Antidepressant Use and Lung Cancer Risk and Survival: A Meta-analysis of Observational Studies

Recent preclinical studies have linked antidepressants (AD) to their potential anticancer effects in multiple cancers, but the impact on lung cancer remains unclear. This meta-analysis examined the associations between ADs and lung cancer incidence and survival. The Web of Science, Medline, CINAHL, and PsycINFO databases were searched to identify eligible studies published by June 2022. We conducted a meta-analysis using a random-effects model to compare the pooled risk ratio (RR) and 95% confidence interval (CI) in those treated with or without ADs. Heterogeneity was examined using Cochran Q test and inconsistency I2 statistics. The methodologic quality of the selected studies was assessed using the Newcastle-Ottawa Scale for observational studies. Our analysis, including 11 publications involving 1,200,885 participants, showed that AD use increased lung cancer risk by 11% (RR = 1.11; 95% CI = 1.02–1.20; I2 = 65.03%; n = 6) but was not associated with overall survival (RR = 1.04; 95% CI = 0.75–1.45; I2 = 83.40%; n = 4). One study examined cancer-specific survival. Subgroup analysis showed that serotonin and norepinephrine reuptake inhibitors (SNRIs) were associated with an increased lung cancer risk by 38% (RR = 1.38; 95% CI = 1.07–1.78; n = 2). The quality of selected studies was good (n = 5) to fair (n = 6). Our data analysis suggests that SNRIs were associated with an elevated risk of lung cancer, raising concerns regarding the use of AD treatment in patients vulnerable to lung cancer. The effects of ADs—particularly SNRIs—and their interplay with cigarette use and lung cancer risk in vulnerable patients merits further study. Significance: In this meta-analysis of 11 observational studies, we found evidence of a statistically significant association between the use of certain ADs and lung cancer risk. This effect merits further study, particularly as it relates to known environmental and behavioral drivers of lung cancer risk, such as air pollution and cigarette smoke.


Introduction
Lung cancer is the second most common cancer and the leading cause of cancer-related deaths worldwide, accounting for approximately 11% of newly diagnosed cancers and 18% of cancer deaths in 2020 (1). Current lung cancer treatments include surgery, chemotherapy, radiotherapy, and targeted therapy depending on several factors including the stage and the type of lung cancer suppressed the growth of NSCLC, gastric cancer, melanoma, and hepatocellular carcinoma by inhibiting the mTOR activity (8)(9)(10)(11)(12).
While these tumor-suppressing effects of ADs are encouraging, tumorpromoting results have also been reported in the literature. For example, amitriptyline and fluoxetine promoted the development of fibrosarcoma and melanoma in rodents at clinically relevant doses (13). However, conflicting results have been obtained from epidemiologic studies. A meta-analysis for colorectal cancer reported a reduced risk with TCA use (14), while another metaanalysis for ovarian cancer showed no associations with either SSRI or TCA use (15). In addition, the effect of these ADs on lung cancer risk in humans is still unclear, and none of the studies evaluated the association between serotonin and norepinephrine reuptake inhibitors (SNRIs) and lung cancer outcomes.
Considering the growing use of ADs for the treatment of other conditions such as chronic pain (16) and smoking cessation (17) as well as depression and the high prevalence of depression among patients with lung cancer (18,19), we performed a meta-analysis to summarize the currently available evidence on the relationship between AD use and lung cancer risk in the general population, and survival in patients with lung cancer.

Materials and Methods
This meta-analysis was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) and was registered in the International Prospective Register of Systematic Reviews (PROSPERO; University of York; York, United Kingdom) platform with the registration number CRD42022350719. As this study is a summary of previously published studies, no ethical review is required.

Search Strategy
A comprehensive search was conducted using key terms to maximize the identification of the studies examining the association between AD use and lung cancer incidence and survival. Online search platforms, EBSCOhost (Medline, CINAHL, and PsycINFO) and Web of Science, were searched from inception to June 2022. The search terms included: lung cancer or lung neoplasms or lung tumor or lung adenocarcinoma or NSCLC or non-small cell lung cancer or SCLC or small cell lung cancer AND antidepressant(s) or anti-depressant(s) or AD medication or SSRI or selective serotonin reuptake inhibitor(s) or SNRI or serotonin-norepinephrine reuptake inhibitor(s) or TCA or tricyclic AND mortality or mortality rate or death or death rate or survival rate or survival outcomes or clinical outcomes or incidence or risk. Filters included human studies and articles published in peer-reviewed journals. We also reviewed the reference lists of the included studies and prior meta-analyses and reviews to identify additional publications that could be eligible. The full search strategies are shown in Supplementary Table S1.

