Letters totheEditor
8. 9.
11.
12.
13.
14. Stevens RG. Electric power use and breast cancer a hypothesis. Am J Epidemiol 1987;125:556-61. 15. Matanoski GM, Breysse PN, Elliott EA. Electromagnetic field exposure and male breast cancer. (Letter). Lancet 1991 ;337:737.
Paul A. Demers Department of Environmental Health (SC34) University of Washington School of Public Health Seattle, WA 98195 David B. Thomas Program in Epidemiology (MP474) Fred Hutchinson Cancer Research Center Seattle, WA 98104 Karin A. Rosenblatt Department of Health and Safety Studies University of Illinois at Urbana-Champaign 121 Huff Hall Champaign, IL 61820
RE: "COCAINE USE DURING PREGNANCY: PERINATAL OUTCOMES" In a recent issue of the Journal, Handler et al. (1) reported on perinatal outcomes of cocaine use during pregnancy. The authors say many studies have been hampered by difficulties in exposure ascertainment. Unfortunately, their study suffers from similar biases. A perinatal registry, the source of subjects for their study, does not eliminate bias in information elicitation which occurs if staff only discuss illicit drug use with the pregnant women who seem "likely" candidates for using illicit drugs. In their study, cocaine exposure was determined only if documented on the chart; it would only be noted on the chart if there was a reason for the staff to ask the respondent. The authors noted that women who reported exposure to cocaine were more likely to be black, young, multigravida, smokers, and have no history of prenatal care. However, these might be the very factors which would alert a physician to interrogate the respondent about illicit drug use. Additionally, the authors do not give any details on when and who was asked to give a urine specimen, and it appears that not all women were asked to do so. Again, those who would be asked for a specimen might be those who were thought to be drug users. In order to remedy this situation, all women would need to be asked about illicit drug use and all women would need to be asked to provide a urine specimen so that nonuse would be noted on every chart! The authors felt that their large sample would overcome such bias; however, sample size alone cannot compensate for lack of information. Other flaws in the study include a problem in
defining and assessing "congenital abnormalities." The description and the range of anomalies assessed is not specified. Many congenital anomalies are not recognized at birth, but become apparent after months or years. If cocaine use were associated with later recognized congenital anomalies, the authors wouldn't find such an association since they presumably recorded only those anomalies apparent at birth. To resolve this issue of misclassification of outcome, a longerterm follow-up would be necessary. Statistical tests are performed on data without appropriate justification. For example, the Breslow-Day test of homogeneity is calculated but is referenced to material that has nothing to do with this test. In addition, no justification is given for setting the level of statistical significance at the 0.10 level instead of the usual 0.05 level. Finally, a puzzling finding of greater risk of adverse outcome for nonsmokers highlights the bias that can occur when exposure data are misclassified. The recency of smoking was not discussed, and one explanation for this odd finding is that ex-smokers may have been classified as nonsmokers. In summary, we feel that such an important topic deserves a more thorough study. For us, this study raised more questions than it answered. It should be considered a pilot project and also one which illustrates bias in epidemiologic studies. Assessing quantity and frequency of substance use, consequences of use, and severity of use disorders is a difficult task—one which many of us have been working on steadily (2-5). Peri-
Downloaded from https://academic.oup.com/aje/article-abstract/135/12/1425/88815 by University of Otago user on 18 January 2019
10.
Franco-Swiss case-control study. Int J Cancer 1990;45:661-5. Rosenbaum P, Vena J, Zielezny M, et al. Risk factors for male breast cancer in western New York. (Abstract). Am J Epidemiol 1990; 132:790. McLaughlin JK, Malker HSR, Blot WJ, et al. Occupational risks for male breast cancer in Sweden. BrJIndMedl988;45:275-6. Lerchl A, Honaka KO, Reiter RJ. Pineal gland "magnetosensitivity" to static magnetic fields is a consequence of induced electric currents (eddy currents). J Pineal Res 1991,10:109-16. Tamarkin L, Cohen M, Roselle D, et al. Melatonin inhibition and pinealectomy enhancement of 7,12dimethylbenz(a)anthracene-induced mammary tumors in the rat Cancer Res 1981;41:4432-6. Shah PN, Mhatre MC, Kothari LS. Effect of melatonin on mammary carcinogenesis in intact and pinealectomized rats in varying photoperiods. Cancer Res 1984;44:3403-7. Stevens RG, Davis S, Thomas DB, et al. Electric power, pineal function, and the risk of breast cancer. FASEBJ 1992,6:853-60.
