You are here

Article Text

Article menu
Workplace
Original article
Paternal occupation and birth defects: findings from the National Birth Defects Prevention Study
  1. Tania A Desrosiers1,2,
  2. Amy H Herring1,3,
  3. Stuart K Shapira4,
  4. Mariëtte Hooiveld5,6,
  5. Tom J Luben7,
  6. Michele L Herdt-Losavio8,
  7. Shao Lin8,
  8. Andrew F Olshan1,3,
  9. the National Birth Defects Prevention Study
  1. 1Department of Epidemiology, Gillings School of Global Public Health, University of North Carolina, Chapel Hill, North Carolina, USA
  2. 2Birth Defects Monitoring Program, State Center for Health Statistics, North Carolina Division of Public Health, Raleigh, North Carolina, USA
  3. 3Carolina Population Center, Gillings School of Global Public Health, University of North Carolina, Chapel Hill, North Carolina, USA
  4. 4National Center on Birth Defects and Developmental Disabilities, Centers for Disease Control and Prevention, Atlanta, Georgia, USA
  5. 5Radboud University Nijmegen Medical Center, Nijmegen, The Netherlands
  6. 6Netherlands Institute for Health Services Research, Utrecht, The Netherlands
  7. 7National Center for Environmental Assessment, US Environmental Protection Agency, Research Triangle Park, North Carolina, USA
  8. 8Center for Environmental Health, New York State Department of Health, Troy, New York, USA
  1. Correspondence to Dr Tania A Desrosiers, North Carolina Center for Birth Defects Research and Prevention, Gillings School of Global Public Health, Department of Epidemiology, CB#7435, University of North Carolina, Chapel Hill, NC 27599, USA; ta_desrosiers{at}unc.edu

Abstract

Objectives Several epidemiological studies have suggested that certain paternal occupations may be associated with an increased prevalence of birth defects in offspring. Using data from the National Birth Defects Prevention Study, the authors investigated the association between paternal occupation and birth defects in a case–control study of cases comprising over 60 different types of birth defects (n=9998) and non-malformed controls (n=4066) with dates of delivery between 1997 and 2004.

Methods Using paternal occupational histories reported by mothers via telephone interview, jobs were systematically classified into 63 groups based on shared exposure profiles within occupation and industry. Data were analysed using Bayesian logistic regression with a hierarchical prior for dependent shrinkage to stabilise estimation with sparse data.

Results Several occupations were associated with an increased prevalence of various birth defect categories, including mathematical, physical and computer scientists; artists; photographers and photo processors; food service workers; landscapers and groundskeepers; hairdressers and cosmetologists; office and administrative support workers; sawmill workers; petroleum and gas workers; chemical workers; printers; material moving equipment operators; and motor vehicle operators.

Conclusions Findings from this study might be used to identify specific occupations worthy of further investigation and to generate hypotheses about chemical or physical exposures common to such occupations.

  • Bayes theorem
  • congenital abnormalities
  • occupational exposure
  • occupations
  • paternal exposure
  • epidemiology
  • female reproductive effects and adverse pregnancy outcomes
  • congenital anomalies
  • clinical medicine
  • genetic susceptibility
  • health surveillance
  • primary care
  • investigation of outbreaks of illness
  • air pollution
  • exposure assessment
  • asthma
  • public health
  • climate
  • pollution
  • pesticides
  • metals
  • indoor air
  • environment
  • respiratory
  • renal
  • cardiovascular
  • cancer
  • polymorphisms
  • male reproduction

https://doi.org/10.1136/oemed-2011-100372

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    What this paper adds

    • Previous epidemiological investigations of paternal occupation and birth defects in offspring have grouped aetiologically distinct phenotypes together, which may introduce aetiological heterogeneity and dilute associations. Likewise, job titles with potentially different chemical and physical exposure profiles are often loosely grouped together by major industry rather than by shared exposures, leading to exposure misclassification.

    • This large population-based study was conducted to explore the relation between multiple paternal occupations and over 60 types of birth defects using Bayesian analytic methods that specifically address the statistical challenges associated with analysis of sparsely distributed data across numerous exposures and outcomes.

    • Results from this study indicate that paternal work in a number of occupations may be associated with an increased prevalence of various birth defects in offspring. Findings can be used to inform future investigation of specific paternal occupations found to be associated with birth defects or to generate hypotheses about chemical or physical exposures and exposure mixtures common to such occupations.

    Birth defects are a leading cause of infant mortality and developmental disabilities in the USA, yet the causes of most birth defects are unknown.1 ,2 Previous epidemiological studies have suggested that certain paternal occupations and workplace exposures may be associated with an increased prevalence of birth defects in offspring.3–5 Occupations found to be associated with various defects include agricultural and groundskeeping workers, electronic industry workers, forestry and logging workers, janitors and cleaners, laboratory workers, painters, printers, vehicle manufacturers and mechanics, welders and woodworkers.6–21

    Investigations of occupation as a risk factor for birth defects face a number of methodological challenges.4 Many of these challenges stem primarily from small sample size, which is impacted by low exposure prevalence (eg, diversity of occupations) and the rarity of individual phenotypes. To recover statistical power, investigators may group occupations or birth defects into larger categories, tolerating increased heterogeneity in order to increase sample sizes within groups. Conversely, study designs with more finely categorised (and thus more homogeneous) exposures and outcomes are prone to other issues, such as imprecision and multiple hypotheses testing.