Inclusion and Exclusion Criteria
The search results from four databases were first uploaded to Covidence software (Veritas Health Innovation; ref. 20), and two reviewers independently screened articles according to the Participants, Intervention/Exposure, Comparison, Outcome, and Study design guidelines outlined in Supplementary   Table S2. The population included adults (adults without cancer for lung cancer risk studies or adult lung cancer survivors for survival outcomes) with exposure to AD use. The outcomes included lung cancer incidence or survival from observational studies, including case-control or cohort studies. We excluded review literature, meta-analysis, newsletters, conference abstracts, and animal studies. We also excluded literature evaluating survival relating to AD use in the general population, not from lung cancer survivors.

Data Extraction and Methodologic Quality Evaluation
The quality of the studies was assessed using the Newcastle-Ottawa Scale for observational studies (21). This tool evaluates the quality of case-control and cohort studies on the following three domains regarding (i) the adequacy of the recruitment and selection of study participants, (ii) the comparability of comparison groups, and (iii) the ascertainment of exposure and outcomes. Studies that received a score of 7-9 were considered good, 4-6 as fair, and 0-3 as poor.
Two reviewers extracted the data using a predesigned data extraction form created by authors on Covidence. Disagreements were resolved through discussion by consensus with all authors. The following data were extracted from each study. Only information relevant to lung cancer was extracted when the analysis included multiple cancers.

Statistical Analysis
The meta-analysis was performed by combining the multivariable-adjusted OR/RR/HR of the association between AD use and lung cancer from each study using the Bayesian random-effects pooling model. A Bayesian approach reflects uncertainty in the estimation of between-study heterogeneity and is a better approach when the number of studies in the meta-analysis is small compared with a classic approach using the DerSimonian-Laird method (22). When the study reported the OR/RR/HR for each AD, a combined effect and 95% CI for that study were first obtained using the fixed-effects pooling model. The analysis was conducted separately for lung cancer risk and survival to evaluate whether the associations with AD use were the same for two different outcomes.
Subgroup analysis was conducted to investigate potential sources of heterogeneity. First, the extent to which AD type was associated with lung cancer risk or survival was assessed. Second, study characteristics (study location, design, quality, and comparison groups) were evaluated to examine whether the associations were consistent for different study conditions. Third, AD use with FIGURE 1 Flow diagram of the study selection process.
respect to lung cancer diagnosis was evaluated to assess whether prediagnosis and postdiagnosis AD use had a differential association with survival. Heterogeneity between studies was quantitatively assessed using the Cochran Q test and inconsistency I  test. Egger test and the funnel plot were used to determine the potential existence of publication bias in meta-analysis. Statistical analyses were performed using the Comprehensive Meta-Analysis software (version 3, Biostat Inc.) and Stata -Release 17 (StataCorp). A two-tailed P < 0.05 was considered statistically significant.

Data Availability
The data generated in this study for the meta-analyses are available upon request from the corresponding author. The data generated in this study, apart from the data used for the meta-analyses, are available within the article.

Literature Search Results
We screened 545 abstracts, 12 of which were reviewed in full for eligibility. We excluded one full-text publication due to no results being reported separately for AD use (23), leaving 11 for data extraction and quality assessment. The detailed steps of the systematic search and selection processes are shown in Fig. 1 of a PRISMA flow chart.

