Document LpYRe5rVNLEq972GqweVdxbwb

American Journal of Industrial Medicine 25:361-383 (1994) Review of Epidemiologic Studies of Paternal Occupational Exposure and Spontaneous Abortion David A. Savitz, pud, Nancy L Sonnenfeld, msph, and Andrew F. Olshan, PhD The question of whether paternal exposures influence risk of spontaneous abortion is of great public interest, with the possibiiity supported by laboratory investigations. Thirtynine studies of male occupational exposure and risk of spontaneous abortion were examined, with the methods and results tabulated. Many of those reports were limited by exposure data based on maternal report of the father's job title or by potentially inaccurate paternal reports of spontaneous abortion, though the quality of more recent studies is markedly enhanced. Mercury has been implicated most strongly based on recent studies that included quantitative exposure estimates: a number of studies show ing associations for exposure to anesthetic gases. Suggestive associations have also been found inconsistently for exposure to lead, rubber manufacturing, selected solvents, and some pesticides. Further study is encouraged, but with more intensive effort to measure accurately both spontaneous abortion and occupational exposures. O 1994 Wiley-Uss. Inc. Key words: fathers, occupational exposure, mercury, pesticides, solvents, spontaneous abortion, paternal reproduction risk INTRODUCTION Over the past decade, there has been a growing interest in the potential role of male occupational exposures in reproductive health. A male contribution to infertility is now well accepted, with at least one agent (dibromochloropropane) an established cause of male infertility [Whorton et al., 1977], The phenomenon of concern in this review is the possible impact of paternal workplace exposures on the risk of spon taneous abortion. Such an association might be mediated by maternal exposure through contamination of the home environment [Knishkowy and Baker, 1986], concentration of the agent in semen (Stachel et al., 1989], and passage through sexual intercourse, or transmission of the agent directly on the sperm [Yazigi et al., 1991]. Of particular interest here is a more direct paternal effect on the fetus. Genet ically damaged sperm that remain capable of fertilization might produce an impaired conceptus that is more likely to result in a pregnancy loss. Paternally mediated ab normalities in development are the presumed phenomenon underlying such concerns Department of Epidemiology, School of Public Health, and Carolina Population Center, University of North Carolina. Chapel Hill. NC. Address reprint requests to David A. Savitz, Ph.D., Department of Epidcmtology, CB #7400. School of Public Health, University of North Carolina, Chapel Hill. NC 27599. Accepted for publication February 24, 1993. 1994 Wiley-Liss, Inc. NOTICE: THIS M.AT-Bl M CClPVaiQi^r u.vv-(1.71' MAV I'll PROTECTED BY - 1 P> ~ em 362 Savitz et al. as male exposure to Agent Orange and birth defects in their offspring [Erickson et al., 1984], and ionizing radiation and childhood leukemia [Gardner et al., 1990]. Al though there is not yet a well-established male exposure proven to influence post conception reproductive events in general, or spontaneous abortion in particular, there are several lines of evidence that strongly support the possibility, including the sizable body of epidemiologic data reviewed here. Experimental studies of male laboratory animals (mostly rodents) have clearly indicated the potential for developmental toxicity following paternal exposure to ionizing radiation and mutagenic chemicals. These studies have shown that the full spectrum of developmental endpoints appear to be responsive to paternal exposure, including gene mutations, chromosomal aberrations, lethality, congenital anomalies, tumors, neurobehavioral deficits, and growth retardation [Olshan and Faustman, 1993]. In vivo assays used to assess paternal genetic effects include the specific locus test, dominant skeletal, dominant cataract, and dominant lethal tests, and the heritable translocation test [Russell and Shelby, 1985]. The most relevant test to human spontaneous abortion is probably the dominant lethal test [Green et al., 1985], A dominant lethal mutation is defined as a generic defect occurring in the gamete that permits fertilization but results in embryonic death. The genetic basis for these deaths is believed to be structural or numerical chromosome abnormalities in the germ cell of the treated male, with death occurring before or after implantation. An EPA-Genetox working group summarized the pub lished studies using this assay [Green et al., 1985] and found that, of the 140 chemicals tested, 65 (46%) were determined to be positive. Although the dominant lethal test is not a direct analog of human spontaneous abortion, it does demonstrate that the agent affects the paternal genetic material. For the sizable fraction of human spontaneous abortions that are due to chromosomal abnormalities (estimated at 40% by Kline et al. [1989]), the assay is more directly applicable. Litter size reduction in laboratory animal studies is also relevant to human spontaneous abortion. Reduced litter size has been found following male exposure to ionizing radiation and ethylnitrosurea [Russell and Hunsicker, 1988; Selby and Russell, 1985], Occupational exposures are clearly capable of adversely influencing human semen [Wyrobek