ICH Risk With Concurrent Use of Antidepressants and NSAIDs
ICH Risk With Concurrent Use of Antidepressants and NSAIDs
We used the Korean Health Insurance Review and Assessment Service database for this study. The National Health Insurance programme started in Korea in 1977 and achieved universal coverage of the population by 1989. All Koreans are covered by the programme. Accordingly, the database contains all information on healthcare use and prescribed drugs for approximately 50 million Koreans.
We obtained the claims data for the patients who were prescribed at least one antidepressant drug from 1 January 2009 to 31 December 2013. The database included an unidentifiable code representing each patient together with age, sex, diagnosis, ambulatory care, hospital admissions, and dates of visits. In addition, prescribed drug information included the generic name, prescription date, and duration. The diagnosis was coded according to the international classification of disease, 10th revision (ICD-10). A previous validation study compared the diagnoses derived from the database with the actual diagnoses in the patients' medical records. The overall positive predictive value of the diagnoses was 83.4%.
There was no patient involvement in this study. The study population was composed of antidepressant treated patients. We included new users of antidepressants who took antidepressants for the first time between 1 January 2010 and 31 December 2013 (index date) without a history of having received a prescription for antidepressants during the preceding year. By including only new users, we could ignore the influence of previous antidepressant treatment. We excluded patients who had been diagnosed as having cerebrovascular diseases (ICD-10: I60-I68, G45, G46) as their primary or secondary diagnosis within a year before the index date. We also excluded patients who were over the age of 99, had a diagnosis of intracranial haemorrhage on the index date, or took prescriptions for more than one antidepressant on the index date and those whose index date was the last day of the study. In addition, we excluded patients whose index date came after the date of death (ICD-10: I46.1, I46.9, R96, R98, R99) (figure). Antidepressants included tricyclic antidepressants, selective serotonin reuptake inhibitors, serotonin-norepinephrine reuptake inhibitors, monoamine oxidase inhibitors, and others. Antidepressants included in "others" were bupropion, hypericin, mirtazapine, tianeptine, and trazodone.
(Enlarge Image)
Figure.
Selection of study participants from Health Insurance Review and Assessment Service database in retrospective cohort design. NSAI D=non-steroidal anti-inflammatory drug
Among antidepressant treated patients, we obtained their NSAID prescriptions by using the Anatomical Therapeutic Chemical codes (M01A, N02BA). We defined combined use of antidepressants and NSAIDs as the prescription of at least one NSAID during the defined 30 day follow-up of antidepressants.
We defined the outcome as time to first hospital admission with intracranial haemorrhage (ICD-10: I60–62) as the primary or secondary diagnosis within 30 days' follow-up after the index date. The index date was the date of newly prescribed antidepressants. We assumed follow-up of antidepressant to last for seven days after the final prescription in a continuous course of treatment. We considered follow-up to have started on the index date and to have ended on the date of first hospital admission with intracranial haemorrhage within 30 days, the date the patient switched to another antidepressant, the date of discontinuation, or the last day of the study. We treated death as a competing risk.
Age, sex, comorbidity, and co-medication are all possible confounders of the association between antidepressant use and intracranial haemorrhage. We defined information on comorbidity and co-medication according to previous diagnoses and the use of drugs within one year before the index date. We calculated the modified Charlson index to estimate the severity of disease according to previous diagnoses within one year before the index date. We selected as confounders any comorbidities that may influence the risk of intracranial haemorrhage, which included diabetes, chronic obstructive pulmonary disease, hypertension, osteoarthritis, rheumatoid arthritis, osteoporosis, alcohol related disorder, ischaemic heart disease, chronic kidney disease, peptic ulcer, dementia, non-alcoholic liver disease, schizophrenia, neoplasm, HIV infection, transplantation, atrial fibrillation, heart failure, disease of arteries, and disease of veins. Low dose acetylsalicylic acid (Anatomical Therapeutic Chemical code: B01AC06), steroids (H02AB), warfarin (B01AA03), heparin (B01AB), platelet aggregation inhibitors (B01AC), antithrombotic enzymes (B01AD), direct thrombin inhibitors (B01AE), direct factor Xa inhibitors (B01AF), and other antithrombotic agents (B01AX) were also selected as confounders because they might increase the risk of intracranial haemorrhage through their action on haemostasis.
