<< Back  

 Can't open the PDF? Click here for help.


A simulation model to estimate 10-year risk of coronary heart disease events in patients with schizophrenia spectrum disorders treated with second-generation antipsychotic drugs

Josep  Darbà, PhD

Department of Economics, University of Barcelona, Barcelona, Spain

Lisette  Kaskens, MSc

BCN Health Economics & Outcomes Research, Barcelona, Spain

Pedro  Aranda, MD

Hypertension and Cardiovascular Unit, Carlos Haya Hospital, Málaga, Spain

Celso  Arango, MD

Department of Psychiatry, Hospital General Universitario Gregorio Marañón, Centro de Investigación Biomédica en Red de Salud Mental (CIBERSAM), Madrid, Spain

Julio  Bobes, MD

CIBERSAM, Madrid, Spain , Medicine Department, Psychiatry Area, University of Oviedo, Oviedo (Asturias), Spain

Rafael  Carmena, MD

Department of Endocrinology, Valencia University Clinic Hospital, Valencia, Spain

Javier  Rejas, MD

Health Outcomes Research Department, Medical Unit, Pfizer España, Alcobendas (Madrid), Spain

BACKGROUND: The risk for cardiovascular (CV) events has been shown to be considerably higher among schizophrenia patients than the general population.

OBJECTIVE: The aim of this study was to describe a general stochastic simulation model for the treatment of schizophrenia related to CV-associated risks of second-generation antipsychotics (SGAs).

METHODS: A model to simulate the expected 10-year incidence of all types of coronary heart disease (CHD) events in patients treated with SGAs was developed from the Cardiovascular, Lipid and Metabolic Outcomes Research in Schizophrenia (CLAMORS) study to reproduce baseline conditions, The CHD event risk was estimated through a locally adjusted Framingham risk function using the expected mean change in the CV risk factors from the Clinical Antipsychotic Trials in Intervention Effectiveness (CATIE) study.

RESULTS: The 10-year CHD event rate after treatment with SGAs was 0.181, 0.179, 0.176, and 0.172 for olanzapine, quetiapine, risperidone, and ziprasidone, respectively. Relative risk was calculated relative to no treatment, and values were as follows: olanzapine, 1.03 ± 1.05 (95% CI, 0.74 to 1.42), quetiapine, 1.02 ± 1.05 (95% CI, 0.74 to 1.41), risperidone, 1.00 ± 0.99 (95% CI, 0.73 to 1.36), and ziprasidone, 0.97 ± 0.95 (95% CI, 0.72 to 1.31). There were approximately 25,269 CHD events over a 10-year period in schizophrenia patients treated with olanzapine, 25,157 events with quetiapine, 24,883 with risperidone, and 24,514 events with ziprasidone.

CONCLUSIONS: The estimated outcomes suggest that each SGA shows a different level of CV event risk, with ziprasidone showing the lowest rate without any association for increased risk of CHD.

KEYWORDS: cardiovascular events, coronary heart disease, Framingham function, modeling schizophrenia, second-generation antipsychotics



Independent risk factors, such as tobacco consumption, hypertension, hypercholesterolemia, and diabetes mellitus, are direct causes of ischemic cardiopathy.1-4 Mortality from any cause has been shown to be considerably greater among patients with schizophrenia than among the corresponding general population in different public health settings.5-8 In addition to lifestyle factors and increased suicidal tendencies, other reasons, such as premature development of cardiovascular (CV) disease, currently are considered to have an effect on excess mortality in such patients. In recent years, many studies have demonstrated that the prevalence and incidence of CV risk factors, such as high cholesterol level and other metabolic disorders (eg, hyperglycemia, diabetes, obesity, metabolic syndrome), are higher in patients with schizophrenia than in the general population, independent of any concomitant antipsychotic therapy these patients might receive.9-13 A number of hypotheses have been proposed to explain this association, among them the direct effect of antipsychotic drugs on lipid and carbohydrate metabolism.14 In the Clinical Antipsychotic Trials in Intervention Effectiveness (CATIE) study, the mean risk of serious fatal and nonfatal coronary heart disease (CHD) in 10 years, according to the Framingham function, was estimated to be 9.4% for males and 6.3% for females.15,16 These figures are higher than those reported for the general population in the United States—8.5% for males and 5.2% for females17—but are in line with those observed in the Cardiovascular, Lipid and Metabolic Outcomes Research in Schizophrenia (CLAMORS) study of schizophrenia patients in Spain, which found an overall risk of CHD (Framingham Function [FF]) of 6.8% (FF: 6.9), with a significantly higher risk in males than in females—7.5% and 4.5%, respectively.9

Taking into account all of these CV risk factors and their association with treatment with second-generation antipsychotics (SGAs), it could be of interest to estimate the development of CHD events for the clinical community when such drugs are taken over a long time period (ie, 10 years). Because there have been no long-term trials exploring this possibility until now, modeling techniques are important tools for approaching these problems. The aim of this study was to describe a general stochastic simulation model for the treatment of schizophrenia relative to CV-associated risks of SGAs.


