Association of Benzodiazepines With SARS-CoV-2 Infection and Clinical Outcomes: a Nationwide Cohort Study

The evidence for the impact of benzodiazepine (BZD) use on infection or clinical outcomes of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is limited. We evaluated the association of BZD use with SARS-CoV-2 infection and the clinical outcomes of coronavirus disease 2019 (COVID-19) using a nationwide COVID-19 database from South Korea. This nationwide cohort study was performed using the COVID-19 database from the Health Insurance Review and Assessment Service of Korea, and SARS-CoV-2 positivity was investigated according to BZD use. SARS-CoV-2-positive adult patients were assessed in three groups, those who needed hospitalization, those with severe symptoms requiring intensive care, and those who died. A multivariate logistic regression model was used for all the analyses. After adjusting for potential confounding factors, there was no association between BZD use and SARS-CoV-2 positivity. SARS-CoV-2-positive patients with BZD use showed an increased risk of need for hospitalization from COVID-19 compared to those without BZD use (odds ratio [OR]: 1·33, 95% condence interval [CI]: 1·07–1·65). In addition, there was a higher risk for long-term users (OR: 2·64, 95% CI: 1·08–6·47). Chronic BZD use contributed to a higher risk of the need for hospitalization among COVID-19 patients, whereas BZD use did not increase the risk of SARS-CoV-2 test positivity, severe outcomes, or mortality. (COPD), asthma, hypertension, cardiovascular disease, chronic kidney disease, diabetes, obesity, malignancy, immunosuppressant use and transplantation, and chronic human immunodeciency virus (HIV) infection have been identied. 3 Benzodiazepines (BZD) and BZD receptor agonists have been known to increase the risk of pneumonia and death due to pneumonia. 4,5 There is also concern about the risk of respiratory depression by BZD use in people with pre-existing respiratory problems, although reports have been conicting; an increased risk of respiratory exacerbations was reported for patients with COPD, 6 while associations of BZD use with hospital admission or impaired blood gases were not signicant in COPD patients. 7 A few recent studies on the characteristics of severe COVID-19 cases 8–10 have focused on the use of BZD. However, the results have been inconsistent. positivity was examined in a total of ve models. Model 1 was a crude model for the association between BZD use and SARS-CoV-2 positivity; model 2 was a minimally adjusted model adjusted for age and sex; model 3 was the fully adjusted model, which was adjusted for age, sex, residential area, type of insurance coverage, CCI, and disease history (diabetes, cardiovascular disease, cerebrovascular disease, COPD, asthma, hypertension, and chronic kidney disease); model 4 was the fully adjusted model for the association between the chronic use of BZD for 90 days and SARS-CoV-2 positivity; and model 5 was the fully adjusted model for the association between the chronic use of BZD for 180 days and SARS-CoV-2 positivity. Among patients who tested positive for SARS-CoV-2, Pearson's chi-square tests were used to compare the sociodemographic and clinical characteristics with respect to BZD use and chronic BZD use. To examine the association of BZD use and the chronic BZD use with the risk of SARS-CoV-2 positivity and clinical outcomes, we used multivariable logistic regression models after adjusting for sex, age, residential area, CCI, and disease history. COVID-19 signicantly BZD BZD superinfections in inuenza-infected the of


Introduction
The coronavirus disease  pandemic is causing a global crisis. COVID-19 is caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), with consequences ranging from asymptomatic disease to death. 1 However, 14% of SARS-CoV-2-positive patients show severe disease, and 5% show critical health conditions; 2 the risk of mortality (0.1%) associated with COVID-19 is much higher than that associated with seasonal in uenza. 3 Various risk factors such as age ≥ 65 years, chronic obstructive pulmonary disease (COPD), asthma, hypertension, cardiovascular disease, chronic kidney disease, diabetes, obesity, malignancy, immunosuppressant use and transplantation, and chronic human immunode ciency virus (HIV) infection have been identi ed. 3 Benzodiazepines (BZD) and BZD receptor agonists have been known to increase the risk of pneumonia and death due to pneumonia. 4,5 There is also concern about the risk of respiratory depression by BZD use in people with pre-existing respiratory problems, although reports have been con icting; an increased risk of respiratory exacerbations was reported for patients with COPD, 6 while associations of BZD use with hospital admission or impaired blood gases were not signi cant in COPD patients. 7 A few recent studies on the characteristics of severe COVID-19 cases 8-10 have focused on the use of BZD. However, the results have been inconsistent.
Given the high prevalence (2·6-23·7%) of BZD use, [11][12][13][14][15][16] investigating the impact of BZD use on infection or clinical outcomes of SARS-CoV-2 is bene cial for public health. Here, we assessed the association of BZD use with SARS-CoV-2 positivity and clinical outcomes in three groups of patients, those in need of hospital admission, those with severe symptoms requiring intensive care, and those who died, using a nationwide cohort data from South Korea.

