Statin-Fenofibrate Combination Therapy in an ACCORD Population: A Retrospective Analysis

Cardiometabolic risk (CMR) refers to the increased cardiovascular disease (CVD) risk associated with a group of clinical conditions that includes insulin resistance, hyperglycemia, obesity, hypertension, and dyslipidemia^.1 Mixed dyslipidemia characterized by low levels of high-density lipoprotein cholesterol (HDL-C), elevated triglycerides (TGs), and elevated levels of small low-density lipoprotein cholesterol (LDL-C) particles has been shown to be associated with CMR, and has been termed atherogenic dyslipidemia.^1,2 Evidence that therapy to lower LDL-C significantly reduces CVD risk in patients with diabetes mellitus (DM)^3-13 has prompted the American Diabetes Association to suggest that most, if not all, patients with DM should receive statin therapy.¹⁴ However, despite statin treatment, DM patients continue to have higher rates of CVD events.^{4,5,8,9,13,15} Thus, significant CVD residual risk remains even after LDL-C has been lowered suffi ciently, suggesting the potential relevance of the other lipid components in CVD risk assessment and treatment.

Fibrate therapy is one option in the treatment of patients with low levels of HDL-C and elevated TGs. Several fibrate outcome studies and/or the results of their subanalyses have suggested that fi brates lower CVD risk in patients with features of atherogenic dyslipidemia by 25% to 40%.^16-24 These exploratory results coupled with epidemiologic observational studies have infl uenced several clinical guidelines^1,25,26 to support the addition of fenofi brate or niacin to statin monotherapy to reduce non—HDL-C and TGs as well as to treat low HDL-C, particularly in patients with CMR. However, prospective randomized clinical trial data in support of fibrate and/or niacin adjunctive therapies have been lacking.

The lipid arm of the Action to Control Cardiovascular Risk in Diabetes trial (ACCORD lipid trial)²⁷ was the fi rst completed randomized prospective trial to examine the clinical impact of statin-fenofi brate combination therapy (SFCT) in patients with type 2 diabetes mellitus (T2DM). The ACCORD lipid trial evaluated whether administration of SFCT was more effective in reducing CVD events than simvastatin monotherapy in a cohort of 5518 patients with T2DM. Compared with statin monotherapy patients, SFCT patients had lower rates of death from cardiovascular causes, nonfatal myocardial infarction, and nonfatal stroke (primary end point), though the results were not statistically signifi cant. However, a prespecifi ed analysis of a subgroup of patients with TGs in the highest tertile (>204 mg/dL) and HDL-C in the lowest tertile (<34 mg/dL) demonstrated that coadministration of fenofi brate and simvastatin resulted in statistically significant CVD benefi t compared with simvastatin monotherapy. These mixed results may be interpreted as support for the use of fenofi brate as an adjunct to statin monotherapy in reducing CVD events among special populations such as T2DM patients with mixed dyslipidemia, a population with many of the markers for significant CMR.²⁷

Given the mixed results from the ACCORD lipid trial, this study’s purpose was to understand what findings are most germane to the population using SFCT in a real-world managed care population. Our primary objective was to compare demographics and lipid profiles of patients from the ACCORD lipid trial with those of patients from a nonrandomized real-world clinical setting who were selected using inclusion criteria similar to those of the ACCORD lipid trial. Other objectives were to describe and compare the profiles of the treatment cohorts (ie, lipid profi les, demographics, comorbidities, time to therapy) and identify factors predicting the treatment cohort in a real-world setting.

METHODS

We performed a retrospective database analysis using the Clinformatics DataMart, a product of OptumInsight Life Sciences. Clinformatics DataMart is an administrative claims database of the medical and pharmacy claims for approximately 42 million patients enrolled in a large US managed care plan. The individuals included in the claims data are geographically diverse across the United States. The health plan provides fully insured coverage for physician, hospital, and prescription drug services, and the providers of these services submit their claims for payment directly to the health plan. All data used in this study were de-identified as per Health Insurance Portability and Accountability Act requirements.