Model description

The model simulates the occurrence of CHD events over a 10-year period that occur as a consequence of metabolic effects that schizophrenia spectrum disorder patients may experience as a result of ongoing SGA treatment. Four hypothetical cohorts were generated, 1 for each of the SGAs included in the model—olanzapine, quetiapine, risperidone, and ziprasidone. In each of these cohorts, an attempt was made to simulate the different health states that a patient may experience as a result of the possible increase in CHD event risk due to antipsychotic therapy (FIGURE 1). Each patient in the hypothetical cohort was “run through” the model to depict his or her CHD event experience over time. Before patients were run through the model, they were first assigned to an SGA treatment. Next, each patient was assigned an 18-month CHD event probability for every 18-month cycle the model was run, based on his or her assumed average CV risk factor values, personal characteristics (eg, age, sex), and assumed cycle-to-cycle variability in CV risk factor values around this mean. In this fashion, a history of cycle CV risk values could be constructed for each patient in the model. The other probabilities necessary to build the model were estimated from the published literature or calculated by difference.

FIGURE 1: Schematic of CV event simulation model for the study of schizophrenia patients treated with SGAs
Graphic representation of the different health statuses included in the stochastic Markov simulation model.
CV: cardiovascular; SGAs: second-generation antipsychotics.

The increase in CHD event risk may be attributed to higher values of some risk factors. To analyze this model, it was necessary to estimate the probability of having a CHD event in each cohort. This probability was estimated from the risk of undergoing a CHD event. Different models for predicting the risk of CV disease have been developed to determine the possible mechanisms that affect the increase in risk. Although many studies have investigated CV disease risks, the Framingham study constitutes a basic pillar of such research and has been widely used for therapeutic decision making, because it provides risk estimations based on the characteristics of individual patients.18 Variations of the FF consider different risk factors for their prediction. For this study, the Framingham-Wilson function2 was selected. This function consists of a simplified coronary prediction model, including blood pressure, cholesterol, low-density lipoprotein cholesterol categories, and patient characteristics (eg, smoking, diabetes status, age). The study population in the original Framingham-Wilson function was from Massachusetts. However, when extrapolated to European populations, particularly those of Mediterranean countries, the Framingham-Wilson function overestimated the CHD event risk.19-21 Because our study population was from Spain, we used an adaptation of the Framingham equation adjusted to the Spanish population.20

To estimate the CHD risk for a specific patient or cohort, data on this individual or cohort as well as data on the general population are required, because 2 functions are evaluated. First, a function is constructed with the values of the individual patient, and second, a function is constructed with average values of the risk factors of interest. With the exponential estimation of the difference of both functions, the 10-year probability of a CHD event can be estimated. To estimate the values of each cohort, it is necessary to know the variation that takes place in the CV risk factors over time with each antipsychotic treatment. In this study, the 10-year probability of a CHD event was extrapolated to a probability of 18 months. This probability was estimated for each cycle, as the initial age of the cohort changed with each cycle. The other probabilities that were necessary to estimate for the model included: probability of death due to CHD, probability of death due to other causes (based on age and sex), and probability of continuing antipsychotic treatment without a CHD event. A history of CHD event risk was constructed for each patient. The constructed model was used to project the expected impact of antipsychotic treatment on CHD events for patients with schizophrenia spectrum disorders.

Model estimation

Patient data required to estimate the model included the following: 1) the β coefficient by sex, 2) the distribution of mean CV risk factor values among patients with schizophrenia spectrum disorders, 3) the distribution of the expected change in CV risk factor values by treatment cycle for any given antipsychotic, 4) the mortality data from all causes excluding CHD events, by age and sex, 5) the mortality data from CHD events, and 6) the expected utility for each health state. The principal data were obtained from 3 clinical studies. The original function in the model was adapted to incorporate the prevalence of risk factors in Spain by using the results of the DORICA (Dyslipidemia, Obesity, and Cardiovascular Risk) study.20 This study reported the results of a pooled analysis of a regional, random cross-sectional population survey, with 14,616 patients (6,796 men, 46.4%). The DORICA study allowed us to obtain the β coefficients needed to compute a linear function for calculating the 10-year probability of a CHD event in men and women. Among women, an age-squared term was found to be significant, which was incorporated into the linear function2 (TABLE 1). The β coefficients in the DORICA study were estimated as discrete variables. To obtain better results with differences between each SGA, these coefficients have been reestimated for this model as continuous variables. In this model, the average values for risk factors used were reported in the CLAMORS study, a retrospective, cross-sectional, multicenter study that enrolled 1,452 patients (863 men, 60.9%), age 40.7 ± 12.2, who were receiving oral antipsychotic treatment for ≥12 weeks.22 The objective of this study was to assess the prevalence of CHD and metabolic syndrome in patients treated with 6 different antipsychotics in the Spanish population. The prevalence of the main CV risk factors was estimated by sex and by antipsychotic treatment (TABLE 2).

