A Health Economic Model to Assess the Long-Term Effects and Cost-Effectiveness of Orlistat in Obese Type 2 Diabetic Patients

Cost-effectiveness was analyzed using a Markov state transition model (15) and decision analytical software (DATA 3.5) (16). The Markov process technique simulates the evolution of a cohort of patients over time, when therapy is initiated with a specific treatment option. In Markov processes, time progresses in units of fixed length, called stages. At each stage, 6 months in our analysis, a patient is in a given health/treatment state. At each new stage, the patient can move from one state to another. In total, 20 stages (equal to 10 years) were considered. We applied this 10-year horizon because complications related to obesity and diabetes only become apparent after many years and because clinical guidelines in cardiology systematically refer to 10 years as a relevant period for assessing a patient’s cardiovascular risk (17).

The model evaluates the two possible scenarios in the considered patient population: orlistat versus no orlistat. All patients enter the model in a health state defined as “obese, type 2 diabetic, and free of complications.”

Complications of diabetes can be divided into micro- and macrovascular. Microvascular complications include end-stage renal disease, retinopathy, neuropathy, and foot ulcers. Macrovascular complications include coronary heart disease and ischemic stroke. In the model, every 6 months, diabetic patients have a risk to progress to one of these complication clusters. In other words, patients will be able to stay in the same state of health (without complications), move to another state with complications (micro- or macrovascular or both), or die. It was assumed that patients could not evolve directly to the “micro- and macrovascular complications” state. This is because no clinical data on this transition are available. In other words, we assume that within the same 6 months, a patient can, for example, not have a myocardial infarction (MI) and become blind, which is a conservative assumption.

The target population consists of obese type 2 diabetic patients without micro- or macrovascular complications. At start, four subgroups were defined depending on the presence or absence of hypertension and/or hypercholesterolemia.

To assess the effect of orlistat, ideally, a regression equation should be applied to predict the independent effect of the change in BMI on the incidence of complications and death (10,11). Unfortunately, for obese diabetic patients, such a regression equation does not exist. Therefore, the effect on mortality and micro- and macrovascular complications of treating patients with orlistat was assessed in two steps: 1) the effect of weight loss with orlistat on risk factors, and 2) the effect of risk factors on morbidity and mortality. It is thereby assumed that improving risk factors leads to a reduction of the number of complications, independently of the way these improvements are achieved (10).

Effects of weight loss with orlistat (first step) on the considered cardiovascular risk factors were published by Hollander et al. (5). In this double-blind placebo-controlled trial, 1 year of orlistat and hypocaloric diet was compared with placebo and hypocaloric diet in 391 obese type 2 diabetic patients with a mean BMI at study entry of 34.5 kg/m².

For every subgroup, a similar approach was followed, but baseline levels of risk factors and effects of orlistat may obviously differ. This is discussed in the following paragraphs.

The baseline data (no orlistat) regarding risk on micro- and macrovascular complications are based on the UKPDS substudy in obese patients treated with metformin (8).

Clark reported that over the range of HbA_1c concentrations from 7 to 11%, the relative risk of complications is independently reduced by 40% for every 10% reduction in HbA_1c concentration (18). Hollander et al. (5) reported HbA_1c levels of 7.50% at the time of randomization in both groups. After the 1-year study period, a decrease of 0.28 ± 0.09% was observed in the orlistat group compared with an increase of 0.18 ± 0.11% in the placebo group (P < 0.001) (5), i.e., an absolute difference of 0.46% in HbA_1c between the two treatment groups or a relative improvement of 6.0% was observed. Thus, if a 10% relative reduction of HbA_1c results in a 40% risk reduction in a linear way (18), a 6% relative reduction of HbA_1c is estimated to result in a 24% risk reduction on complications. The relation between the risk reductions of different complications obtained with metformin, published in the UKPDS (8), were then applied in order to calculate the relative reduction of each individual complication.

As an example, in Table 1, for this subgroup, the yearly event rate is shown for both treatment options. Because 6-month stages are used, transition probabilities had to be calculated for this time period using the following formula: Transition_{6 months} = 1 − (1 − Transition_{12 months})^6/12 (19).

In a patient group with LDL cholesterol ≥130 mg/dl treated with orlistat, a mean decrease of 17.05 mg/dl (mean base value 150 mg/dl) in LDL cholesterol was observed compared with 2.32 mg/dl in the placebo group (5). In comparison, in the diabetic patients included in the Helsinki Heart Study (HHS) (20), a decrease in LDL cholesterol of 21.9 mg/dl reduced the incidence of MI and cardiac death 68%. Building on recent evidence from Stratton et al. (21) and Adler et al. (22), we assumed that the effects of orlistat on LDL cholesterol were independent of effects on HbA_1c. Therefore, the risk reduction due to the effect on cholesterol was added to the risk reduction due to the effect on HbA_1c. Patients with diabetes and hypercholesterolemia will have a higher base risk for fatal and nonfatal MI than diabetic patients without hypercholesterolemia. Therefore, the base risk from the HHS, and not the base risk for the general obese diabetic population from the UKPDS, was applied.

