|
Adherence to Fixed Dose Combination 3 Times Better
|
|
|
found by study published in AIDS Care Nov 2005
The main study outcome was medication adherence as assessed by prescription refill data.
"Adherence to combined Lamivudine+Zidovudine versus individual components: A community-based retrospective medicaid claims analysis"
AIDS Care, Vol. 17, No. 8, November 2005, pp. 938-948 | Year: 2005 | PP: 938-948
A. Legorreta MD, MPHa,*, A. Yub, H. Chernicoffc, A. Gilmorec, J. Jordand and J. C. Rosenzweigd
aUCLA School of Public Health, Department of Health Services, Los Angeles, California; bSchool of Pharmacy, University of Southern California, Los Angeles, California; cHealth Benchmarks Inc. (HBI), Woodland Hills, California; dGlaxoSmithKline, Research Triangle Park, North Carolina.
"NUMBER OF PILLS AFFECTS ADHERENCE TO HIV TREATMENT REGIMENS"
press release from GSK
Patients Taking Combination Pills Three Times More Likely to Refill Prescriptions on Time
Research Triangle Park, NC - November 30, 2005 - Fixed-dose combination therapies for HIV may increase the likelihood of treatment adherence among patients, according to a recently published study of 2,112 patients conducted by GlaxoSmithKline. Patients taking a fixed-dose combination of lamivudine (3TC) 150 mg and zidovudine (ZDV) 300 mg (COMBIVIR) achieved ≥ 95 percent adherence more than twice as often as patients whose regimens included 3TC and ZDV as separate pills. The results were published in the November issue of AIDS Care.
"Pill burden is an important factor in achieving desirable adherence rates. Fixed-dose combination products have been available for almost a decade, with COMBIVIR being the first to market. However, until now, limited data have existed demonstrating the ability of these products to facilitate adherence. That's why we are encouraged by the results of this study," said Mark Shaefer, Pharm. D., acting vice president, HIV, Infectious Disease Medicine Development Center at GSK.
The retrospective, longitudinal cohort study analyzed 60-day prescription refill Medicaid
claims data from 1995-2001. The data combined to cover more than 3.3 million patient lives.
To adjust for the time prior to the approval of a fixed-dose combination therapy
(COMBIVIR) in 1997, two sets of adjusted analysis were conducted, one for the total population (n=2,112) over the entire study period (1995-2001) and another for the population subgroup (n=1,427) that received separate pills or combination therapy from September 1997 to December 2000. The investigators focused on the results for the subgroup analysis to minimize the potential influence of other time-related variables on the outcome.
The primary study outcome was medication adherence as assessed by prescription refill
data. The adherence ratio was calculated for each prescription during the follow up period; for example, a patient with a prescription claim for a 30-day supply who did not refill the prescription until 45 days later would be considered 67 percent adherent (30/45 = 67 percent). The threshold for adherence was set at 95 percent; existing evidence suggests that this level of adherence is associated with a lower risk of virologic failure. A patient would have been considered non-adherent on or after the 32nd day following a 30-day prescription refill.
Mean adherence was higher for patients taking COMBIVIR (85 percent) compared to
patients taking the separate components, lamivudine and zidovudine, (75 percent) when
considering the population subgroup (p<0.001). Similarly, in measuring adherence rates for the subpopulation, patients taking COMBIVIR achieved ≥ 95 percent adherence more than twice as often as patients who took the separate components (36 percent vs. 16 percent respectively).*
Researchers also reported that findings from analyses conducted with the entire population during the complete study period (1995-2001) were similar to those found in the subgroup analysis.
"Fixed-dose combinations like COMBIVIR can reduce treatment complexity when
compared to their separate components. Treatment complexity is a common barrier to adherence and fixed-dose combination products should be considered for patients with difficulty adhering to their medication regimen," said Shaefer.
COMBIVIR was the first two-drug combination tablet approved for HIV treatment and is one of the most studied and widely prescribed dual nucleoside combination products. One out of five people currently taking antiretrovirals in the United States is prescribed COMBIVIR.
From published study
In general, findings from the multivariate analyses conducted with the entire population during the complete study period (1995 to 2001) were similar to those found in the subgroup analyses. For example, the risk of nonadherence was similarly lower for patients taking combination therapy (fixed dose AZT/3TC) compared to patients in the separate pills (AZT+3TC) group (IRR = 0.63, CI: 0.55, 0.73; HR = 0.31; 0.29, 0.33). The significance of other model covariates differed between the entire population and population subgroup analyses.
