Warfarin Therapy in the HIV Medical Home Model: Low Rates of Therapeutic Anticoagulation Despite Adherence and Differences in Dosing Based on Specific Antiretrovirals

Clinic background and definitions

The Grady Health System Infectious Diseases Program (IDP) in Atlanta, Georgia is one of the largest centers for HIV care in the United States with an active census of over 4500 HIV-infected patients. The IDP has been designed to care for patients with advanced HIV infection, meaning that the majority of patients have either CD4⁺ T-cell count less than 200 cells per microliter or an AIDS-defining disease at enrollment. Pharmacy, mental health, and other services are also provided as part of a patient-centered medical home model. ¹⁹ The IDP has two full-time infectious disease pharmacists who direct the management of patients on anticoagulation therapy. Anticoagulation clinic visits are in addition to patient visits with primary medical providers (physician or physician extender). The framework for this management comes from the American College of Chest Physicians (ACCP) guidelines for treatment of thrombosis and vitamin K antagonist therapy. ^13,20 It includes clinic visits at least once every 4 weeks in stable patients (more frequently if changes are needed) in which the pharmacist assesses warfarin adherence, asks about bleeding events, and monitors the blood INR to ensure therapeutic anticoagulation. Changes in warfarin dosing and use of vitamin K were based on the ACCP guidelines, and warfarin dosing changes were made in increments of 10–20%. INR measurements were performed by the Grady Health System central coagulation laboratory using an FDA-approved commercial system. Point of care INR testing was not in place during the time of the study.

For this study, we reviewed all individuals who were followed for anticoagulation therapy at the IDP from 2004 (time of implementation of current electronic laboratory record) through 2009. Demographic parameters were established at baseline and laboratory and medications parameters were followed over time. Patients who were seen at the IDP for HIV care but were followed for anticoagulation by an outside warfarin clinic were not included in the study. Patients were categorized as adherent with warfarin at pharmacy clinic visits if the patient confirmed that the warfarin was taken as prescribed daily and the patient was able to accurately communicate the dosing regimen. If the patient received warfarin directly from the IDP (as opposed to an outside pharmacy), the IDP pharmacy system was also reviewed to ensure that warfarin was picked up once monthly. Adherence to other medications, including antiretrovirals, was not rigorously assessed at these pharmacy visits, but a list of all other active medications was reviewed verbally with each patient.

For this analysis, the diagnosis of thrombosis or other indication for anticoagulation required radiographic, echocardiographic, or electrocardiographic (if cardiac arrhythmia) confirmation to be included. Mental illness was defined by the diagnosis of depression, bipolar disorder, or schizophrenia at baseline. Other baseline characteristics included alcohol abuse, cocaine use, or injection drug use as documented by the patient's regular medical provider. Hospitalization was defined as inpatient admission for at least 2 days in the previous 4 weeks prior to anticoagulation initiation. Hepatitis B virus (HBV) infection was defined by surface antigen positivity at baseline with no documentation of clearance during follow-up. Hepatitis C virus (HCV) infection was defined by serum antibody positivity at baseline with either positive or unknown serum HCV RNA level and no history of HCV treatment prior to or during follow-up. Per U.S. guidelines, therapeutic INR for thromboembolic disease and cardiac arrhythmia was defined as 2.0–3.0 and therapeutic INR for mechanical heart valve replacement was defined as 2.5–3.5. ^13,21,22

