Hum. Reprod. Advance Access originally published online on July 26, 2007
Human Reproduction 2007 22(12):3265-3267; doi:10.1093/humrep/dem230
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Letters to the Editor |
Vital statistics: a poor source of data for investigating the association between paternal age and birth defects
Department of Maternal and Child Health, School of Public Health, University of Alabama at Birmingham, 320 Ryals Building, 1530 3rd Avenue South, Birmingham, Alabama 35294-0022, USA
1 Correspondence address. Tel: 205-934-7161/2985; Fax: 205-934-8248; E-mail: rkirby{at}uab.edu
The design and conduct of epidemiologic investigations incorporates aspects both of art and science. Selecting the appropriate data sources, statistical methods, tabular and graphical presentation of results and assessment of study limitations and strengths requires both creativity and objectivity. All too often, researchers allow preconceived notions concerning the quality of secondary data sources to influence their study designs, selection of study subjects and operationalization of study variables, analytical methods and interpretation of study findings.
Readers should treat the results of Yang et al.'s (2007)
report on the association between paternal age and birth defects, based on a large population-based dataset from US birth certificates, with caution. Three major issues require careful consideration: (i) specification of the dependent variable, (ii) measurement of the main independent variable and (iii) overall study generalizability. All of these issues stem directly from the use of the linked birth-infant death certificate public use files made available by the National Center for Health Statistics (NCHS) as the sole data source for their study.
By far the biggest concern is the choice of certificates of live birth as the source of data for the dependent variable, birth defects. Although there is an extensive literature on the poor quality of these data (Watkins et al., 1996; Kirby, 1997; Kirby and Salihu, 2003; inter alia), naïve researchers continue to persist in designing studies utilizing these data rather than obtaining data on birth defects prevalence from active case-find, multi-source birth defects surveillance programs like the Texas Birth Defects Registry (e.g. Archer et al., 2007). Given the overall sensitivity of birth certificate reporting of birth certificates (in the range of 20–30%) and less than stellar predictive value positive (in the range of 70–80%) with considerable variation across the non-specific categories, the potential for misclassification may overwhelm any statistically significant findings identified by Yang et al. (2007). Add to that the fact that the proportion of cases for specific birth defects included in live birth prevalence data varies considerably across the range of conditions, yet Yang et al. had access only to data on live birth events, and the completeness of ascertainment situation becomes even more clouded. But that's not all—the NCHS categories, faithfully reproduced by Yang et al. (2007) in Table 3, involve some distinct clinical entities but in many cases both clinically and etiologically heterogeneous conditions. For example, while the birth certificate has a single category for omphalocele/gastroschisis, clinicians and epidemiologists alike recognize these as distinct birth defects with highly heterogeneous epidemiologic features (Salihu et al., 2003). Incidentally, although the US NCHS adopted a new standard certificate of live birth in 2003, the revisions fail to remedy the deficiencies in reporting and classifying birth defects on American birth certificates (Kirby and Salihu, 2006).
The second area of concern focuses on the primary independent variable, paternal age. While demographic information concerning mothers of live born infants is almost universally recorded in vital records documents, paternal data elements contain significant numbers of missing values. In study under scrutiny, 945 237 records were excluded due to missing data on paternal age (14.5% of all births in the study dataset). While the authors note that exclusion of these records may represent selection bias, and that in a previous study mothers of infants whose birth certificates had missing paternal age data were significantly more likely to be teens, unmarried, black, less educated and tobacco smokers (Tan et al., 2004), they suggest that the only result of excluding records with missing paternal age was to reduce the estimates of the effect of paternal age on birth defects. This assertion requires more careful scrutiny, because missing paternal information is strongly associated with mothers who are unmarried (Gaudino et al., 1999) and may be associated with adequacy of prenatal care and access to perinatal services including ultrasonography and prenatal genetics clinics. The possibility of an interaction between missing paternal age and presence of birth defects not reported on the birth certificate cannot be discounted.
