Forecasting Patient Outcomes in the Management of Hyperlipidemia FREE

CORONARY HEART disease (CHD) is the leading cause of death in men and women in the United States. Although CHD risk in women lags approximately 10 years behind that of men, the incidence of CHD in women increases progressively after menopause until ultimately as many women as men die of CHD.¹ While gender is one differential risk factor in the onset of CHD, strong evidence exists that other risk factors contribute to the prevalence and incidence of CHD.^2- 6 Risk factors identified as modifiable are cigarette smoking, hypertension, low high-density lipoprotein (HDL) levels, and diabetes mellitus. Risk factors identified as nonmodifiable are age and family history of premature CHD. Both modifiable and nonmodifiable risk factors can be present for many years before clinical CHD develops. This long lead time presents an opportunity to modify the risk factors to avoid or delay morbidity and mortality. However, a substantial investment in disease management is required before the long-term benefits are realized.

Epidemiological evidence supports a direct relationship between the level of total blood cholesterol, specifically low-density lipoprotein (LDL) concentration, and the rate of CHD.^3- 5,7 In addition, clinical trials have shown that a reduction in the LDL level leads to a reduction in the number of CHD events.^8- 19 Recently, several trials using angiographic assessment have demonstrated that lowering serum cholesterol levels slows progression and produces regression of coronary atherosclerosis in some patients.^17,20- 26 In 1993, the National Cholesterol Education Program's (NCEP's) Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults^1- 2 updated its 1988 report on cholesterol management and recommended linking treatment decisions to the patient's risk for CHD. Nonlipid risk factors, such as cigarette smoking, hypertension, obesity, physical inactivity, and diabetes mellitus, can be modified to help reduce CHD risk. Although all risk factors contribute to the clinical condition of CHD, the focus of this analysis is the modi fication of LDL cholesterol levels. Modification of LDL cholesterol levels are used by NCEP as a surrogate outcome of decreased cardiac morbidity and mortality. The purpose of this analysis is to approximate the number of patients from a population perspective that would reach NCEP target LDL levels based on drug intervention.

Many drug interventions are available to reduce LDL levels in patients. Whether a patient reaches an NCEP target goal for risk reduction from CHD depends on several factors, including the potency of the drug intervention, the starting LDL level of the patient, the target LDL level, genetic variations, and potential differences in drug metabolism. For this evaluation, appropriate patient identification strategies, interventions, and counseling on lifestyle modifications are assumed to have occurred; whatever benefit there was from diet and counseling has been achieved; and the drug therapy is the cause of further LDL reduction. The target goals for modification of the LDL risk factor are defined by the second report of the NCEP's Expert Panel¹ and are stratified for primary prevention (patients without established CHD) and secondary prevention (patients with established symptomatic CHD). Although there is evidence of the beneficial effects of reducing total cholesterol and/or increasing HDL levels,²⁷ total cholesterol/HDL-cholesterol ratios or LDL-HDL ratios were not included in the NCEP guidelines,^1- 2 and thus were not evaluated in this analysis.

Patients were stratified by risk categories, and target reductions in LDL concentration to reach treatment goals were based on data from the National Health and Nutritional Examination Survey III (NHANES III) (National Center for Health Statistics, unpublished data, 1993) and NCEP guidelines, which assume an LDL reduction of 10% with diet therapy alone.^1,28- 29 Primary and secondary prevention patients were assumed to respond to drug intervention with a similar percentage decrease in LDL level.³⁰

A MEDLINE search was conducted of the English-language medical literature published from 1984 to 1998, with emphasis on articles published since 1991. The specific intent of the search was to gather data on drug therapies classified as "major classes" by NCEP to evaluate the effect of the drug therapy on LDL levels in patients. Data on the response of different subgroups of patients, ie, those with CHD vs a general population cohort, were reviewed for evidence of a difference in LDL reduction.

Search terms included hyperlipidemia, hyperlipidemia/drug therapy, coronary heart disease, HMG-CoA reductase inhibitors, niacin, and bile acid sequestrants. The search was limited to the English language, human subjects, randomized controlled trials, and drug therapy. Articles meeting these criteria were reviewed for potential data extraction.