et al., 1983a,b], yet few studies have attempted to evaluate whether abnormal semen characteristics are associated with spontaneous abortion. Case-con trol studies from the 1960s [Furuhjelm et al., 1962; Joel, 1966] found that concep tions ending in spontaneous abortion were more likely to be the product of `^abnormal" semen than pregnancies that were successful. The semen abnormalities linked to poor pregnancy outcome were reduced sperm concentration, increased proportion with abnormal morphology, elevated sperm concentration, and reduced DNA content of the sperm. Given the era in which these studies were conducted, the lack of technological sophistication in semen evaluation, as well as methodological deficien cies concerning confounding, analytic methods, and selection of control groups are expected. The question of whether agents that influence sperm grossly may adversely affect the health of the resulting conceptus remains largely unexarained. In this review, we will summarize the epidemiologic evidence concerning as sociations between paternal exposures and the risk of spontaneous abortion. /The methodologic strategies of the past studies will be critically examined to note the key limitations and to recommend specific improvements needed to,extend knowledge on this topic. Of particular concern in this research area is the challenge in accurately 3J CO Paternal Exposure and Spontaneous Abortion 363 ascertaining-the occurrence of spontaneous abortion, the validity of exposure assign ment based on job title, and the potential for confounding by lifestyle factors of the mother or father that are correlated with paternal occupation, all discussed in more detail afterthe presentation of the results. Finally, given the array of studies to date, the most promising avenues for further research will be noted. t * METHODS OF REVIEW We attempted to identify all studies in which the father's occupation was ex t. amined in relation to the risk of spontaneous abortion in his partner. Literature T searches, cross-checking reference lists, and consultation with experts were used to h% make this compilation of evidence as complete as possible. In many instances, spon taneous abortion was only one among many outcomes considered, or paternal occu pational exposures were one among many potential risk factors considered, raising questions about whether our search was completely successful.'Studies that contained fewer than five exposed cases in all the associations they reported [e.g.. Goldsmith et al., 1984] were excluded from the review. Studies were not excluded on the basis ii of qualitative criteria, but we attempted to provide sufficient methodologic informa tion for the reader to make independent evaluations of the quality of the evidence. i Information abstracted for this review is presented in nine text tables and an appendix, with the studies arranged in chronological order within the tables. Table 1 presents key methodological data on the studies, including the source population, sources of exposure and outcome data, number of spontaneous abortions, and re sponse proportion, in order to provide some insight into the potential biases and limitations of the various studies. The final column in Table I reports the categories of exposure examined in each study, serving as an index to the following seven tables, which are organized to present results for a specific category of exposure. The source population, delineated as "general population" or "occupational cohort" was noted since, in general, occupational cohorts provide the opportunity for more refined exposure assessments than studies of the general population, but may be less generaiizable. The number of spontaneous abortions reported in Table I includes all observations of spontaneous abortion used in any relevant analysis within the paper, even though the number of spontaneous abortions included in any specific i analysis was often smaller. The response proportion noted in Table I is that reported by the original author for the total study population, including both the exposed and unexposed. In a number of studies, the reported response proportion provides a rather favorable assessment of the representativeness of the sample population, since indi viduals who were excluded because of migration or incomplete records were omitted from the denominator. The response proportion typically reports only the proportion who had complete data in records or agreed to the interview, and does not take into account unusable data or missing values on specific items of interest. Tables II--VIII present the results from the studies that were reviewed, divided into seven broad categories of exposure: metals, solvents, anesthetic gases, pesti cides, physical agents, hydrocarbons and automobile exhaust, and miscellaneous industries, occupations, and agents. The heterogeneity within some of these catego ries should be noted, with solvents including a variety of exposures in the plastics and rubber industry, and pesticides broadened to include sterilizing agents and wood preservatives. The results tables include the exposed group used in each analysis and 30 fio C/J u CroO o CO f 364 Savitz et al. TABLE I. Reports on Paternal Occupational Exposure and Spontaneous Abortion: Study Designs and Methods Reference Source population Sourcn of exposure data Sources of outcome data Number of spontaneous abortions Response proportion (%> Exposures Studied4 Cohen ei al. 119741 Boueei Ml. (19731 Cohen cl al. 119751 KmJMoneset al. |I975| Infante* al. (1976) Sanotskit (19761 Rosenberg and Vamuncn |I97*| Tomlin (1979| Cohen *1.