    The objective of this study was to use data from a large national case–control study of birth defects to explore the relation between paternal occupation and various birth defects using analytic methods that specifically address the statistical challenges associated with the analysis of sparsely distributed data across numerous occupational groups and birth defect categories. A complementary analysis of the association between maternal occupation and birth defects in this study population has been previously published.22

    Methods

    All data are from the National Birth Defects Prevention Study (NBDPS), an ongoing, multicenter, population-based case–control study designed to investigate a range of potential risk factors for major birth defects. Detailed study methods have been previously published.23 Briefly, eligible cases with one or more major birth defect were identified by 10 participating state birth defect surveillance systems (Arkansas, California, Georgia, Iowa, Massachusetts, New Jersey, North Carolina, New York, Texas, Utah) and included live births, fetal deaths and prenatally diagnosed elective terminations. Controls were randomly selected in each state among live births without major defects from either hospital records or birth certificates, with an approximate overall case:control ratio of 3:1. Between 6 weeks and 24 months after the estimated delivery date (EDD), mothers were enrolled and interviewed by telephone in either English or Spanish using a structured questionnaire that covers numerous demographic, behavioural and clinical factors before and during pregnancy.

    This analysis includes cases and controls with EDDs between 1 October 1997 and 31 December 2004. Overall participation in the interview among case and control mothers during this time period was 69.4% and 66.1%, respectively.

    Outcome classification

    Clinical geneticists at each centre performed standardised case review and coding to determine eligibility.24 Eligible cases of major defects (those typically considered to have major medical or surgical significance) were centrally reviewed again by NBDPS clinicians to confirm eligibility and to classify each case as having only one major birth defect (‘isolated’), more than one major birth defect (‘multiple’), or a pattern of defects that represent a complex developmental sequence. In general, only cases with non-syndromic isolated and multiple defects were considered for this analysis, which includes more than 60 distinct birth defect categories. The birth defects of interest in this study (table 1) are grouped by the primary organ system affected; this grouping is for convenience of presentation only and is not indicative of shared aetiology or embryological development.

    View this table:
    Table 1

    Distribution of major birth defect categories eligible for analysis, by primary organ system, National Birth Defects Prevention Study, USA, 1997–2004

    Occupational classification

    During the interview, mothers reported the job title, main activities and other details for each job held by the infant's father for at least 1-month duration from 3 months preceding the estimated date of conception (EDC) through the EDD. Jobs were then coded using the Standard Occupational Classification manual and North American Industry Classification System (NAICS).25 ,26

    We restricted the exposure period of interest to 3 months preceding the EDC through the first month of pregnancy, which corresponds to the primary critical window of susceptibility for male-mediated mechanisms of teratogenesis.5 ,27 All jobs held during this period were further grouped by SOC and NAICS codes into a modified occupational classification scheme, which combines occupations considered to have similar physical and chemical exposure profiles.22 ,28 For example, the occupational group ‘farmers and farm workers’ included agricultural workers, farmers and ranchers, and floral designers. Fathers who held more than one job during the critical period were assigned to more than one occupational group as appropriate (up to two groups per job). For example, Air Force pilots were assigned to ‘armed forces’ and ‘aircraft operators’.

    Study sample

    The overall study population included fathers of 14 920 singleton cases and 5771 singleton controls conceived without donor sperm or embryo. We further excluded fathers with missing occupational histories (498 (3.3%) cases; 182 (3.2%) controls), fathers who did not work at all between 3 months before the EDC and the EDD (1024 (6.9%) cases; 321 (5.6%) controls), fathers who did not have a job or were students during the exposure period of interest (3318 (22.2%) cases; 1180 (20.4%) controls) and fathers who worked during the relevant period but whose job descriptions were insufficient for classification (82 (0.5%) cases; 22 (0.4%) controls). The final sample for analysis consisted of fathers of 9998 cases and 4066 controls.