Articles were reviewed by one reviewer (K.L.B.). A determination of study quality was made on the basis of scientific rigor, study design, and publication in peer-reviewed journals, with the exception of data reviewed by the Food and Drug Administration as a component of the drug-approval process. General principles for reporting results of clinical trials in biomedical literature were applied to the studies.³¹ Large studies with randomization, blinding of investigators, and parallel placebo control, or crossover designs were preferred over small open-label studies. Data were collected on sample size, LDL reduction, mean LDL reduction, drug therapy, serial measurements of LDL reduction if a dose-titration study, and length of study. A weighted mean LDL reduction was calculated and used as a variable for the percentage of patients who reach target goal. The weighted mean was calculated from the total number of patients per study, multiplied by the mean LDL reduction. The number was summed with the result of similar calculations for other studies examining the same drug intervention. The resultant number was divided by the total (pooled) number of patients. An expert panel of practicing clinical consultants consisting of a cardiologist, an internist, a family practitioner, 2 pharmacists, and a dietician reviewed the reference selection to validate an appropriate representation of the literature. The physician consultants were selected from a list of consultants maintained by the military services offices of the surgeons general. The pharmacists and dietician were selected based on clinical reputation within the Department of Defense.

Data from NHANES III (phase I) (1988-1991) were obtained from the US Department of Health and Human Services, Public Health Services. These data contained risk factor information and LDL values for persons older than 20 years. The risk factors were stratified by presence of CHD, 2 or more risk factors without CHD, fewer than 2 risk factors without CHD, sex, and ethnicity. A cumulative frequency curve of LDL values for each risk category was constructed (Figure 1). This curve was used to estimate the percentage of patients expected to achieve the minimum LDL reduction to meet the NCEP guidelines.

To forecast patient outcomes, a spreadsheet (Lotus 1-2-3, release version 4.01 for Windows; Lotus Development Corp, Cambridge, Mass) was constructed to determine the maximum LDL level a person may have and still reach NCEP targets given the efficacy of a therapeutic intervention (Table 1). The formula was as follows:

Grahic Jump Location

The mean percentage LDL reduction was obtained from Table 2,^32- 93 and the maximum LDL level to reach NCEP target was read from Table 1. This maximum LDL level was then plotted in Figure 1, the intersection of the LDL level distribution curve was determined, and the cumulative percentage of patients reaching NCEP target for a particular drug therapy was determined.

For example, from Table 2, the mean percentage of LDL reduction for 20 mg of pravastatin sodium is 24.8%. In Table 1, the 24.8% (round the percentage to the nearest whole number) reduction in LDL corresponds to an approximate maximum LDL level of 4.45 mmol/L (172 mg/dL) to reach the 3.36-mmol/L (130-mg/dL) target NCEP LDL level. On the x-axis of Figure 1, 4.45 mmol/L (172 mg/dL) is noted and, moving vertically, the intersection of the "High-Risk Male" curve is determined. A horizontal line is drawn to the y-axis and the cumulative percentage of high-risk males who will achieve a target NCEP LDL level with 20 mg of pravastatin is noted to be approximately 87%. This process was conducted for all drug interventions and patient stratifications reported in Table 3.

Published data from the clinical trials on the LDL reduction provided by a drug therapy, merged with data from NHANES III, provide a useful scale for estimating the percentage of a patient population who will reach the target LDL level with a given therapy under ideal conditions (Table 3).

The anticipated outcome varies based on the intervention strategies for the patient risk category. While male and female have some small numerical differences in the percentage of population reaching target goal of therapy, it cannot be determined at this time if this is a true difference, an artifact of the imprecise risk factor classification of the NCEP guidelines, difference in drug therapy, or some other reason. The studies included in our analysis did not necessarily report data by sex of patients, and some studies did not include women in their protocols.

In addition to directly reading forecast outcomes for monotherapy from Table 3, it can also be used to forecast the outcomes of combination therapy. When combination therapy is required for lipid lowering, the LDL reduction with drug therapy is additive within normal dosing ranges.^{21,35- 36,45- 51,63,73,78,88,91,94- 98} For example, 20 mg of fluvastatin provides an approximate 20% LDL reduction as monotherapy.^36,45- 55 Niacin, 3 g, provides an approximate 21% LDL reduction as monotherapy.^19,51,76 When these drugs are administered as combination therapy, this combination provides approximately a 41% LDL reduction.^48,50- 51 A 41% LDL reduction corresponds to approximately 99% of both male and female patients in the primary prevention categories reaching NCEP target LDL levels, and about 81% and 86% of male and female patients, respectively, in the secondary prevention category reaching NCEP target LDL level.

