Greedy caliper propensity score matching can yield variable estimates of the treatment-outcome association-A simulation study.

PURPOSE: Greedy caliper propensity score (PS) matching is dependent on randomness, which can ultimately affect causal estimates. We sought to investigate the variation introduced by this randomness. METHODS: Based on a literature search to define the simulation parameters, we simulated 36 cohorts of different sizes, treatment prevalence, outcome prevalence, treatment-outcome-association. We performed 1:1 caliper and nearest neighbor (NN) caliper PS-matching and repeated this 1000 times in the same cohort, before calculating the treatment-outcome association. RESULTS: Repeating caliper and NN caliper matching in the same cohort yielded large variations in effect estimates, in all 36 scenarios, with both types of matching. The largest variation was found in smaller cohorts, where the odds ratio (OR) ranged from 0.53 to 10.00 (IQR of ORs: 1.11-1.67). The 95% confidence interval was not consistently overlapping a neutral association after repeating the matching with both algorithms. We confirmed these findings in a noninterventional example study. CONCLUSION: Caliper PS-matching can yield highly variable estimates of the treatment-outcome association if the analysis is repeated.

A novel method for predicting the budget impact of innovative medicines: validation study for oncolytics.

BACKGROUND: High budget impact (BI) estimates of new drugs have led to decision-making challenges potentially resulting in restrictions in patient access. However, current BI predictions are rather inaccurate and short term. We therefore developed a new approach for BI prediction. Here, we describe the validation of our BI prediction approach using oncology drugs as a case study. METHODS: We used Dutch population-level data to estimate BI where BI is defined as list price multiplied by volume. We included drugs in the antineoplastic agents ATC category which the European Medicines Agency (EMA) considered a New Active Substance and received EMA marketing authorization (MA) between 2000 and 2017. A mixed-effects model was used for prediction and included tumor site, orphan, first in class or conditional approval designation as covariates. Data from 2000 to 2012 were the training set. BI was predicted monthly from 0 to 45 months after MA. Cross-validation was performed using a rolling forecasting origin with e^|Ln(observed BI/predicted BI)| as outcome. RESULTS: The training set and validation set included 25 and 44 products, respectively. Mean error, composed of all validation outcomes, was 2.94 (median 1.57). Errors are higher with less available data and at more future predictions. Highest errors occur without any prior data. From 10 months onward, error remains constant. CONCLUSIONS: The validation shows that the method can relatively accurately predict BI. For payers or policymakers, this approach can yield a valuable addition to current BI predictions due to its ease of use, independence of indications and ability to update predictions to the most recent data.

Performance of prior event rate ratio adjustment method in pharmacoepidemiology: a simulation study.

PURPOSE: Prior event rate ratio (PERR) adjustment method has been proposed to control for unmeasured confounding. We aimed to assess the performance of the PERR method in realistic pharmacoepidemiological settings. METHODS: Simulation studies were performed with varying effects of prior events on the probability of subsequent exposure and post-events, incidence rates, effects of confounders, and rate of mortality/dropout. Exposure effects were estimated using conventional rate ratio (RR) and PERR adjustment method (i.e. ratio of RR post-exposure initiation and RR prior to initiation of exposure). RESULTS: In the presence of unmeasured confounding, both conventional and the PERR method may yield biased estimates, but PERR estimates appear generally less biased estimates than the conventional method. However, when prior events strongly influence the probability of subsequent exposure, the exposure effect from the PERR method was more biased than the conventional method. For instance, when the effect of prior events on the exposure was RR = 1.60, the effect estimate from the PERR method was RR = 1.13 and from the conventional method was RR = 2.48 (true exposure effect, RR = 2). In all settings, the variation of the estimates was larger for the PERR method than for the conventional method. CONCLUSION: The PERR adjustment method can be applied to reduce bias as a result of unmeasured confounding. However, only in particular situations, it can completely remove the bias as a result of unmeasured confounding. When applying this method, theoretical justification using available clinical knowledge for assumptions of the PERR method should be provided.

Reporting of covariate selection and balance assessment in propensity score analysis is suboptimal: a systematic review.

OBJECTIVES: To assess the current practice of propensity score (PS) analysis in the medical literature, particularly the assessment and reporting of balance on confounders. STUDY DESIGN AND SETTING: A PubMed search identified studies using PS methods from December 2011 through May 2012. For each article included in the review, information was extracted on important aspects of the PS such as the type of PS method used, variable selection for PS model, and assessment of balance. RESULTS: Among 296 articles that were included in the review, variable selection for PS model was explicitly reported in 102 studies (34.4%). Covariate balance was checked and reported in 177 studies (59.8%). P-values were the most commonly used statistical tools to report balance (125 of 177, 70.6%). The standardized difference and graphical displays were reported in 45 (25.4%) and 11 (6.2%) articles, respectively. Matching on the PS was the most commonly used approach to control for confounding (68.9%), followed by PS adjustment (20.9%), PS stratification (13.9%), and inverse probability of treatment weighting (IPTW, 7.1%). Balance was more often checked in articles using PS matching and IPTW, 70.6% and 71.4%, respectively. CONCLUSION: The execution and reporting of covariate selection and assessment of balance is far from optimal. Recommendations on reporting of PS analysis are provided to allow better appraisal of the validity of PS-based studies.