    Statistical analysis

    Despite the large sample size overall, the data were sparsely distributed since the cross-classification of occupations and outcomes yielded over 5000 combinations, many of which were represented by very few or no father–infant pairs. Given this distribution as well as the underlying assumptions about common workplace exposures within occupational groups, we assumed that an occupation associated with one defect may be more likely to be associated with other defects and likewise a birth defect associated with one occupation may be more likely to be associated with other occupations. Using Bayesian logistic regression with a heavy-tailed scale mixture of normals prior, we incorporated this data structure in a hierarchical prior for dependent shrinkage across coefficients for occupation. The prior corresponded to a t-distribution with low degrees of freedom, which is expressed as a gamma (Ga) precision mixture of normals centred at zero. In particular, we specify βkj∼N(0,λj−1νk−1), over the occupations k and birth defects j, with λj and νk assigned Ga(5,5) hyperpriors. Note that by fixing these parameters at their prior mean values λ j=ν k=1, this prior is simply a N(0,1) prior, which does not borrow any information across occupations or outcomes. An N(0,1) prior is relatively non-informative for log ORs, expressing our prior belief that ORs >7.4 or <0.14 are possible, but unlikely to be observed in our study. Closely related priors are widely used due to their tendency for strong shrinkage of small coefficients towards zero, leading to a parsimonious model while limiting shrinkage of larger coefficients to reduce bias in estimating the signal.29 These priors exhibit good performance in simulation studies and have previously been used to borrow information across related outcomes in high-dimensional regression.30 In our study, shrinkage occurs in two dimensions: towards the typical OR for each birth defect and also towards the typical OR for each occupational group, accounting for multiplicities and dimensionality through an intrinsic Bayes correction via the hierarchical structure of the model. Thus, an advantage of this flexible modelling approach is that it is directly informed by the observed data: the degree of shrinkage is adaptively increased or decreased depending on the underlying data structure. Regression analyses were conducted using MATLAB version 7.7.0 (MathWorks Inc, Natick, Massachusetts, USA; 2008).

    An a priori set of potential confounding factors obtained from the interview included maternal residence at delivery (ie, study centre), maternal age, maternal race/ethnicity, maternal education, use of supplemental folic acid or prenatal vitamins, maternal smoking and alcohol use. Though some paternal characteristics were available (eg, paternal race/ethnicity), maternal characteristics were chosen for adjustment because paternal information was often missing. For these covariates, we used independent N(0,2) prior distributions, which are relatively non-informative.

    The primary referent group was a fixed referent consisting of the combined occupational groups ‘managers, administrators’ and ‘salesworkers’, which were considered to have little or no chemical exposure. Analyses were repeated using an alternative referent for each index occupational group consisting of all other groups combined, such that this revolving referent was different for each occupational group. Analyses were conducted first with all cases (cases with isolated or multiple defects) and then with cases of isolated defects only since isolated and non-isolated defects may differ aetiologically depending on the exposure and phenotype of interest. Adjusted posterior median ORs with 95% credible intervals (95% CI) are reported.

    Results

    Table 2 presents the distribution of occupational groups among fathers by case–control status. The majority of fathers (90%) held only one job during the critical period of interest; thus, most fathers (also 90%) were assigned to only one occupational group. Managers and administrators (10%), salesworkers (9%) and construction workers (9%) were the most common occupational groups for both cases and controls. Thirty-three groups were sparsely populated, each representing <1% of cases or controls. Of the 63 occupational groups in the coding scheme, only one group, metal miners, was unrepresented. Two groups, ‘managers and administrators’ and ‘salesworkers’, were combined for subsequent analyses because job descriptions were similar and often resulted in assignment to both occupational groups (eg, manager of an auto part sales department was assigned to both groups).

    View this table:
    Table 2

    Distribution of assigned occupational groups among fathers with one or more jobs during the critical exposure period,* National Birth Defects Prevention Study, USA, 1997–2004

    The main occupational analysis yielded over 20 000 effect measure estimates. A complete set of results is available from the corresponding author. To facilitate the presentation and interpretation of results, effect measure estimates are reported in table 3 only for occupation–defect combinations with any exposed cases for which the 95% CI excluded the null (1.0) before rounding to two significant digits or for which the observed effect estimate was ≥2.0 or ≤0.5 in either analyses with all cases (isolated and multiple defects) or with only cases of isolated defects. Thus, occupations were considered to be potentially associated with a defect if the effect estimate met these specified criteria. Since very similar results were observed using either referent group, results are presented only for analyses using the fixed referent category consisting of managers, administrators and salesworkers.

    View this table:
    Table 3

    Adjusted posterior median ORs and 95% credible intervals (CI) for birth defects associated with occupational groups assigned to fathers with one or more jobs during the critical exposure period,* National Birth Defects Prevention Study, USA, 1997–2004

    Several occupations were positively associated with three or more birth defect categories: mathematical, physical and computer scientists; artists; photographers and photo processors; food service workers; landscapers and groundskeepers; hairdressers and cosmetologists; office and administrative support workers; sawmill workers; petroleum and gas workers; chemical workers not elsewhere classified; printers; material moving equipment operators and motor vehicle operators.

    Occupations associated with several different defects within the same anatomic system include artists, for which large effect estimates were observed for several defects of the oral cavity, eyes and ears, gastrointestinal system, limbs and heart. Photographers and photo processors were associated with three different eye defects: cataracts, anophthalmos/microphthalmos and glaucoma/anterior chamber defects. Motor vehicle operators were associated with anophthalmos/microphthalmos and glaucoma/anterior chamber defects. Landscapers and groundskeepers were associated with three gastrointestinal defects: oesophageal atresia, duodenal atresia/stenosis and biliary atresia.