Targeting the at-risk patients for prevention and treatment is potentially more efficient than offering population-wide efforts as evidenced by data from the Framingham Study⁷ and the Helsinki Heart Study.^10,99 Although the NCEP patient stratification is not as precise as other stratification models,¹⁰⁰ the NCEP treatment guidelines are relatively easy to use.

Patients with established CHD should be aggressively treated to prevent further cardiac events. With the exception of high-risk females, patients with established CHD apparently have a lower mean LDL level than do patients in most other cholesterol groups (Figure 1). However, since the LDL target is lower than in the other patient groups, the overall percentage of LDL reduction required is greater. To reach the treatment goal in this group, the use of a 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) reductase inhibitor is required. In addition, the HMG-CoA reductase inhibitors have been shown to decrease mortality in patients with CHD.^14,24,101

Patients with 2 or more risk factors for CHD, but without a diagnosis of CHD (primary prevention), generally do not need to be as aggressively treated as those with established CHD. A wide range of therapeutic interventions are available to the clinician.

Most patients with moderate primary hypercholesterolemia, with fewer than 2 risk factors, and without a diagnosis of CHD or without heterozygous familial hypercholesterolemia can be effectively treated with virtually any intervention. Clinical judgment should dictate whether diet therapy alone can reach treatment goals before initiating individualized drug therapy.

Patients who require a substantial reduction in LDL may not achieve it with monotherapy. This is because there is a diminishing return in LDL reduction expected from increasing the dose of a drug (Table 2). The lipid-lowering effects of many drugs, especially the HMG-CoA reductase inhibitors, tend to follow a log-linear relationship and reach a dose plateau. As the dose is increased, less of a response is realized.^32,79,91,102

In applying the results of this analysis, the limitations of the methods and the extrapolation into clinical practice must be considered. The first limitation is the trials evaluated for inclusion in the analysis. Since we primarily focused on randomized controlled trials from 1991 to the present, earlier trials are underrepresented. Seven years of clinical trials may not provide sufficient patient sampling to draw conclusions since some trials had very few patients. The second limitation is using a mean percentage LDL reduction to estimate the percentage of population who will reach target NCEP goal. In this study, we made the assumption that the observed mean percentage of LDL reduction approximated a true population mean percentage of LDL reduction, and there are minimal deviations of observations from the mean. The appropriateness of using a mean percentage of LDL reduction increases as the amount of deviation from the mean decreases. The third limitation is the extrapolation of sex difference from the studies. Many of the studies used in developing Table 3 are based on male-only data. We assumed patients in either of the 3 NCEP risk categories respond with a similar LDL level lowering with the same drug therapy. It is not clear if both male and female patients will have similar LDL level lowering with the same drug therapy. In a recent review, Walsh and Grady¹⁰³ noted that women have been underrepresented in clinical trials of the treatment of hyperlipidemia. The authors conclude that since the natural history of CHD differs in men and women, data from studies in men should not simply be extrapolated to women. A fourth limitation is that data were extracted by one reviewer (K.L.B.). While this makes the extraction process consistent, perhaps a better method would be to have 2 or more reviewers extract data from each study, then compare for deviations between reviewers as a control for observer bias. A fifth limitation is that new, more potent treatment regimens continued to be developed.^55,104

Population data and drug efficacy data can be used to forecast patient outcomes. This study used NCEP targets as goals of therapy. Meeting LDL level goals is recognized as a surrogate for a decrease in morbidity and mortality from coronary disease. Forecasting methods as presented here can be used to assess the potential burden of illness such as coronary artery disease in a population. While this analysis focuses on hyperlipidemia, other chronic disease states, such as hypertension, diabetes mellitus, or acid-peptic disorders, may be amenable to similar analysis depending on available data. Great potential exists to use forecasting methods in cost-effectiveness analysis and the development of disease-state treatment guidelines.

Accepted for publication June 10, 1998.

The views in this article are those of the authors and do not necessarily reflect the official policy or position of the US Departments of the Air Force, Army, or Navy or the Department of Defense of the US government.

Corresponding author: John R. Downs, MD, Department of Defense Pharmacoeconomic Center, 1750 Greeley Rd, Bldg 4011, Room 217, Fort Sam Houston, TX 78234-6190.