Clinical usefulness of therapeutic concentration monitoring for imatinib dosage individualization: results from a randomized controlled trial.

PURPOSE: This study assessed whether a cycle of "routine" therapeutic drug monitoring (TDM) for imatinib dosage individualization, targeting an imatinib trough plasma concentration (C min) of 1,000 ng/ml (tolerance: 750-1,500 ng/ml), could improve clinical outcomes in chronic myelogenous leukemia (CML) patients, compared with TDM use only in case of problems ("rescue" TDM). METHODS: Imatinib concentration monitoring evaluation was a multicenter randomized controlled trial including adult patients in chronic or accelerated phase CML receiving imatinib since less than 5 years. Patients were allocated 1:1 to "routine TDM" or "rescue TDM." The primary endpoint was a combined outcome (failure- and toxicity-free survival with continuation on imatinib) over 1-year follow-up, analyzed in intention-to-treat (ISRCTN31181395). RESULTS: Among 56 patients (55 evaluable), 14/27 (52 %) receiving "routine TDM" remained event-free versus 16/28 (57 %) "rescue TDM" controls (P = 0.69). In the "routine TDM" arm, dosage recommendations were correctly adopted in 14 patients (median C min: 895 ng/ml), who had fewer unfavorable events (28 %) than the 13 not receiving the advised dosage (77 %; P = 0.03; median C min: 648 ng/ml). CONCLUSIONS: This first target concentration intervention trial could not formally demonstrate a benefit of "routine TDM" because of small patient number and surprisingly limited prescriber's adherence to dosage recommendations. Favorable outcomes were, however, found in patients actually elected for target dosing. This study thus shows first prospective indication for TDM being a useful tool to guide drug dosage and shift decisions. The study design and analysis provide an interesting paradigm for future randomized TDM trials on targeted anticancer agents.

Clinical usefulness of therapeutic concentration monitoring for imatinib dosage individualization: Results from the randomized controlled I-COME Trial

Introduction: Imatinib trough plasma concentrations (Cmin) have been correlated with treatment response in chronic myeloid leukemia (CML) patients. The use of Cmin monitoring for optimizing imatinib dosage (therapeutic drug monitoring [TDM]) is therefore proposed for patients with unsatisfying response or tolerance ("rescue TDM"). A cycle of "routine TDM" for dosage individualization could also be beneficial to prevent unfavorable events, yet its clinical usefulness has not been evaluated. We aimed to assess prospectively whether a "routine TDM" intervention targeting imatinib Cmin of 1000 ng/mL (tolerance, 750-1500 ng/mL) could improve efficacy, tolerance, and persistence on treatment compared with "rescue TDM" use only. Patients (or Materials) and Methods: The Swiss Imatinib COncentration Monitoring Evaluation (I-COME) study was a multicenter randomized controlled trial (ISRCTN31181395). Adult patients in chronic or accelerated phase CML receiving imatinib ≤5 years were eligible. Patients were randomly (1:1) allocated to receive "routine TDM" intervention or to serve as controls with access only to "rescue TDM". All had 1-year follow-up. The primary endpoint was a combined efficacy-safety outcome (failure- and toxicity-free survival without imatinib discontinuation), analyzed in intention-to-treat. Results: Among 56 CML recruited patients, 55 had their molecular and cytogenetic response measured. 14/27 of patients receiving "routine TDM" (52% [33%-71%]) remained event-free versus 16/28 of control patients with "rescue TDM" only (57% [39%-75%]; P=0.69). In the "routine TDM" group, dosage recommendations were adopted entirely in 50% of patients (median Cmin at study end, 895 ng/mL; CV = 33%). These patients had fewer unfavorable events (28% [5%-52%]) compared with patients not receiving the advised dosage (77% [54%-99%]; P = 0.03; median Cmin at study end, 648 ng/mL; CV = 38%). Conclusion: This first prospective target concentration intervention trial could not formally demonstrate a benefit of "routine TDM" of imatinib, especially due to a small patient number and limited prescriber's adherence to dosage recommendations. Nevertheless, the patients receiving the advised dosage more often met target concentrations and the combined outcome (efficacy, tolerance, and persistence). A cycle of routine TDM could thus be favorable, at least in patients eligible for dosage adjustment. Its usefulness should, however, be further confirmed in larger trials.

