Detection of unknown ototoxic adverse drug reactions: an electronic healthcare record-based longitudinal nationwide cohort analysis

Ototoxic medications can lead to significant morbidity. Thus, pre-marketing clinical trials have assessed new drugs that have ototoxic potential. Nevertheless, several ototoxic side effects of drugs may remain undetected. Hence, we sought to retrospectively investigate the potential risk of ototoxic adverse drug reactions among commonly used drugs via a longitudinal cohort study. An electronic health records-based data analysis with a propensity-matched comparator group was carried out. This study was conducted using the MetaNurse algorithm for standard nursing statements on electronic healthcare records and the National Sample Cohort obtained from the South Korea National Health Insurance Service. Five target drugs capable of causing ototoxic adverse drug reactions were identified using MetaNurse; two drugs were excluded after database-based analysis because of the absence of bilateral hearing loss events in patients. Survival analysis, log-rank test, and Cox proportional hazards regression models were used to calculate the incidence, survival rate, and hazard ratio of bilateral hearing loss among patients who were prescribed candidate ototoxic drugs. The adjusted hazard ratio of bilateral hearing loss was 1.31 (1.03–1.68), 2.20 (1.05–4.60), and 2.26 (1.18–4.33) in cimetidine, hydroxyzine, and sucralfate users, respectively. Our results indicated that hydroxyzine and sucralfate may cause ototoxic adverse drug reactions in patients. Thus, clinicians should consider avoiding co-administration of these drugs with other well-confirmed ototoxic drugs should be emphasized.


Introduction
Drug-induced ototoxicity is de ned as a functional impairment, such as hearing and/or balance disorders, depending on the involvement of the cochlear and/or vestibular system, respectively, or both [1][2][3], characterized by temporary or permanent inner hair cell damage caused by therapeutic agents [1]. Almost all ototoxicity events spontaneously resolve after treatment discontinuation, although several events often result in permanent serious consequences, and detract from the patient's quality of life [1,4].
Several monitoring protocols for drug-induced ototoxicity exist, although their practical use is questionable due to adherence issues [5][6][7]. Generally, pre-marketing studies of new pharmaceuticals are important to detect adverse drug reactions (ADR), although these studies sometimes cannot reliably detect ADR [8]. Post-marketing surveillance could facilitate a full understanding of the risk-bene ts of drugs [9]. Thus, both pre-marketing clinical trials and post-marketing data are necessary to detect unknown information on ototoxicity.
To date, the commonest ototoxic drugs in clinical use include aminoglycoside antibiotics and platinumbased chemotherapeutic agents (cisplatin and carboplatin) which have well-documented e cacy against various infections and malignancies, respectively, in both children and adults [10]. The panorama of druginduced ototoxicity has expanded in the last few decades from aminoglycoside antibiotics and platinumbased chemotherapeutic agents to include non-steroidal anti-in ammatory drugs, loop diuretics, and other antimicrobial or antineoplastic agents [11]. Several reports indicate that more than 600 drugs have ototoxic potential. Nevertheless, the ototoxic effects of several drugs may remain undetected. Thus, researching the possibility of drug-induced ototoxicity is meaningful in clinical elds [1,4,5,12]. Therefore, we used MetaNurse, an algorithm for detecting ADRs based on a meta-analysis technique normalized on a yearly basis and an improved patient-sampling and comparison group creation strategy, in an initial attempt to investigate candidate drugs that can possibly induce ototoxicity [13]. To con rm the ototoxic effect of candidate drugs selected from MetaNurse, we used the National Sample Cohort (NSC) from South Korea to evaluate the incidence of bilateral hearing loss (as a representative diagnosis of ototoxicity) in patients who were prescribed the candidate drugs. We veri ed the ndings through an external validation process using the United States Food and Drug Administration Adverse Event Reporting System (FAERS).