(19*01 Khanviet at. (19*01 CarmeUi et al. (19*11 Lrenreryi a al-11911] Qocfcmin A Noidmom (19*21 Kcmill al. (19*21 Kline al. 119821 Smithctal. (|982| Townsend B al. (19*21 Cnoil. iwni* Hemminfct et al. (19*31 Nordstrom B al. (19*21 Undbohm et al. (19*4) Morgan et al. (19*41 Rom Bal. (19*41 Savitz et al. (19*4) Suftkind A Heiubtff |IS4| Brodsky B al. (19*3} Dmtell A Vaughan |I9**1 Stellman et al. (198*) CDC|I9S9| Akser et al. |199| Aschenerau A Monson (19*91 McDonald es al. |I9*9| Taskincn al. (19*9) Guirgfiii el al. (1990] Restrepo et al. (1990) Coidierctal. (19911 Ltndbohm B al. (1991a} Occupational cohort General population Occupational cohon Occupational cohort OocupBioaai cohort Occupational cohort Occupational cohort OccupMoui a*m OcnpMnaal otnt Occupational cnbon Occupational cohort Occupational cohort Occupational cohort Occur*mall General populBion Occupational cohort Occupational cohort Gcncnl popaMn General pqpulaiinw Occupational cohort General pupulmon Occupational cohort Occupational cohort OccupBional cohort Occupational cohort Ompualotai General population Occupational cohort Occupational cohort OccupBional cohort General population General population Occupational cohort Occupation* cohort Occupational oBon Occupational cohort General population Paternal report Maternal report Paternal report Paternal report Company record! Workplace monitonnf Paternal report Paternal report Paternal report paternal report Paternal report Paternal report Preernal report Workplace monitonnf PMoulitpon Maternal report paternal report Employer records Company records NA Census records Company records Paternal report Ceneus records Company records Paternal report Paternal report Union records Paternal report Work location Company records Paternal report Birth certificate Paternal report Military records Paternal report Company records Biological monoonnt' Medical records Military records Maternal report Solvent rtfwry Budofical monitonnf Paternal report Employer information Paternal report Workplace roonnonnf Company records Biological monitonnf Census records Paternal report XaiyayiK Paternal report Paternal report Paternal report Paternal report Paternal repon Paternal report Paternal repon Paternal report Hospital records Maternal report Paternal report Paternal report Hospital records Paternal report Karyotype of fetus Prenatal clinic records Maternal report Medical records Maternal report Maternal report Hospital registry Hospital records Paternal report Hospital registry Maternal report Paternal report Hospital records Maternal report Matcrnal/peternal reports Paternal report Paternal report 8mh certificate Paternal repon Paternal report Paternal repon Hospital records Hospital records Maternal report Hospital registry Outpatient records Paternal report Paternal report Maternal report Hospital registry Undbotun al. 11991b] Rupactal. (I99l| Occupational cohort Occupational cohort Purmil report Biolofical monitonnf NA Hospital registry Clinic records NA 153 544* 2*1 1.452 73 na* 62 54 6*0 29 134 95 134 48 991 92 29* 9.419 210 68 3.399 22 103 220 120 672 NA 427 1.75* 170 201 10.893 120 4*8 *1 70 8.731 213 1.535 54 5K 48 70 62-77 NA 69 92 74 61 56 47 *9 83 79 A P* A A 5 S A A A P P A M S M Pa H Ms 86 P 62 P NA P 83 MMi 89 Pa 91 MSHMs S3 Ms 2* 35 53* 73 91 NA S4-S7 81 94 NA 75 P S P M S P P M P MSPPa HMs 5 Ms 81 A 95 P 77 M NA MSPPa K Ms 74 MSPaH NA P *M = metals, S = solvents: A = anesthetic gases; P = pesticides; Pa = physical agents; H = hydrocarbons and automobile exhaust; Ms = miscellaneous industries, occupations, and agents. bKaryotypically abnormal case-control. `NA = not available. ``Cited in Sterling and Arundel (1986]. 'Excludes losses because of mortality. v l I t : i: i ! R&S 143210 Paternal Exposure and Spontaneous Abortion 365 an estimate of the risk ratio with 95% confidence interval (Cl) contrasting exposed to unexposed men. The desired 95% CIs were often not provided and had to be derived from the data using test-based methods^Rothman, 1986] for riskratios orodds ratios as required [Alcser et al., 1989; Beckman arid Nordstrom, 1982; Boue et al., 1975; Brodsky etal., 1985; Cohen etal., 1974, 1975;Hamilletal., 1982; Hemminkietal., 1983; Infante et alw 1976; Kharrazi et al., 1980; Knill-Jones et al., 1975; Lauwerys et al., 1981; Morgan et al., 1984; Nordstrom et al., 1983; Rosenberg and. Vanttinen, 1978; Roan et al., 1984; Rupa et al., 1991; Savitz et al., 1984; Smith et al., 1982; Sterling and Arundel, 1986; Suskind and Hertzberg, 1984]. Exposure classification was often based on a combination of job title or em ployment setting and an inference regarding the specific agents of interest. The author's terminology and intentions were retained in describing the exposures. When a broad array of exposures were analyzed, only exposures for which there were five or more exposed, cases were tabulated. Papers frequendy included multiple analyses of closely related groups. To restrict the size of the tables and to retain clarity, it was necessary to present these results selectively. When multiple unexposed groups were analyzed and all comparisons produced similar results, results were presented for the largest unexposed group or the one thought least likely to introduce bias. When multiple unexposed