Discussion
This retrospective observational analysis using Medicaid claims data from two states compared patient refill adherence to a fixed dose combination of 3TC and ZDV versus the same medications given as separate pills. Subgroup analyses essentially standardized patient index dates and were conducted within a limited exposure period to account for historical biases associated with the combination therapy's entry into the antiretroviral market. Across multiple multivariate regression models conducted for the subgroup, including logistic regression, Poisson regression, and Cox proportional hazards, patients taking the combination therapy were associated with a higher level of adherence (defined at a 95% threshold) compared to patients taking the individual components as separate pills. The likelihood of refill adherence among combined therapy patients was three times that of patients in the separate pills group. Also, combination therapy patients had 33% lower incidence of adherence failure per patient month of follow-up and nearly double the time to adherence failure compared to separate pills patients. Results were similar for analyses considering the entire population (i.e. the combined therapy group was consistently associated with better study therapy adherence than the separate pills group) and are supported by findings from a previous study, which conducted similar analyses between combined therapy and separate pills cohorts using administrative data from private health plans (Jordan et al., 2004).
Despite the modest effect and marginal significance of other model parameters compared to treatment group, certain factors are worth noting. For instance, male gender and PI use were associated with reduced risk of nonadherence in all regression models. Similarly NNRTI use was significantly associated with a lower risk of adherence failure in two of the three analyses. One explanation for the findings associated with PI and NNRTI use as significant predictors of adherence to study therapy is that the high correlation between adherence to non-study antiretroviral therapies and adherence to study therapy might allow both to be associated with a reduced risk of nonadherence. For NNRTIs and PIs, the measure could be capturing some unobserved component associated with the nature of the drug. For example, the literature suggests that the patient/provider relationship is an important source of information for medication effectiveness and adherence (Stone, 1998). Thus, if a provider promotes the use of PIs differently or with more urgency, the patient might be more inclined to take the drug as directed.
Analyses of non-randomized administrative data require statistical management of patient-level differences in order to minimize the likelihood that study results are confounded by selection bias. In the subgroup baseline analysis for example, patients taking combined therapy or separate pills were similar on many characteristics, but they differed on non-study antiretroviral therapy use (i.e. PIs and NNRTIs) measured by the number of prescription claims, age, and average daily pill count. In this study, we included several patient-level characteristics to control for inter-cohort differences that might threaten internal validity. However it must be acknowledged, as with all non-randomized studies, that unmeasurable differences between treatment group characteristics may not have been controlled for and this could have biased our study results.
Another limitation of observational studies involves the comparability of study results to those from randomized controlled trials given that patients tend to be more adherent to treatment regimens under controlled conditions (Bergeron et al., 1998; Heinasmaki et al., 1998). On the other hand, the primary advantage of using administrative claims data is its collection from a real world setting as opposed to a clinical trial environment. This strengthens study findings by increasing external validity or generalizability to other populations operating in an uncontrolled, clinical practice environment. Another advantage is the longitudinal nature of the data, which in this case, allowed for a long-term adherence measure that could account for market entry.
Several measures of adherence other than prescription refill (Hogg et al., 2002; Turner, 2000) have been employed in previous studies reported in the literature. They include provider assessment (Bangsberg et al., 2001), patient self-report (Hill, 2003; Nieuwkerk et al., 2001; Ostrop, 2000) and computer assisted electronic monitoring systems (Golin et al., 2002; Farley, 2003) and have resulted in a wide reported range (28 to 84%) of adherence to antiretroviral therapy. Each method of measuring adherence embodies certain advantages and limitations. One additional advantage of prescription refill data however, is the tendency toward a conservative bias. For example, if prescription refill data overestimates true adherence (given that the patients' actual medication use might not be congruent with prescription refills), the subsequent estimated effect on favorable health outcomes will likely be underestimated (Hogg et al., 2002).
Overall, this study presented consistent findings that use of a fixed-dose combination of 3TC and ZDV in Medicaid insured patients is associated with better adherence to medication regimens compared to use of its components as separate pills. These results are commensurate with those found by two similar studies, an observational analysis that assessed privately insured patients (Jordan et al., 2002) and a randomized trial (Eron, 2000), and suggest that fixed dose combination therapies should be considered when prescribing an appropriate HIV treatment that includes a dual nucleoside.
Results
Of the entire study population, 1,605 and 507 patients were in the combined and separate pills treatment groups respectively. There were 1,363 patients receiving combined therapy and 64 patients receiving separate pills that initiated study therapy on or after September 1997 and were included in subgroup analyses. At a statistical significance level of p ≦ 0.05, baseline subgroup comparisons indicated that patients in the combination therapy subgroup were slightly younger, more likely to be receiving NNRTIs, less likely to be receiving PIs, and had a lower daily pill count than the separate pills subgroup (Table I).