Changes in risk factors due to metabolic changes induced by the use of antipsychotics were analyzed in the CATIE study,23 a double-blind, active-control clinical trial comparing the effectiveness of second-generation and conventional antipsychotics—olanzapine, perphenazine, quetiapine, risperidone, and ziprasidone—administered to 1,493 patients (1,080 men, 74%), age 40.6 ± 11.1, for ≤18 months. The prevalence of the CV risk factors after treatment with an antipsychotic was estimated based on data from the CLAMORS and CATIE studies. Data on risk factor values from the CLAMORS study were chosen as baseline, and changes in these values due to changes in metabolic variables induced by antipsychotic treatment were taken from the CATIE study to obtain the prevalence of the main CV risk factors associated with each antipsychotic treatment of interest (TABLE 3). With the CATIE and CLAMORS data, we estimated the 2 functions (one calculated from the average values in the total population with schizophrenia and the other from the individual values in each antipsychotic treatment cohort) that allowed us to estimate the 10-year CV event risk.

Mortality data by age and sex from all causes of death, excluding CHD, were obtained from the Spanish National Statistics Institute, published in the Health Information System of the National Health System of Spain.24 Mortality data from CV events were obtained from the Anglo-Scandinavian Cardiac Outcomes Trial–Blood Pressure Lowering Arm (ASCOT-BPLA), a multicenter, prospective, randomized controlled trial in 19,257 patients with hypertension, age 40 to 79, who had ≥3 other CV risk factors.25


β coefficients from the DORICA study

Risk factors Female Male
  Age 0.3377 0.0438
  Age2 –0.0027
Cholesterol (mg/dL)    
  <160 –0.2614 –0.0066
  160 to 199 0.0000 0.0000
  200 to 239 0.2077 0.1769
  240 to 279 0.2439 0.5054
  ≥280 0.5351 0.6571
HDL cholesterol (mg/dL)    
  <35 0.8431 0.4974
  35 to 44 0.3780 0.2431
  45 to 49 0.1979 0.0000
  50 to 59 0.0000 –0.0511
  ≥60 –0.4295 –0.4866
Blood pressure (mm Hg)    
  Good (<120/<80) –0.0534 0.0023
  Normal (120 to 129/80 to 84) 0.0000 0.0000
  Normal to high (130 to 139/85 to 89) –0.0677 0.2832
  Grade I (140 to 159/90 to 99) 0.2629 0.5217
  Grade II (≥160/≥100) 0.4657 0.6186
  Yes 0.5963 0.4284
  Yes 0.2925 0.5234
Age2: square mean age; DORICA: Dyslipidemia, Obesity, and Cardiovascular Risk; HDL: high–density lipoprotein.


Primary CV risk factors from the CLAMORS study

CV risk factor Female Male Total
Mean SD Mean SD Mean SD
Age (years) 42.5 12.6 39.3 11.6 40.7 12.2
Sex (%) 60.9%   39.1%   100%  
Triglycerides (mg/dL) 134.3 83.9 154.1 93.4 145.9 89.5
Cholesterol, total (mg/dL) 203.5 43.5 200.3 42.4 201.9 42.8
HDL cholesterol (mg/dL) 52.7 16.4 48 15.4 49.9 15.9
Blood pressure  
  Systolic 124.4 16.7 128.8 15 127.1 15.8
  Diastolic 75.8 11.2 78.3 9.4 77.4 10.2
Diabetes 6.8% 4.7% to 8.9% 5.8% 4.2% to 7.4% 6.3% 5% to 7.5%
Smoker 40.4% 36.3% to 44.4% 62.3% 59.1% to 65.6% 53.7% 51.1% to 56.2%
Baseline glycemia 96.3 19.8 95.2 22.1 95.6 21.2
CLAMORS: Cardiovascular, Lipid and Metabolic Outcomes Research in Schizophrenia; CV: cardiovascular; HDL: high–density lipoprotein; SD: standard deviation.