Statistical analysis description

Proportions were compared between the two anticoagulation groups (deep venous thrombosis [DVT]/pulmonary embolus [PE] versus others) using the χ² or Fisher's exact test. Continuous variables were compared between anticoagulation groups using the Wilcoxon rank-sum test or the t test. An extension of the generalized estimating equations (GEE) introduced by Liang and Zeger ^23,24 was used to analyze the correlated nominal INR outcome data (subtherapeutic, therapeutic, and supratherapeutic). These three INR multinomial categories were assumed to have no intrinsic ordering. The GEE methodology, using an exchangeable correlation structure with a polytomous logistic marginal distribution, was used to model INR status at each clinical visit and relate the INR status to set covariables (baseline: alcohol abuse, cocaine abuse, injection drug use, mental illness, hepatitis B or C diagnosis, enrollment age; time dependent covariate: warfarin adherence was assessed at each clinical visit). Univariable and multivariable analyses of INR status were performed using these 7 risk factors. The odds ratio and its 95% confidence interval were calculated for each risk factor in the presence of others in the final model. The odds ratio was provided for subtherapeutic INR status relative to therapeutic INR status and for supratherapeutic status relative to therapeutic INR status in both the univariable and multivariable analyses. A standard GEE univariable analysis was performed for the repeated binary data within patient for viral load (≥100 copies per milliliter or <100 copies per milliliter; exchangeable binomial-logit model) and a similar longitudinal GEE univariable analysis was performed for the warfarin adherence data. Longitudinal analyses of average weekly warfarin dose were performed for the subset of 31 patients on one of three PI/NNRTI-based regimens who were adherent at each clinic visit and achieved therapeutic INR. The average weekly warfarin dose was analyzed with a means model using SAS Proc Mixed (version 9, SAS, Cary, NC) providing estimates of the mean by treatment group (lopinavir/ritonavir; efavirenz or atazanavir/ritonavir). A compound-symmetry form in the longitudinal measurements was assumed for the warfarin dose, and robust estimates of the standard errors of parameters were used to perform statistical tests and construct 95% confidence intervals. As an exploratory analysis, the model was then refitted and included trimethoprim-sulfamethoxazole (yes or no at each clinic visit) as an additional covariable.

Demographics

A total of 73 patients were identified (see Table 1 for baseline demographics). The majority of patients were male (89.0%) and either African American or African (83.6%). 60 of the 73 patients (82.2%) had either DVT, PE, or a combination of the two as the indication for anticoagulation. One third of these 60 patients were diagnosed during or immediately after a hospitalization. 15 of the 60 patients (25.0%) had at least one active opportunistic infection (OI) at the time of diagnosis. The most common OIs were Pneumocystis carinii pneumonia (8/60: 13.3%) and mucocutaneous candidiasis (5/60: 8.3%), followed by disseminated Mycobacterium avium infection, central nervous system toxoplasmosis, and cytomegalovirus retinitis (each 2/60: 3.3%). Six of 60 patients (10.0%) had a concurrent malignancy at the time of DVT or PE diagnosis (2 lymphoma, 2 squamous cell cancer of the anus, 1 Kaposi's sarcoma, and 1 lung cancer). Two patients had documented low protein S activity levels on the day of thrombosis diagnosis. At some point in their care, 8 other patients with thrombosis had documented low protein S activity levels and 3 patients had documented low protein C activity levels. However, these 11 patients were either on warfarin or had an unknown anticoagulation status when the level was drawn, making the results difficult to interpret. None of the patients with thrombosis had abnormal test results for antithrombin III deficiency, Factor V Leiden, antiphospholipid antibodies, or homocysteine, though the majority of patients were not checked for these.

Of the 60 patients, 34 (56.7%) were receiving prescriptions for HAART at the time of DVT/PE. Plasma HIV RNA levels were available for 33 of these 34 patients at baseline. The median level was 230 copies per milliliter (minimum < 50 copies per milliliter, maximum >500,000 copies per milliliter). Of these 33 patients, 15 (45.5%) had plasma HIV RNA levels of <100 copies per milliliter at the time of DVT/PE. In terms of age, gender, race/ethnicity, HBV/HCV status, CD4⁺ T-cell count, and HIV RNA level, patients with DVT/PE were not significantly different from patients with other indications for anticoagulation at baseline.