![combined-1.gif](../images/120505/combined-1.gif)
Only the NNRTI and pill count results were similar when assessing the whole study population. Overall, more significant differences were observed between the two cohorts during this longer time period. For example, combination therapy patients were more likely to be female, African American, receiving PIs and receiving non-study NRTIs. On the other hand, patients taking 3TC and ZDV in separate pills were more likely to be from the West Coast, have a history of symptomatic HIV, visit an infectious disease specialist and experience higher outpatient utilization.
Mean adherence across all prescriptions for study therapy was higher for patients on the combined therapy (85%) compared to patients taking the drugs separately (75%) when considering the population subgroup (p < 0.001). This relationship was also observed for the whole population, however in light of baseline differences in other factors, we chose to focus on the multivariate findings for the subgroup with similar time periods..
The subgroup difference in mean adherence remained after including other covariates. Being in the combined therapy group reduced the risk of adherence failure as defined by the 95% adherence threshold (adjusted OR = 0.33, 95% CI: 0.16, 0.63) (Figure 1). In other words, combined therapy patients were 3 times more likely than patients taking separate pills to refill their prescription on time. Other regressors that were associated with a decreased likelihood of adherence failure included male gender (OR = 0.82, 95% CI: 0.66, 0.99), use of NNRTIs (OR = 0.83, 95% CI: 0.75, 0.91), and use of PIs (OR = 0.79, 95% CI: 0.70, 0.88).
Poisson regression model results for the subgroup analysis indicated that combined therapy patients had a lower risk of nonadherence compared to patients taking 3TC and ZDV in separate pills (IRR = 0.67, 95%CI: 0.56, 0.80). This rate difference translated to about 1.4 fewer adherence failures per person year for combined therapy patients compared to patients in the separate pills group (Figure 2). PI and NNRTI use (IRR = 0.93, CI: 0.88, 0.98, and IRR = 0.92, CI: 0.87, 0.97, respectively) were also associated with a lower risk of nonadherence, whereas female gender (IRR = 1.10, CI: 1.01, 1.20) and having a medical history of Hepatitis B (IRR = 1.55, CI: 1.01, 2.27) were associated with increased risk of adherence failure.
Results from the Cox proportional hazards model in the subgroup analysis indicated that the risk of adherence failure was much lower for combined therapy patients compared to patients in the separate pills group (HR = 0.49, CI: 0.44, 0.55) (Table II). Male gender (HR = 0.91, CI: 0.86, 0.96) and PI use were also associated with a lower risk (HR = 0.97, CI: 0.94, 0.99), whereas history of alcohol and drug abuse was associated with a higher risk of nonadherence (HR = 1.19, CI: 1.06, 1.33).
For any chronic condition, higher rates of adherence are associated with more favorable health outcomes. However, achieving the best possible HIV-related outcomes requires high, sustained levels of medication adherence throughout a patient's lifetime. An adherence rate of 95% or greater has been associated with lower virologic failure, fewer hospital days, and reduced morbidity and mortality in HIV-infected patients. The need exists for simpler regimens such as combination therapies that may effectively boost patient medication adherence.
A recent clinical study conducted over 16 weeks found significantly greater adherence to a fixed dose combination of lamivudine (3TC) 150 mg/zidovudine (ZDV) 300 mg and a trend toward better virologic response when compared to its individual components taken as separate pills (Eron et al., 2000). However, evidence suggests that medication adherence in a clinical trial is greater than that observed in a real-world setting.
In general, pharmacy claims data are a useful measure of adherence because they include standardized information for a population of individuals based on all medications paid for by a particular source. Medicaid is a US state and federally funded program that provides health insurance to underserved populations including low-income and disabled individuals. Because Medicaid is the single payer for all of its beneficiary's medical service and pharmacy use, it is an ideal setting for studies of adherence assuming that pharmacy claims are complete for each patient. Few studies have examined antiretroviral adherence among HIV-infected Medicaid beneficiaries, all of which have focused on certain subgroups such as women either during or after pregnancy (Laine, 2000; Turner, 2000) or women with a comorbid mental illness (Walkup, 2004). The observational analysis presented here adds to the previous literature by evaluating antiretroviral medication adherence in two whole Medicaid populations where patients received either a fixed dose combination of lamivudine (3TC) 150 mg/zidovudine (ZDV) 300 mg or its components taken as separate pills. This study also addresses a potential confounding factor of previous studies by examining subcohorts of patients started on combined therapy and separate pills within the same time frame.