groups produced different results and none was clearly superior to the others, all analyses were tabulated. Adjusted results were provided when confounding was present and the adjusted results were available. Indications of effect modification by age or other factors were not routinely included in the results tables but are discussed in the text where relevant. The Appendix table presents additional information on study location, study type, confounders addressed, definition of cases, and baseline risk of spontaneous abortion. The potential confounders listed for a given study include all confounders addressed in any exposure examined within that study but not all results provided *- statistical adjustments for the entire list of factors. Footnotes are used in the results tables to indicate potentially important discrepancies between the confounders listed for a given study and the confounders addressed for a given exposure. Included among the confounders addressed are those risk factors which the authors claimed to I have accounted for, sometimes based on inappropriate methods such as statistical testing of differences between study groups or relying only on an impression that the * groups were not notably different. The baseline risk of spontaneous abortion was computed for the unexposed when possible, but in some studies only the overall rate of spontaneous abortion among exposed plus unexposed was available. RESULTS OF REVIEW *As summarized.in .Table I, 39 studies published over the last 18 years have -provided data to evaluate.4he^potential link between paternal occupation and spon-. taneous abortion. Only 8 appeared in the 1970s, with notable growth to the present and 4 appearing in 1991 alone. The attributes of those studies vary markedly and are noted as relevant in discussing the agents which they address. Most studies have relied on maternal or paternal self-report for both exposure and pregnancy outcome data. Study sizes vary markedly, from around 20 to nearly 10,000 spontaneous abortions. Response proportions range from under 50% to around 90%, with an obvious potential for influencing the study's validity. R&S 143211 366 Savitz et al. TABLE II. Results of Stndta of Paternal Exposures to Heavy Metals and Spontaneous Abortion Reference Beckman & Nordstrom 11982] Kline el al. (1982) Hemminki et al. [1983] Lindbohm et al. 11984) Brodsky et al. [1983] Alcser etal. (1989) McDonald et al. [1989] Cordier al. [1991] Lindbohm et al. [1991a] Lindbohm et al. [1991b] Ageni. industry, or occupation Copper smelter (lead, arsenic, mercury. cadmium) Metals Metallurgical industry (zinc, cobalt) Metal industry Metals manufacturing Metal-plate and construction steel Dentistry (mercury) Urinary mercury levels 2,000-3,999 V*tJ1 Urinary mercury levels 4.000-9.000 Pgfl Ore, metal, stone processing Metal machining Metal fabricating Urinary mercury levels 1-19 p.g/1 Urinary mercury levels 20-49 p.g/1 Urinary mercuty levels >49 p.g/1 Lead Nickel and nickel oxides Chromium/chramium compounds Metals Wire & pipe drawers Blood lead levels 1.0-1.4 p.mol/1 Blood lead levels 1.3-1.8 p.mol/1 Blood lead levels > 1.8 pmol/l Blood lead levels 1.0-1.4 p.mol/1 during spermatogenesis Blood lead levels > 1.4 p.mol/1 during spermatogenesis Cadmium Chromium Nickel Zinc Copper RR 1.5 0.5* l.0b l.l 0.7 0.9 1.4 0.9 1.1 1.7 1.2 1.0 0.9 1.3 1.7 2.3 0.9 1.0 1.0 1.0 1.2 1.0 1.3 1.6 0.7 3.8 1.4 0.9 1.1 2.2 2,8 95% Cl 0.9-2.3 0.2-1.1 0.4-2.2 0.8-1.5 0.4-1.5 0.8-1.0 1.0--1.9 0.7-1.1 0.8-1.6 1.1-2.5 0.9-1.5 0.9-1.2 0.7-1.2 1.0-1.7 1.0-3.0 1.0-5.2 0.8-1.0 0.9-1.1 0.9-1.1 0.8-1.1 0.5-3.1 0.6-1.7 0.5-3.4 0.6-4.0 0.3-1.9 1.2-12.0 0.6-3.4 0.5-1.6 0.6-1.9 0.8-5.5 1.1-7.0 `Public clinic patients, normal karyotype. `'Public clinic patients, abnormal karyotype. Exposure to Heavy Metals Occupational agents as potential causes of spontaneous abottion have been examined in a number of mostly recent studies (Table I). Ten studies contain infor mation that address paternal exposure to lead, mercury, and other metals in relation to spontaneous abottion (Table I and II), Many are community-based surveys, with exposure defined by inferences based on employment sector or job title rather than direct measurement. However, four studies focused directly on cohorts exposed to the metals of interest, namely, lead [Lindbohm et al., 1991b] and mercury [Brodsky et al., 1985; Alcser et al., 1989; Cordier et al., 1991]. With these exceptions, the Paternal Exposure and Spontaneous Abortion 367 I#**'*:-'-' ' magnitude of metal exposure is not known and may not even be present for some of the "exposed" groups (e.g., work in the "metal industry"). Also, many of these exposure sources (e.g., copper smelter) contain a range of potentially hazardous agents, in addition to metals. A number of studies provide evidence of positive associations, including smelter workers (RR = 1.5) [Beckman and Nordstrom, 1982], metal-plate and steel industry workers (RR = 1.4) [Lindbohm et al., 1984], men with elevated estimated lead exposure (RRs = 1.0, 1.3, and 1.6 for low, moderate, and high levels, respec tively), zinc exposure (RR = 2.2), and copper exposure (RR = 2.8) [Lindbohm et al., 1991b]. Most notable is the association for high measured mercury exposure (RR = 1.7) [Alcser et al., 1989] and estimated mercury exposure (RRs = 1.3, 1.7, and 2.3 for low, moderate, and high levels, respectively) [Cordieret al., 1991] (Table II). Unique among the studies was Lindbohm