We estimated the propensity scores for adding NSAIDs to antidepressants without regard to outcomes by multiple logistic regression analysis using the following variables: age category, sex, Charlson index category, comorbidity, and co-medication ( Table 1 ). We assessed model discrimination with the c statistic. Matching was done using the Greedy 5→1 digit matching macro with the estimated propensity score. We used a standardised difference to compare baseline characteristics between patients who were treated with antidepressants without NSAIDs and those treated with antidepressants and NSAIDs. We calculated Cohen's d as the difference between two sample means divided by a pooled standard deviation for the data. We defined imbalance as an absolute value greater than 0.1.
We calculated the incidence rate per 1000 person years by dividing the number of intracranial haemorrhage events by the total number of person years at risk and multiplying the result by 1000 and calculated the 95% confidence interval assuming a Poisson distribution. For construction of the multivariable model, we included variables that achieved statistical significance in the likelihood ratio test. The final model included dementia, warfarin, heparin group, and steroids as the adjusting variables. We assessed the status of combined use of NSAIDs and covariates on a daily basis during the follow-up period for the time varying covariates. We used matched Cox regression models to estimate hazard ratios and their 95% confidence intervals for intracranial haemorrhage with time varying covariates in the propensity based matched cohort. By using this model, we could obtain an unbiased estimate of the change in the hazard of intracranial haemorrhage because of the concomitant use of antidepressants and NSAIDs. Competing risks arise when patients are exposed to several causes and failure due to one cause excludes failure due to other causes. In our study, we treated death as a competing risk rather than censoring it owing to its potential causal effect on the outcome of interest.
We also did a subgroup analysis according to antidepressant class, age category, sex, type of intracranial haemorrhage, comorbidity, and co-medication. We did subgroup analysis using a single model with interaction terms to see whether the association with the concurrent use of NSAIDs among antidepressant users differed significantly. We used the SAS statistical application program (release 9.3) for all statistical analyses. We considered a two tailed value of P<0.05 to be statistically significant.
Methods
Data Source
We used the Korean Health Insurance Review and Assessment Service database for this study. The National Health Insurance programme started in Korea in 1977 and achieved universal coverage of the population by 1989. All Koreans are covered by the programme. Accordingly, the database contains all information on healthcare use and prescribed drugs for approximately 50 million Koreans.
We obtained the claims data for the patients who were prescribed at least one antidepressant drug from 1 January 2009 to 31 December 2013. The database included an unidentifiable code representing each patient together with age, sex, diagnosis, ambulatory care, hospital admissions, and dates of visits. In addition, prescribed drug information included the generic name, prescription date, and duration. The diagnosis was coded according to the international classification of disease, 10th revision (ICD-10). A previous validation study compared the diagnoses derived from the database with the actual diagnoses in the patients' medical records. The overall positive predictive value of the diagnoses was 83.4%.
Patient Involvement and Study Population
There was no patient involvement in this study. The study population was composed of antidepressant treated patients. We included new users of antidepressants who took antidepressants for the first time between 1 January 2010 and 31 December 2013 (index date) without a history of having received a prescription for antidepressants during the preceding year. By including only new users, we could ignore the influence of previous antidepressant treatment. We excluded patients who had been diagnosed as having cerebrovascular diseases (ICD-10: I60-I68, G45, G46) as their primary or secondary diagnosis within a year before the index date. We also excluded patients who were over the age of 99, had a diagnosis of intracranial haemorrhage on the index date, or took prescriptions for more than one antidepressant on the index date and those whose index date was the last day of the study. In addition, we excluded patients whose index date came after the date of death (ICD-10: I46.1, I46.9, R96, R98, R99) (figure). Antidepressants included tricyclic antidepressants, selective serotonin reuptake inhibitors, serotonin-norepinephrine reuptake inhibitors, monoamine oxidase inhibitors, and others. Antidepressants included in "others" were bupropion, hypericin, mirtazapine, tianeptine, and trazodone.
(Enlarge Image)
Figure.