    A number of occupations were associated with reduced odds of certain birth defects. For example, mathematical, physical and computer scientists were associated with reduced odds of anorectal atresia/stenosis as well as coarctation of the aorta. However, no occupation was associated with reduced odds of more than one defect within the same anatomic system.

    Nearly one-third of the occupational groups were not associated with any birth defect: architects, drafters and designers; biological scientists; entertainers and athletes; healthcare practitioners; dentists and dental assistants; firefighters; fishers, hunters and trappers; vehicle manufacturers; foundry and smelter workers; stone, glass and concrete workers; painters; paper workers; semiconductor processors; electronic equipment operators; plant and system operators; rail transportation workers; armed forces and commercial divers.

    Discussion

    This analysis provides a broad examination of the relation between paternal work in over 60 occupations and numerous birth defects. We observed over 100 occupation–defect combinations with effect measure estimates that met our prespecified criteria for significance. Given the breadth of results and the potential for common aetiological pathways, one helpful summary approach is to examine the pattern of associated birth defects within each occupation. Several occupations were associated with more than one defect, suggesting the potential for diverse effects from a potential exposure or mixture of exposures represented by these paternal occupation categories. Some occupations were even associated with more than one defect within the same anatomic system, suggesting the possibility of varied effects from an early insult on morphogenesis, or perhaps differences in effects on components of an anatomic system depending on timing or dose of a teratogenic exposure. For example, photographers and photo-processing workers were associated with three distinct eye defects. Landscapers and groundskeepers were associated with three categories of gastrointestinal defects. Artists were associated with the most number of individual defects, including several eye/ear defects, oral clefts and defects of the gastrointestinal system. The artist group is particularly interesting as there are no exposed controls, which is reflected in the strikingly large ORs and upper limits of the credible intervals. To our knowledge, an increased prevalence of birth defects among offspring of artists has not been previously reported. However, some artists' media may contain organic solvents or lead, both of which have been previously associated with birth defects in offspring of exposed fathers.31 ,32

    An alternative interpretive approach is to examine the pattern of associations across occupations that are thought to share specific exposures in the workplace. However, we did not have any information about specific agents to which fathers in our study population may have been occupationally exposed. Therefore, to better elucidate any potential patterns in exposures across occupational groups found to be associated with birth defects in these data, and to offer an additional perspective on interpreting our results, we employed an existing classification scheme determined by industrial hygienist review of job titles to identify occupations commonly exposed to solvents, wood and wood products, heavy metals and pesticides, which have each been associated with certain birth defects in previous studies of paternal occupation.6 Occupations considered exposed to wood (eg, sawmill workers) or metals (eg, vehicle mechanics) were not observed to be associated with any consistent pattern of birth defects in our study. However, solvent-exposed occupations were significantly associated with an increased prevalence of numerous phenotypes among neural tube defects (NTDs), eye defects, oral clefts, gastrointestinal defects, limb deficiencies and heart defects. The following occupations were considered potentially exposed to solvents: artists, chemical workers, pharmacists, chemical engineers, electricians and electrical workers, janitors, mechanics, nurses, painters, dry cleaning and laundry workers, printers and plumbers. Associations between solvent-exposed paternal occupations and NTDs12 ,13 ,32 ,33 and other defects6–8 ,34 have been previously reported in many but not in all11 ,15 ,35 studies. We also observed an association between defects of the neural tube and both food processing workers and landscapers/groundskeepers, which is consistent with at least one other study reporting an association between pesticide-exposed occupations and NTDs.16 In contrast to some previous studies, we did not observe an association with pesticide-exposed occupations and limb deficiencies.21

    Only two population-based case–control studies in North America have previously examined a similar cross-classification of multiple paternal occupations and multiple birth defects. Olshan et al6 linked cases of birth defects from a Canadian surveillance registry to birth certificates, from which matched controls and information about paternal occupation were obtained. An occupational classification scheme similar to the one in our study was used, but there were important differences in the scope and definition of defects considered eligible. Standard statistical methods were applied, and all results were presented without restriction by statistical significance or magnitude of the effect estimates. We did not observe any of the noted occupation–defect associations reported in the Canadian study, such as janitors and ventricular septal defects (OR 2.45, 95% CI 1.10–5.45; 13 exposed cases); forestry and logging workers and cataracts (OR 2.28, 95% CI 1.29–4.02; 30 exposed cases) and atrial septal defects (OR 2.03, 95% CI 1.35–3.05; 54 exposed cases); and painters and spina bifida (OR 3.21, 95% CI 0.91–11.36; seven exposed cases) and cleft palate (OR 3.36, 95% CI 1.19–9.46; nine exposed cases). However, as observed by Olshan et al, our study yielded similar results for other elevated effect estimates (OR ≥1.5), such as shippers, messengers and cleft lip (OR 1.50, 95% CI 0.44–5.12; five exposed cases); chemical workers not elsewhere classified and cleft lip (OR 6.00, 95% CI 0.62–57.81; three exposed cases); motor vehicle operators and cleft lip + cleft palate (OR 1.49, 95% CI 1.04–2.11; 58 exposed cases); and photographers, photo processors and hypospadias (OR 2.00, 95% CI 0.28–14.23; two exposed cases). These results, though consistent, are based on small numbers of exposed cases in both studies.