Anticoagulation clinic visits, adherence, and virologic control

As shown in Table 1, there were 15 patients (20.5% of total cohort) diagnosed with an indication for anticoagulation in the inpatient setting, while the rest of the cohort were referred to the clinic after already being on warfarin therapy as outpatients. Twenty-two patients were on enoxaparin at their first warfarin clinic visit. The median time between first and last warfarin clinic visits for the 73 patients was 182 days, with a maximum length of 1838 days. There were 911 warfarin clinic visits and the median number of visits was 10 per patient. Patients were judged to be adherent to warfarin at 78.0% of all visits (711/911). Ten patients were judged to be adherent at less than 50% of visits. At baseline, 3 of these 10 patients were positive for both cocaine use and injection drug use and 1 of the 10 was positive for alcohol abuse. In the univariable analysis of risk factors for study defined warfarin nonadherence, there were no variables with a statistically significant odds ratio (Table 2). The combined risk factors of cocaine use or injection drug use (OR 0.9, 95% CI 0.5–1.7, p=0.78) as well as alcohol abuse or cocaine use or injection drug use (OR 1.0, 95% 0.5–1.8, p=0.92) were also not associated with study defined warfarin nonadherence.

On average, patients were prescribed HAART at 87.5% of visits, with 56 of 73 patients prescribed HAART at 100% of visits. Plasma HIV RNA levels were obtained at 28.0% of pharmacy clinic visits (255/911) and the result was <100 copies per milliliter at 59.2% of visits. In univariable analysis, adherence to warfarin therapy was associated with a lower risk of HIV RNA level ≥100 (OR 0.6, 95% CI 0.4–0.9, p=0.01) during the follow-up period. However, the baseline characteristics of alcohol abuse, cocaine use, injection drug use, hepatitis B or C, mental illness, or age at enrollment (per 10-year increase) were not associated with a statistically significant risk of HIV RNA ≥100 during follow-up pharmacy visits (Table 3). The combined risk factors of cocaine use or injection drug use as well as alcohol abuse or cocaine use or injection drug use were also not associated with HIV RNA ≥100 during the follow-up period.

Warfarin effectiveness and complications

Ten patients (16.7% of subjects with DVT/PE as initial indication for anticoagulation) experienced clinical recurrence of DVT with radiographic confirmation (9 patients with 1 recurrence and 1 patient with 2 recurrences). All 10 were African American/African men with a median age of 36 years, median CD4⁺ count of 232 cells per microliter, and median log HIV RNA of 2.9 at the time of recurrence. The median time to first recurrence was 1190 days (minimum, 41 days; maximum, 4031 days) and 7 of the 10 patients were on HAART at the time of recurrence. Hospitalization was required in 4 recurrences (median stay 8 days, maximum stay 25 days), but there were no deaths during recurrences. There were 2 recurrences in patients taking warfarin as directed but with subtherapeutic INR levels, 2 recurrences in patients who had stopped warfarin prematurely, and 2 recurrences in the setting of unknown anticoagulation (diagnosed at outside facilities). DVT recurred after completion of an appropriate course (at least 3–6 months) of anticoagulation in 4 patients. There was only 1 patient who developed a thrombosis recurrence despite being therapeutic on warfarin (INR 2.0 at diagnosis). This patient was treated with low molecular weight heparin and warfarin dosing was increased to a goal INR of 3.0 (range, 2.5–3.5). There were no subsequent thromboses in this patient.

Three subjects experienced a bleeding event while on warfarin (median dose 57.5 mg/week, median INR 6.3 at time of event). Each of the three episodes involved hematuria requiring hospital observation. None of the episodes required blood transfusion and none resulted in death. One of these patients was not taking any warfarin-interacting medications, while one experienced an INR increase after a 14-day course of fluconazole. The third bleeding event occurred in a patient taking a stable regimen of four potentially warfarin-interacting drugs: atazanavir, ritonavir, clarithromycin, and trimethoprim-sulfamethoxazole.