In the UKPDS 38 trial, where intensive blood pressure control was studied in hypertensive diabetic patients, a reduction of diastolic blood pressure of 5 mmHg resulted in a significant decrease in complications (23). With orlistat, in hypertensive diabetes, a reduction of 7.29 mmHg was obtained (5). However, this was lower (but not significantly) than the reduction obtained with placebo (13.0 mmHg). Thus, compared with the placebo group, no extra effect of orlistat on blood pressure can be added. Only the impact on HbA_1c (0.48% absolute reduction) is accounted for in this population, starting from a base risk typical for this hypertensive subpopulation. In UKPDS 38, however, the general hypertensive diabetes population was described (23), so in order to calculate the base risk for obese hypertensive patients, a correction factor for obesity had to be applied. This correction factor was the ratio of risk for a given event in the conventionally treated obese UKPDS population (UKPDS 34) and the conventionally treated general UKPDS population (UKPDS 33) (8,24). The correction factors are 1.03 for MI, 1.10 for stroke, 1.09 for all-cause mortality, and 0.81 for microvascular disease.

Here, effects of orlistat on cholesterol are combined with the effect of orlistat in the previous subgroup. We calculated the base risk using the values of the hypertensive UKPDS cohort adapted for obesity, and the nonfatal and fatal MI correction factor was added for hypercholesterolemia (risk with hypercholesterolemia [HHS]/risk without hypercholesterolemia [UKPDS 34]) (20,24). The correction factors are 1.99 for nonfatal MI and 1.98 for fatal MI.

In the Hollander trial (5), an extra 4.2% of patients treated with orlistat were able to stop oral antidiabetes treatment. Because the blood glucose levels of these patients are not known, it is impossible to predict the risk of events for these patients, so, for reasons of conservatism, these patients remained in the diabetes group. Of course, for the calculation of resource use, drug use will be calculated for this small patient group (discussed further in detail).

According to European registration, orlistat can only be administered for 2 years. Thus, for the model, a 2-year administration of orlistat was assumed. It has been reported that 5 years after a weight reduction program is stopped, almost all weight is regained (6). If all weight is regained, risk on complications will be the same as in the not-treated group. Hence, in the model, risk for complications and mortality returns to placebo values in function of weight catch-up over the 5 years after discontinuation of orlistat. As a result, 7 years after study start, the placebo value is reached.

The perspective taken is that of the health care consumer. The cost of micro- and macrovascular complications in Belgium was studied in the Cost of Diabetes in Europe-Type 2 (CODE-2) study (25). The total health care cost was 1,726 Euro, 2,578 Euro, 3,844 Euro, and 5,443 Euro for 1 year for patients without complications, with microvascular complications, with macrovascular complications, and with both micro- and macrovascular complications, respectively. These costs were reported for the year 1998 and were adapted to year 2000 values using a 3% inflation rate. In the analysis, future costs have been discounted at 3%. There is no clear consensus in Europe with regard to the discounting of life years (LY). In our model, we applied a 0% discount rate for LY in the base case calculation and varied this figure in sensitivity analyses.

The drug cost for orlistat per patient per year from the public insurance perspective would be 881 Euro. Metformin, the base treatment for obese diabetic patients, costs 119 Euro per patient per year. As mentioned before, in the trial of Hollander et al. (5), an extra 4.2% of patients could stop taking oral antidiabetics, whereas an extra 10.1% had an average reduction in medication intake of 24.8%.

The results for the different subgroups are shown in Table 2. In obese type 2 diabetic patients without arterial hypertension (AHT) and without hypercholesterolemia, 0.08 LY can be gained over the study period. The incremental cost-effectiveness was 19,986 Euro per LY gained, i.e., approximately what is generally accepted as good value for money (26). The estimated mortality rate over the 10-year study period was 11.7% in the orlistat group compared with 12.7% in the control group. In other words, for every 100 patients treated with orlistat, one life can be saved.

In patients with hypercholesterolemia, 0.204 LY can be gained, resulting in a cost-effectiveness of 7,407 Euro/LYG.

The best results are obtained in diabetic patients with AHT and hypercholesterolemia. Here, a cost per LYG of 3,462 Euro has been found.

Different sensitivity analyses were applied in the different subgroups. Only the results in the group without AHT or hypercholesterolemia and the group with both AHT and hypercholesterolemia are shown, except for the effect of orlistat on cholesterol, where the two analyses in the hypercholesterolemic patients are shown (Table 3).

The effects of a 3% discount per year would increase cost per LYG from 19,986 to 23,522 Euro for obese diabetic patients without other risk factors and from 3,462 to 4,062 Euro for patients with AHT and hypercholesterolemia.

