Our results indicate that self-reported measures of substance use in this cohort of HIV-infected women are fairly reliable when compared to meconium test results. We also observed that inclusion of passive exposure reports is important for the reliability of tobacco exposure reporting. Sensitivity of self-reported cocaine use compared to meconium was similar to that previously reported36
, while sensitivity for tobacco use in our study was higher than for a previous study (61%) which did not incorporate passive smoke exposure44
. In addition, our results suggest that meconium may not be a gold standard for tobacco exposure, particularly when this exposure is light or intermittent, as with passive exposure. When we included reports of passive exposure to tobacco smoke during the second or third trimester the sensitivity of self-reports relative to meconium increased. However, the positive predictive value of self-reported information was greatly reduced (from 75% to 29%) because the majority of women who reported passive exposure to tobacco smoke did not have a positive meconium assay.
The agreement between meconium assays and self-report is reassuring since the only measure of first trimester exposures in SMARTT is self-reported. However, while our self-report measures indicated moderate to high agreement with meconium the agreement was not complete. Our results show that the incorporation of meconium or other biological confirmation of self-report is worthwhile in research studies where accurate measurement of substance use is essential.
For the subset of subjects who had urine or blood toxicology tests, we observed low agreement with self-report for cocaine, marijuana and opiates. For the small number with both a positive toxicology result and an available meconium sample, overall agreement between meconium and positive toxicology was high for cocaine but low for marijuana and opiates. Clinic sites have indicated that these tests are most often done when there is suspicion of recent substance use. Because of the small numbers with both types of biological samples these comparisons should not be over-interpreted, and it is also difficult to draw conclusions regarding the positive toxicology results since we do not know whether they are from a screening assay or confirmatory testing. However, it is worthwhile to point out that the application of urine or blood toxicology tests for suspicion of current use may explain the negative meconium results for marijuana, since sporadic substance use can result in positive urinalysis but negative meconium analysis39
. Also, false positive screens, particularly for opiates, can result from cross-reactivity with several other compounds45,46
While our data show that illicit substance use during pregnancy is low, 21% of women did report using either alcohol or tobacco in the first trimester. Also, 41% of women reported passive tobacco smoke exposure inside or outside the home. Prenatal exposure to passive tobacco smoke has been linked to preterm birth, low birth weight, and behavioral problems7,12,47,48
. Prenatal exposures to alcohol and tobacco may be as harmful as exposure to cocaine or opiates49-51
. There is also a potential concern for perinatal HIV transmission since alcohol consumption can lead to ARV non-adherence, due to a belief among some HIV-positive adults of adverse interactions between alcohol and HIV medications52,53
. Biomarkers of fetal exposure to alcohol have been recently developed54
; while their sensitivity and specificity are still being evaluated, future studies might benefit from incorporating these biomarkers, since information on the dangers of alcohol use are widely known and pregnant women may have strong incentive to underreport the extent of their use55
Our results also support other evidence that use of cocaine, marijuana, heroin and other opiates, as well as injected drug use, has substantially decreased relative to use reported by earlier cohorts of HIV-infected pregnant women. Among 876 women enrolled in the Women and Infants Transmission Study between 1990 and 2000, 39% used hard drugs (cocaine, heroin/opiates, street methadone and injected drugs) during pregnancy as measured by self-report or urinalysis29
. In the dynamic SMARTT cohort, nearly a decade later, fewer than 4% of women reported use of hard drugs, and no women reported injection drug use. In the SMARTT cohort self-reported rates of trimester-specific use, including a decline in use over trimester, are similar to those in a recent general population study of pregnant women56
Similar to previous studies, we found that substance use during pregnancy was associated with race or ethnicity and lower education level40-42
. However, in contrast to an earlier study of uninfected women showing younger age associated with use40
, we found older age associated with more reporting of use. This finding also may reflect the changing epidemic, with younger women constituting a more recently-infected segment of the cohort. There is some evidence that teenage women constitute a group less likely to use substances while pregnant57
There were several strengths to our study. The first is collection of trimester-specific self-report information. This will be important when controlling for confounding in studies of the effects of prenatal exposure to ARV, which also can vary by timing25
, and also illustrates the decrease in substance use over trimesters, suggesting that once women found out they were pregnant many stopped their use. Secondly, meconium has a favorable profile relative to other biological specimens, including a longer detection window than urine, and fewer collection difficulties than blood or neonatal hair. Also, while incomplete disclosure of substance use is well-documented, efforts were made to maximize disclosure by securing a federal confidentiality certificate and by emphasizing to all women the protections this certificate confers.
Some limitations must be acknowledged. While our biological measure of substance use was prospectively collected, self-reports were only recorded after delivery, which might lead to problems with recall, particularly for the first and second trimesters. Secondly, meconium does not detect use during the entire second trimester. For this second trimester meconium is therefore not a highly sensitive measure. Also, since both self-reports and positive meconium results for marijuana and cocaine were low it is important to note that this limits the ability to test and interpret the sensitivity and specificity for these measures, which in this study were both observed to be high.
We did not have meconium results for all infants. To date meconium has only been tested for infants born by December 31, 2008. However there were no differences in maternal characteristics or self-reported substance use for mothers of infants born on or before this date and those born after. In addition, we were not able to obtain meconium for 27% of infants born by the above date. If women concerned about disclosing use were both less likely to report this use and more likely to refuse sample collection, this could have biased the results. However, when we compared self-reported use for women with a meconium sample to use for women without a sample, there was no difference in prevalence other than the suggestion that women who refused meconium collection were actually more likely to report use.
Collection of meconium for research purposes is challenging. Ostrea et al (1992) reported 8% missed meconium samples and Derauf et al (2003) reported 16% missed samples. For both studies, infants were enrolled at only one site – the hospital in which they were delivered. The Maternal Lifestyle Study36
with a higher missed collection rate of 25.5% included 4 enrollment sites. In our study, there are 22 sites, with infants from several sites born at unaffiliated hospitals, where meconium collection is more difficult. Primary challenges to collection include hospital staff unaware of study protocols, infants delivered at outside hospitals after unexpected labor, quicker transition from meconium to stool, preterm infants that do not produce sufficient meconium, and meconium samples exhausted for standard of care testing. While some of these difficulties are unavoidable, clinic sites have developed innovative methods of maximizing collection, including: maintaining consistent communication with hospital nurseries; instructing mothers to save diapers or to remind hospital staff about collection; adding written instructions to mothers' and infants' medical charts; providing meconium collection kits and research clinic phone numbers to nurseries in advance of delivery; and showing appreciation to hospital staff for the extra efforts entailed by this collection.
The dynamic SMARTT cohort enrolls only HIV-uninfected infants and their infected mothers. In addition, substance use has been associated with risk of perinatal HIV transmission29
and still remains associated with viral load even in the current era of highly-effective antiretroviral therapy (HAART)43
; therefore SMARTT is likely not enrolling women who are the heaviest substance users. Our results are therefore generalizable to women receiving adequate prenatal care including HAART.
It will be important in SMARTT to examine possible associations between the substance use detected here and adverse outcomes in children, and also to adjust for substance use as a potential confounder.
In summary, self-reported substance use, as confirmed with meconium assay, is fairly reliable in this cohort of HIV-infected women. Our results indicate that use of illicit substances is low but that alcohol and tobacco exposures are relatively common and could pose health risks for infants born to these women.