    Schnitzer et al7 used self-reported job information from fathers in the Atlanta Birth Defects Case–Control Study, applied the same occupational classification scheme as Olshan et al, but considered additional birth defect categories. The authors presented results from conditional logistic regression analyses with an OR of >1.5 and at least three exposed cases. We observed only one association consistent with this study: food processors and hydrocephaly (OR 3.3, 95% CI 1.2–8.8; six exposed cases). A large Norwegian study linking national birth and occupation registries also examined multiple occupations and defects; the only common finding with this study was for vehicle mechanics and hypospadias (OR 5.19, 95% CI 1.31–14.24; three exposed cases).36

    A recent analysis of maternal occupation and birth defects in the NBDPS (1997–2003) employed a similar study design and analytic approach.22 Though the analysis of maternal occupation examined slightly different occupational groupings (given the different distribution of occupations between sexes), common findings are nevertheless potentially informative as they may point to teratogens encountered in the shared workplace. In both studies, solvent-exposed occupations were associated with defects of the neural tube, eye, limb, heart and gastrointestinal system as well as orofacial clefts. For example, the observed prevalence of amniotic bands and orofacial clefts was higher among offspring of both mothers and fathers who worked as janitors or cleaners. Admittedly, caution is warranted in making direct comparisons between studies of maternal and paternal occupation and occupational exposures. Given the expected variability in exposure patterns—even of the same physical or chemical agent—between men and women employed in the same occupation,37 an association may be observed between a particular occupation and defect for one gender (eg, the gender with the typically ‘higher’ exposure level) but not the other. Furthermore, because agents may demonstrate different mechanisms of teratogenesis and impact different target tissues following exposure during different time periods of susceptibility for men and women, it would not be unexpected if a particular occupational exposure had a true causal effect for one gender but not the other.

    Direct comparison with other studies of paternal occupation and birth defects is also complicated by differences in study population, source and classification of occupation, and grouping of birth defects. For example, we cannot directly compare our results with studies that lump all birth defects together or that use very different occupational classifications. Furthermore, this study is not designed to investigate any particular occupation–defect or exposure–defect relationship in depth, which would require an exposure assessment strategy beyond exclusive use of job titles. However, the results of our broad screening analysis suggested that paternal work in solvent-exposed occupations may be associated with an increase in the prevalence of several birth defects among offspring, including defects of the eye, NTDs and oral clefts. We also observed that a number of occupations (ie, artists; photographers and photo processors; motor vehicle operators; landscapers and groundskeepers) were positively associated with several defects within the same anatomic system, which may suggest heterogeneity in effects depending on unmeasured exposure parameters, such as timing or intensity.

    Using the Bayesian shrinkage approach, we were able to improve upon many limitations of previous studies. For example, to reduce misclassification, we used finer more homogeneous groupings of both occupation and birth defects. Preserving the aetiological diversity of individual phenotypes is a major analytic challenge in birth defects research, and our adaptive methods use shrinkage in a manner that allows examination of more homogeneous defect categories while also borrowing information across exposures and outcomes according to the presumed underlying structure of the observed data. Furthermore, these methods address the problematic issues related to using a large number of finer exposure and outcome categories, such as decreased precision and multiple hypothesis testing.

    Nonetheless, caution in the interpretation of the results is warranted. We found that when there were sufficient numbers of cases within a particular occupation, the model would allocate the coefficient for that occupation to the tails of the distribution and avoid shrinkage. However, when there were sufficient individuals in an occupation to rule out a zero log-odds ratio but not enough for reliable estimation of the coefficient, large point and interval estimates were sometimes obtained. For example, all artists in the study were fathers of cases, leading to inflated effect measure estimates and large upper interval bounds for this occupation. Furthermore, although large effect measure estimates (OR≥2.0) were observed for several occupation-defect combinations, most were based on few exposed cases.

    Other limitations associated with exposure assessment in our study should be considered when interpreting the results. First, paternal occupational histories were reported by mothers via interview. Although use of maternal reports likely introduced some error in our occupational classification, agreement between maternal and paternal report of paternal occupation has been shown to be high (up to 80%) within 2 years after birth.38 ,39 Furthermore, such misclassification has been demonstrated to be non-differential with respect to case status and thus will generally bias the effect measure estimates towards the null.40 Despite the potential error associated with maternal report of paternal job, the quality and completeness of paternal occupational histories obtained by maternal interview exceeds that obtained from birth certificates.41 ,42 Second, the process of coding reported jobs by occupation and industry, and the further classification into 63 aggregate occupational groups, likely introduced non-differential misclassification that may have resulted in attenuation of observed effects for any given occupation. Ultimately, observed associations should be interpreted with caution, as our study is limited to using occupational groups as a surrogate measure for workplace exposures and exposure mixtures potentially encountered at each job.

