included a leave-one-out analysis, indicating that no study was solely responsible for the observed effect. We limited the variability in the definition of time window of antibiotic exposure in the subgroup analysis. Nevertheless, the variability in reported patients’ characteristics, medical history, and treatment characteristics having various combinations of immunotherapies and chemotherapies made the studies difficult to compare, which may have created bias. As an illustration of the high level of heterogeneity, in the study from Rounis et al.,
Rounis K, Papadaki C, Makrakis D, et al. Correlation of various clinical, imaging and laboratory parameters with outcome in patients with metastatic non-small cell lung cancer (NSCLC) treated with immune checkpoint inhibitors (ICIs): results from a prospective, observational, single institution study. Poster presented at: ELCC 2019- European Lung Cancer Congress; April 10–13, 2019; Geneva, Switzerland.
which reported one of the highest HR for OS, patients seemed to have more severe than the average (only stage IV NSCLC, 25% with an ECOG PS ≥ 2), and only antibiotic courses lasting more than 14 days were considered. In addition, in Bagley et al.,
only 61 patients (3%) out of 1960 received antibiotics, which was the lowest rate found in the systematic review. Furthermore, the level of information on the disease state of these patients was lacking. In Derosa et al.,
Negative association of antibiotics on clinical activity of immune checkpoint inhibitors in patients with advanced renal cell and non-small-cell lung cancer.
hospitalizations were more common for patients on antibiotics, which could suggest that those patients’ states were more severe. As a final example, the prospective work of Pinato et al.
revealed strikingly different results than those of other studies, with a difference in OS much larger (2.5 versus 26 months). One may legitimately question whether antibiotic use before immunotherapy initiation could alone explain such a huge shift in outcomes, even when considering the long-lasting nature of antibiotics’ impact on microbiota. More generally, heterogeneity of anticancer treatment was also observed, with some studies including patients treated with ICI in combination with chemotherapy (four studies mentioned it explicitly) or with anti-VEGF therapy (two studies). Nevertheless, this bias was mitigated, because the vast majority of patients included were treated with ICI as a monotherapy. Among the 12 studies offering details on anticancer treatments, 10 included more than 85% of patients having ICI as a monotherapy.
Another limit was the retrospective nature of most studies included for analysis and the lack of individual data. Retrospective studies sometimes lack control over variables that can alter outcomes. The retrospective evaluation of response or PFS by the investigators, and not by an independent panel, could be also highly biased, whereas OS is considered to be a reliable variable. Nevertheless, it could be altered by the influence of subsequent lines of therapies, which were highly variable across countries and studies. Furthermore, most studies included a limited number of patients, a restricted geographic scope with often only one center of recruitment, and a lack of information about other potential gut modulators, such as diet, geography, or comedications other than antibiotics that could be confounding.
In conclusion, this study suggested that the use of antibiotics around ICI treatment initiation was correlated with a decreased survival for patients having NSCLC, which was hypothesized to be a loss of efficacy of ICI. The impact of antibiotics seemed to depend on the timeframe of their administration with respect to ICI treatment initiation, and it seemed to be the strongest for limited time windows shortly before and shortly after ICI treatment initiation. This meta-analysis highlights a need for larger and well-conducted prospective studies evaluating patients’ survival and changes of the intestinal microbiota to further explore this hypothesis and reject the hypothesis of confounding factors that may not have been identified. Basic research is also needed on the possible mechanism of action of hypothesized interactions. Future studies should also further explore the antibiotic effects among patients treated with ICI and chemotherapy during the first line, which is becoming the norm and among patients treated exclusively with chemotherapies or targeted therapies to check whether the correlation between antibiotic use and worse outcomes persists in the absence of immunotherapy. This should further highlight the potential role of confounding factors. Only with more robust medical data, the medical community should be able to proffer the best recommendations regarding antibiotic use for patients having cancer treated with ICI.
Acknowledgments
The study was funded by Da Volterra, France. The authors would like to thank Enago (www.enago.com) for the English language review. Dr. Lurienne, Dr. Cervesi, Dr. de Gunzburg, Dr. Buffet, Mr. Bandinelli, and Prof. Andremont designed the study. Dr. Lurienne, Mrs. Duhalde, Dr. Cervesi, Dr. Buffet, and Mr. Bandinelli ran the systematic search, collected the data, and performed the analyses. Dr. Lurienne, Mrs. Duhalde, Dr. Cervesi, Dr. de Gunzburg, Dr. Buffet, Mr. Bandinelli, Prof. Andremont, and Prof. Zalcman discussed the results and reviewed critically the manuscript.
Relative efficacy of checkpoint inhibitors for advanced NSCLC according to programmed death-ligand-1 expression: a systematic review and network meta-analysis.
Pembrolizumab versus docetaxel for previously treated, PD-L1-positive, advanced non-small-cell lung cancer (KEYNOTE-010): a randomised controlled trial.
Atezolizumab versus docetaxel in patients with previously treated non-small-cell lung cancer (OAK): a phase 3, open-label, multicentre randomised controlled trial.
Atezolizumab versus docetaxel for patients with previously treated non-small-cell lung cancer (POPLAR): a multicentre, open-label, phase 2 randomised controlled trial.
Barrón F, Arrieta OG, Cardona A, et al. Relevance of antibiotic use on clinical activity of immune checkpoint inhibitors in Hispanic patients with advanced NSCLC (CLICAP-ABs). Poster presented at: 2019 World Conference on Lung Cancer; September 7–10, 2019; Barcelona, Spain.