Results
Search for drug-induced ototoxicity based on the MetaNurse algorithm We rst selected 101 target drugs based on the activity-enabled pharmacovigilance MetaNurse algorithms, including Beers criteria medications (n = 107), medicines with pediatric use restrictions (n = 79), and the United Nations list of banned drugs (n = 28) (Fig. 1). We excluded 61 drugs due to previously known ototoxic effects. Thus, we assessed 40 drugs that have signi cant associations with ear and labyrinth disorders, and only 12 of the drugs showed HR > 1.5 (p < .005) in MetaNurse. We extracted ve target drugs (capsaicin, cimetidine, epinephrine, hydroxyzine, and sucralfate) after the exclusion process based on a review of the Side Effect Resource (SIDER 4.1) and ototoxicity-related articles.
Incidence of subsequent bilateral hearing loss in patients who were prescribed candidate drugs A total of ve target drugs were investigated for the possibility of subsequent development of bilateral hearing loss based on the KNHIS-NSC database. Among the ve target drugs, capsaicin and epinephrine were excluded from the nal analysis because no bilateral hearing loss event had occurred until 2013.
This cohort study included patients who were prescribed each candidate drug (i.e., cimetidine, hydroxyzine, and sucralfate) for more than 14 days during the index period (January 2002 to December 2013, Supplementary Fig. 1). A schematic description of the study ow is presented in Fig. 2. Each of the three cohorts (prescribed cimetidine, hydroxyzine, and sucralfate) showed similar distributions of sex, age, residential area, household income, and the CCI between the medication and non-medication groups because these variables were used for sample matching, which indicates that group matching was performed appropriately (supplementary tables 1-3).
In this study, 155039.9 and 122629.5, 33551.4 and 23384.7, and 36481.9 and 13825.6 person-years in the cimetidine non-users and users, hydroxyzine non-users and users, and sucralfate non-users and users, respectively, were evaluated for bilateral hearing loss events. We used univariate and multiple Cox regression models to analyze the HR for the development of bilateral hearing loss. After adjusting for sociodemographic and comorbidities factors, the cimetidine, hydroxyzine, and sucralfate users had an associated prospective bilateral hearing loss development with an adjusted HR of 1.31 (95% CI, 1.03-1.68), 2.20 (95% CI, 1.05-4.60), and 2.26 (95% CI, 1.18-4.33), respectively (Table 1). Kaplan-Meier survival curves with log-rank tests for the 12-year follow-up period are presented in Fig. 3, and indicate that patients who were prescribed cimetidine, hydroxyzine, and sucralfate developed the bilateral hearing loss more frequently than patients who were not prescribed those medicines. External validation of the possibility for drug-induced ototoxicity with cimetidine, hydroxyzine, and sucralfate To con rm our ndings, we performed external validation using FAERS data from 2012 to 2018 (Table 2) wherein we quanti ed the total number of otologic ADR, including ototoxicity, deafness, and tinnitus. The PRR and ROR with 95% CI for hydroxyzine versus total drugs in otologic adverse events were 1.