1425
1426
Letters to the Editor
natal outcome studies especially deserve systematic, controlled, and standardized assessments so that information can be adequately classified and complex associations understood.
Linda B. Cottier Wilson M. Compton, III Department of Psychiatry Washington University School of Medicine St. Louis. MO 63110 THE A UTHORS REPL Y We read with interest the letter of our colleagues Cottier and Compton (1) and share their concerns about the use of registries to examine the relation between maternal substance abuse and perinatal outcomes. It is our belief that we have carefully and explicitly addressed the biases that might impact the results of our study (2), particularly those related to the difficulties in exposure ascertainment. Cottier and Compton suggest that the remedy to exposure ascertainment problems is both to ask all women about their cocaine use and to provide a urine specimen, so that nonuse would be noted on every chart. We concur and conclude that: "Universal screening during the prenatal period which is nonpunitive and is accompanied by informed consent would eliminate both the research bias and the social stigma that result when screening is based on neonatal status or racial or socioeconomic group" (2, p. 824). We do not believe that large sample size compensates for lack of good information. Iafact, we explicitly state: T h e use of a large perinatal registry does not eliminate difficulties related to
REFERENCES
1. Cottier LB, Compton WM III. Re: "Cocaine use during pregnancy: perinatal outcomes." (Letter).
Downloaded from https://academic.oup.com/aje/article-abstract/135/12/1425/88815 by University of Otago user on 18 January 2019
REFERENCES
1. Handler A, Kistin N. Davis F, et al. Cocaine use during pregnancy: perinatal outcomes. Am J Epidemiol !99l;l33:8l8-25. 2. Cottier LB, Keating SK. Operationalization of alcohol and drug dependence criteria by means of a structured interview. Recent Dev Alcohol l990;8: 69-83. 3. Mitchell AA, Cottier LB, Shapiro S. The effect of questionnaire design on recall of drug exposure in pregnancy. Am J Epidemiol 1986; 123:670-6. 4. Cottier LB. The CIDI and CIDI substance abuse module (SAM): cross-cultural instruments for assessing DSM-III, III-R, and ICD-10 criteria. In: Problems of drug dependence, 1990. National Institute on Drug Abuse Research Monograph Series 105. Rockville, MD: US Department of Health and Human Services, National Institute on Drug Abuse, 1991:220-6. (DHHS publication no. (ADM)91-1753). 5. McLellan AT, Luborsky L, Cacciola J, et al. New data from the Addiction Survey Index. Reliability and validity in three centers. J Nerv Ment Dis 1985; 173:412-23.
exposure ascertainment; however, a large sample size does permit an evaluation of the effect of any cocaine use on perinatal outcomes while adjusting for other risk factors" (2, p. 819). We have some comments about more specific concerns raised by Cottier and Compton. The University of Illinois Perinatal Database (19821989) tracked 52 single congenital anomalies that were ascertained during the neonatal period. While we agree that additional congenital anomalies might be identified through long-term follow-up, we explicitly state that our study is a study of perinatal outcomes rather than all potential effects of cocaine use. We reference Rothman (3) when we discuss the Breslow-Day test of homogeneity because of his elegant presentation of the methodology for evaluating effect modification. On the basis of our evaluation of the stratum-specific relative risks, we discovered two variables for which there was effect modification by smoking (small for gestational age and head circumference). Because of small sample size, the Breslow-Day test was marginally significant; we did not use p < 0.10 a priori. However, the clear difference in relative risks between strata warranted the entry of interaction terms for cocaine and smoking into models for each of these outcome variables. We concur that our finding of greater risk of adverse outcomes for nonsmokers is unusual and discuss alternative explanations including misclassification of smoking data. Further, we were able to investigate this potential misclassification. We found: "The method of detection of cocaine use was similar in smoking and nonsmoking cocaine users; a trend toward a greater percentage of nonsmokers who were cocaine users to have had their cocaine use detected by urine screen was not evident" (2, p. 823). We conclude that regardless of smoking status, cocaine use increases the risk of small for gestational age births. We look forward to large studies of the impact of maternal cocaine use which are based on complete and standardized exposure ascertainment and which utilize long-term follow-up. However, we contend that studies based on registries in which data collection is not influenced by the hypothesis under study further our understanding of the relation between many exposures and health status outcomes and should be encouraged to provide direction and support for more elaborate and expensive studies. Finally, we continue to believe that our data provide support for the expansion of substance abuse treatment programs for pregnant women as a way to reduce adverse pregnancy outcomes and improve public health.