Characteristics of Included Studies
The characteristics of included studies are summarized in Table 1. Four studies (24)(25)(26)(27) were conducted in the United Kingdom, two studies (28,29) in the United States, two studies (30,31) in Finland, and one in Hungary (32), Israel (33), and Taiwan (34), respectively. Seven studies (27-33) used a retrospective cohort study design, while four studies (24)(25)(26)34) used a case-control study design. The sample size varied from 174 patients with lung cancer in a single medical center study (28) to 418,588 pairs in a population-based retrospective cohort study (31). The proportion of male participants also varied from 35.3% (33) to 100% (30), and smoking prevalence ranged from 0.15% in a population-based lung cancer incidence study (34) to 92% in a clinic-based lung cancer survival study (29). Only five studies (24,25,30,33,34) reported the prevalence of depression in their study sample, which ranged from 3.5% in a population-based case-control study (34) to 31% in a population-based study but only including those with at least one AD purchase (33). Only three studies specified the histology of the lung cancer cases they included: one study (28) examined the association between ADs and overall survival among patients with NSCLC, another study (32)

AD Use
Regarding the types of ADs used, two studies (30, 33) did not specify them. One study (24) focused on TCAs, another study (27)
Furthermore, the results varied by the definition of exposure and the reference group. For example, Shoval and colleagues (33) compared AD users who were good adherents to those who were nonadherents, while Boursi and colleagues (27) compared continuous AD users with past AD users. Continuous AD use from prediagnosis to postdiagnosis was associated with lower survival (27), while postdiagnosis AD use was associated with increased survival (RR = 0.82; 95% CI = 0.69-0.98; n = 3). Studies conducted in Hungary (32) and Israel (33) showed positive associations between survival and AD use, while studies from the United Kingdom (27) and the United States (28) implied inverse associations. One study (29) reported increased lung cancer-specific survival with AD use (RR = 0.68; 95% CI = 0.49-0.95).

Publication Bias
Funnel plots ( Supplementary Fig. S1) do not indicate noteworthy evidence of publication bias. The P values from Egger tests were 0.676 for studies included in the analysis of lung cancer risk and 0.827 for the analysis of overall survival.