NSAID-antihypertensive drug interactions: which outpatients are at risk for a rise in systolic blood pressure?

BACKGROUND: Management guidelines for drug-drug interactions between non-steroidal anti-inflammatory drugs (NSAIDs) and antihypertensives recommend blood pressure monitoring in hypertensive patients. We measured the short-term effect of initiating NSAIDs on systolic blood pressure (SBP) in users of antihypertensives, aiming to investigate which outpatients are at risk for an increase in SBP in daily clinical practice. DESIGN: A cohort study with a nested case-control design in Dutch community pharmacies. METHODS: Patients with a drug-drug interaction alert for a newly initiated NSAID and antihypertensive were interviewed and their SBP was measured at T0, after one week (T1) and after two weeks (T2). We evaluated risk factors for exceeding a predefined limit of change (PLoC) in SBP (≥ 10 mmHg to ≥ 140 mmHg) at T1 and T2 versus T0. RESULTS: For 112 patients the SBP at T0 was measured. Two patients were excluded (T0 SBP ≥ 180 mmHg). PLoC was exceeded in 10 patients (10.4%) at T1 and in seven patients (8.0%) at T2. Patients using etoricoxib (odds ratio (OR), 21.0; 95% confidence interval (CI), 3.7-120.6) and patients using >1 defined daily dose of an NSAID (OR, 3.3; 95% CI, 1.1-10.0) were at increased risk of a rise in SBP. CONCLUSIONS: A newly initiated NSAID has an immediate clinically relevant effect on SBP in some users of antihypertensives. Management guidelines for NSAID-antihypertensive drug-drug interactions should advise SBP monitoring before and after initiation of an NSAID or intensification of NSAID therapy. Monitoring is especially relevant in patients prescribed high dosages of NSAIDs. Etoricoxib should not be used in hypertensive patients.

Propensity score balance measures in pharmacoepidemiology: a simulation study.

BACKGROUND: Conditional on the propensity score (PS), treated and untreated subjects have similar distribution of observed baseline characteristics when the PS model is appropriately specified. The performance of several PS balance measures in assessing the balance of covariates achieved by a specific PS model and selecting the optimal PS model was evaluated in simulation studies. However, these studies involved only normally distributed covariates. Comparisons in binary or mixed covariate distributions with rare outcomes, typical of pharmacoepidemiologic settings, are scarce. METHODS: Monte Carlo simulations were performed to examine the performance of different balance measures in terms of selecting an optimal PS model, thus reduction in bias. The balance of covariates between treatment groups was assessed using the absolute standardized difference, the Kolmogorov-Smirnov distance, the Lévy distance, and the overlapping coefficient. Spearman's correlation coefficient (r) between each of these balance measures and bias were calculated. RESULTS: In large sample sizes (n ≥ 1000), all balance measures were similarly correlated with bias (r ranging between 0.50 and 0.68) irrespective of the treatment effect's strength and frequency of the outcome. In smaller sample sizes with mixed binary and continuous covariate distributions, these correlations were low for all balance measures (r ranging between 0.11 and 0.43), except for the absolute standardized difference (r = 0.51). CONCLUSIONS: The absolute standardized difference, which is an easy-to-calculate balance measure, displayed consistently better performance across different simulation scenarios. Therefore, it should be the balance measure of choice for measuring and reporting the amount of balance reached, and for selecting the final PS model.

A 10-year analysis of the effects of media coverage of regulatory warnings on antidepressant use in The Netherlands and UK.

BACKGROUND: In 2003-2004 and 2007-2008, the regulatory banning of SSRI use in pediatrics and young adults due to concerns regarding suicidality risk coincided with negative media coverage. SSRI use trends were analyzed from 2000-2010 in the Netherlands (NL) and the UK, and whether trend changes might be associated with media coverage of regulatory warnings. METHODS: Monthly SSRIs sales were presented as DDDs/1000 inhabitants/day. SSRI-use trends were studied using time-series segmented regression analyses. Timing of trend changes was compared with two periods of media coverage of warnings. Annual Dutch SSRI prescription data were analyzed by age group. RESULTS: Trend changes in SSRI use largely corroborated with the periods of media coverage of warnings. British SSRI use declined from 3.9 to 0.7 DDDs/month (95%CI 3.3;4.5 & 0.5;0.9, respectively) before the first warning period (2003-2004). A small decrease of -0.6 DDDs/month (-1.2; -0.05) was observed in Dutch SSRI use shortly after 2003-2004. From 2007-2008, British SSRI use stabilized, whilst Dutch SSRI use diminished to -0.04 DDDs/month (-0.4;0.3). Stratified analyses showed a rapid decrease of -1.2 DDDs/month (-2.1; -1.7) in UK paroxetine use before 2003-2004, but only a minimal change in Dutch paroxetine use (-0.3 DDDs/month -0.8;0.2). Other SSRI use, especially (es)citalopram, increased during 2003-2004 in both countries. Significant reductions in Dutch paroxetine use were observed in pediatrics, adolescents, and young adults after 2003-2004. CONCLUSION: Changes in SSRI use (NL & UK) were associated with the timing of the combined effect of media coverage and regulatory warnings. Our long-term assessment illustrates that changes in SSRI use were temporal, drug-specific and more pronounced in pediatrics and young adults. The twofold increase in SSRI use over one decade indicates that regulatory warnings and media coverage may come and go, but they do not have a significant impact on the overall upward trend of SSRI use as a class in both countries.