8. 9.
11.
12.
13.
14. Stevens RG. Electric power use and breast cancer a hypothesis. Am J Epidemiol 1987;125:556-61. 15. Matanoski GM, Breysse PN, Elliott EA. Electromagnetic field exposure and male breast cancer. (Letter). Lancet 1991 ;337:737.
Paul A. Demers Department of Environmental Health (SC34) University of Washington School of Public Health Seattle, WA 98195 David B. Thomas Program in Epidemiology (MP474) Fred Hutchinson Cancer Research Center Seattle, WA 98104 Karin A. Rosenblatt Department of Health and Safety Studies University of Illinois at Urbana-Champaign 121 Huff Hall Champaign, IL 61820
RE: "COCAINE USE DURING PREGNANCY: PERINATAL OUTCOMES" In a recent issue of the Journal, Handler et al. (1) reported on perinatal outcomes of cocaine use during pregnancy. The authors say many studies have been hampered by difficulties in exposure ascertainment. Unfortunately, their study suffers from similar biases. A perinatal registry, the source of subjects for their study, does not eliminate bias in information elicitation which occurs if staff only discuss illicit drug use with the pregnant women who seem "likely" candidates for using illicit drugs. In their study, cocaine exposure was determined only if documented on the chart; it would only be noted on the chart if there was a reason for the staff to ask the respondent. The authors noted that women who reported exposure to cocaine were more likely to be black, young, multigravida, smokers, and have no history of prenatal care. However, these might be the very factors which would alert a physician to interrogate the respondent about illicit drug use. Additionally, the authors do not give any details on when and who was asked to give a urine specimen, and it appears that not all women were asked to do so. Again, those who would be asked for a specimen might be those who were thought to be drug users. In order to remedy this situation, all women would need to be asked about illicit drug use and all women would need to be asked to provide a urine specimen so that nonuse would be noted on every chart! The authors felt that their large sample would overcome such bias; however, sample size alone cannot compensate for lack of information. Other flaws in the study include a problem in
defining and assessing "congenital abnormalities." The description and the range of anomalies assessed is not specified. Many congenital anomalies are not recognized at birth, but become apparent after months or years. If cocaine use were associated with later recognized congenital anomalies, the authors wouldn't find such an association since they presumably recorded only those anomalies apparent at birth. To resolve this issue of misclassification of outcome, a longerterm follow-up would be necessary. Statistical tests are performed on data without appropriate justification. For example, the Breslow-Day test of homogeneity is calculated but is referenced to material that has nothing to do with this test. In addition, no justification is given for setting the level of statistical significance at the 0.10 level instead of the usual 0.05 level. Finally, a puzzling finding of greater risk of adverse outcome for nonsmokers highlights the bias that can occur when exposure data are misclassified. The recency of smoking was not discussed, and one explanation for this odd finding is that ex-smokers may have been classified as nonsmokers. In summary, we feel that such an important topic deserves a more thorough study. For us, this study raised more questions than it answered. It should be considered a pilot project and also one which illustrates bias in epidemiologic studies. Assessing quantity and frequency of substance use, consequences of use, and severity of use disorders is a difficult task—one which many of us have been working on steadily (2-5). Peri-
Downloaded from https://academic.oup.com/aje/article-abstract/135/12/1425/88815 by University of Otago user on 18 January 2019
10.
Franco-Swiss case-control study. Int J Cancer 1990;45:661-5. Rosenbaum P, Vena J, Zielezny M, et al. Risk factors for male breast cancer in western New York. (Abstract). Am J Epidemiol 1990; 132:790. McLaughlin JK, Malker HSR, Blot WJ, et al. Occupational risks for male breast cancer in Sweden. BrJIndMedl988;45:275-6. Lerchl A, Honaka KO, Reiter RJ. Pineal gland "magnetosensitivity" to static magnetic fields is a consequence of induced electric currents (eddy currents). J Pineal Res 1991,10:109-16. Tamarkin L, Cohen M, Roselle D, et al. Melatonin inhibition and pinealectomy enhancement of 7,12dimethylbenz(a)anthracene-induced mammary tumors in the rat Cancer Res 1981;41:4432-6. Shah PN, Mhatre MC, Kothari LS. Effect of melatonin on mammary carcinogenesis in intact and pinealectomized rats in varying photoperiods. Cancer Res 1984;44:3403-7. Stevens RG, Davis S, Thomas DB, et al. Electric power, pineal function, and the risk of breast cancer. FASEBJ 1992,6:853-60.