    Finally, we acknowledge that the length of time between the infant's birth and the maternal interview could be a potential source of recall bias. Women were asked between 6 weeks and 24 months after the EDD to report on jobs held by their baby's father during their entire pregnancy and the 3 months before conception. Although a pregnancy calendar was used to aid recall, an extended time-to-interview could contribute to increased errors in reporting and could further lead to recall bias if systematic differences in reporting accuracy due to time-to-interview exist between cases of different defects. However, the average time-to-interview in our study was <1 year (11 months for cases and 9 months for controls) and ranged from only 9 to 14 months among cases of different defects. Furthermore, mothers were asked about the fathers' job title and general description of tasks, which is likely less susceptible to recall error than questions about specific chemical agents or other workplace exposures. Therefore, we do not expect time-to-interview to be a significant source of recall error or bias in our study.

    Our exploratory study has a number of notable strengths that make it an important contribution to the existing literature on paternal occupation as a risk factor for birth defects. We used data from a national population-based case–control study with systematic case review and classification for a wide variety of birth defects among live born infants, stillbirths and electively terminated pregnancies. We were able to examine more than 60 defect categories, many of which have not been previously investigated in relation to paternal occupation. Despite the aforementioned limitations related to classification of occupation, our study improves upon previous studies that have relied on data from birth certificates because we obtained detailed occupational histories by interview on multiple jobs held during the time window most aetiologically relevant to male-mediated teratogenesis. We were able to account for potential confounding by several important covariates collected during the maternal interview. Furthermore, consistency in observed effects across our two comparison groups—a fixed referent comprised of all managers, administrators and salesworkers, representing a set of jobs considered unlikely to have substantial chemical exposure; and a revolving referent consisting of all occupations other than the index occupation, an approach which is frequently used in similar studies—minimises concern that our observed results are affected by residual confounding by factors such as socioeconomic status.15 Finally, our adaptive Bayesian analytic approach allows for the estimation of associations between numerous relatively homogeneous occupational groups and aetiologically distinct categories of birth defects, thereby addressing common small sample analytic limitations and avoiding the need for further aggregation of either exposure or outcomes into larger less homogeneous groups.

    Our study contributes evidence to the growing body of epidemiological literature on male-mediated teratogenesis. Findings from this broad screening analysis can be used to inform further investigation of specific paternal occupations found to be associated with birth defects and to generate hypotheses about chemical or physical exposures and exposure mixtures common to such occupations.

    Acknowledgments

    The authors thank Joanna Smith (University of North Carolina, Gillings School of Global Public Health, Chapel Hill, North Carolina) for her invaluable programming support.

