Clinical Consequences of Generic Warfarin Substitution: An Ecological Study

To the Editor: Generic substitution policies are financially attractive.¹ Although several studies have found no clinically important effects of switching to generic warfarin formulations,² some observational studies have found an increase in warfarin dosages³ or frequent and substantial changes in international normalized ratio (INR).⁴ Few studies have been sufficiently large to examine clinical outcomes, and none have done so in a large-population setting.

On June 7, 2001, the provincial drug benefit plan in Ontario, Canada, instituted a policy requiring pharmacists to substitute either of 2 generic warfarin formulations (Apo-warfarin and Taro-warfarin) for Coumadin (Bristol-Myers Squibb Canada Inc, Montreal, Quebec). Physicians were advised of the policy a week before implementation.⁵ Patients could still receive Coumadin by paying the difference between it and the lowest-priced generic or by providing evidence of an adverse event during treatment with generic warfarin.

We performed an ecological study of trends in warfarin prescribing, INR testing, and hospitalization for major hemorrhage or stroke before and after introduction of this policy.

We performed a population-based, cross-sectional, time-series analysis of all persons aged 66 years or older in Ontario (comprising approximately 1.4 million individuals), using administrative health data from April 1, 1997, to March 31, 2002. Information about outpatient prescription drug use was obtained from the Ontario Drug Benefit Plan Database, which has an overall error rate of less than 1%.⁶ Data for laboratory services were obtained from the Ontario Health Insurance Plan Database. Hospitalization records were obtained from the Canadian Institute for Health Information Discharge Abstract Database, providing information on clinical outcomes documented to have high positive predictive values.^{7 - 9}

The study period was divided into monthly intervals: 40 months before policy implementation, the month during which the policy was initiated, and 9 months thereafter. Only data through March 2002 were included because of a change to a nonvalidated diagnostic coding system. We anticipated that adverse effects, if present, would be evident within weeks to months of implementation.

We identified every prescription for warfarin and other drugs that might influence clinical outcomes, including other anticoagulants, antiplatelet agents, nonsteroidal anti-inflammatory drugs, corticosteroids, selective serotonin reuptake inhibitors, gastric acid suppressants, and estrogens. We identified all outpatient INR tests and hospital admissions involving a primary diagnosis of upper gastrointestinal tract hemorrhage (International Classification of Diseases, Ninth Revision [ICD-9] codes 531, 532, 533, 534, 578.0, 578.1, and 578.9), subarachnoid hemorrhage (ICD-9 430.0), intracerebral hemorrhage (ICD-9 431.0), thrombotic stroke (ICD-9 434.0), or cerebral embolism (ICD-9 434.1).

Time-series analysis was used to examine patterns in monthly rates of warfarin prescription and rates of INR testing and hospitalization among those dispensed warfarin. We used autoregressive integrated moving average models incorporating a step function at time of policy initiation¹⁰ and compared observed and predicted postpolicy event rates and numbers. The results of multivariable models adjusting for demographic variables were not different and are not reported herein. All analyses were performed using SAS for Windows, version 8.02 (SAS Institute Inc, Cary, NC) with a 2-tailed type I error rate of .05. The study was approved by the Research Ethics Board at Sunnybrook and Women's College Health Sciences Centre, Toronto.

The number of patients treated with warfarin increased steadily, from 22 926 in April 1997 to 36 724 in March 2002. Corresponding prescription rates are shown in Figure 1. By 3 months postpolicy, generic warfarin formulations constituted 87% of all warfarin prescriptions dispensed.

There was no change in rates of INR testing (P = .93) or hospitalization for either major hemorrhage (P = .97) or cerebral thromboembolism (P = .89; Figure 2). These results did not change appreciably after adjustment for temporal trends in the use of other relevant drugs. Observed and predicted numbers of INR tests and hospitalizations were similar: tests, 579 638 vs 571 249 (95% confidence interval [CI], 369 783-618 150); hemorrhage, 346 vs 318 (95% CI, 202-434); cerebral thromboembolism, 58 vs 41 (95% CI, −28 to 110), respectively.

These results indicate that Ontario's generic warfarin substitution policy was not associated with any discernable change in rates of INR testing or hospitalization for major hemorrhage or cerebral thromboembolism. Because the drug benefit price for generic warfarin was 70% of that for Coumadin, the policy has likely resulted in substantial cost savings.

Some limitations of our study merit emphasis. We focused on major clinical events resulting in hospitalization but excluded other important outcomes with less reliable coding, such as lower gastrointestinal hemorrhage, deep vein thrombosis, and pulmonary embolism. We excluded younger patients and INR tests performed in-hospital. Given the ecological nature of the study, we could not determine whether some patients were more vulnerable to adverse consequences of substitution or whether outcomes differed between patients who received generic formulations and those who continued taking Coumadin. Further research is needed to identify and characterize those who may be at particular risk for adverse events.