Study design and population
In this study, National Health Information Database (NHID)-COVID data provided by the National Health Insurance Service (NHIS) were used; data pertaining to the period from 2015 to 2020 were obtained. The dataset included data from patients with COVID-19 who tested positive from January 1 to May 30, 2020.
Control groups included general controls and subjects who showed negative SARS-CoV-2 test results.
There are two different study population in this study. Study population 1 is to examine the association between benzodiazepine use and SARS-CoV-2 test positivity, and study population 2 is to investigate the association between benzodiazepine use and clinical outcomes in SARS-CoV-2-positive patients. The initial population for the year 2020 included a total of 351,377 subjects. These were categorized into 8,070 patients with SARS-CoV-2, and 343,307 controls. After excluding those with missing values, 328,373 subjects remained. They were divided into two different groups according to whether BZD was used or not: 52,151 subjects used BZD and 276,222 subjects did not use BZD. Using the 1:1 propensity score matching (PSM) method, 52,151 participants each from the case group and the control group were selected as the nal study subjects (out of a total of 104,302 subjects). In addition, a total of 7,596 patients with COVID-19 were divided into two groups according to whether BZD used or not: 1,074 subjects used BZD and 6,522 subjects did not use BZD (Fig. 1). All methods were performed in accordance with the relevant guidelines and regulations.

Outcome variables
In this study, we set SARS-Cov-2 test positivity and clinical outcomes as the outcome variables. The clinical outcomes for SARS-CoV-2-positive patients consisted of three variables: hospital admission, severe symptoms requiring intensive care or invasive ventilation, and mortality. Patients who died before receiving hospital care were also included in the rst group, as it was assumed that they needed hospital admission.

Variables of interest
The study subjects were divided into two groups according to BZD use. Use of BZD was de ned based on claim history in 2019, one year before the outbreak of COVID-19 in 2020.
BZD categories were based on the following main ingredient codes: clobazam (135702ATB), clonazepam (136401ATB), chlordiazepoxide (131201ATB, 131202ATB, 255800ATB), diazepam (142930BIJ), unitrazepam (160601ATB), urazepam (161801ATB), ethyl lo azepate (156201ATB, 156202ATB), alprazolam (105501ATB, 105502ATB, 105504ATB, 105505ATB, 105507ATB), bromazepam (118501ATB), clotiazepam (137302ATB), etizolam (156501ATB, 156502ATB, 156503ATB), lorazepam (185501ATB, 185504ATB), to sopam (241201ATB), and triazolam (243501ATB, 243502ATB). In addition, chronic BZD use was categorized based on a 90-day usage period and a 180-day usage period within a year. 16,17 Covariates In this study, we adjusted for variables that could directly or indirectly affect outcomes, including basic independent variables such as sex, age, and residential area. The residential areas were grouped based on the major regions and the metropolitan areas where the incidence of COVID-19 was high in Korea at that time (Seoul, Gyeonggi-do, Daegu, and Gyeongbuk); the rest of the regions were combined into a single group. Type of insurance coverage was classi ed into three different groups: workplace, local, and medical bene ts. Participants' clinical baseline characteristics were also considered as covariates. The Charlson Comorbidity Index (CCI) was used to con rm the patients' comorbidity status in 2019, which was the year prior to the COVID-19 outbreak. Based on CCI data, the severity of comorbidity was categorized as 0, 1, or 2+. Additionally, we reviewed the disease records for the period from 2015 to 2018 for diabetes, cardiovascular disease, cerebrovascular disease, COPD, asthma, hypertension, and chronic kidney disease, which could be associated with worse clinical outcomes and BZD use.

Statistical analysis
We investigated the results after 1:1 PSM; in this method, the case and control participants who have a similar propensity score values are matched. 18 We matched case and control groups by including age, sex, and CCI variables as parameters in the propensity score model. The association between BZD use and SARS-CoV-2 positivity was examined in a total of ve models. Model 1 was a crude model for the association between BZD use and SARS-CoV-2 positivity; model 2 was a minimally adjusted model adjusted for age and sex; model 3 was the fully adjusted model, which was adjusted for age, sex, residential area, type of insurance coverage, CCI, and disease history (diabetes, cardiovascular disease, cerebrovascular disease, COPD, asthma, hypertension, and chronic kidney disease); model 4 was the fully adjusted model for the association between the chronic use of BZD for 90 days and SARS-CoV-2 positivity; and model 5 was the fully adjusted model for the association between the chronic use of BZD for 180 days and SARS-CoV-2 positivity. Among patients who tested positive for SARS-CoV-2, Pearson's chi-square tests were used to compare the sociodemographic and clinical characteristics with respect to BZD use and chronic BZD use. To examine the association of BZD use and the chronic BZD use with the risk of SARS-CoV-2 positivity and clinical outcomes, we used multivariable logistic regression models after adjusting for sex, age, residential area, CCI, and disease history.
All the statistical analyses were performed using SAS statistical software version 9.4 (Statistical Analysis System Institute, Cary, NC, USA).