    References

    1. ↵
      1. Jelliffe-Pawlowski LL,
      2. Shaw GM,
      3. Nelson V,
      4. et al
      . Risk of mental retardation among children born with birth defects. Arch Pediatr Adolesc Med 2003;157:545–50.
      OpenUrlCrossRefPubMedWeb of Science
    2. ↵
      1. Matthews TJ,
      2. MacDorman MF
      . Infant mortality statistics from the 2005 period linked birth/infant death data set. Natl Vital Stat Rep 2008;57:1–32.
      OpenUrlPubMed
    3. ↵
      1. Sever LE
      . Congenital malformations related to occupational reproductive hazards. Occup Med 1994;9:471–94.
      OpenUrlPubMed
    4. ↵
      1. Chia SE,
      2. Shi LM
      . Review of recent epidemiological studies on paternal occupations and birth defects. Occup Environ Med 2002;59:149–55.
      OpenUrlAbstract/FREE Full Text
    5. ↵
      1. Olshan AF,
      2. Faustman EM
      . Male-mediated developmental toxicity. Annu Rev Public Health 1993;14:159–81.
      OpenUrlCrossRefPubMedWeb of Science
    6. ↵
      1. Olshan AF,
      2. Teschke K,
      3. Baird PA
      . Paternal occupation and congenital anomalies in offspring. Am J Ind Med 1991;20:447–75.
      OpenUrlPubMedWeb of Science
    7. ↵
      1. Schnitzer PG,
      2. Olshan AF,
      3. Erickson JD
      . Paternal occupation and risk of birth defects in offspring. Epidemiology 1995;6:577–83.
      OpenUrlPubMedWeb of Science
    8. ↵
      1. Hooiveld M,
      2. Haveman W,
      3. Roskes K,
      4. et al
      . Adverse reproductive outcomes among male painters with occupational exposure to organic solvents. Occup Environ Med 2006;63:538–44.
      OpenUrlAbstract/FREE Full Text
    9. ↵
      1. Chia SE,
      2. Shi LM,
      3. Chan OY,
      4. et al
      . Parental occupations and other risk factors associated with nonchromosomal single, chromosomal single, and multiple birth defects: a population-based study in singapore from 1994 to 1998. Am J Obstet Gynecol 2003;188:425–33.
      OpenUrlCrossRefPubMedWeb of Science
    10. ↵
      1. Nguyen RH,
      2. Wilcox AJ,
      3. Moen BE,
      4. et al
      . Parent's occupation and isolated orofacial clefts in norway: a population-based case-control study. Ann Epidemiol 2007;17:763–71.
      OpenUrlCrossRefPubMedWeb of Science
    11. ↵
      1. Blatter BM,
      2. Hermens R,
      3. Bakker M,
      4. et al
      . Paternal occupational exposure around conception and spina bifida in offspring. Am J Ind Med 1997;32:283–91.
      OpenUrlCrossRefPubMedWeb of Science
    12. ↵
      1. Brender JD,
      2. Suarez L
      . Paternal occupation and anencephaly. Am J Epidemiol 1990;131:517–21.
      OpenUrlAbstract/FREE Full Text
    13. ↵
      1. Magnusson LL,
      2. Bonde JP,
      3. Olsen J,
      4. et al
      . Paternal laboratory work and congenital malformations. J Occup Environ Med 2004;46:761–7.
      OpenUrlPubMedWeb of Science
    14. ↵
      1. Sung TI,
      2. Wang JD,
      3. Chen PC
      . Increased risks of infant mortality and of deaths due to congenital malformation in the offspring of male electronics workers. Birth Defects Res A Clin Mol Teratol 2009;85:119–24.
      OpenUrlCrossRefPubMed
    15. ↵
      1. Shaw GM,
      2. Nelson V,
      3. Olshan AF
      . Paternal occupational group and risk of offspring with neural tube defects. Paediatr Perinat Epidemiol 2002;16:328–33.
      OpenUrlCrossRefPubMedWeb of Science
    16. ↵
      1. Lacasana M,
      2. Vazquez-Grameix H,
      3. Borja-Aburto VH,
      4. et al
      . Maternal and paternal occupational exposure to agricultural work and the risk of anencephaly. Occup Environ Med 2006;63:649–56.
      OpenUrlAbstract/FREE Full Text
    17. ↵
      1. Fear NT,
      2. Hey K,
      3. Vincent T,
      4. et al
      . Paternal occupation and neural tube defects: a case-control study based on the oxford record linkage study register. Paediatr Perinat Epidemiol 2007;21:163–8.
      OpenUrlCrossRefPubMedWeb of Science
    18. ↵
      1. Bradley CM,
      2. Alderman BW,
      3. Williams MA,
      4. et al
      . Parental occupations as risk factors for craniosynostosis in offspring. Epidemiology 1995;6:306–10.
      OpenUrlPubMedWeb of Science
    19. ↵
      1. Pierik FH,
      2. Burdorf A,
      3. Deddens JA,
      4. et al
      . Maternal and paternal risk factors for cryptorchidism and hypospadias: a case-control study in newborn boys. Environ Health Perspect 2004;112:1570–6.
      OpenUrlCrossRefPubMedWeb of Science
    20. ↵
      1. Garcia AM,
      2. Benavides FG,
      3. Fletcher T,
      4. et al
      . Paternal exposure to pesticides and congenital malformations. Scand J Work Environ Health 1998;24:473–80.
      OpenUrlPubMedWeb of Science
    21. ↵
      1. Hanke W,
      2. Jurewicz J
      . The risk of adverse reproductive and developmental disorders due to occupational pesticide exposure: an overview of current epidemiological evidence. Int J Occup Med Environ Health 2004;17:223–43.
      OpenUrlPubMed
    22. ↵
      1. Herdt-Losavio ML,
      2. Lin S,
      3. Chapman BR,
      4. et al
      . Maternal occupation and the risk of birth defects: an overview from the National Birth Defects Prevention Study. Occup Environ Med 2010;67:58–66.
      OpenUrlAbstract/FREE Full Text
    23. ↵
      1. Yoon PW,
      2. Rasmussen SA,
      3. Lynberg MC,
      4. et al
