William P. Warburton

Major congenital malformations following prenatal exposure to serotonin reuptake inhibitors and benzodiazepines using population‐based health data

BACKGROUND To determine a population-based incidence of congenital anomalies following prenatal exposure to serotonin reuptake inhibitor (SRI) antidepressants used alone and in combination with a benzodiazepines (BZ). METHODS Population health data, maternal health, and prenatal prescription records were linked to neonatal records, representing all live births (British Columbia, Canada, N=119,547) during a 39-month period (1998-2001). The incidence and risk differences (RD) for major congenital anomalies (CA) and congenital heart disease (CHD), including ventricular and atrial septal defects (VSD, ASD), from infants of mothers treated with an SRI alone, a benzodiazepine (BZ) alone, or SRI+BZ in combinationcompared to outcomesno exposure. RESULTS Risk for a CA or CHD did increase following combined SRI+BZ exposure compared with no exposure. However, using a weighted regression model, controlling for maternal illness characteristics, combination therapy risk remained significantly associated only with CHD. The risk for an ASD was higher following SRI monotherapy compared with no exposure, after adjustment for maternal covariates. Dose/day was not associated with increased risk. CONCLUSIONS Infants exposed to prenatal SRIs in combination with BZs had a higher a incidence of CHD compared to no exposure, even after controlling for maternal illness characteristics. SRI monotherapy was not associated with an increased risk for major CA, but was associated with an increased incidence of ASD. Risk was not associated with first trimester medication dose/day.

Effects of timing and duration of gestational exposure to serotonin reuptake inhibitor antidepressants: population-based study

BACKGROUND Late-gestational serotonin reuptake inhibitor (SRI) exposure has been linked to adverse neonatal outcomes; however, the impact of timing and duration of exposure is unknown. AIMS To determine whether late-gestational exposure to an SRI is associated with increased risk of adverse neonatal outcome relative to early exposure. METHOD Population-based maternal and neonatal health records were linked to prenatal maternal prescription records for an SRI medication (n=3500). RESULTS After controlling for maternal illness and duration of exposure, using propensity score matching, neonatal outcomes did not differ between late and early exposure (P>0.05). After controlling for maternal illness, longer prenatal exposure increased the risks of lower birth weight, respiratory distress and reduced gestational age (P<0.05). CONCLUSIONS Using population health data, length of gestational SRI exposure, rather than timing, increased the risk for neonatal respiratory distress, lower birth weight and reduced gestational age, even when controlling for maternal illness and medication dose. These findings highlight the importance of distinguishing the specific impact of medication exposure from exposure to maternal illness itself.

A register study of the impact of stopping third trimester selective serotonin reuptake inhibitor exposure on neonatal health

Warburton W, Hertzman C, Oberlander TF. A register study of the impact of stopping third trimester selective serotonin reuptake inhibitor exposure on neonatal health.

Tightening a welfare system: the effects of benefit denial on future welfare receipt

We derive Local Average Treatment Effect estimates of the impact of welfare benefit denial on future receipt using a unique experiment involving reassessment of some applicants who were originally slated to receive benefits. We find evidence of considerable heterogeneity among applicants. Our results support a model with a peripheral group who exhibit scarring effects from being granted benefits and a core group who do not. The core group moves quickly back onto welfare when they are denied benefits. Even for the peripheral group, benefit denial has intermediate term but not permanent impacts.

Canada Needs Better Data for Evidence-Based Policy: Inconsistencies Between Administrative and Survey Data on Welfare Dependence and Education

This study compares administrative and survey data on BC welfare (social assistance) recipients, to test whether survey data is sufficiently accurate for use in policy-oriented research. BC welfare and education data is compared to the 1994 Public Use Microdata (BC sample) of Statistics Canadas Survey of Labour and Income Dynamics (SLID). BC 1994 SLID significantly understates welfare dependence, and overstates education levels of BC welfare recipients. Statistics Canada should lead a national initiative to make provincial administrative datasets available for research; and should use these data to improve key national longitudinal social research surveys such as SLID, NLSCY, and NPHS.

Population-Level Associations between Preschool Vulnerability and Grade-Four Basic Skills

Background This is a predictive validity study examining the extent to which developmental vulnerability at kindergarten entry (as measured by the Early Development Instrument, EDI) is associated with childrens basic skills in 4th grade (as measured by the Foundation Skills Assessment, FSA). Methodology/Principal Findings Relative risk analysis was performed on a large database linking individual-level EDI ratings to the scores the same children obtained on a provincial assessment of academic skills (FSA – Foundation Skills Assessment) four years later. We found that early vulnerability in kindergarten is associated with the basic skills that underlie populations of childrens academic achievement in reading, writing and math, indicating that the Early Development Instrument permits to predict achievement-related skills four years in advance. Conclusions/Significance The EDI can be used to predict childrens educational trends at the population level and can help select early prevention and intervention programs targeting pre-school populations at minimum cost.