Abstract
Adherence to a fixed dose combination of dual nucleoside antiretroviral therapy was compared between human immunodeficiency virus (HIV)-infected patients newly started on a fixed dosed combination of lamivudine (3TC) 150 mg/zidovudine (ZDV) 300 mg versus its components taken as separate pills. Medicaid pharmacy claims data were used for analyses. To examine the association between treatment group and medication adherence, three types of multivariate regressions were employed. In addition, all regressions were conducted for the whole population using data from 1995 to 2001 as well as a subpopulation, which excluded data prior to September 1997. Model covariates included patient characteristics, healthcare utilization, and non-study antiretroviral therapy use. The likelihood of ≥95% adherence among patients on combination therapy was three times greater than patients taking 3TC and ZDV in separate pills. Also, combination therapy patients had on average 1.4 fewer adherence failures per year of follow-up and nearly double the time to adherence failure compared to the separate pills group. Consistency among study results suggests that fixed dose combination therapies such as lamivudine (3TC) 150 mg/ zidovudine (ZDV) 300 mg should be considered when prescribing HIV treatment that includes an appropriate dual nucleoside.
Methods
Study population
Medicaid claims data used in this study were from two sources, a West Coast state and a Southeast state. These data combined cover more than 3.3 million patient lives. A retrospective, longitudinal cohort design using data from 1995 to 2001 was employed. All patients with at least one outpatient pharmacy prescription claim for combined therapy or 3TC and ZDV as separate pills during the study period were identified for potential study inclusion. The index date, defined as the date of first prescription claim for any of these medications was determined for each patient. Cohort assignment, combined therapy or separate pills, was based on index date prescription. Patients that were prescribed both combined therapy and one of the separate pill components during the baseline period were not considered for this study. Patient follow-up time began 60 days subsequent to the index date and continued to the end of the study period unless either Medicaid enrollment ended or the patient switched to an alternative form of antiretroviral therapy.
For study inclusion, patients had to be at least 18 years old, newly started on antiretroviral therapy (i.e. have no antiretroviral therapy claim for 12 months prior to index date despite confirmed plan eligibility), and have at least 1 prescription refill for the index drug(s) in addition to the index prescription within the first 60 days post index date. Patients in the separate pills group were subject to the additional requirement that the first prescription for each drug (3TC and ZDV) be received within 7 days of the other in order to capture intent to treat with a combination of separate pills. Patients with insufficient follow-up time (fewer than 60 days) were excluded from the study.
Baseline characteristics
Patient characteristics including age, sex, race/ethnicity and region were measured at the index date, while prior healthcare utilization and medical condition history were examined during the 60 days prior to index date. Measures of prior utilization included whether or not the patient visited an infectious disease specialist and total number of outpatient visits. Health conditions considered in the patient's medical history were identified through International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) diagnostic codes and Current Procedural Terminology (CPT) codes, including symptomatic HIV, alcohol or drug abuse, other psychiatric illness, and hepatitis B. Symptomatic HIV included diagnoses for opportunistic infections and other HIV-related illnesses. Alcohol or drug abuse included psychoses, dependence, and nondependent use as well as psychotherapy, counseling, rehabilitation and other services, whereas psychiatric illnesses were defined by mental disorders and therapy services unrelated to alcohol or drug abuse. A full list of ICD-9 codes used in this analysis is available upon request.
Patient characteristics measured during the first 60 days of follow-up included additional non-study antiretroviral therapy use measured as the number of prescription claims as well as total pill burden calculated as the number of pills per day. Additional non-study antiretroviral therapies included protease inhibitors (PIs), non-nucleoside reverse transcriptase inhibitors (NNRTIs), and other nucleoside reverse transcriptase inhibitors (NRTIs). Additional non-study antiretroviral therapy variables were not mutually exclusive, meaning patients could have had prescriptions for more than one non-study antiretroviral therapy category.
Medication adherence
The main study outcome was medication adherence as assessed by prescription refill data. Central to this evaluation is the determination of the adherence ratio - the ratio of the number of therapy-days supplied on the prescription to the number of calendar days until the patient refilled the prescription. The adherence ratio was calculated for each study therapy prescription during the follow-up period, excluding the last prescription. For example, a patient with a prescription claim for a 30-day supply who did not refill the prescription until 45 days later would be considered 67% (30/45 = 0.67) adherent for the period covered by the prescription. For the separate pills group, adherence ratios for ZDV and 3TC prescriptions were calculated separately and then averaged together. Any remaining supply from previous prescriptions was carried over to subsequent prescription periods for the calculation of therapy days supplied. Also, hospital days were excluded from the calculation of days elapsed between prescriptions.