et al.'s [1991b] evaluation of lead exposures around the time of spermatogenesis, which yielded a markedly elevated but imprecise risk ratio. Contradictory null results or even inverse associations are also present in Table II, including the absence of an increased risk for dentists preparing mercury amal gams [Brodsky et al., 1985] and workers with job titles thought to indicate exposure to lead [Lindbohm et al., 1991a], However, the most methodologically sophisticated in terms of exposure assessment [Alcser et al.. 1989: Lindbohm et al., 1991b; Cordier et al., 1991] provide strong suggestions of a link between exposure to heavy metals, especially mercury, and spontaneous abortion. A' '* Exposure to Rubber, Plastics, and Solvents The array of agents under this rubric is quite broad, with a number of studies related to rubber, plastics, and related industries, as well as specific solvents and other agents often encountered in those work settings (Table III). In addition to the community studies that contribute results on a wide range of exposures [Lindbohm et al., 1984; McDonald etal., 1989; Lindbohm et al., 1991a],(a number of studies have targeted particular exposures in this category, most notably vinyl chloride [Infante et al., 1976], toluene diamine and dinitrotoluene [Hamill et al., 1982], several specific solvents such as toluene and trichloroethylene [Taskinen et al., 1989], and perchloroethylene [Eskenazi et al., 1991]. Regardless of the results, the effort to design studies capable of implicating or exonerating specific agents is commendable. The initial study of Infante et al. [1976] provided the first suggested link between males' occupational chemical exposures and reproductive outcome. An overall elevation in risk of spontaneous abortion was found among spouses of men working with vinyl chloride (RR = 1.8), pronounced among younger (but not older) fathers (RR = 3#)*(TabIe III). Maternal age was not ascertained so that it could be examined only indirectly through the closely correlated father's age. Sanotskii [1976] reported a strong association of spontaneous abortion with chloroprene exposure, but insufficient methodological details are available to criti cally evaluate that observation. Risk ratios of 1.5 or greater were reported for organic solvents in general (RR = 2.3), toluene (RR = 1.5), and xylene (RR = 1.8) by Taskinen et al. [1989], in a study with unusual precision in exposure classification to specific solvents in specific industrial settings. Lindbohm et al.'s (199la] survey found sizable associations with gasoline or benzene exposure in petroleum refineries (RR = 2.2), trichloroethane and methylene chloride (RR = 1.9) in rubber manu- %e*r' fa 368 Savitz et al. TABLE IIL Result! fat Studies of Paternal Exposure to Robber, Plastics, or Solvents and Spontaneous Abortion J* Is Reference Agent, industry, or occupation RR 95% Cl Infante et al. (1976] Sanotskii (1976] Hamill et al. (1982] Lindbohm et al. (1984] Vinyl chloride, polyvinyl chloride Chloroprene and manufacturing Toluene diamine, dinitrotoluene Solvents 1.8* >3.0 1.0 0.9 1.0-3.1 NAb 0.6-1.8 0.7-1.1 Savitz etal. [1984] Daniell & Vaughan (1988) Halogenated hydrocarbons Autobody workers Painters in maintenance, construction Fiberglass, bathtub, boat, plastics workers Printers, pressmen, lithographers 1.2 0.8-1.8 1.0 0.8-1.2 0.9 0.8-1.1 0.9 0.7-1.2 1.1 0.8-1.3 McDonald et al. [1989] Laundry/dry cleaning Rubber, plastic fabricating Printing operations 0.9 0.6-1.5 1.0 0.8-1.5 0.9 0.8-1.2 Taskinen et al. [1989] Organic solvents Aromatic hydrocarbons Styrene Toluene Xylene Halogenated hydrocarbons Trichloroethylene 1,1.1 -Trichloroethane Aliphatic hydrocarbons Acetone Miscellaneous solvents 2.3 . 1.6 1.3 1.5 1.8 1.1 1.0 0.9 1.5 1.0 1.7 1.1-5.0 1.0-2.4 0.8-2.1 0.9-2.5 1.1-3.2 0.6-1.8 0.6-2.0 0.3-2.3 0.9-2.5 0.6-1.7 1.1-2.6 Lindbohm et al. (1991a) Petroleum refinery solvents (gasoline, benzene) Benzene Trichloroethylene Rubber manufacturing solvents (trichloroethylene, methylene chloride) Rubber chemicals Styrene Plastic manufacturing solvents (styrene) Plastic monomers Rubber product workers 2.2 1.3-3.8 1.0 0.7-1.3 0.9 0.3-2.1 1.9 1.2-2.8 1.5 1.1-2.2 0.9 0.5-1.8 1.0 0.7-1.5 0.8 0.4-1.4 1.5 1.1-2,2 Lindbohm et al. [1991b] Organic solvent exposure 1.1 0.7-1.5 Postemployment risk for exposed vs. unexposed, bNot available. facturing, and trichloroethane in general (RR = 1.5). It is difficult to juxtapose results given the diversity of agents of interest, but there are replicated positive associations present for rubber workers, and to a lesser extent for petroleum refinery products and individual solvents ^No attempts to replicate the observations for vinyl chloride (Infante et al., 1976], chloroprene [Sanotskii, 1976], -andpecific solvents implicated by Taskinen et al. [1989].have been reported. Many of the negative studies are limited in their ability to exonerate any ex posures"^ven the weak methods of exposure classification combined with poor re sponse proportions [Savitz et al., 1984] or questionably high rates of spontaneous' abortion [Daniell and Vaughan, 1988]. However, some of fheniegative results are more credible. Hamill et al.'s [1982] study was small but included a strong effort to R&S 143214 I Paternal Exposure and Spontaneous Abortion 369 TABLE IV. Results In Studies of Paternal Exposure to Anesthetic Gases and I Spontaneous Abortion_________________________________________________ Reference Agent, industry, or occupation RR 95% a Cohen et il. (1974] Anesthesiologists Anesthetic nurses Operation room technicians 0.9 0.8-1.1 1.2 0.3-2.6 1.8 0.8-4.1 Cohen et al. [197S] Knill-Jones et ai. (1973] Rosenberg A Vanttinen (1978] Dentists working with anesthetics >3 hr/week Physicians working in openting room Anesthesiologists 1.8 1.4-2.2 1.1* 0.7-1.7 0.8 0.5-1.4 Tomlin (1979] Anesthesiologists working in operating room 20 or more bouts per week 1.6 0.9-3.0 Cohen et al. (1980] Dentists working with anesthesia 1-8 hr/week Dentists working with anesthesia 8 -1- hr/week 1.1 1.0--1.4 1.3 1.3-1.8 Lauwerys et al. (1981] Guirguis et al. (1990] Physicians, nurses working in OR Physicians, nurses, other staff in OR 1.1 0.7-1.8 2.3 1.7-3.1 OR - operating room. 