Changes in the use of broad-spectrum antibiotics after withdrawal of cefepime from the market: an interrupted time series analysis

Background and objective: Cefepime was one of the most used broad-spectrum antibiotics in Swiss public acute care hospitals. The drug was withdrawn from market in January 2007, and then replaced by a generic since October 2007. The goal of the study was to evaluate changes in the use of broad-spectrum antibiotics after the withdrawal of the cefepime original product. Design: A generalized regression-based interrupted time series model incorporating autocorrelated errors assessed how much the withdrawal changed the monthly use of other broad-spectrum antibiotics (ceftazidime, imipenem/cilastin, meropenem, piperacillin/ tazobactam) in defined daily doses (DDD)/100 bed-days from January 2004 to December 2008 [1, 2]. Setting: 10 Swiss public acute care hospitals (7 with\200 beds, 3 with 200-500 beds). Nine hospitals (group A) had a shortage of cefepime and 1 hospital had no shortage thanks to importation of cefepime from abroad. Main outcome measures: Underlying trend of use before the withdrawal, and changes in the level and in the trend of use after the withdrawal. Results: Before the withdrawal, the average estimated underlying trend (coefficient b1) for cefepime was decreasing by -0.047 (95% CI -0.086, -0.009) DDD/100 bed-days per month and was significant in three hospitals (group A, P\0.01). Cefepime withdrawal was associated with a significant increase in level of use (b2) of piperacillin/tazobactam and imipenem/cilastin in, respectively, one and five hospitals from group A. After the withdrawal, the average estimated trend (b3) was greatest for piperacillin/tazobactam (+0.043 DDD/100 bed-days per month; 95% CI -0.001, 0.089) and was significant in four hospitals from group A (P\0.05). The hospital without drug shortage showed no significant change in the trend and the level of use. The hypothesis of seasonality was rejected in all hospitals. Conclusions: The decreased use of cefepime already observed before its withdrawal from the market could be explained by pre-existing difficulty in drug supply. The withdrawal of cefepime resulted in change in level for piperacillin/tazobactam and imipenem/cilastin. Moreover, an increase in trend was found for piperacillin/tazobactam thereafter. As these changes generally occur at the price of lower bacterial susceptibility, a manufacturers' commitment to avoid shortages in the supply of their products would be important. As perspectives, we will measure the impact of the changes in cost and sensitivity rates of these antibiotics.

Example-based support vector machine for drug concentration analysis

Machine learning has been largely applied to analyze data in various domains, but it is still new to personalized medicine, especially dose individualization. In this paper, we focus on the prediction of drug concentrations using Support Vector Machines (S VM) and the analysis of the influence of each feature to the prediction results. Our study shows that SVM-based approaches achieve similar prediction results compared with pharmacokinetic model. The two proposed example-based SVM methods demonstrate that the individual features help to increase the accuracy in the predictions of drug concentration with a reduced library of training data.

Cost-effective experimental design to support modeling of concentration-response functions

Modeling concentration-response function became extremely popular in ecotoxicology during the last decade. Indeed, modeling allows determining the total response pattern of a given substance. However, reliable modeling is consuming in term of data, which is in contradiction with the current trend in ecotoxicology, which aims to reduce, for cost and ethical reasons, the number of data produced during an experiment. It is therefore crucial to determine experimental design in a cost-effective manner. In this paper, we propose to use the theory of locally D-optimal designs to determine the set of concentrations to be tested so that the parameters of the concentration-response function can be estimated with high precision. We illustrated this approach by determining the locally D-optimal designs to estimate the toxicity of the herbicide dinoseb on daphnids and algae. The results show that the number of concentrations to be tested is often equal to the number of parameters and often related to the their meaning, i.e. they are located close to the parameters. Furthermore, the results show that the locally D-optimal design often has the minimal number of support points and is not much sensitive to small changes in nominal values of the parameters. In order to reduce the experimental cost and the use of test organisms, especially in case of long-term studies, reliable nominal values may therefore be fixed based on prior knowledge and literature research instead of on preliminary experiments