The threshold for adherence was set at 95% in congruence with existing evidence that a lower risk of virologic failure is associated with maintaining adherence at or above this level (Paterson et al., 2000). Therefore patients were classified as 'non-adherent' and assigned an 'adherence failure' each time adherence dropped below the 95% threshold value. In the example provided above, the patient would have been considered non-adherent on or after the 32nd day following the date of the first prescription (30/0.95 = 32) if the next prescription had not yet been filled.
Patients were considered 'at-risk' of adherence failure from 60 days post index to receipt of the last prescription for study therapy during follow-up. An adherence failure marked the end of one's at-risk status beginning with the day following the date of failure. Only a subsequent prescription claim for study therapy warranted the reassignment of one's at-risk status.
Therefore, adherence failures were measured for each medication refill prior to the last refill date and patients were allowed to have multiple, non-sequential 'risk intervals' throughout the study period. For patients in the separate pills cohort, risk intervals were calculated separately for each drug. Patients had to be at risk of non-adherence for both therapies in order to be considered part of the at-risk pool. Mean adherence across all prescriptions for study therapy, total number of adherence failures and exposure time were calculated for each patient. Exposure time was the total number of 'at-risk' person-months from 60 days post index date to the end of the last risk interval.
Statistical analysis
Statistical methods for this study included both unadjusted, two variable comparisons and adjusted analyses, which examined the relationship between two variables, controlling for several covariates. Patient characteristics were examined at baseline and compared between cohorts using Wilcoxon tests for continuous or count variables and Chi-square tests for categorical variables.
One potential confounding factor of previous studies was that 3TC and ZDV as separate pills were more commonly used prior to 1997 when combined therapy was introduced to the market. Thus the separate pills cohort is more likely to be observed during an earlier time period than the combined therapy cohort. In an effort to control for this potential confounding factor and to increase the robustness of study findings, two sets of adjusted analyses were conducted, one for the total population over the entire study period (1995-2001) and another for the population subgroup that received separate pills or combination therapy during the same period of time. Subgroup inclusion was determined by study entry (September 1997 to December 2000) in order to account for the release of combined therapy into the antiretroviral therapy market in late 1997. For analysis of the total population, a dummy variable for time prior and subsequent to market entry was included in regression models.
Three types of multiple regression analyses were conducted including logistic regression, Poisson regression and Cox proportional hazards regression, based on the nature of the outcome variables (e.g., binary, count, or continuous). Multiple logistic regression was used to model the likelihood of adherence failure (coded '1' for failure and '0' otherwise) for each patient with treatment group as the primary variable of interest. Other independent variables included patient age, sex, race/ethnicity, region, medical history of symptomatic HIV, alcohol/drug abuse, psychiatric disorder, and hepatitis B, prior utilization of outpatient services and visits to an infectious disease specialist, daily antiretroviral therapy pill count (including study therapy), and use of other classes of antiretroviral therapies, with results reported as odds ratios. Adjusted odds ratios were computed for treatment group alternatively assuming that all patients received combined therapy or separate pills respectively. Nonparametric bootstrapping techniques with 500 repetitions were used to compute 95% confidence intervals around the odds ratio point estimates (Carpenter et al., 2000). Bootstrapping methods use random samples of the study population to generate repeated estimates of the statistic of the interest (e.g., odds ratio) in order to construct a distribution and assign levels of statistical significance and are commonly employed in the health sciences literature (Mooney, 1993).
Number of adherence failures per month, a count variable, was modeled by Poisson regression analysis (Long, 1997). We corrected for overdispersion - a common problem in Poisson models that occurs when the variance of adherence failures is greater than the mean - by setting the scale parameter to the deviance divided by the residual degrees of freedom. Incidence rate ratios representing the risk of adherence failure were obtained from model results by exponentiating the parameter estimates. Covariates included in the model were the same as those mentioned above.
We examined the association between treatment group and time to nonadherence using Cox proportional hazards regression with multiple risk intervals (Prentice, 1981). Cox proportional hazards regression is typically used to examine the time until an event such as nonadherence occurs. The hazard function is useful for modeling the potential that an event occurs at a specific time, given that time is a continuous variable. Pill burden, use of non-study antiretroviral therapies, and history of symptomatic HIV were included as time-dependent variables measured at the beginning of each 60-day interval subsequent to the index date. Hazard ratios representing the risk of nonadherence were used to report model results.
|
|
|
|
|
|
|