'Father exposed but mother not exposed. accurately classify workplace exposure. McDonald et al.'s [1989] study did not identify associations for iaundry/dry cleaning or rubber workers, and a number of solvents, including benzene and carbon disulfide, were not associated with sponta neous abortion in Lindbohm et al.'s [1991 a] study.^The most promising leads to pursue would be exposures in the rubber industry and exposures to vinyl chloride and specific solvents associated with spontaneous abortion (Table III). Exposure to Anesthetic Gases Given the observation that maternal exposure to anesthetic gases could increase the risk of spontaneous abortion [Tannenbaum and Goldberg, 1985], a number of studies in the 1970s addressed the possibility that paternal exposure to anesthetic gases could produce a similar effect (Table IV). All but one of the eight studies of this issue were published between 1974 and 1981, and all were based on surveys of male health professionals, generally dentists or physicians. All were questionnaire surveys of self-reported exposure and pregnancy outcome, yet the reported risk of spontane ous abortion varied rather markedly from 5.1% [Guirguis et al., 1990] to 13.2% [Rosenberg and Vanttinen, 1978] in spite of the similar methods (Appendix table). Such variation may be due to true heterogeneity in the populations, selection bias due i to nonresponse patterns, or erroneous reporting. Response proportions did, in fact, vary quite markedly (41-92.4%), but there was not an obvious association between response proportion and reported risk of spontaneous abortion. The comparability of exposed and unexposed groups is unusually favorable, in that subgroups of poten tially exposed health professionals (physicians, dentists, or technicians) were com pared to other (presumably sociologically similar) subgroups from the same profes sion who were not likely to be exposed. Most of these studies have found increased risks for the partners of anestheticexposed men (Table IV). Risk ratios of 1.5 or greater have been reported for operating room technicians (RR = 1.8) [Cohen et al.. 1974], dentists working with anesthesia over 3 hours per week (RR = 1.8) [Cohen et al., 1975], anesthesiologists working 1 R&S 143215 370 Savitz et al. I half-time or more in the operating room (RR = 1.6) [Tomlin, 1979], dentists working^ with anesthetics 8 or more hours per week (RR = 1.5) [Cohen et al., 1980], and hospital workers exposed to the operating room (RR = 2.3) [Guirguis et al., 1990], Other studies of physicians in particular, including those with briefer exposure peri ods, have not found increased risk (Knill-Jones et al., 1975; Rosenberg and Vanttinen, 1978]. The suggestive evidence for dentists is noteworthy, although the domi nance of the literature by a single group of investigators (with a particular approach to study methods, questionnaire design, etc.) diminishes the value of the independent ^replications.' The traditionally intensive and uncontrolled use of anesthetics, espe cially nitrous oxide, in dental practice [Cohen et al., 1980] warrants further exami nation. Exposure to Pesticides and Related Products Fifteen studies have provided data to assess risks of spontaneous abortion from male exposure to pesticides, antimicrobials and related products encountered in chem ical manufacturing, agricultural application, and from herbicide use in the war in Vietnam (Tables I and V). The diversity of agents and exposure settings is apparent. Nearly all studies established exposure through paternal report, sometimes sup plemented by company records [Smith et al., 1982; Townsend et al., 1982; Suskind and Hertzberg, 1984] or military records [Stellman et al., 1988; Aschengrau and Monson, 1989]. Overall, substantial attention was given to specific exposures in these studies, with most studies explicitly designed to address the agent(s) of interest. However, the degree of success in pinpointing actual exposures to Agent Orange, for example, or pesticides in agricultural applications is still quite limited. Reproductive outcomes were typically ascertained through maternal or paternal report. In reviewing the methodological details (Table I and Appendix table), notable outliers are the baseline risk of spontaneous abortion of 1.9% in Restrepo et al.'s [1990] study and the response of only 26% in the survey of Roan et al. [1984], seriously diminishing the credibility of their results. Usually, but not always, major potential confounders such as maternal age and cigarette smoking were addressed. Even without the clearly invalid result of Restrepo et al. [1990], a twofold range in spontaneous abortion risks (5.8-11.9%) was