Eligibility criteria were designed to replicate as closely as possible the eligibility criteria of the ACCORD lipid trial while giving consideration to the limitations inherent in administrative claims data. Patients 40 to 64 years of age with the following baseline lipid profiles were identified during the index period (January 2004—March 2008): LDL-C 60-180 mg/dL, TGs <750 mg/dL, and HDL-C <55 mg/dL for men and <50 mg/dL for women. The date of the baseline lipid panel during the index period was set as the index date for each selected patient. Other eligibility requirements were (1) continuous enrollment in the managed care plan for at least 180 days before and after the index date; (2) absence of lipid treatment during the 180-day preindex period; and (3) presence of at least 2 medical claims containing a diagnosis for DM (International Classification of Diseases, Ninth Revision, ClinicalModification [ICD-9-CM] code 250.xx) or evidence of 1 or more prescribed DM medications as defined by the American Hospital Formulary System classifi cation system28 (alpha-glucosidase inhibitors, amilynomimetics, dipeptidyl peptidase-4 inhibitors, incretin mimetics, biguanides, insulins, meglitinides, sulfonylureas, thiazolidinediones, and glycogenolytic agents) for at least 90 days any time during the 180-day preindex period. For the sake of sample size, all DM patients were included, unlike the ACCORD lipid trial which included only T2DM patients.

All pharmacy claims for the patients selected by the aforementioned criteria were examined 180 days after the index date to determine whether any lipid therapies were dispensed using American Hospital Formulary System classification²⁸ (antilipemic agents, bile acid sequestrants, cholesterol absorption inhibitors, fibric acid derivatives, HMG-CoA reductase inhibitors [statins], miscellaneous antilipemic agents). All lipid therapies were classifi ed as a statin, fenofibrate, gemfi brozil, niacin, ezetimibe, bile acid sequestrant, or omega-3 fatty acid. Patients with no pharmacy claims for lipid therapies were categorized in the no-therapy cohort. Patients with at least 2 pharmacy claims for lipid therapy from a single medication classification without a 30-day overlap with a different lipid medication classifi cation were categorized in the monotherapy cohort, and patients with at least 2 pharmacy claims for lipid therapy from 2 or more medication categories started on the same day or with at least 30 days of overlap were categorized in the combination therapy cohort. Patients with any such overlaps between a statin and fenofi brate were subcategorized in the SFCT cohort; all others were categorized in the other combination therapy (OCT) cohort. The fixed combinations simvastatin-ezetimibe (Vytorin) and statin-niacin (Simcor, Advicor) were assigned to the OCT cohort. To best capture prescriber practice patterns, the initial prescribed lipid therapy was used to subcategorize the monotherapy cohort into the statin monotherapy and other monotherapy cohorts. Patients were categorized in a combination therapy cohort if they received combination therapy for any contiguous 30-day period in the 180 days after the index date.

Data Analyses

Descriptive analyses were performed to examine the differences between the baseline clinical characteristics of the cohorts. An analysis of the baseline lipid profi le case mix of the cohorts enabled a comparison with the case mix of the population enrolled in the ACCORD lipid trial. We also analyzed demographics (age, sex), associa ted comorbidities, CVD-related comorbidities, and the use of hypertension and DM medications (by the American Hospital Formulary System classifi cation28). ICD-9-CM codes were used to determine the presence of comorbidities. Associated comorbidities were hypertension (401.xx-405.xx), dyslipidemia (272.xx), and depression (296.2x, 296.3x, 300.4, and 311). The CVD comorbidities we identifi ed were acute myocardial infarction (410.xx, 412.xx), angina (413.xx), all other acute (411.xx), and chronic (414.xx) ischemic heart diseases, nonfatal stroke/ transient ischemic attack (433.xx-435.xx, 437.1), coronary insufficiency (429.2), peripheral vascular disease (443.9), heart failure (428.xx), atherosclerosis (440.xx), atrial fibrillation (427.3x-427.5x), and abdominal aortic aneurysm (441.3-441.4). The CVD and non-CVD comorbidities were chosen a priori if there was supporting evidence in the literature linking them as a risk factor or representative of a risk factor for the ACCORD lipid primary and secondary outcomes.

Means, medians, standard deviations, and percentiles were reported where appropriate. Logistic regression analyses were performed to determine the degree to which demographics and clinical characteristics including lipid profile and comorbidities (CVD and non-CVD) predicted the choice of lipid therapy in our study. Time from index date to therapy initiation were compared using an analysis of variance.