Change in CV risk factors in the CATIE studya

Change from baseline Olanzapine Quetiapine Risperidone Ziprasidone
Blood glucose (mg/dL)        
  Mean 15.0 (2.8) 6.8 (2.5) 6.7 (2.0) 2.3 (3.9)
  Adjusted mean 13.7 (2.5) 7.5 (2.5) 6.6 (2.5) 2.9 (3.4)
Cholesterol (mg/dL)        
  Mean 9.7 (2.1) 5.3 (2.1) –2.1 (1.9) –9.2 (5.2)
  Adjusted mean 9.4 (2.4) 6.6 (2.4) –1.3 (2.4) –8.2 (3.2)
Triglycerides (mg/dL)        
  Mean 42.9 (8.4) 19.2 (10.6) –2.6 (6.3) –18.1 (9.4)
  Adjusted mean 40.5 (8.9) 21.2 (9.2) –2.4 (9.1) –16.5 (12.2)
aValues expressed as mean (SD).
CATIE: Clinical Antipsychotic Trials in Intervention Effectiveness; CV: cardiovascular; SD: standard deviation.

The model was developed in Microsoft Excel. Estimations were based on a probabilistic model, using probability distributions and Monte Carlo simulation techniques—which allowed estimations with numerous repetitions in large sample sizes—in order to reflect the uncertainty associated with each parameter (FIGURE 2).26 The probability distributions used depended on the collected observational data, the type of parameter estimated, and the estimation process used. TABLE 4 shows the stochastic parameters of the model. The first step was to develop a cohort with 100,000 subjects for each treatment. Each cohort started in the “schizophrenia treatment” state, projecting the expected impact of each antipsychotic treatment on CHD events for a 60-year temporal horizon. With this simulation, we obtained a mean cost and effectiveness for each cohort. This process of running the trial for each cohort and averaging their final values was repeated for 1,000 samples. Estimations for a dynamic model for a great number of trials and high number of samples reflect first- and second-order uncertainty, respectively. By running a large number of simulated cases, uncertainty with respect to model estimates (eg, mean CV risk factor values or expected change in CV risk factor values) is minimized and reflected in standard error calculations.

FIGURE 2: Graphic representation of the different health statuses included in the stochastic Markov simulation model
CV: cardiovascular; p: probability; SGA: second-generation antipsychotic.


Stochastic parameters and type of distribution used in the base case and alternative scenarios in the model

Stochastic parameters Probability distribution used in base case Probability distribution used in alternative case
  Female Normal Gamma
  Male Normal Gamma
  Total Normal Gamma
HDL cholesterol
  Female Normal Gamma
  Male Normal Gamma
  Total Normal Gamma
Cholesterol change
  Olanzapine Normal Normal
  Quetiapine Normal Normal
  Risperidone Normal Normal
  Ziprasidone Normal Normal
  Female Normal Beta
  Male Normal Beta
  Female Normal Beta
  Male Normal Beta
Probability of death due to CV event Beta Beta
CV: cardiovascular; HDL: high–density lipoprotein.

This model may be used to generate a variety of clinical outcomes that are of interest for the patients in 1 of the hypothetical cohorts that were assumed to receive 1 of the antipsychotic treatments of interest, such as: 1) the 10-year probability of a CHD event by sex in each cycle, 2) the 1-year probability of a CHD event by sex in each cycle, 3) the CHD event over a lifetime course of schizophrenia, and 4) relative risk (RR) of CHD relative to no treatment presented as mean ± standard deviation (95% CI).

Sensitivity analysis

To assess the robustness of the economic model, it is necessary to observe the changes in the results of the analysis when key variables are varied over a specified range. To evaluate this uncertainty, a sensitivity analysis was performed in an alternative scenario to the base-case analysis. In this alternative scenario, the change in base-case values was calculated using the adjusted mean, and the probability distribution was replaced by other statistical distributions that might also fit the study parameters. The probabilistic distributions of all parameters in the model were changed except for cholesterol and probability of death due to CV events.


The results of this model show estimations in CHD events over a 10-year period in a lifetime course of schizophrenia. In the base-case scenario with the mean change in base-case values (TABLE 3) and a normal probability distribution for all stochastic parameters, with the exception of “probability of death due to CHD event” that was characterized by a beta probability distribution, the 10-year mean risk probability of a CHD event following antipsychotic treatment was 0.181 for olanzapine, 0.179 for quetiapine, 0.176 for risperidone, and 0.172 for ziprasidone patients. The RR for a CHD event was calculated relative to no treatment and relative to ziprasidone, the latter because it was the drug showing the lower incremental risk.

RR values relative to no treatment were as follows: olanzapine, 1.03 ± 1.05 (95% CI, 0.74 to 1.42); quetiapine, 1.02 ± 1.05 (95% CI, 0.74 to 1.41); risperidone, 1.00 ± 0.99 (95% CI, 0.73 to 1.36); and ziprasidone, 0.97 ± 0.95 (95% CI, 0.72 to 1.31). RR values relative to ziprasidone were as follows: olanzapine, 1.06 ± 1.10 (95% CI, 0.75 to 1.48); quetiapine, 1.05 ± 1.10 (95% CI, 0.74 to 1.47); and risperidone, 1.03 ± 1.04 (95% CI, 0.74 to 1.42) (TABLE 6). Results by sex also are shown in TABLE 6.