Finally, how generalizable are the findings reported by Yang et al. (2007)?a Due to the exclusion of live births occurring in the states of California, Indiana, South Dakota and New York because these states did not collect information concerning maternal smoking during pregnancy, the findings represent only 75–80% of all births in the USA. In their multivariable analyses, the authors control for maternal age, smoking and alcohol consumption during pregnancy, race, education, marital status, parity and trimester initiation of prenatal care. Interestingly, the authors did not control for plurality, a factor clearly associated with prevalence of birth defects and also associated with maternal age (Alexander et al., 2005; Tang et al., 2006). Numerous studies demonstrate the poor quality of data on alcohol use reported on birth certificates (Northam and Knapp, 2006), so much so that this variable can hardly be said to be controlled in this analysis. More concerning are the limitations imposed by the combination of a poorly sensitive measure of the outcome variable, birth defects and differentially biased under-reporting of the primary exposure variable, paternal age. Additionally, because the authors limit their analysis to live births, no information concerning birth defects among spontaneous abortions, fetal deaths, or pregnancy terminations are included in the analysis. For selected conditions, the proportion of all cases that survive to a live birth can be quite small, and undoubtedly varies with paternal age just as it does with maternal age, socio-economic status, access to perinatal care services, cultural and other factors. In the final analysis, we believe that readers should view the results of Yang et al. (2007) with caution, and await a larger population-based study that utilizes case data from birth defects surveillance programs and carefully assesses any bias imparted from incomplete data on paternal demographic characteristics. While vital statistics play many important roles in public health, descriptive epidemiology, advocacy and related endeavors (Schoendorf and Branum, 2006), these data must always be used with caution (Kirby, 2001; Ananth, 2005), and researchers must never assume that the mere fact of widespread availability of public use databases absolves them of the need to question the accuracy, reliability and validity of the data contained therein. Sadly, this is especially true in the case of birth defects reported on US birth certificates.
References
Alexander GR, Slay WM, Salihu H, Kirby RS. Fetal and neonatal mortality risks of multiple births. Obstet Gynecol Clin North Am (2005) 32:1–16.[CrossRef][Web of Science][Medline]
Ananth CV. Perinatal epidemiologic research with vital statistics data: validity is the essential quality. Am J Obstet Gynecol (2005) 193:5–6.[CrossRef][Web of Science][Medline]
Archer NP, Langlois PH, Suarez L, Brender J, Shanmugam R. Association of paternal age with prevalence of selected birth defects. Birth Defects Res A Clin Mol Teratol (2007) 79:27–34.[CrossRef][Web of Science][Medline]
Gaudino JA Jr, Jenkins B, Rochat RW. No fathers' names: a risk factor for infant mortality in the State of Georgia, USA. Soc Sci Med (1999) 48:253–265.[CrossRef][Web of Science][Medline]
Kirby RS. The quality of data reported on birth certificates (Letter). Am J Public Health (1997) 87:301.
Kirby RS. Invited commentary: using vital statistics databases for perinatal epidemiology: does the quality go in before the name goes on? Am J Epidemiol (2001) 154:889–890.
Kirby RS, Salihu HM. Birth malformations and other adverse perinatal outcomes: available data sources pose a dilemma. Environ Health Perspect (2003) 111:A868–A869.[Web of Science][Medline]
Kirby RS, Salihu HM. Back to the future? A critical commentary on the 2003 U.S. National standard certificate of live birth. Birth (2006) 33:238–244.[CrossRef][Web of Science][Medline]
Northam S, Knapp TR. The reliability and validity of birth certificates. J Obstet Gynecol Neonatal Nurs (2006) 35:3–12.[CrossRef][Web of Science][Medline]
Salihu HM, Pierre-Louis BJ, Druschel CM, Kirby RS. Omphalocele and gastroschisis in the State of New York, 1992–1999. Birth Defects Res A Clin Mol Teratol (2003) 67:630–636.[CrossRef][Web of Science][Medline]
Schoendorf KC, Branum AM. The use of United States vital statistics in perinatal and obstetric research. Am J Obstet Gynecol (2006) 194:911–915.[CrossRef][Web of Science][Medline]
Tan H, Wen SW, Walker M, Demissie K. Missing paternal demographics: a novel indicator for identifying high risk population of adverse pregnancy outcomes. BMC Pregnancy Childbirth (2004) 4:21.[CrossRef][Medline]
Tang Y, Ma CX, Cui W, Chang V, Ariet M, Morse SB, Resnick MB, Roth J. The risk of birth defects in multiple births: a population-based study. Matern Child Health J (2006) 10:75–81.[CrossRef][Web of Science][Medline]
Watkins ML, Edmonds L, McClearn A, Mullins L, Mulinare J, Khoury M. The surveillance of birth defects: the usefulness of the revised US standard birth certificate. Am J Public Health (1996) 86:731–734.
Yang Q, Wen SW, Leader A, Chen XK, Lipson J, Walker M. Paternal age and birth defects: how strong is the association? Hum Reprod (2007) 22:696–701.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||