RESULTS

A total of 141,595 patients met study criteria. A large majority (n = 124,464, 87.9%) were in the no-therapy cohort, 14,850 (10.5%) were in the monotherapy cohort, and 2281 (1.6%) were in the combination therapy cohort. The majority of monotherapy patients initiated statin monotherapy (n = 12,136, 81.7%). Other monotherapies in order of prevalence were fenofi brate (n = 1135, 7.6%), ezetimibe (n = 680, 4.6%), and niacin (n = 413, 2.8%). Only a minority of monotherapy patients (n = 704, 4.7%) were initially prescribed 1 lipid medication and later changed to another from a different category. Among those receiving combination therapy, 181 (7.9%) comprised the SFCT cohort. The majority of these patients (n = 159, 87.8%) were prescribed only SFCT, while the remaining patients were prescribed a third lipid medication. Whereas 82 of the 181 SFCT patients (45.3%) received both therapies on the same day, the remaining 99 patients started on 1 therapy and added on the other at a later date within the first 180 days after the index date. The mean time to add-on therapy was54.5 ± 39 days (median 50 days). Among the 2100 OCT patients, most received statin-ezetimibe (n = 1579, 75.1%) and statin-niacin (n = 321, 15.3%) combination therapies.

Table 1

lists the demographics and baseline comorbidities of all the different cohorts. The average age of patients in our study (52 years) was a decade younger than that of the ACCORD lipid trial population (62 years), though the proportion of men enrolled in the ACCORD lipid trial (69.3%) was similar to the proportion of men in the SFCT cohort (71.7%). Among all patients in this study, the top 3 associated comorbidities in order of prevalence were hypertension, dyslipidemia, and depression. A significantly greater proportion of patients in both the SFCT and OCT cohorts were diagnosed with dyslipidemia (SFCT 59.1%, OCT 48.4%) and hypertension (SFCT 48.6%, OCT 45.8%) compared with those in the no-therapy cohort (dyslipidemia 26.6%, hypertension 34.2%, both P <.05). A signifi cantly greater proportion of patients in the SFCT cohort were diagnosed with dyslipidemia compared with those in the monotherapy cohort (59.1% vs 46.6%, P <.05).

The time from index lab assessment to lipid therapy initiation was signifi cantly shorter (P <.05) in the SFCT cohort than it was in any other cohort (

Table 2

). The median time to therapy was 19 days for the monotherapy cohorts, whereas it was 15 and 20 days for the SFCT and OCT cohorts, respectively.

The largest intercohort differences were noted with TGs (

Table 3

). Specifically, the highest and lowest median TG levels were observed in the SFCT (253.0 mg/dL) and no-therapy (138.0 mg/dL) cohorts, respectively. All cohorts with the exception of the no-therapy cohort had a higher median TG level (OCT 188.0 mg/dL, monotherapy 176.0 mg/dL, statin monotherapy 167.0 mg/dL, other monotherapy 233.0 mg/dL) than that observed in the ACCORD lipid trial (162.0 mg/dL). In contrast to the TG levels, all lipid therapy cohorts had similar HDL-C levels. The highest mean HDL-C level was observed among women in the no-therapy cohort (46.2 ± 5.9 mg/dL); the lowest was observed among men in the SFCT cohort (38.3 ± 5.7 mg/dL). Men in the ACCORD lipid trial had lower mean HDL-C than women (36.6 mg/dL vs 41.4 mg/ dL). Wide variation between the cohorts was observed with LDL-C. The lipid therapy cohorts sorted by mean LDL-C level in decreasing order were as follows: monotherapy (133.3 ± 29.6 mg/dL), OCT (133.1 mg/dL ± 31.4), SFCT (122.7 ± 31.7 mg/dL), and no therapy (118.8 ± 26.8 mg/dL). The mean LDL-C level in the statin monotherapy cohort (136.3 ± 28.7 mg/dL) was higher than the mean LDL-C level in the other monotherapy cohort (120.1 ± 29.8 mg/dL). By contrast, the ACCORD lipid trial population had a mean baseline LDL-C level of 100.6 mg/dL, lower than that of any of the cohorts in the current study.

Table 4

reports the results of the logistic regression analyses of the clinical characteristics predictive of SFCT. Patients with a diagnosis of dyslipidemia and patients with other chronic ischemic heart disease had 2.8 (95% confi dence interval [CI] 2.059-3.851) and 2.4 (95% CI 1.422-4.146) times greater odds, respectively, of being prescribed SFCT versus no therapy, and 1.6 (95% CI 1.18-2.236) and 1.7 (95% CI 1.158-3.082) greater odds, respectively, of being prescribed SFCT versus statin monotherapy. Women were less likely than men to receive SFCT (odds ratio [OR] 0.63, CI 0.450-0.907). Patients with elevated LDL-C levels had greater odds of being prescribed SFCT versus no therapy (OR 1.009; 95% CI 1.003-1.014) and greater odds of being prescribed statin monotherapy versus SFCT (OR 0.987; 95% CI 0.982-0.992). Elevated TGs were weakly though significantly associated with increasing odds of being prescribed SFCT versus monotherapy (OR 1.007, CI 1.005- 1.008) or statin monotherapy (OR 1.008, CI 1.007-1.010).