There were 25,269 CHD events over a lifetime course of schizophrenia in patients treated with olanzapine; 25,157 events with quetiapine; 24,883 with risperidone; and 24,514 events with ziprasidone (TABLE 5). There were 9,389 deaths due to CHD over a lifetime course of schizophrenia in patients treated with olanzapine, and 9,341 deaths in patients treated with quetiapine, while the number of deaths associated with the use of risperidone was 9,228, and 9,074 with ziprasidone.

In the alternative scenario, the 10-year mean risk probability of a CHD event in patients following antipsychotic treatment was 0.152 for olanzapine, 0.152 for quetiapine, 0.151 for risperidone, and 0.148 for ziprasidone. The RR for CHD events was calculated relative to no treatment and relative to ziprasidone. RR values relative to no treatment were as follows: olanzapine, 1.01 ± 1.06 (95% CI, 0.72 to 1.40); quetiapine, 1.01 ± 1.04 (95% CI, 0.73 to 1.39); risperidone, 1.00 ± 1.00 (95% CI, 0.73 to 1.36); and ziprasidone, 0.98 ± 0.97 (95% CI, 0.73 to 1.33). RR values relative to ziprasidone were as follows: olanzapine, 1.02 ± 1.09 (95% CI, 0.73 to 1.44); quetiapine, 1.02 ± 1.07 (95% CI, 0.73 to 1.42); and risperidone, 1.02 ± 1.02 (95% CI, 0.74 to 1.40).

CHD events occurring over a lifetime course of schizophrenia in patients treated with olanzapine were 24,514 events; 24,314 events with quetiapine; 23,973 events with risperidone; and 23,667 events with ziprasidone. Deaths due to CHD events occurring over a lifetime course of schizophrenia in patients treated with olanzapine were 9,389, and 9,341 with quetiapine, whereas deaths associated with the use of risperidone were 9,228, and 9,074 with ziprasidone.


Excess in CHD events, all types combined, according to type of SGA relative to ziprasidone, base case

  Events per 100,000 Excess events per 100,000 vs ziprasidone Excess events per 100,000 vs baseline
Olanzapine 25,269 755 361
Quetiapine 25,157 643 249
Risperidone 24,883 369 –25
Ziprasidone 24,514 –386
Baseline 24,900 386
CHD: coronary heart disease; SGA: second–generation antipsychotic.


Relative risk of CV event vs no treatment and ziprasidone, pooled and according to sex, base–case

Comparison RR of CV event vs no treatment
  Women Men Pooled
Olanzapine 1.01 ± 1.01
(95% CI, 0.74 to 1.38)
1.07 ± 1.06
(95% CI, 0.77 to 1.48)
1.03 ± 1.05
(95% CI, 0.74 to 1.42)
Quetiapine 1.01 ± 1.01
(95% CI, 0.74 to 1.37)
1.04 ± 1.04
(95% CI, 0.76 to 1.44)
1.02 ± 1.05
(95% CI, 0.74 to 1.41)
Risperidone 1.00 ± 1.00
(95% CI, 0.73 to 1.36)
1.00 ± 1.00
(95% CI, 0.73 to 1.36)
1.00 ± 0.99
(95% CI, 0.73 to 1.36)
Ziprasidone 0.98 ± 0.98
(95% CI, 0.72 to 1.32)
0.97 ± 0.97
(95% CI, 0.72 to 1.31)
0.97 ± 0.95
(95% CI, 0.72 to 1.31)
Comparison RR of CV event vs ziprasidone
  Women Men Pooled
Olanzapine 1.03 ± 1.03
(95% CI, 0.75 to 1.42)
1.10 ± 1.09
(95% CI, 0.79 to 1.54)
1.06 ± 1.10
(95% CI, 0.75 to 1.48)
Quetiapine 1.03 ± 1.03
(95% CI, 0.75 to 1.42)
1.08 ± 1.07
(95% CI, 0.77 to 1.50)
1.05 ± 1.10
(95% CI, 0.74 to 1.47)
Risperidone 1.02 ± 1.02
(95% CI, 0.75 to 1.41)
1.03 ± 1.02
(95% CI, 0.75 to 1.41)
1.03 ± 1.04
(95% CI, 0.74 to 1.42)
CV: cardiovascular; RR: relative risk.