Discussion
Using a nationwide cohort database from South Korea, in this study, we showed that the chronic use of BZD contributed to an increase in the risk of the need for hospitalization among COVID-19 patients. However, BZD use did not signi cantly in uence the risk of SARS-CoV-2 positivity, severe outcomes, or mortality.
In animal studies, BZD increased mortality due to a variety of bacterial infections [19][20][21][22] and bacterial superinfections related to in uenza. 19 In human subjects, controversy persists regarding a causal connection between BZD use and infections. 23 The association between BZD use and the increased need for hospitalization in this study may be in line with previous studies which showed increased susceptibility to superinfections in in uenza-infected animals 24 and humans. 5 The underlying mechanism may be related to the effects of BZD on the immune system; BZD ampli es the effect of the gamma-aminobutyric acid receptor in immune cells, which may lead to an immune-suppressant pro le. 24 Regarding the long term use of BZD, chronic consumption of BZD was related to the appearance of modi ed lymphocyte subsets. 25,26 However, relatively few cases of severe COVID-19 and inaccessible variables in our data, such as the dosage of BZD, may require replication of pharmacoepidemiologic research on the relationship between BZD and COVID-19.
Recent studies on the association between mental illness and COVID-19 outcomes have shown a higher risk for severe COVID-19 outcomes in patients with a mental illness, though the analyses did not include adjustment for BZD use. 27,28 Since BZD is frequently prescribed for anxiety symptoms and sleep disturbances, our ndings suggest that BZD use should be considered in further studies on the relationship between mental illness and COVID-19 outcomes. Likewise, adjusting for mental illness in future studies in the association between BZD and COVID-19 outcomes would uncover the risk of BZD use regardless of psychiatric diagnoses. However, it is interesting to note that 87·7% of BZD prescriptions were related to non-psychiatric diagnoses in a nationwide cohort study from South Korea. 16 Due to limited medical resources, especially with reference to negative pressure beds, policies on the priority for hospitalization among COVID-19 patients have been amended. For example, South Korea has introduced a residential treatment center to isolate asymptomatic patients or patients who do not need hospital care. Therefore, patients with moderate-to-severe symptoms may be hospitalized rst. 29 In this regard, the results of this nationwide cohort study may be applied to e ciently set strategies for managing COVID-19 patients, based the nding that patients with chronic BZD use need to be monitored frequently due to a high risk of need for hospitalization; however, the strategies should also consider our nding that BZD use does not imply increased severity of clinical outcomes related to COVID-19.
Some limitations of this study should be acknowledged. Although a validation study showed the overall agreement of diagnosis at 82·0%, 30 outcomes were identi ed by diagnostic and procedural codes, and possible misclassi cations cannot be ruled out. Moreover, data about the indication and BZD dose, as well as the hospitalization period were unavailable, which precluded a full assessment. The use of nationwide longitudinal data strengthens the causal relationship established in our study, and the generalizability of our ndings. However, we could not include all the COVID-19 patients up to the present, and thus may have missed analyzing new clinical outcomes that may have resulted due to various mutations in SARS-CoV-2.
In summary, BZD use was not associated with the risk of SARS-CoV-2 positivity, severe outcomes, or mortality. However, BZD use, especially for more than 180 days, conferred a higher risk of need for hospitalization among COVID-19 patients. Health professionals and public health authorities need to be alert about patients with long-term use of BZD, and these patients need to be closely monitored even if they currently do not need hospital care.

Declarations Ethics approval and consent to participate
The data were anonymized before they were obtained; thus, informed consent was not required. The Yonsei University Institutional Review Board approved this study (Approval number: 4-2020-1240).

Consent for publication
Not applicable

Availability of data and materials
The data that support the ndings of this study are available from the National Health Insurance Service in South Korea but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data are however available from the authors upon reasonable request and with permission of the National Health Insurance Service in South Korea.

Competing interests
There are no con icts of interest to declare.  2017R1A2B3008214). The funder of the study had no role in the study design; data collection, analysis, and interpretation; writing the report; and the decision to submit the paper for publication.
Authors' contributions H.Y.P. and J.K. designed the study and drafted the manuscript. H.Y.P. is responsible for study concept and design. J.K. took responsibility for acquisition, analysis, or interpretation of data. H.Y.P., J.K., S.K.A., and E.C.P. contributed to the discussion and reviewed and edited the manuscript. S.K.A. and E.C.P. are the guarantors of this work and as such, had full access to all study data. S.K.A. and E.C.P. assume responsibility for the integrity of the data and the accuracy of the data analysis.

History of Hypertension
No

History of Chronic Kidney Disease
No  NHI: National health insurance, CCI: Charlson Comorbidity Index, COPD: chronic obstructive pulmonary disease, ICU: intensive care unit † Hospital admission comprised admission, admission to the intensive care unit, invasive ventilation, or mortality. ‡ Severe outcome comprised admission to the intensive care unit or invasive ventilation. Figure 1 Flowchart showing the selection of the study population. NHID-COVID: National Health Information Database-Coronavirus disease; SARS-CoV-2: severe acute respiratory syndrome coronavirus 2.