observed across studies. Results of pesticide studies were mixed (Table V), but again it is difficult to define which results address truly analogous exposures. DBCP exposure was found to increase spontaneous abortion risk in one study (RR = 3.0) [Kharrazi et al., 1980]. The results for 2,4-D, 2,4,5-T, dioxin, and Agent Orange show sporadic increases in risk, for example, among forestry workeis(RR = 1.6) [Carmellietal., 1981] and men potentially exposed to herbicides in Vietnam based on self-perception [CDC, 1988] or imputed potential exposure (Stellman et al., 1988].X)ther studies of related expo sures, equal' or superior in quality, found no increased risk [Smith et al., 1982; Townsend etal., 1982; Suskind and Hettzberg, 1984; Aschengrau and Monson, 1989]. Rupa et al. [1991] found an increased risk of spontaneous abortion in the wives of male cotton workers in India. Although the study lacked some methodological details, the agents and exposure conditions may be worthy of closer examination. The notable associations reported by Restrepo et al. [1990] are greatly tempered by the anomalous spontaneous abortion risk of 1.9%, which suggests serious error in re porting accuracy or completeness, and the report-of marked elevations in risk, even for jobs thought to-involve no exposure to pesticides. Similarly, Roan et al.'s [1984] ' R&S 143216 Paternal Exposure and Spontaneous Abortion 371 TABLE V. Results in Studies of Paternal Exposure to Pesticides, And-Microbials, and Related Products and Spontaneous Abortion Reference Agent, industry, or occupation RR 95* a Khimzi et al. (1980) Dibromochloropropane (DBCP) 3.0 1.3-7.0 Canndlietal. (1981] 2.4-dichlorophenoxyacetic add (2,4-D) Fanning Forestry/commercial applicators 1.0 0.5-2.0 1.6 0.8-3.1 Smith et al. (1982) Townsend et al. (1982) 2,4.5-trichlorophenoxyacetic acid (2,4.5-T) 2.3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) Any dioxin 0.9 1.0 1.0 0.6-1.3 0.7-1.4 0.8-1.4 Canet al. [19831 Goldsmith et al. (1984) Roanet al. (1984) Agent Orange DBCP Agricultural pesticides 1.2 1.2-1.3 1.1 0.3-4.0 1.2 1.0-2.1 Suskind Ar Hertzberg- (1984} CDC [1989] 2,4,5-T and dioxins Herbicides (any perceived exposure) Low Moderate High 0.9* 0.6-1,2 1.3 1.2-1.4 1.2 1.0-1.4 1.4 1.2-1.6 1.7 1.3-2.1 Stellman et al. (1988] Agent Orange Handled herbicides Low exposure Moderate exposure High exposure 1.6 0.8-2.9 1.3 1.0-1.6 1.5 1.1-2.0 1.7 1.3-2.3 Aschcngrau & Monson (19891 McDonald et al. (1989) Service in Vietnam Agricuiture/horttculture 0.9 0.4-1.9 1.0 0.8-1.4 Restrepo et al. (1990) Floricultural industry No pesticide job Low pesticide job Moderate pesticide job High pesticide job 2.4 1.1-5.0 1.6 0.7-3.7 2.0 1.2-3.3 0.7 0.1-5.3 Lindbohm et al. (1991a) Ethylene oxide Imptegnants of wood Formaldehyde (low exposure) Formaldehyde (high exposure) 4.7 1.2-18.4 1.9 0.7-5.0 1.1 0.9-1.4' 1.0 0.8-1.4 Rupa et al. 11991] Pesticides used in cotton production . ... ......1,7.._ .1.6-1.9 'Not adjusted despite differences between' exposed and unexposed in age, education, and duration of smoking. report of a 1.5-fold increased risk for agricultural pilots is diminished in its impor tance by the poor response proportion. ^.mong the diverse results that were reported. the observatfons^most deserving of attempted replication (given study'quality and biologic plausibility) are the increased risk for DBCP [Khanazi et al., 1980], the association of spontaneous abortion with cotton pesticides (RR = 1.7) [Rupa et al., 1991], and the isolated observation of an increased risk from ethylene oxide (RR *= 4.7) [Lindbohm et al., 1991a], Exp sure to Physical Agents Of the six studies providing data on male exposure to physical agents (ionizing and non-ionizing radiation, heat), only one (Nordstrom et al., 1983] was focused on 4 372 Savitz et al. TABLE VI. Remits in Studies of Paternal Exposure to Physical Agents and Spontaneous Abortion Reference Boue et al. (1975) Kline ct al. [1982] Agent, industry, or occupation Ionizing radiation Electrical Heat Nonionizing radiation RR 1.3 1.2* 2.0* 0.7-1,8b Nordstrom et al. [1983] Electromagnetic fields Other jobs (lowest voltage) Switchyard/linc construction (moderate) Switchyard workers (high voltage) 0.6 0.6 1.0 McDonald et al. [1989] Lindbohm et al. (1991a] Lindbohm et al. (1991b] Ionizing radiation Radon exposure X-tays 0.9 1.2 1,5 'Public patients, normal karyotype. ''Range of risk ratios across normal/abnormal karyotype, public/private clinic. 95% Cl 1.1-1.5 0.6-2.3 0.9--4.9 -- 0.3-1.1 0.3-1.1 0.5-1.7 0,6-1.3 0.8-1.8 0.6-3.6 this particular topic, so that the quality of exposure assessment was generally limited (Table VI). Suggestive associations were found for ionizing radiation in two studies [Boue et al., 1975; Lindbohm et al., 1991b), with modestly elevated risk ratios of 1.3-1.5. Nonionizing radiation, a ubiquitous but poorly defined exposure, was found by Kline et al. [1982] to be associated with spontaneous abortions of both normal and abnormal karyotype but only among private patients. Nordstrom et al. [1983] did not find power-frequency electromagnetic fields to be associated with increased risk. Given clear evidence that ionizing radiation causes mutations (BEIR, 1990] and that heat can impair sperm function [Levine et al., 1990]' this avenue appears to be worthy of pursuit in spite of the results currently available. The only data on heat exposure provided a suggestively positive result [Kline et al., 1982]. Exposure to