Validity of maternal recall of prescription drug use during pregnancy

Marleen van Gelder1

Iris van Rooij1, Hermien de Walle2,

Nel Roeleveld1, Marian Bakker2

1) Department of Epidemiology, Biostatistics and HTARadboud University Nijmegen Medical Centre

2) Eurocat Northern Netherlands, Department of Genetics, University Medical Center Groningen

EUROCAT 2011 symposium, June 17 20 2011, Antwerp

Study design in birth defects epidemiology

• Majority of studies used case-control design

• Method of data collection:Medical records

Self-administered questionnaire

Interview

• Often data collection takes place monthsto years after delivery

• Data on the accuracy of recall of prescription drug use during pregnancy are scarce

Research questions

1. What is the validity of a self-administered questionnaire to assess prescription drug use just before and during pregnancy?

2. Do maternal and pregnancy characteristics influence the accuracy of questionnaire data on prescription drug use during pregnancy?

Methods (1): Study population

• Imbedded in Eurocat Northern Netherlands Population-based birth defects registry

Covers provinces of Groningen, Friesland, and Drenthe

Annually 18,000 births (10% of all births in Netherlands)

• Infants younger than 16 years and fetuses with birth defects and genetic disorders eligible for registration

• Validation study: cases registered between 01-01-2009 and 30-06-2010 (n = 1,105)

Methods (2): Data collection

• Extensive questionnaire sent to parents

• Separate questionnaire on prescription drug use Screening question: “Did you use any…”

Questions on specific indications / drug groups

o Specify drug in open-ended question

o Timing of use

• Pharmacy records checked for compliance by maternal interviews

Methods (3): Statistical analyses

• Drugs ordered in three mutually exclusive categories:* For chronic conditions

For occasional and short-term use

Pregnancy-related

• “Gold Standard”: Eurocat database (pharmacy records)

• Validity: sensitivity and specificity

• Only drugs with ≥10 true exposures

• Multivariable logistic regression analysis for selected maternal and pregnancy characteristics

*Bakker et al. BJOG 2006

Results (1): Descriptives

• Response rate questionnaire 72% (n=777)

• 560 women were included Pharmacy records not available for 55 women (8%)

Gold Standard not yet available for 162 women (21%)

• Prescription drug use: Questionnaire:

o 233 women (42%)

o 129 different individual drugs / drug groups

Gold Standard:

o 389 women (69%)

o 221 different individual drugs / drug groups

Results (1): Descriptives

• Response rate questionnaire 72% (n=777)

• 560 women were included Pharmacy records not available for 55 women (8%)

Gold Standard not yet available for 162 women (21%)

• Prescription drug use: Questionnaire:

o 233 women (42%)

o 129 different individual drugs / drug groups

Gold Standard:

o 389 women (69%)

o 221 different individual drugs / drug groups

Sensitivity: 59%

Results (2): Validity and reliability comparisons

• Sensitivity drug categories: Corticosteroids: 12% (systemic corticosteroids: 50%)

Antihypertensives / thyroid therapy: 88%

• Sensitivity individual drugs: Naproxen: 0%

Salbutamol: 73%

• Specificity: 93-100%

• Sensitivity highest for chronic drugs

• Sensitivity first 4 months pregnancy generally comparable or slightly better than for complete pregnancy period

Results (3): Characteristics influencing disagreement

• Most maternal and pregnancy characteristics did not influence level of disagreement

• Disagreement occurred more often among: Younger women (<25 years of age)

Women who had fertility problems

Women who smoked during pregnancy

Women who had miscarriage / stillbirth

Women who completed questionnaire >1 year after delivery

Discussion

• Strengths and limitations:+ Choice of gold standard

+ High consent rate to obtain pharmacy records

– only mothers of infants with major birth defects

– no information on OTC drugs, low prevalence individual drugs

– screening question

• Validity comparable with studies conducted in the ‘80s and ‘90s

• Use of infants with genetic disorder as control group does not seem to introduce recall bias

Conclusion

• Validity of self-administered questionnaire on prescription drug use was generally moderate to poor for most drugs

• Future retrospective studies need additional data sources for prescription drug use

• Drug-specific questions could increase sensitivity

• Data collection should be completed before or as shortly after delivery as possible

Questions?

Thank you for your attention!