Chalabi M, Cardona A, Nagarkar D, et al. Effects of antibiotics and proton pump inhibitors in NSCLC patients treated with atezolizumab and docetaxel: pooled analysis of the OAK and POPLAR. Poster presented at: 2018 ESMO Immuno-Oncology Congress; December 13–16, 2018; Geneva, Switzerland.
Negative association of antibiotics on clinical activity of immune checkpoint inhibitors in patients with advanced renal cell and non-small-cell lung cancer.
Fidler MJ, Hogue C, Kuzel T, et al. Impact of antibiotic usage on survival during checkpoint inhibitor treatment of non-small cell lung cancer (NSCLC). Poster presented at: 2019 World Conference on Lung Cancer; September 7–10, 2019; Barcelona, Spain.
Baseline absolute lymphocyte count and ECOG performance score are associated with survival in advanced non-small cell lung cancer undergoing PD-1/PD-L1 blockade.
Khan U, Pena C, Brouwer J, et al. Impact of antibiotic use on response to treatment with immune checkpoint inhibitors. Poster presented at :2019 Gastrointestinal Cancers Symposium; January 17-19, 2019; San Francisco, California.
Kulkarni A, Kumar M, Fellows Pease D, Wang Y, DeFor TE, Patel M. Impact of antibiotics and proton pump inhibitors on clinical outcomes of immune check point blockers in advanced non-small cell lung cancers and metastatic renal cell cancer. Poster presented at: 2019 ASCO Annual Meeting; May 31–June 4, 2019; Chicago, IL.
Metges J, Michaud E, Gabbas M, et al. Impact of anti-infectious and glucocorticoids on immunotherapy : nivolumab and pembrolizumab follow-up for lung cancer in a French study. Poster presented at: ESMO 2018 Congress; October 19–23, 2018; Munich, Germany.
Mielgo Rubio X, Chara L, Sotelo-Lezama M, et al. Antibiotic use and PD-1 inhibitors: shorter survival in lung cancer, especially when given intravenously. Type of infection also matters. Poster presented at: 2018 World Conference on Lung Cancer; September 23–26, 2018; Toronto, Canada
Mielgo Rubio X, Aguado C, Sereno M, et al. Early antibiotic use affects the efficacy of first line immunotherapy in lung cancer patients but route of administration seems to be decisive. Poster presented at: 2019 World Conference on Lung Cancer; September 7–10, 2019; Barcelona, Spain
Riudavets M, Mosquera J, Garcia R, et al. Impact of corticosteroids and antibiotics on efficacy of immune-checkpoint inhibitors in patients with advanced non-small cell lung cancer. Poster presented at: 2019 World Conference on Lung Cancer; September 7–10, 2019; Barcelona, Spain
Rounis K, Papadaki C, Makrakis D, et al. Correlation of various clinical, imaging and laboratory parameters with outcome in patients with metastatic non-small cell lung cancer (NSCLC) treated with immune checkpoint inhibitors (ICIs): results from a prospective, observational, single institution study. Poster presented at: ELCC 2019- European Lung Cancer Congress; April 10–13, 2019; Geneva, Switzerland.
Schett A, Rothschild S, Mauti L, et al. Prognostic impact of the use of antibiotics in patients with advanced non-small cell lung cancer (NSCLC) receiving PD-(L)1 targeting monoclonal antibodies. Poster presented at: ELCC 2019- European Lung Cancer Congress April 10–13, 2019; Geneva, Switzerland.
Thompson J, Szabo A, Arce-Lara C, Menon S. Microbiome & immunotherapy: antibiotic use is associated with inferior survival for lung cancer patients receiving PD-1 inhibitors. Poster presented at: 2017 IASLC 18th World Congress; October 15–18, 2017; Yokohama, Japan.
Tien Phuc D, Madnukeshwar Hedge A, Cherry CR, et al. Antibiotic use and overall survival in lung cancer patients receiving nivolumab. Poster presented at: 2018 ASCO Annual Meeting; June 1–5, 2018; Chicago, IL.
Disclosures: Dr. Lurienne, Mrs. Duhalde, Dr. Cervesi, and Mr. Bandinelli report personal fees from Da Volterra and may own or receive shares of Da Volterra, during the conduct of the study. Dr. De Gunzburg reports personal fees from Da Volterra and is a shareholder of Da Volterra, during the conduct of the study, and is an inventor of a European patent, submitted and pending. Prof. Andremont reports personal fees from Da Volterra, during the conduct of the study, and personal fees from BIOASTER, outside of the submitted work, and is a coinventor in multiple pending and issued patents around technologies developed by Da Volterra. Prof. Zalcman reports nonfinancial support from Da Volterra, during the conduct of the study; grants, personal fees, and nonfinancial support from ROCHE; personal fees from MSD; personal fees and nonfinancial support from Bristol-Myers Squibb, Inventiva, and AstraZeneca Pharmaceuticals; and nonfinancial support from Pfizer, AbbVie, and Da Volterra, outside of the submitted work. Dr. Buffet reports personal fees from Da Volterra, during the conduct of the study, and personal fees from ALFA Collaborative group, outside of the submitted work. Mrs. Duhalde, Dr. Lurienne, Dr. Cervesi, Dr. Buffet, and Mr. Bandinelli are employees of Da Volterra. Dr. De Gunzburg, Prof. Andremont, and Prof. Zalcman are consultants for Da Volterra.
30194-5&id=gr1.jpg){kind=link}
30194-5&id=gr2.jpg){kind=link}
30194-5&id=gr3.jpg){kind=link}
30194-5&id=gr4.jpg){kind=link}
30194-5&id=gr5.jpg){kind=link}