Reducing and increasing effectiveness of orlistat will alter risk for an event (death and micro- and macrovascular complications). Hollander observed a 0.46 ± 0.14% difference in HbA_1c between the two treatment groups. This is a relative difference of 6.0 ± 1.8%. A sensitivity analysis was performed using this standard deviation as interval. Thus, the decrease of risk on complications ranges between 16.7 and 31.3%. The resulting incremental cost-effectiveness ranges between 28,164 and 14,751 Euro/LYG for patients without other cardiovascular risk factors (Table 3).

If the risk reduction due to the effect on LDL cholesterol is only 50% of the value applied in the base case, incremental cost-effectiveness does not change significantly.

Finally, in the base case, we assumed that catch-up of weight and thus evolution to risk of the no weight loss group takes 5 years. If this catch-up period was only 2.5 years, incremental cost-effectiveness ratios would increase but remain very good for patients with AHT and hypercholesterolemia (Table 3).

The objective of this study was to predict long-term health and economic outcomes of orlistat in obese type 2 diabetic patients.

Therefore, we examined the impact of weight loss due to orlistat on different cardiovascular risk factors and extrapolated this impact to morbidity and mortality outcomes based on results of long-term trials in diabetic patients.

A 40% risk reduction obtained for every 10% relative reduction in HbA_1c was observed in patients with type 1 diabetes (18) and extrapolated to type 2 diabetes by other authors (10,11,18,27). This risk reduction was confirmed for type 2 diabetes by a more recent publication of the UKPDS investigators (21).

A debatable aspect of our analysis is that for patients with hypercholesterolemia, we independently added the beneficial effects on HbA_1c associated with weight loss caused by orlistat to the effects on LDL cholesterol. This may be an overestimation of the effects of orlistat. However, besides the evidence from Stratton et al. (21) and Adler et al. (22), a study on 290 diabetic patients with obesity and hypercholesterolemia showed that the effect of orlistat on cholesterol is independent of weight loss (28).

We applied results of the HHS to calculate risk reduction obtained with orlistat. Although both the HHS and the Hollander trial were in type 2 diabetic patients, there were some differences in the baseline characteristics (20). Mean BMI was higher in the Hollander trial (34.5 vs. 28.5 kg/m²), but mean LDL cholesterol level was lower (150 vs. 200 mg/dl). A higher baseline value facilitates the absolute decrease in LDL cholesterol. Hence, applying HHS data is disadvantageous for orlistat, and is thus a conservative approach.

The comparison in our evaluation was performed versus the placebo group of Hollander et al. Because both groups received the same lifestyle modifications (hypocaloric diet), improvements in cardiovascular risk factors were due to orlistat. In real practice, the improvements versus baseline may be weaker because of poor compliance with diets (29). The improvement of the active drug versus the comparator may be stronger in real life because the comparison is then between orlistat and “no drug for treating obesity” and not between orlistat and placebo.

Another element is that decreasing cholesterol will also decrease the number of strokes and retinopathy (microvascular disease). This is not taken into account because the HHS did not provide those data (20,30,31). Also, the study did not examine potential noncoronary effects of weight loss, such as the effect on cancer and certain musculo-skeletal disorders, e.g., knee osteoarthritis, and lower back pain.

Subgroup analyses were conducted related to hypertension and hypercholesterolemia. Unfortunately, no data were available to draw conclusions based on different baseline BMI levels in this population. Not only is the effect of BMI in diabetic patients poorly studied, but age also seems to be a factor that influences the impact of BMI, which would further complicate such an analysis (32).

It should finally be noted that the costs of treatment may vary widely from country to country, and thus the results shown here may not be directly applicable to other countries. On the other hand, Belgium is associated with rather low costs of complications compared with other countries, hence our results may again be conservative (33).

Incremental cost-effectiveness of orlistat for 2 years in patients with diabetes and obesity, free of events, and without hypertension or hypercholesterolemia is estimated at 19,986 Euro/LYG. This is approximately what is considered cost-effective (26), but sensitivity analysis showed that the conclusion of cost-effectiveness lacks robustness for these patients. Treating only obese diabetic patients with hypercholesterolemia or hypertension, and especially those with combined hypertension and hypercholesterolemia, widely falls within the range of what is generally accepted as good value for money. Sensitivity analyses prove the robustness of these results.

In conclusion, our results suggest that orlistat is cost-effective in the management of obese diabetic patients, especially with presence of hypercholesterolemia and/or hypertension. As with all new drugs, exercises such as this are important for timely assessment of the potential cost-effectiveness. Observational evidence on longer-term benefits of orlistat (>2 years) would be of importance for future decision-making and validation of our results.

This study was funded by a research grant from Roche Pharmaceuticals, Brussels, Belgium.