      . The national birth defects prevention study. Public Health Rep 2001;116(Suppl 1):32–40.
      OpenUrlCrossRefPubMedWeb of Science
    24. ↵
      1. Rasmussen SA,
      2. Olney RS,
      3. Holmes LB,
      4. et al
      . Guidelines for case classification for the national birth defects prevention study. Birth Defects Res A Clin Mol Teratol 2003;67:193–201.
      OpenUrlCrossRefPubMedWeb of Science
    25. ↵
      Executive Office of the PresidentOffice of Management and Budget. Standard Occupational Classification Manual. Lanham, MD: Bernan Press, 2000.
    26. ↵
      Executive Office of the PresidentOffice of Management and Budget. North American Industry Classification System: United States, 1997. Lanham, MD: Bernan Press, 1998.
    27. ↵
      1. Cordier S
      . Evidence for a role of paternal exposures in developmental toxicity. Basic Clin Pharmacol Toxicol 2008;102:176–81.
      OpenUrlCrossRefPubMedWeb of Science
    28. ↵
      1. Schnitzer PG,
      2. Teschke K,
      3. Olshan AF
      . A classification scheme for aggregating U.S. Census occupation and industry codes. Am J Ind Med 1995;28:185–91.
      OpenUrlPubMedWeb of Science
    29. ↵
      1. Tipping ME
      . Sparse bayesian learning and the relevance vector machine. J Machine Learn Res 2001;1:211–44.
      OpenUrlCrossRef
    30. ↵
      1. Liu F,
      2. Dunson D,
      3. Zou F
      . High-dimensional variable selection in meta-analysis for censored data. Biometrics 2011;67:504–12.
      OpenUrlCrossRefPubMedWeb of Science
    31. ↵
      1. Bellinger DC
      . Teratogen update: lead and pregnancy. Birth Defects Res A Clin Mol Teratol 2005;73:409–20.
      OpenUrlCrossRefPubMed
    32. ↵
      1. Logman JF,
      2. de Vries LE,
      3. Hemels ME,
      4. et al
      . Paternal organic solvent exposure and adverse pregnancy outcomes: a meta-analysis. Am J Ind Med 2005;47:37–44.
      OpenUrlCrossRefPubMedWeb of Science
    33. ↵
      1. Aguilar-Garduno C,
      2. Lacasana M,
      3. Blanco-Munoz J,
      4. et al
      . Parental occupational exposure to organic solvents and anencephaly in Mexico. Occup Environ Med 2010;67:32–7.
      OpenUrlAbstract/FREE Full Text
    34. ↵
      1. Correa-Villasennor A,
      2. Frencz C,
      3. Loffredo C,
      4. et al
      . Paternal exposures and cardiovascular malformations: the Baltimore-Washington infant study group. J Expo Anal Environ Epidemiol 1993;3:173–85.
      OpenUrlPubMed
    35. ↵
      1. Brender J,
      2. Suarez L,
      3. Hendricks K,
      4. et al
      . Parental occupation and neural tube defect-affected pregnancies among Mexican Americans. J Occup Environ Med 2002;44:650–6.
      OpenUrlCrossRefPubMedWeb of Science
    36. ↵
      1. Irgens A,
      2. Kruger K,
      3. Skorve AH,
      4. et al
      . Birth defects and paternal occupational exposure. Hypotheses tested in a record linkage based dataset. Acta Obstet Gynecol Scand 2000;79:465–70.
      OpenUrlCrossRefPubMedWeb of Science
    37. ↵
      1. Eng A,
      2. t'Mannetje A,
      3. McLean D,
      4. et al
      . Gender differences in occupational exposure patterns. Occup Environ Med 2011;68:888–94.
      OpenUrlAbstract/FREE Full Text
    38. ↵
      1. McKean-Cowdin R,
      2. Preston-Martin S,
      3. Pogoda JM,
      4. et al
      . Reliability of demographic, smoking and occupational data provided by mothers vs. fathers in a childhood cancer study. Paediatr Perinat Epidemiol 2000;14:257–62.
      OpenUrlCrossRefPubMedWeb of Science
    39. ↵
      1. Tagiyeva N,
      2. Semple S,
      3. Devereux G,
      4. et al
      . Reconstructing past occupational exposures: How reliable are women's reports of their partner's occupation? Occup Environ Med 2011;68:452–6.
      OpenUrlAbstract/FREE Full Text
    40. ↵
      1. Schnitzer PG,
      2. Olshan AF,
      3. Savitz DA,
      4. et al
      . Validity of mother's report of father's occupation in a study of paternal occupation and congenital malformations. Am J Epidemiol 1995;141:872–7.
      OpenUrlAbstract/FREE Full Text
    41. ↵
      1. Brender JD,
      2. Suarez L,
      3. Langlois PH
      . Validity of parental work information on the birth certificate. BMC Public Health 2008;8:95.
      OpenUrlCrossRefPubMed
    42. ↵
      1. Shaw GM,
      2. Malcoe LH,
      3. Croen LA,
      4. et al
      . An assessment of error in parental occupation from the birth certificate. Am J Epidemiol 1990;131:1072–9.
      OpenUrlAbstract/FREE Full Text

    Supplementary materials

    • Press release

      Press release

      Files in this Data Supplement:

    Footnotes

    • The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.

    • Funding This work was supported by the Centers for Disease Control and Prevention (cooperative agreement number U50CCU422096) and the US National Institute of Environmental Health Sciences (grant number P30ES10126). This manuscript has been approved for submission to Occupational and Environmental Medicine by the National Birth Defects Prevention Study and the National Center on Birth Defects and Developmental Disabilities.

    • Competing interests None.

    • Ethics approval Ethics approval was provided by UNC Institutional Review Board and by CDC Institutional Review Board.

    • Provenance and peer review Not commissioned; externally peer reviewed.