In developing countries, CV diseases are becoming the leading cause of death due to alarming increases in obesity, sedentary lifestyles, smoking, and improvements in prevention and treatment of malnutrition and infection.2 Patients with schizophrenia have a 20% shorter life expectancy (ie, age 76 vs age 61) than non-schizophrenia patients.11 Approximately 50% of people in the general population die from CV diseases compared with approximately 75% of patients with schizophrenia (RR = 1.5).11 Schizophrenia patients have considerably higher rates of CV risk factors such as obesity, dyslipidemia, hypertension, diabetes, and cigarette smoking than the general population.4,11,27,28 The extent of CV risk observed in schizophrenia patients is evident from mortality studies that show they experience a mortality rate from CV causes twice as high as the general population.29,30 Data from the CATIE study found that schizophrenia patients have a substantially higher 10-year risk for major CV events16 and twice the prevalence of metabolic syndrome31 compared with the demographic characteristics of the general population. There is little debate about the fact that the high prevalence of metabolic syndrome is a major public health problem for schizophrenia patients in many Western countries.22,32-34 Moreover, the CATIE study revealed the marked undertreatment of common metabolic problems such as diabetes, hypertension, and dyslipidemia in schizophrenia patients.

Different models have been developed to determine the cost and effectiveness of antipsychotic treatment. These models have focused on classic effectiveness measures, such as reducing negative symptoms, preventing relapse, improving quality of life in psychosis, cognitive aspects, and improvement in total Positive and Negative Syndrome Scale score over time.35-39 Despite existing clinical evidence confirming the potential CV risk to schizophrenia patients, the adverse events measured in this model have not been the subject of previous modeling. In this article, a general, stochastic simulation model was developed to show the CHD event risk associated with treatment with the most frequently used SGA drugs in the Spanish National Health System. The main inputs of the model also may be adapted to different populations. The necessary information for adaptation of the model and the ß coefficients to different European populations20 or American populations2 is published. Therefore, the model can be used to project the expected impact of antipsychotic treatment on CHD events for different patient populations. Moreover, by adding the resource utilization used and utilities for each health state, it is possible to estimate the incremental cost-effectiveness ratios and the cost-effectiveness acceptability curves (CEACs). Given the stochastic nature of the model, extra consideration may be given to the CEACs. Cost-effectiveness for each antipsychotic drug could be estimated for a certain willingness to pay (WTP) threshold value; thus, the most cost-effective treatments could be obtained for different WTP values.

The results of this simulation suggest that differences in metabolic changes observed in the CATIE study translate into varying risk levels for CV events over a 10-year period for the 4 SGAs included in this study. Ziprasidone showed the lowest level due to medication-attributable risk for CV events owing to its reduction in patient weight, total cholesterol, and triglyceride levels observed in the CATIE study. Our model showed that olanzapine, quetiapine, and risperidone were associated with an increased CV event risk compared with ziprasidone. These results are consistent with the observed changes in weight and lipid profile observed in the CATIE study. Of interest is the observed risk for ziprasidone, lower than the theoretical no-treatment option, which translates into a reduction of expected CV events when applied to a hypothetical cohort of 100,000 schizophrenia patients. This reduction also was observed for risperidone, but to a lesser extent, whereas quetiapine and olanzapine produced higher numbers of CV events when applied to the hypothetical cohort. The sensitivity analysis performed in the alternative scenario showed results similar to the base-case scenario. Results of the simulation seem robust.

These results may have implications for understanding excess mortality in schizophrenia. Further research is needed to estimate the economic burden of medication-related comorbid medical conditions, as well as the possible reduction in total medical and societal burden linked to greater use of metabolically neutral antipsychotics. The stochastic nature of the model allows for any interpatient and intrapatient variability in CV risk. In this way, the model can serve as the basis for new studies designed to determine other clinical outcomes. Adding health care resources and health state utilities to such trials would provide economic results and offer a foundation for formal evaluations of the cost-effectiveness and cost-utility of alternative antipsychotic drugs for the treatment of schizophrenia spectrum disorders.


This modeling may have some limitations. First, we used a modified risk equation for the Spanish environment, resulting in a lower incidence of CV events when applied to other geographical contexts, such as Northern European countries where higher risk of CV deaths have been shown. The use of international studies as sources for input parameters in the model also could influence the outcomes of the model. In the absence of clinical trial data on the effectiveness of second-generation and conventional antipsychotics and CV event mortality data within the Spanish setting, we used data from international studies such as the CATIE study and the ASCOT-BPLA study with similar study populations. However, it should be noted that differences between the patient populations of these studies and Spanish patients may exist and could affect the outcomes. Another possible limitation is inherent in the mathematical nature of this modeling and simulates future events rather than follow a cohort of patients during the simulated time period. In that regard, the model may have under- or over-estimated the real rate of CHD events, particularly if we take into account that the model assumes that patients will be exposed to a particular SGA for a period of 10 years, which may not reflect actual clinical practice. However, the flexibility of the model allows for further modeling of different situations in patient management and treatment adherence. A further limitation could be that we do not have data on a naÏve cohort of patients not exposed to SGAs; such data would allow us to check cholesterol, glucose, triglycerides, and weight, and then precisely estimate the differential risk of CHD events vs no treatment. Finally, the results of this work apply only to the SGAs analyzed in this model.