Patterns of asthma medication use: early asthma therapy initiation and asthma outcomes at age 8.

Wheeze has many underlying pathophysiologies in childhood, but is the main reason for anti-asthma drugs prescription. This study was conducted to describe asthma medication use patterns among children in their first eight years of life. Longitudinal medication use data from 777 children participating in the PIAMA study were used. Medication patterns were described for four groups that started therapy before the third birthday, when the peak in prescriptions occurred in our cohort; short-acting ß-agonists (SABA), inhaled corticosteroids (ICS), SABA + ICS or none of these. One third (n = 255) of the children received a first SABA or ICS prescription before age 8. Only three children (1.2%) used medication continuously during follow-up. Of the children who started SABA, 53.8% discontinued within 1-2 years. Of the children who started ICS before age 3, 42.1% discontinued within 1-2 years and 31.6% received additional SABA. 41.5% of the children who started SABA + ICS used this short-term (≤1 -2 years) and 21.5% long-term (≥ 3 years). Fifteen percent of children who did not start asthma therapy in their first 3 years of life did receive prescriptions between age 3 and 8. Children prescribed SABA + ICS before age 3 had the highest prevalence of hyper responsiveness at age 8, and similar prevalence of atopy as the other groups. Asthma medication is prescribed frequently in the first 8 years of life, particularly before age 3, and only few children use it continuously. ICS and SABA prescription occurs especially in those who were more likely to develop signs of asthma at age 8.

Variable access to clopidogrel in a harmonized EU market.

OBJECTIVES: This study focuses on the different national coverage and reimbursement strategies and their consequences for access to clopidogrel, a drug with a central European Union (EU) registration. Our objectives are 1) to assess whether changes in reimbursement policies in EU member states influenced clopidogrel prescribing; and 2) to determine whether clopidogrel-specific policy characteristics, general characteristics of the health system, or indicators for the amount of cardiovascular care delivered were associated with the level of clopidogrel prescribing. METHODS: Data were collected in Austria, Belgium, Denmark, Germany, Hungary, Portugal, Slovenia, The Netherlands, and the United Kingdom (England). Utilization rates were expressed as defined daily doses (DDDs)/1000 persons/day. To determine whether changes in reimbursement policies influenced clopidogrel utilization, a segmented linear regression approach was used. RESULTS: Clopidogrel prescribing varied widely in the studied countries, from 2.76 (The Netherlands) to 6.83 (Belgium) DDDs/1000 persons/day (March 2005). Six countries had therapeutic indication restrictions to clopidogrel use. Health system characteristics did not explain variation in clopidogrel prescribing. CONCLUSION: A disconnect will be indicated in this study between the concept of a harmonized EU pharmaceuticals market and the reality in an individual member state. Although clopidogrel was centrally registered in the EU, policy measures at the national level result in different roles in clinical practice for this drug.

Distinguishing patterns in the dynamics of long-term medication use by Markov analysis: beyond persistence.

BACKGROUND: In order to accurately distinguish gaps of varying length in drug treatment for chronic conditions from discontinuation without resuming therapy, short-term observation does not suffice. Thus, the use of inhalation corticosteroids (ICS) in the long-term, during a ten-year period is investigated. To describe medication use as a continuum, taking into account the timeliness and consistency of refilling, a Markov model is proposed. METHODS: Patients, that filled at least one prescription in 1993, were selected from the PHARMO medical record linkage system (RLS) containing >95% prescription dispensings per patient originating from community pharmacy records of 6 medium-sized cities in the Netherlands.The probabilities of continuous use, the refilling of at least one ICS prescription in each year of follow-up, and medication free periods were assessed by Markov analysis. Stratified analysis according to new use was performed. RESULTS: The transition probabilities of the refilling of at least one ICS prescription in the subsequent year of follow-up, were assessed for each year of follow-up and for the total study period.The change of transition probabilities in time was evaluated, e.g. the probability of continuing ICS use of starters in the first two years (51%) of follow-up increased to more than 70% in the following years. The probabilities of different patterns of medication use were assessed: continuous use (7.7%), cumulative medication gaps (1-8 years 69.1%) and discontinuing (23.2%) during ten-year follow-up for new users. New users had lower probability of continuous use (7.7%) and more variability in ICS refill patterns than previous users (56%). CONCLUSION: In addition to well-established methods in epidemiology to ascertain compliance and persistence, a Markov model could be useful to further specify the variety of possible patterns of medication use within the continuum of adherence. This Markov model describes variation in behaviour and patterns of ICS use and could also be useful to investigate continuous use of other drugs applied in chronic diseases.