Discussion
To our knowledge, the current study is the first meta-analysis of observational studies examining the associations of AD use with lung cancer risk and survival.
Our study demonstrates that AD use could increase the risk of lung cancer by 11%, with the most increased risk with SNRI use. Among patients with lung cancer, AD use was not associated with survival.
First, AD use increased the risk of lung cancer, which is somewhat contrasting with the results from previous meta-analyses in other cancers. Meta-analyses for breast cancer (35), colon cancer (36), and epithelial ovarian cancer (15) showed that AD use was not associated with an increased cancer risk. In contrast, another meta-analysis showed that AD use was associated with a reduced risk of colon cancer (14). The discrepancy in the results may be due to the differ-ences in the types of ADs, cancer type, study design, and the number of studies included in the meta-analysis. For example, the subgroup analysis of Li and colleagues (35) including only the results from cohort studies showed an increased breast cancer risk with AD use, which is consistent with our results. The results varied by type of ADs. SSRI use was not associated with colon cancer risk (36), but TCA use was associated with a lower risk (14). In addition, as shown in our subgroup analysis, the definition and timing of AD use and the study location could be the source of heterogeneity in observed results. Considering that AD use was linked to an increased risk for stroke (RR In our analysis, only SNRIs showed an increased risk for lung cancer. SNRIs are commonly prescribed for depressive disorders, anxiety, and chronic pain for their ability to inhibit the reuptake of serotonin and norepinephrine, increasing the concentration of these neurotransmitters in the brain (16,39). Because the management of depressive disorders, anxiety, and chronic pain may require long-term treatments and high cumulative exposure to SNRI use, the genotoxic effect of these drugs has been evaluated in several preclinical experiments. Treatment in mice with duloxetine, a commonly prescribed SNRI, led to increased oxidative DNA damage (40) and production of micronucleated erythrocytes (41), a key indicator of genetic damage. In addition, long-term administration of duloxetine to female mice resulted in the formation of liver tumors (42). Exposure of human peripheral blood lymphocytes to venlafaxine, another SNRI, generated chromosomal abnormalities such as chromatid breaks and sister chromatid unions producing an elevated concentration of micronucleated cells (43). Collectively, these studies suggest a potential role of SNRIs in increasing cancer risk by inducing DNA damage; however, considering that our results were based on a small number of studies, more studies with SNRIs and lung cancer are needed to confirm our findings.
Although many preclinical studies have reported tumor suppressive effects of AD use (8)(9)(10)(11)(12), our meta-analysis for overall survival from four observational   epidemiologic studies did not show that AD use with either SSRIs or TCAs increased survival among lung cancer survivors, which is consistent with Chen and colleagues meta-analysis (14) for colorectal cancer reporting no associations between AD use and overall survival. Among the aforementioned four studies included in the analysis of survival outcome, one study (33) suggested that mortality could be reduced by 21% in patients with lung cancer with high adherence to AD prescription compared with the least adherent group, which may be due to increased compliance to cancer treatments with the management of depression with AD use in patients with cancer (44). When the authors restricted the analysis to those with the diagnosis of depression, the effect size was greater than the analysis with a whole sample (HR = 0.52 vs. HR = 0.79). For lung cancer-specific survival, meta-analysis was not possible because only one study reported this outcome. Zingone and colleagues study (29) reported the result with three different ADs (SSRIs, SNRIs, and TCAs) among 1,097 patients with lung cancer with 5-year follow-up, among which only TCA use before cancer diagnosis was associated with improved survival, which is supported by Chen and colleagues study, showing only TCA use being associated with a lower colon cancer risk (14), and many aforementioned preclinical studies. More epidemiologic research is needed to precisely estimate the association of AD use with lung cancer-specific survival.
There are several limitations of the current analysis that should be considered. First, there was substantial heterogeneity in key study features among the published reports, including study setting, study populations, the definition of exposure, and study designs, which probably contributed to the summary measures of the meta-analysis. Second, the number of studies included in the current meta-analysis is relatively small, so the meta-analysis effect estimates are estimated with low precision. Therefore, the results of our study should be interpreted cautiously because the estimation of heterogeneity can be challenging in meta-analyses where the number of studies included in the analysis is small (n ≤ 5; ref. 45). As more studies are published in the future, an updated meta-analysis is warranted. Third, although we extracted the results from the multivariable-adjusted analysis, there may be residual confounding because many studies did not include important confounders such as smoking, depression, and cancer treatment (46). Smoking history/status, the most significant risk factor for lung cancer, was adjusted in seven studies (24,26,27,29,30,33,34); however, the number of pack-years was adjusted in only one study (29). As the dose-response relationship between pack-years of smoking and lung cancer outcomes is well established (47), any residual confounding by smoking could influence the outcome of the current study. In addition, the conditions in which AD use was indicated were not specified. Although ADs are primarily prescribed to treat major depressive disorders, they are also used to manage anxiety disorders, chronic pain, and smoking cessation. Because the presence of depression directly affects the incidence of cancer and negatively impacts cancer prognosis, any residual confounding by coexisting depression could alter the result of the current analysis (44). In cases where ADs, such as bupropion or nortriptyline, are prescribed for smoking cessation, changes in cancer-related behaviors need to be monitored to assess lung cancer risk more accurately (17). Furthermore, the prognosis of lung cancer varies widely among tumor subtypes, disease stages, and treatment modalities. For example, a significant difference exists between the 5-year survival rate for NSCLC (26%) and SCLC (7%), which is more strongly associated with smoking (3). Only three studies either restricted analysis to patients with histologically confirmed cases or adjusted for tumor subtypes (28,29,32). Therefore, survival data also need to be interpreted with caution, because they were not adjusted for variables such as duration of cancer treatment or availability of targeted therapy. Future studies to evaluate the effect of ADs on lung cancer survival would benefit from including the number of pack-years, an indication for AD prescription, tumor subtypes, and cancer treatment.
There are several strengths of the current analysis to be noted. First, this study represents the first meta-analysis of lung cancer risk and survival with AD use. Next, in nine (24-28, 31-34) out of 11 studies, AD use was obtained from medical/prescription records. Cancer diagnosis and death were also accessed from medical records or national/governmental vital statistics, which are less likely to be influenced by recall bias. However, exposure misclassification is still possible because adherence was only assessed in one study. Nonetheless, misclassification was more likely to be nondifferential, biasing the results toward the null. Finally, most studies were based on a large population-based sample with a sufficient follow-up period, which increases the generalizability of study findings.

Conclusions
In summary, the number of studies assessing the associations between AD use and lung cancer risk and survival is deficient despite the high prevalence of depression and high usage of ADs in people with lung cancer. This first metaanalysis for lung cancer suggested that AD use could increase the risk of lung cancer. To increase the evidence, well-designed prospective cohort studies are needed, where commonly used ADs (SSRIs, SNRIs, and TCAs) are evaluated on cancer outcomes in individuals with lung cancer and with or without a diagnosis of depression.