Discussion
Using MetaNurse and KNHIS-NSC, this study investigated certain drugs with the potential for causing drug-induced ototoxicity [13][14][15][16][17]. To our knowledge, this is the rst study to evaluate the possibility of ototoxicity with the use of cimetidine, hydroxyzine, and sucralfate. We used MetaNurse as a postmarketing pharmacovigilance tool based on EHR and the South Korean nationwide population insurance data. We initially selected several candidate drugs for ototoxicity, and then conducted a retrospective cohort study to evaluate the risk of bilateral hearing loss in patients who were prescribed the candidate drugs. External validation was performed to verify the novel ndings using the spontaneous reports FAERS database.
Ototoxicity is a pharmacological adverse reaction affecting the inner ear or auditory nerve, characterized by cochlear or vestibular dysfunction [1,2]. To date, aminoglycoside antibiotics and platinum-based chemotherapy, mainly cisplatin, are the most important ototoxic drugs that are commonly used in clinical practice [10]. These medications play an important role in modern medicine although they cause harmful effects and lead to signi cant patient morbidity. Thus, the early detection of ototoxicity through prospective monitoring is very important because it allows the consideration of treatment modi cations to minimize or prevent permanent hearing loss and balance impairment [18][19][20]. However, some pharmacotherapy associated with the development of ototoxicity could remain undetected for a long time because physicians do not recognize the need for drug concentration monitoring for these drugs.
In this study, we detected capsaicin, cimetidine, epinephrine, hydroxyzine, and sucralfate as possible candidate ototoxic drugs using the MetaNurse algorithm, which is based on nurse statements because nurses reportedly play a more important role in discovering and spontaneously reporting ADR than doctors and pharmacists [24][25][26][27]. However, on a limitation is that the result of the MetaNurse algorithm extracts the candidate drug list as the result of analysis for one hospital. To overcome this issue, we further veri ed candidate drugs obtained from the MetaNurse algorithm through a nationwide cohort sample because this dataset could accurately represent the characterization of South Koreans. In the analysis of KNHIS-NSC data, we used the diagnostic codes of bilateral hearing loss (H90.0, H90.3, and H90.6). There is no diagnostic code for ototoxicity and, therefore, we used the diagnostic code for bilateral hearing loss, which is the commonest phenotype of ototoxicity. Interestingly, we observed an increased incidence of bilateral hearing loss in patients who received cimetidine, hydroxyzine, and sucralfate when compared with those in the propensity score-matched comparator group developed based on sociodemographic factors, comorbidities, and enrollment year. After the analysis of KNHIS-NSC data, external validation was performed and showed that the values of PRR and ROR in hydroxyzine and sucralfate were positive indicators of ototoxicity because their 95% CI values were statistically signi cant; however, there was no signi cance in the PRR and ROR of cimetidine. Collectively, this means that these novel study ndings, which indicate that hydroxyzine and sucralfate have a potential ototoxic effect, may be reliable.
Hydroxyzine is a histamine receptor (H 1 ) antagonist that has been widely used to treat allergic skin reactions, such as pruritus and urticaria, and it is also used as an anti-emetic and anti-cholinergic medication. Additionally, hydroxyzine can reduce central nervous system activity. Thus, in psychiatric disorders, hydroxyzine has been used widely as a sedative to relieve anxiety and tension in generalized anxiety disorder since the 1980s [28,29]. The commonest side effect of hydroxyzine is dry mouth because of the anticholinergic effect. Drowsiness and involuntary motor activity could also occur when patients take considerably higher than recommended doses. Drug allergic reactions, headaches, and hallucinations are rarely reported in post-marketing surveillance data, but we could not nd any information on hydroxyzine-induced ototoxicity. Sucralfate has been commonly used to treat and prevent duodenal or stress ulcers because it adheres to damaged ulcer tissue and protects against acid and enzyme damage, leading to better healing. Treatment with other medications, such as antibiotics, may be needed to treat and prevent the ulcer recurrence caused by some types of bacteria (e.g., H. pylori) [30]. Sucralfate can sometimes cause some side effects, including constipation, hives, rash, itching, di culty in breathing or swallowing, and angioedema. Post-marketing surveillance reports revealed hypersensitivity reactions and hyperglycemia as side effects, but there were no reports of sucralfaterelated ototoxicity.
This study has several limitations. First, to date, mechanisms of drug-induced ototoxicity with hydroxyzine and sucralfate have not been de ned previously. Thus, further mechanistic studies are needed to determine their ototoxic effects. Second, we used bilateral hearing loss as a diagnostic code; however, in some cases, bilateral hearing loss may have developed due to non-ototoxicity-related causes. However, ototoxicity is the commonest cause of bilateral hearing loss, which suggests that this issue may be an acceptable limitation. Third, data on the degree of hearing impairment were not available in the registry (KNHIS-NSC); therefore, we could not ascertain the degree of hearing di culty effected by the candidate drug and whether that side effect was transient or permanent. Finally, we could not access other health data, including body mass index, lipid pro le, and information on behavioral risk factors, such as smoking or alcohol consumption. Therefore, these possible confounding factors could not be controlled in the study analyses.
This study investigated the detection of unknown ototoxic drugs among widely used drugs using MetaNurse and the KNHIS-NSC. We found that patients who were prescribed cimetidine, hydroxyzine, and sucralfate had a signi cant risk of bilateral hearing loss events. Additionally, the possibility of ototoxicity with hydroxyzine and sucralfate drugs was con rmed by an external validation based on FAERS. This study provides new insight into the potential risk of ototoxic ADR with hydroxyzine and sucralfate. Further studies are needed to elucidate the underlying pathophysiological mechanisms mediating the drug-related ototoxicity identi ed for these three drugs.