Overall, the frequency of therapeutic INR levels during warfarin therapy was low, even when only including visits at which patients were judged to be adherent to warfarin. A therapeutic INR was recorded at 245 of these 711 adherent visits (34.5%, 95% CI 28.9–40.1%). The median percentage of visits with therapeutic INR while warfarin adherent was 30.8%. The median percentage of visits with subtherapeutic INR while warfarin adherent was 50.0%, while the median percentage of visits with supratherapeutic INR while warfarin adherent was 16.7%. The median number of warfarin dose changes per patient was 4 and the mean number of warfarin dose changes per patient was 5.7. For the total cohort, the mean INR at first warfarin clinic visit was 2.33 and the median was 1.89. For the 15 diagnosed as inpatients, the mean INR at first warfarin clinic visit was 2.56 and the median was 2.15. Changes in mean INR over time were not statistically significant (p=0.17). However, there was a trend in which the mean INR approaches 2.0 in months 3 through 6 during the follow-up period (Table 4). An analysis of INR longitudinal variability revealed that 92% of the estimated INR variance could be attributed to within-patient variance (as opposed to between-patient variance).

A univariable analysis yielded multiple factors that were associated with non-therapeutic INR (Table 5). A subsequent multivariable analysis (Table 6) showed that injection drug use at baseline was an independent risk factor for subtherapeutic INR (OR 2.4, 95% CI 1.3–4.7, p=0.01). Additionally, warfarin adherence was protective of both sub-therapeutic (OR 0.4, 95% CI 0.2–0.6, p<0.0001) and supratherapeutic (OR 0.5, 95% CI 0.3–0.9, p=0.02) INR status.

There were three PI/NNRTI- based regimens that were prescribed during at least 20 warfarin adherent clinic visits with goal INR 2.0–3.0 and thus were the focus of the warfarin dosing analysis: Lopinavir/ritonavir (LPV/RTV) 800 mg/200 mg total daily dose plus at least two nucleoside reverse transcriptase inhibitors (≥2 NRTIs), efavirenz (EFV) 600 mg daily dose plus ≥ 2NRTIs, and atazanavir/ritonavir (ATV/RTV) 300 mg/100 mg daily dose plus ≥ 2NRTIs. LPV/RTV plus ≥ 2 NRTIs was prescribed for 18 patients for a total of 71 warfarin adherent visits, EFV plus ≥ 2 NRTIs was prescribed for 11 patients for a total of 35 warfarin adherent visits, and ATV/RTV plus ≥ 2 NRTIs was prescribed for 6 patients for a total of 21 warfarin adherent visits. Two of these patients received EFV followed by ATV/RTV. These two patients were excluded from the warfarin dosing analysis due to failure of the mixed effects linear model. Of the remaining 31 patients, those on LPV/RTV plus ≥ 2 NRTIs patients had therapeutic INR at 38.8% of visits, sub-therapeutic INR at 49.2% of visits, and supratherapeutic INR at 12.0% of visits. Patients on EFV plus ≥ 2NRTIs had therapeutic INR at 33.3% of visits, subtherapeutic INR at 37.2% of visits, and supratherapeutic INR at 29.5% of visits. Patients on ATV/RTV plus ≥ 2 NRTIs had therapeutic INR at 20.0% of visits, subtherapeutic INR at 45.0% of visits, and supratherapeutic INR at 35.0% of visits. The average weekly warfarin dose to maintain a therapeutic INR of 2.0–3.0 was significantly lower for patients on EFV plus ≥ 2 NRTIs compared to regimens based on either boosted PI. Specifically, the mean warfarin dose for therapeutic INR was 46 mg/week (95% CI 35–56 mg) for patients on EFV plus ≥ 2 NRTIs compared to 68 mg/week (95% CI 54–81 mg) for patients on LPV/RTV plus ≥2 NRTIs (p=0.01) and 71 mg/week (95% CI 57–85 mg) for patients on ATV/RTV plus ≥ 2 NRTIs (p=0.007).