1425
1426
Letters to the Editor
natal outcome studies especially deserve systematic, controlled, and standardized assessments so that information can be adequately classified and complex associations understood.
Linda B. Cottier Wilson M. Compton, III Department of Psychiatry Washington University School of Medicine St. Louis. MO 63110 THE A UTHORS REPL Y We read with interest the letter of our colleagues Cottier and Compton (1) and share their concerns about the use of registries to examine the relation between maternal substance abuse and perinatal outcomes. It is our belief that we have carefully and explicitly addressed the biases that might impact the results of our study (2), particularly those related to the difficulties in exposure ascertainment. Cottier and Compton suggest that the remedy to exposure ascertainment problems is both to ask all women about their cocaine use and to provide a urine specimen, so that nonuse would be noted on every chart. We concur and conclude that: "Universal screening during the prenatal period which is nonpunitive and is accompanied by informed consent would eliminate both the research bias and the social stigma that result when screening is based on neonatal status or racial or socioeconomic group" (2, p. 824). We do not believe that large sample size compensates for lack of good information. Iafact, we explicitly state: T h e use of a large perinatal registry does not eliminate difficulties related to
REFERENCES
1. Cottier LB, Compton WM III. Re: "Cocaine use during pregnancy: perinatal outcomes." (Letter).
Downloaded from https://academic.oup.com/aje/article-abstract/135/12/1425/88815 by University of Otago user on 18 January 2019
REFERENCES
1. Handler A, Kistin N. Davis F, et al. Cocaine use during pregnancy: perinatal outcomes. Am J Epidemiol !99l;l33:8l8-25. 2. Cottier LB, Keating SK. Operationalization of alcohol and drug dependence criteria by means of a structured interview. Recent Dev Alcohol l990;8: 69-83. 3. Mitchell AA, Cottier LB, Shapiro S. The effect of questionnaire design on recall of drug exposure in pregnancy. Am J Epidemiol 1986; 123:670-6. 4. Cottier LB. The CIDI and CIDI substance abuse module (SAM): cross-cultural instruments for assessing DSM-III, III-R, and ICD-10 criteria. In: Problems of drug dependence, 1990. National Institute on Drug Abuse Research Monograph Series 105. Rockville, MD: US Department of Health and Human Services, National Institute on Drug Abuse, 1991:220-6. (DHHS publication no. (ADM)91-1753). 5. McLellan AT, Luborsky L, Cacciola J, et al. New data from the Addiction Survey Index. Reliability and validity in three centers. J Nerv Ment Dis 1985; 173:412-23.
exposure ascertainment; however, a large sample size does permit an evaluation of the effect of any cocaine use on perinatal outcomes while adjusting for other risk factors" (2, p. 819). We have some comments about more specific concerns raised by Cottier and Compton. The University of Illinois Perinatal Database (19821989) tracked 52 single congenital anomalies that were ascertained during the neonatal period. While we agree that additional congenital anomalies might be identified through long-term follow-up, we explicitly state that our study is a study of perinatal outcomes rather than all potential effects of cocaine use. We reference Rothman (3) when we discuss the Breslow-Day test of homogeneity because of his elegant presentation of the methodology for evaluating effect modification. On the basis of our evaluation of the stratum-specific relative risks, we discovered two variables for which there was effect modification by smoking (small for gestational age and head circumference). Because of small sample size, the Breslow-Day test was marginally significant; we did not use p < 0.10 a priori. However, the clear difference in relative risks between strata warranted the entry of interaction terms for cocaine and smoking into models for each of these outcome variables. We concur that our finding of greater risk of adverse outcomes for nonsmokers is unusual and discuss alternative explanations including misclassification of smoking data. Further, we were able to investigate this potential misclassification. We found: "The method of detection of cocaine use was similar in smoking and nonsmoking cocaine users; a trend toward a greater percentage of nonsmokers who were cocaine users to have had their cocaine use detected by urine screen was not evident" (2, p. 823). We conclude that regardless of smoking status, cocaine use increases the risk of small for gestational age births. We look forward to large studies of the impact of maternal cocaine use which are based on complete and standardized exposure ascertainment and which utilize long-term follow-up. However, we contend that studies based on registries in which data collection is not influenced by the hypothesis under study further our understanding of the relation between many exposures and health status outcomes and should be encouraged to provide direction and support for more elaborate and expensive studies. Finally, we continue to believe that our data provide support for the expansion of substance abuse treatment programs for pregnant women as a way to reduce adverse pregnancy outcomes and improve public health.