Methods
All methods were carried out in accordance with relevant guidelines and regulations is required in the method section of the manuscript.

MetaNurse
MetaNurse is an electronic healthcare record (EHR)-based pharmacovigilance algorithm applied to determine the frequency of standard nursing statements (SNS) or ADR symptoms [13] and to quantify symptoms only detected through bedside nurse observations that cannot be observed in a laboratory. Based on SNS in EHR, we identi ed the patients who experienced ADR and determined which drugs to include in a safety review. MetaNurse comprises a meta-analysis technique that is normalized annually with improved patient-sampling and comparison-group creation. In MetaNurse, patients who were unexposed to the drug were de ned as the comparator group, and ADR that were recorded more than twice in SNS after the initial drug administration constituted the signal. Hazard ratios (HR) were used as statistical indicators that were subsequently computed by adjusting age, sex, department, and disease severity. We analyzed all inpatient EHR data that were obtained from January  [31,32]. Strati ed random sampling was performed using 1,476 strata and by age (18 groups), sex (2 groups), and income level (41 groups, including 40 health insurance bene ciaries and 1 medical aid) among the South Korean population (46 million in 2002) [15,16]. This study was approved by the Institutional Review Board of Hallym Medical University Chuncheon Sacred Hospital (IRB No. 2016-05-052). The requirement for written informed consent was waived by the approving authority because the KNHIS-NSC dataset comprises de-identi ed secondary data for research purposes.

Study population
The patient groups for this study included all patients who were prescribed the MetaNurse-proposed candidate ototoxic drugs between January 2002 and December 2013. Each patient was tracked until 2013, and patients diagnosed with bilateral hearing loss (H90.0, H90.3, and H90.6) were identi ed. We excluded the following patients: (1)

External validation
To verify the novel study ndings, we used the publicly available FAERS database, which is a spontaneous reporting system that contains data on adverse events and medication errors, to identify the association between candidate drugs and adverse events by the case/non-case method. All reports that were published in the FAERS database from January 2012 to December 2018 were accessed from the o cial website of the U.S. Food and Drug Administration (http://www. fda.gov/) [33].

Statistical analyses
We employed 1:1 propensity score matching according to age, sex, residential area, household income, and the CCI. Incidence rates per 1,000 person-years for bilateral hearing loss were obtained by dividing the number of patients with bilateral hearing loss by the person-years at risk. The overall disease-free survival rate was determined using Kaplan-Meier survival curves for the entire observation period. To identify whether the candidate drugs increased the risk of bilateral hearing loss, we used Cox proportional hazard regression to calculate the hazard ratio (HR) and 95% con dence intervals (CI), after adjusting for other predictor variables. For external validation, a case/non-case analysis was performed by calculating the proportional reporting ratio (PRR) and the reporting odds ratio (ROR) with 95% CI. All statistical analyses were performed using R version 3.6.0 (R Foundation for Statistical Computing, Vienna, Austria) with a signi cance level of 0.05.

Declarations
Funding source Competing Interests statement: All authors declare that they have no con icts of interest relevant to the work presented in this article.