Despite the above limitations, a general and adaptable model has been developed with stochastic simulation of the CV risk of SGA therapy. The estimated clinical outcomes suggest that each SGA shows a different level of CHD event risk, with ziprasidone showing the lowest rate without an association with an increased risk of CHD. Primary prevention strategies should include the choice of antipsychotic drug regimens that do not adversely affect the major risk of death in such patients.

DISCLOSURES: Dr. Darbà and Ms. Kaskens receive grant or research support from Pfizer, S.L.U. Drs. Aranda, Arango, Bobes, Carmena, and Rejas report no financial relationship with any company whose products are mentioned in this article or with manufacturers of competing products.

Data collection, analysis, and writing assistance were funded by Pfizer, S.L.U. All authors had complete access to the data, participated in the analysis and/or interpretation of results, and drafted the manuscript. All authors were responsible for the design of the model and its internal validity. All authors were responsible for interpretation of data. All authors participated in the manuscript preparation.


  1. Kannel WB. Contributions of the Framingham Study to the conquest of coronary artery disease. Am J Cardiol. 1988;62:1109–1112.
  2. Wilson PW, D’Agostino RB, Levy D, et al. Prediction of coronary heart disease using risk factor categories. Circulation. 1998;97:1837–1847.
  3. Grundy SM, Bazzarre T, Cleeman J, et al. Prevention Conference V: Beyond secondary prevention: identifying the high-risk patient for primary prevention: medical office assessment: Writing Group I. Circulation. 2000;101:E3–E11.
  4. Pocock SJ, McCormack V, Gueyffier F, et al. A score for predicting risk of death from cardiovascular disease in adults with raised blood pressure, based on individual patient data from randomised controlled trials. BMJ. 2001;323:75–81.
  5. Brown S, Inskip H, Barraclough B. Causes of the excess mortality of schizophrenia. Br J Psychiatry. 2000;177:212–217.
  6. Babidge NC, Buhrich N, Butler T. Mortality among homeless people with schizophrenia in Sydney Australia: a 10-year follow-up. Acta Psychiatr Scand. 2001;103:105–110.
  7. Hiroeh U, Appleby L, Mortensen PB, et al. Death by homicide, suicide, and other unnatural causes in people with mental illness: a population-based study. Lancet. 2001;358:2110–2112.
  8. Morgan MG, Scully PJ, Youssef HA, et al. Prospective analysis of premature mortality in schizophrenia in relation to health service engagement: a 7.5-year study within an epidemiologically complete, homogeneous population in rural Ireland. Psychiatry Res. 2003;117:127–135.
  9. Bobes J, Rejas J, Garcia-Garcia M, et al. Weight gain in patients with schizophrenia treated with risperidone, olanzapine, quetiapine or haloperidol: results of the EIRE study. Schizophr Res. 2003;62:77–88.
  10. De Hert MA, van Winkel R, Van Eyck D, et al. Prevalence of the metabolic syndrome in patients with schizophrenia treated with antipsychotic medication. Schizophr Res. 2006;83:87–93.
  11. Hennekens CH, Hennekens AR, Hollar D, et al. Schizophrenia and increased risks of cardiovascular disease. Am Heart J. 2005;150:1115–1121.
  12. Lindenmayer JP, Czobor P, Volavka J, et al. Changes in glucose and cholesterol levels in patients with schizophrenia treated with typical or atypical antipsychotics. Am J Psychiatry. 2003;160:290–296.
  13. Van Gaal LF. Long-term health considerations in schizophrenia: metabolic effects and the role of abdominal adiposity. Eur Neuropsychopharmacol. 2006;16:S142–S148.
  14. Basu R, Brar JS, Chengappa KN, et al. The prevalence of the metabolic syndrome in patients with schizoaffective disorder—bipolar subtype. Bipolar Disord. 2004;6:314–318.
  15. Meyer JM, Nasrallah HA, McEvoy JP, et al. The Clinical Antipsychotic Trials Of Intervention Effectiveness (CATIE) Schizophrenia Trial: clinical comparison of subgroups with and without the metabolic syndrome. Schizophr Res. 2005;80:9–18.
  16. Goff DC, Sullivan LM, McEvoy JP, et al. A comparison of ten-year cardiac risk estimates in schizophrenia patients from the CATIE study and matched controls. Schizophr Res. 2005;80:45–53.
  17.  Centers for Disease Control and Prevention (CDC). Prevalence of coronary heart disease--United States 2006-2010. MMWR Morb Mortal Wkly Rep. 2011;60:1377–1381.