Relative to SFCT, the odds of being treated with statin-ezetimibe (OR 0.992; 95% CI 0.990-0.994) or statinniacin (OR 0.995; 95% CI 0.992-0.997) combination therapy decreased when the patient had elevated TG levels (

Table 5

). In contrast, the odds of being treated with statin-ezetimibe combination therapy increased in the presence of elevated LDL-C levels (OR 1.012; 95% CI 1.007-1.018), and the odds of statin-niacin combination therapy increased in the presence of low HDL-C levels (OR 0.938; 95% CI 0.906-0.971). A diagnosis of dyslipidemia increased the odds being treated with SFCT versus statin-niacin combination therapy (OR 0.542; 95% CI 0.355- 0.826).

DISCUSSION

The baseline lipid panel of SFCT patients in a realworld setting who were selected by ACCORD lipid trial entry criteria did not resemble the baseline lipid panel of SFCT patients in the ACCORD lipid trial. In the clinical practice setting, patients receiving SFCT had higher TG levels (median 253 mg/dL) prior to start of the combination therapy compared with TG levels of patients in the ACCORD lipid trial27 (median 162 mg/dL).

The ACCORD lipid trial was designed to study T2DM patients with low levels of HDL-C and elevated TGs who had LDL-C at or within reach of expected statin-treated LDL-C treatment guidelines goal. The LDL-C levels of all cohorts in the managed care population of this study were higher than the LDL-C levels of patients in the ACCORD lipid trial²⁷ (100 mg/dL), possibly as a result of the majority of the patients in the trial (59%) being treated with statin monotherapy at baseline, whereas the managed care population was lipid therapy—naïve. In contrast, the mean HDL-C levels of all treatment groups in the real-world managed care cohort were slightly higher than the mean HDL-C levels for men and women in the ACCORD lipid trial.²⁷ Finally, the median TG levels in the ACCORD lipid trial (162.0 mg/dL)²⁷ were much lower than the TG levels of any of the lipid therapy cohorts in the real-world managed care population, and nearly 100 mg/dL lower than that of the SFCT cohort specifi cally. As atherogenic dyslipidemia associated with CMR includes both reduced HDL-C and elevated TGs, the median TG levels at a level close to

normal (150 mg/dL as defi ned by the National Cholesterol Education Program’s Adult Treatment Panel III²⁵) indicate that the ACCORD lipid trial population had less pronounced cases of classic atherogenic dyslipidemia compared with those seen in actual practice.

A prespecified subgroup analysis of the ACCORD lipid trial evaluated patients who had baseline TG levels in the highest tertile (>204 mg/dL) and HDL-C in the lowest tertile (<34 mg/dL).²⁷ This subgroup, which more closely represented patients with atherogenic dyslipidemia than did the overall ACCORD lipid trial population, were shown to derive cardiovascular benefit from the addition of fenofi brate to a statin. Consistent with these observations, SFCT in our study was prescribed preferentially to patients with a lipid case mix resembling that of the elevated TG and low HDL-C ACCORD lipid trial subgroup. This suggests that in the real world, patients who were treated with SFCT were those who would derive the most benefi t from combination therapy. The overall conclusions of the ACCORD lipid trial might have been overly infl uenced by a population with low (almost normal) TG levels.

Lipid therapy prescription patterns in the managed care cohort were generally consistent with current treatment recommendations of the National Cholesterol Education Program.²⁵ The SFCT cohort had a lipid case mix with higher TG and lower HDL-C levels than those of all other cohorts. In the SFCT cohort, LDL-C was elevated but only modestly so compared with the more signifi cant LDL-C elevation in the monotherapy and statin monotherapy cohorts. Relative to monotherapy, elevated TG levels consistently predicted combination therapy, particularly SFCT, whereas elevated LDL-C predicted use of statin monotherapy versus combination therapy. These results were consistent with current clinical guidelines that recommend statin monotherapy for

elevated LDL-C and more aggressive therapy for elevated non-HDL-C, including TGs.^1,25,26 Male sex and the presence of ischemic heart disease increased the odds of being treated with SFCT. This result was consistent with observations made over a 7-year period in a high-risk referral lipid clinic that studied 136 patients who were given SFCT, 69% of whom had a history of ischemic heart disease and 70% of whom were male.²⁹