The Impact of Placing Adolescent Males into Foster Care on Education, Income Assistance, and Convictions

Understanding the causal impacts of taking atrisk youth into government care is part of the evidence base for policy. Two sources of exogenous variation affecting alternative subsets of the atrisk population provide causal impacts interpreted as local average treatment effects. Placing 16 to18yearold males into care decreases or delays high school graduation, increases income assistance receipt, and has alternative effects on criminal convictions depending upon the instrument employed. This suggests that asking whether more or fewer children should be taken into care is insufficient; it also matters which, and how, children are taken into care.

What Went Wrong in the CETA Evaluations

Experience in estimating the impact of the U.S. Comprehensive Emnployment Training Act (CETA) programs seems to bear out our worst fears about selection bias. Impacts of those programs were estimated by comparing the subsequent incomes of participants with the incomes of non participants who were judged comparable. Different researchers, using essentially the same data came up with very different results. Westat, the primary contractor for the U.S. Department of Labor reported impacts which were positive overall and, for some groups, statistically significant (Bryant and Rupp 1987). Dickinson, Johnson and West reported impacts which were negative overall, and for some groups, statistically significant (Dickinson, Johnson and West 1986). This occurred despite the fact that the data used were exceptionally rich including long individual earnings histories as well as education, age, gender, family income, occupation, and labour force status. These discrepancies have been attributed by others to differences in methods used to control for selection bias (see Stromsdorfer et al. 1985). This note reports on the results of a study carried out with British Columbia data that seems to indicate that the largest source of discrepancy in the different CETA studies arises because estimates of the impact of CETA programs were made using annual data, and, for the most part, ignoring Unemployment Insurance (UI) and welfare dependence (see Warburton 1994 for the full results). As explained in the first section, our study uses information from a wage subsidy program in British Columbia to illustrate the magnitude of the changes in estimates of impact which result when earnings information is aggregated into annual totals, and welfare and UI dependence is largely ignored. The second section reports some of the main empirical findings on the nature of the pre-program earnings dip, and the third section relates ttlese findings to the CETA experience.

Reply: Letter to the editor

During the last decade, a large number of studies have established the characteristics of neonatal withdrawal after late pregnancy exposure to SSRIs and SNRIs. This pattern makes biological sense and, in many ways, follows the pattern of discontinuation syndrome reported in adults receiving these drugs. In 2002, we noted that this syndrome is characterized by respiratory distress, not described in other forms of neonatal withdrawal, such as from opioids or benzodiazepines (1). In the August 2010 issue of Acta Psychiatrica Scandinavica, Warburton and colleagues compared neonatal adverse effects in babies exposed to SSRIs during the last few weeks of gestation to babies whose mothers discontinued SSRIs in late pregnancy (2). They corroborated previous results, showing significantly more respiratory distress and seizures in babies exposed to term. But then the authors conducted multivariate analysis, controlling for what they defined as confounders, and the significance of the findings disappeared. This has led Warburton and colleagues to speculate that not drug discontinuation, but possibly other factors (e.g. maternal morbidity) may cause or contribute to these adverse neonatal effects. I believe there is a major problem in the conceptual framework leading to this analysis and, hence, to these far-reaching conclusions: When one tries to control for confounders, one has to have reasons to assume that those potential factors cause or contribute to the outcomes in questions (i.e. neonatal symptoms in this case). The confounders purportedly controlled for by the authors included total maternal drug exposure and maternal health. None of these has ever been associated with neonatal respiratory distress or seizures. Untreated maternal depression in late pregnancy has not been associated with neonatal respiratory distress. Respiratory distress is of course caused by hyaline membrane disease, aspiration, streptococcus b, and additional large differential diagnosis, but none of these were documented in this series, and none of these is associated with the time of discontinuation of SSRIs SNRIs (3). In other words, the authors took significant and biologically plausible results and diluted them to non-significance by controlling for factors not known to contribute to the phenomenon in question. In their report, Warburton et al. did not quote recent studies explicitly showing correlation between the decreasing levels of SNRIs in the neonate and the emergence of respiratory distress, strengthening the evidence that these are part of neonatal withdrawal.