  18. Dawber TR, Meadors GF, Moore FE Jr. Epidemiological approaches to heart disease: the Framingham Study. Am J Public Health Nations Health. 1951;41:279–281.
  19. Marrugat J, Solanas P, D’Agostino R, et al. Coronary risk estimation in Spain using a calibrated Framingham function [in Spanish]. Rev Esp Cardiol. 2003;56:253–261.
  20. Aranceta J, Pérez Rodrigo C, Foz Sala M, et al. Tables of coronary risk evaluation adapted to the Spanish population: the DORICA study [in Spanish]. Med Clin (Barc). 2004;123:686–691.
  21. Cristóbal J, Lago F, de la Fuente J, et al. Comparison of coronary risk estimates derived using the Framingham and REGICOR equations [in Spanish]. Rev Esp Cardiol. 2005;58:910–915.
  22. Bobes J, Arango C, Aranda P, et al. Cardiovascular and metabolic risk in outpatients with schizophrenia treated with antipsychotics: results of the CLAMORS Study. Schizophr Res. 2007;90:162–173.
  23. Lieberman JA, Stroup TS, McEvoy JP, et al. Effectiveness of antipsychotic drugs in patients with chronic schizophrenia. N Engl J Med. 2005;353:1209–1223.
  24.  Ministry of Health, Social Services and Equality. Series 1981-2007: Mortality by cause of death Spain and regions [in Spanish]. http://www.msc.es/estadEstudios/estadisticas/estadisticas/estMinisterio/mortalidad/seriesTablas.htm. Accessed September 20, 2012.
  25. Dahlöf B, Sever PS, Poulter NR, et al. Prevention of cardiovascular events with an antihypertensive regimen of amlodipine adding perindopril as required versus atenolol adding bendroflumethiazide as required, in the Anglo-Scandinavian Cardiac Outcomes Trial-Blood Pressure Lowering Arm (ASCOT-BPLA): a multicentre randomised controlled trial. Lancet. 2005;366:895–906.
  26. Darbà J. Use of probabilistic methods in economic evaluation of health technologies [in Spanish]. Gac Sanit. 2006;20:74–77.
  27. Altamura AC, Bobo WV, Meltzer HY. Factors affecting outcome in schizophrenia and their relevance for psychopharmacological treatment. Int Clin Psychopharmacol. 2007;22:249–267.
  28. Michelsen JW, Meyer JM. Cardiovascular effects of antipsychotics. Expert Rev Neurother. 2007;7:829–839.
  29. Osby U, Correia N, Brandt L, et al. Mortality and causes of death in schizophrenia in Stockholm county, Sweden. Schizophr Res. 2000;45:21–28.
  30. Colton CW, Manderscheid RW. Congruencies in increased mortality rates years of potential life lost, and causes of death among public mental health clients in eight states. Prev Chronic Dis. 2006;3:A42.
  31. McEvoy JP, Meyer JM, Goff DC, et al. Prevalence of the metabolic syndrome in patients with schizophrenia: baseline results from the Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE) schizophrenia trial and comparison with national estimates from NHANES III. Schizophr Res. 2005;80:19–32.
  32. Cohn T, Prud’homme D, Streiner D, et al. Characterizing coronary heart disease risk in chronic schizophrenia: high prevalence of the metabolic syndrome. Can J Psychiatry. 2004;49:753–760.
  33. Hägg S, Lindblom Y, Mjörndal T, et al. High prevalence of the metabolic syndrome among a Swedish cohort of patients with schizophrenia. Int Clin Psychopharmacol. 2006;21:93–98.
  34. De Hert M, van Winkel R, Van Eyck D, et al. Prevalence of diabetes, metabolic syndrome and metabolic abnormalities in schizophrenia over the course of the illness: a cross-sectional study. Clin Pract Epidemiol Ment Health. 2006;2:14.
  35. Heeg B, Buskens E, Knapp M, et al. Modelling the treated course of schizophrenia: development of a discrete event simulation model. Pharmacoeconomics. 2005;23:17–33.
  36. Eklund M, Bäckström M. A model of subjective quality of life for outpatients with schizophrenia and other psychoses. Qual Life Res. 2005;14:1157–1168.
  37. Morris BJ, Cochran SM, Pratt JA. PCP: from pharmacology to modelling schizophrenia. Curr Opin Pharmacol. 2005;5:101–106.
  38. Woodward TS, Menon M, Hu X, et al. Optimization of a multinomial model for investigating hallucinations and delusions with source monitoring. Schizophr Res. 2006;85:106–112.
  39. Schwegler H. Phenomenological modelling of some mechanisms in schizophrenia. Pharmacopsychiatry. 2006;39:S43–S49.

CORRESPONDENCE: Josep Darbà, PhD, Department of Economics, University of Barcelona, Diagonal 690, 08034 Barcelona, Spain E-MAIL: darba@ub.edu