The National Health and Nutrition Examination Survey 2003-2004 showed that only 35% of patients with dyslipidemia in United States were receiving lipid therapy and that only 16.8% of DM patients had all 3 lipid profi le components at consensus goal levels.³⁰ Our results were consistent with these observations, as the vast majority of the patients in our study received no therapy. Among those who did, the majority were prescribed monotherapy, usually a statin. Patients prescribed SFCT had elevated TG levels, but not all patients with elevated TGs were prescribed SFCT. Further, men in our study were treated with SFCT and other lipid therapies more often than were women. This lack of treatment does not necessarily correspond to a lower risk, however, as the Framingham Heart Study suggested the relative risk of CVD associated with DM was signifi cantly greater in women than in men.³¹ The treatment gap does not appear to be a result of men being treated early due to early CVD risk because the average ages of men and women treated with SFCT in our study were similar (men 51 years, women 52 years).

In our study we focused on newly treated patients (ie, dyslipidemia treatment—naïve patients) in order to identify the baseline laboratory data prior to start of therapy. It would be benefi cial to conduct this study using patient charts or electronic medical records in order to have clinical and family history information. With medical chart data, a future study could evaluate all patients who either had therapy added on or who had their lipid treatment altered. That would give insight into some of the reasoning behind the physician’s decision to add or switch therapy and would provide a fuller treatment paradigm beyond initial therapy. Because the ACCORD lipid trial population was a decade older than the population in the current study, it might be worthwhile evaluating the ACCORD lipid trial results in the Medicare population as well. Although the ACCORD lipid trial was limited to patients with T2DM, future research should consider evaluating dyslipidemia treatment in the rest of the population, especially in those with a prior cardiovascular event.

This study must be viewed within the context of its limitations. The sample size of the managed care cohort may limit the generalizability, and the selected time period of 180 days before and after the index date may have been insuffi cient to capture all relevant data. Lack of available contextual information caused us to select only lipid therapy—naïve patients, whereas the majority of the population in the ACCORD lipid trial had been receiving statin monotherapy before enrollment. Since the claims data would not have included all Medicare claims, we excluded individuals aged >65 years. Because of sample size restrictions, we included all individuals with diabetes (type 1 and type 2). The claims database on which we relied may have contained incomplete or inaccurate diagnostic coding data, a known problem of administrative claims data sets. The study group was a convenience sample based on information available in the claims database (selection bias). We did not have access to patient clinical information such as past medical histories or family histories that may have infl uenced the choice of lipid therapy, and we could not ascertain the consistency with which patients were given appropriate lipid therapy. These kinds of database limitations also prevented an accurate determination of how long patients with DM had that diagnosis, whereas the population enrolled in the ACCORD lipid trial had a median T2DM duration of 10 years. Despite our efforts to limit multicollinearity by excludingvariables such as nondyslipidemia treatment, we did not investigate interaction terms that could increase the odds of being treated with certain lipid therapies.

CONCLUSIONS

In general, the baseline cardiovascular risk characteristics (including lipid profi les) of patients in the ACCORD lipid trial who were randomized to receive treatment do not appear to be similar to the characteristics of patients receiving SFCT in a real-world setting. Patients prescribed SFCT in our study had more comorbidities, lower HDL-C levels, and higher TG levels than did those initiating statin monotherapy. Particularly with respect to elevated TGs, the SFCT cohort represented a lipid case mix that was markedly different from the lipid case mix of the overall population enrolled in the ACCORD lipid trial. However, patients in a real-world setting more closely resemble the ACCORD lipid trial subgroup with elevated TGs and reduced HDL-C that was shown to benefit from the addition of fenofi brate to statin monotherapy. These results were consistent with several previous fi brate studies and/or their subanalyses16-24 and established clinical recommendations25,26 that support the use of SFCT in patients with the atherogenic dyslipidemia lipid profile of reduced HDL-C and elevated TGs. Because very few patients in the ACCORD lipid trial were truly dyslipidemic (low HDL-C and/or elevated TGs), the ACCORD lipid trial may have underrepresented the patient population most likely to benefit from the addition of fenofi brate to statin monotherapy. The results of this study appear to indicatethat in a real-world setting combination therapy is being prescribed for the patient population likely to benefit.