Self-selection and risk selection on the Swiss health insurance market

Abstract This thesis presents three empirical studies in the field of health insurance in Switzerland. First we investigate the link between health insurance coverage and health care expenditures. We use claims data for over 60 000 adult individuals covered by a major Swiss Health Insurance Fund, followed for four years; the data show a strong positive correlation between coverage and expenditures. Two methods are developed and estimated in order to separate selection effects (due to individual choice of coverage) and incentive effects ("ex post moral hazard"). The first method uses the comparison between inpatient and outpatient expenditures to identify both effects and we conclude that both selection and incentive effects are significantly present in our data. The second method is based on a structural model of joint demand of health care and health insurance and makes the most of the change in the marginal cost of health care to identify selection and incentive effects. We conclude that the correlation between insurance coverage and health care expenditures may be decomposed into the two effects: 75% may be attributed to selection, and 25 % to incentive effects. Moreover, we estimate that a decrease in the coinsurance rate from 100% to 10% increases the marginal demand for health care by about 90% and from 100% to 0% by about 150%. Secondly, having shown that selection and incentive effects exist in the Swiss health insurance market, we present the consequence of this result in the context of risk adjustment. We show that if individuals choose their insurance coverage in function of their health status (selection effect), the optimal compensations should be function of the se- lection and incentive effects. Therefore, a risk adjustment mechanism which ignores these effects, as it is the case presently in Switzerland, will miss his main goal to eliminate incentives for sickness funds to select risks. Using a simplified model, we show that the optimal compensations have to take into account the distribution of risks through the insurance plans in case of self-selection in order to avoid incentives to select risks.Then, we apply our propositions to Swiss data and propose a simple econometric procedure to control for self-selection in the estimation of the risk adjustment formula in order to compute the optimal compensations.

Employment risk, unemployment insurance and search strategies: a disaggregated equilibrium approach with application to the Swiss labour market in the 1990-ies

The Organization of the Thesis The remainder of the thesis comprises five chapters and a conclusion. The next chapter formalizes the envisioned theory into a tractable model. Section 2.2 presents a formal description of the model economy: the individual heterogeneity, the individual objective, the UI setting, the population dynamics and the equilibrium. The welfare and efficiency criteria for qualifying various equilibrium outcomes are proposed in section 2.3. The fourth section shows how the model-generated information can be computed. Chapter 3 transposes the model from chapter 2 in conditions that enable its use in the analysis of individual labor market strategies and their implications for the labor market equilibrium. In section 3.2 the Swiss labor market data sets, stylized facts, and the UI system are presented. The third section outlines and motivates the parameterization method. In section 3.4 the model's replication ability is evaluated and some aspects of the parameter choice are discussed. Numerical solution issues can be found in the appendix. Chapter 4 examines the determinants of search-strategic behavior in the model economy and its implications for the labor market aggregates. In section 4.2, the unemployment duration distribution is examined and related to search strategies. Section 4.3 shows how the search- strategic behavior is influenced by the UI eligibility and section 4.4 how it is determined by individual heterogeneity. The composition effects generated by search strategies in labor market aggregates are examined in section 4.5. The last section evaluates the model's replication of empirical unemployment escape frequencies reported in Sheldon [67]. Chapter 5 applies the model economy to examine the effects on the labor market equilibrium of shocks to the labor market risk structure, to the deep underlying labor market structure and to the UI setting. Section 5.2 examines the effects of the labor market risk structure on the labor market equilibrium and the labor market strategic behavior. The effects of alterations in the labor market deep economic structural parameters, i.e. individual preferences and production technology, are shown in Section 5.3. Finally, the UI setting impacts on the labor market are studied in Section 5.4. This section also evaluates the role of the UI authority monitoring and the differences in the Way changes in the replacement rate and the UI benefit duration affect the labor market. In chapter 6 the model economy is applied in counterfactual experiments to assess several aspects of the Swiss labor market movements in the nineties. Section 6.2 examines the two equilibria characterizing the Swiss labor market in the nineties, the " growth" equilibrium with a "moderate" UI regime and the "recession" equilibrium with a more "generous" UI. Section 6.3 evaluates the isolated effects of the structural shocks, while the isolated effects of the UI reforms are analyzed in section 6.4. Particular dimensions of the UI reforms, the duration, replacement rate and the tax rate effects, are studied in section 6.5, while labor market equilibria without benefits are evaluated in section 6.6. In section 6.7 the structural and institutional interactions that may act as unemployment amplifiers are discussed in view of the obtained results. A welfare analysis based on individual welfare in different structural and UI settings is presented in the eighth section. Finally, the results are related to more favorable unemployment trends after 1997. The conclusion evaluates the features embodied in the model economy with respect to the resulting model dynamics to derive lessons from the model design." The thesis ends by proposing guidelines for future improvements of the model and directions for further research.