Selection of study participants from Health Insurance Review and Assessment Service database in retrospective cohort design. NSAI D=non-steroidal anti-inflammatory drug
Combined Use of Antidepressants and NSAIDs
Among antidepressant treated patients, we obtained their NSAID prescriptions by using the Anatomical Therapeutic Chemical codes (M01A, N02BA). We defined combined use of antidepressants and NSAIDs as the prescription of at least one NSAID during the defined 30 day follow-up of antidepressants.
Follow-up to Intracranial Haemorrhage
We defined the outcome as time to first hospital admission with intracranial haemorrhage (ICD-10: I60–62) as the primary or secondary diagnosis within 30 days' follow-up after the index date. The index date was the date of newly prescribed antidepressants. We assumed follow-up of antidepressant to last for seven days after the final prescription in a continuous course of treatment. We considered follow-up to have started on the index date and to have ended on the date of first hospital admission with intracranial haemorrhage within 30 days, the date the patient switched to another antidepressant, the date of discontinuation, or the last day of the study. We treated death as a competing risk.
Potential Confounders
Age, sex, comorbidity, and co-medication are all possible confounders of the association between antidepressant use and intracranial haemorrhage. We defined information on comorbidity and co-medication according to previous diagnoses and the use of drugs within one year before the index date. We calculated the modified Charlson index to estimate the severity of disease according to previous diagnoses within one year before the index date. We selected as confounders any comorbidities that may influence the risk of intracranial haemorrhage, which included diabetes, chronic obstructive pulmonary disease, hypertension, osteoarthritis, rheumatoid arthritis, osteoporosis, alcohol related disorder, ischaemic heart disease, chronic kidney disease, peptic ulcer, dementia, non-alcoholic liver disease, schizophrenia, neoplasm, HIV infection, transplantation, atrial fibrillation, heart failure, disease of arteries, and disease of veins. Low dose acetylsalicylic acid (Anatomical Therapeutic Chemical code: B01AC06), steroids (H02AB), warfarin (B01AA03), heparin (B01AB), platelet aggregation inhibitors (B01AC), antithrombotic enzymes (B01AD), direct thrombin inhibitors (B01AE), direct factor Xa inhibitors (B01AF), and other antithrombotic agents (B01AX) were also selected as confounders because they might increase the risk of intracranial haemorrhage through their action on haemostasis.
Statistical Analysis
We estimated the propensity scores for adding NSAIDs to antidepressants without regard to outcomes by multiple logistic regression analysis using the following variables: age category, sex, Charlson index category, comorbidity, and co-medication ( Table 1 ). We assessed model discrimination with the c statistic. Matching was done using the Greedy 5→1 digit matching macro with the estimated propensity score. We used a standardised difference to compare baseline characteristics between patients who were treated with antidepressants without NSAIDs and those treated with antidepressants and NSAIDs. We calculated Cohen's d as the difference between two sample means divided by a pooled standard deviation for the data. We defined imbalance as an absolute value greater than 0.1.
We calculated the incidence rate per 1000 person years by dividing the number of intracranial haemorrhage events by the total number of person years at risk and multiplying the result by 1000 and calculated the 95% confidence interval assuming a Poisson distribution. For construction of the multivariable model, we included variables that achieved statistical significance in the likelihood ratio test. The final model included dementia, warfarin, heparin group, and steroids as the adjusting variables. We assessed the status of combined use of NSAIDs and covariates on a daily basis during the follow-up period for the time varying covariates. We used matched Cox regression models to estimate hazard ratios and their 95% confidence intervals for intracranial haemorrhage with time varying covariates in the propensity based matched cohort. By using this model, we could obtain an unbiased estimate of the change in the hazard of intracranial haemorrhage because of the concomitant use of antidepressants and NSAIDs. Competing risks arise when patients are exposed to several causes and failure due to one cause excludes failure due to other causes. In our study, we treated death as a competing risk rather than censoring it owing to its potential causal effect on the outcome of interest.
We also did a subgroup analysis according to antidepressant class, age category, sex, type of intracranial haemorrhage, comorbidity, and co-medication. We did subgroup analysis using a single model with interaction terms to see whether the association with the concurrent use of NSAIDs among antidepressant users differed significantly. We used the SAS statistical application program (release 9.3) for all statistical analyses. We considered a two tailed value of P<0.05 to be statistically significant.