Hydrocarbons and Exhaust Several studies provided data on miscellaneous hydrocarbon and motor vehicle exhaust exposures, all as components of broader surveys (Table VII). Nonetheless, perhaps because of the literature linking such agents to childhood cancer in the offspring [Savitz and Chen, 1990], several investigators had considerable interest in these exposures [Kline et al., 1982; Lindbohm et al,, 1984; Lindbohm et al., 1991a]. These studies address diverse exposures that are highly prevalent, such that a number of studies provide evidence of small but statistically precise elevations in risk [Lind bohm et al., 1984; McDonald ct al., 1989], Consideration of the magnitude of increase found across studies yields rather consistently close to the null. Only the risk ratios of 1.4 and 1.5 found for chimney sweeps and refinery workers, respectively [Lindbohm et al., 1991a], would constitute positive results worthy of replication. However, the presence of known mutagenic agents within these exposure groups provides a greater impetus for further study than can be derived from these results. Exposure to Miscellaneous Industries, Occupations, and Agents In addition to the results that fall into the above categories, many reported associations did not fall into any of the specific categories. For example, one prom- R&S 143218 Paternal Exposure and Spontaneous Abortion 373 TABLE VII. Results In Studies of Paternal Exposure to Exhaust and Hydrocarbons and Spontaneous Abortion ___________________________________________ Reference Agent, industry, or occupation RR 95% a Kline etal. (1982) Auto fumes (narrow) Auto fumes (narrow) Auto fumes (narrow) Hydrocarbons (narrow) Hydrocarbons (narrow) 1.2* 0.8-1.7 l.3b 0.4-3.7 t.O* 0.6-1.5 1.2* 0.7-2.3 1.2C 0.6-2.6 Lindbohm et al. (1984) Polycyclic aromatic hydrocarbons Automobile exhaust and fumes Service station attendants 1.0 0.9-1.1 1.0 0.9-1.1 1.2 1.0-1.6 McDonald et al. (1989) Mechanics and repairers Transport operations/matenals handling 1.1 1.0-1.2 1.0 0.9-1.1 Lindbohm et aJ. [1991a] Polycyclic aromatic hydrocarbons (moderate/high) Polycyclic aromatic hydrocarbons (low) Gasoline Chimney sweeps Refinery workers 1.0 1.0 0.9 1.5 1.4 0.9-1.2 0.9-1.0 0.8-1.1 0,9-2.4 0.8-2.5 Lindbohm et at. (1991b) Carbon monoxide l.l 0.8-1.7 `Normal karyotype, public clinic patients. "'Normal karyotype; private clinic patients. `Abnormal karyotypes, public clinic patients. inent study [Morgan et al., 1984] addressed wastewater treatment plant workers who are not easily classified elsewhere (Table VIII). Notably elevated risk ratios (2.0 or greater) have been found for chemical crushers (RR = 2.2), cloth sewers (RR = 2.5), caretakers of fur-bearing animals (RR = 2.3) [Lindbohm et al., 1984], all wastewater treatment plant workers (RR = 2.1), and mechanical, instrument, and electrical workers in the wastewater treatment plant (RR = 4.9) [Morgan et al., 1984], Associations between 1;5 and 2.0 have been reported for leather industry workers (RR = 1.8) [Hemminki et al., 1983] and for watchmakers (RR = 1.5) [Lindbohm et al.. 1991a]. A number of specific agents that could plausibly be linked to spontaneous abortion (e.g., mutagens, miscellaneous industrial chemicals) were queried or im puted by investigators in a number of studies but did not yield any notably positive associations (Table VIII). None of the reported risk ratios reached 1.5, providing little basis for directing future study efforts. The explicit goal of studies such as these is to suggest agents more worthy of detailed examination, and leads from the above results are rather nonspecific in their suggestions for further study of hazardous chemicals in industry and health care. DISCUSSION Summary of Results The literature providing data to evaluate possible associations between paternal occupational exposures and spontaneous abortion is large but, at least until recently, not of high quality. Most studies examining the issue did so without having focused on this particular topic, with limited effort to measure exposure or outcome or both 374 Savitz et al. TABLE VIU. Remits in Studies of Miscellaneous Paternal Exposum/Occupatioas and Spontaneous Abortion ... Reference Agent, industry, or occupation RR 93% Cl Kline ei al. (1982| Hcmmmki et al, |1983| Chemicals Food Chemical industry Textile industry Leather industry 0.7-0.8* 0.7-1.2J 0.8 0.8 1.8 -- -- 0.4-1.6 0.4-1.8 0.8-3.9 Lindbohm ci al. |I984| Chemical crushers Cloth sewers Caretakers of fur-bearing animals Textile dust Animal microorganisms Other chemicals 2.2 1.1-4.3 2.5 1.3-5.0 2.4 1.3-4.4 1.0 0.6-1.5 1.3 0.9-1.9 0.9 0.7-1.2 Morgan et al. (I984| Waste water treatment plant workers All job titles Process operators Mechanical, instrument, electrical 2.1 0.9-4.5 0.9 0.1-6.8 4.9 1.5-15.6 McDonald et al. |I989| Physical scientists Physicians/dcnlists Technicians/xrt Chemical processing Food, beverage, wood, textiles Wood, stone machining Electrical fabricating Wood, textile, leather labncating Construction trades Stationary equipmentimiscellaneous manufacturing ' 1.2 1.0 1.1 0.9 1.0 1.2 0.9 1.0 1.0 1.1 0.9-1.6 0.7-1.2 0.8-1.2 0.6-1.4 0.9-1.2 0.8-1.8 0.7-1.0 0.8-1.1 0.9-1.1 0.9-1.4 Taskinen et al. 11989| Carcinogens Dusts 1.3 0.9-2.1 1.4 0.7-2.6 Lindbohm et al. |I99U| Watchmakers Textile machine seller's operator' Textile finishers and dryers Mutagens (high exposure Mutagens < low exposure* 1.5 0.8-2.7 1.2 0.8-1.9 1.3 0.8-2.4 1.0 0.9-1.2 1.0 0.9-1.0 'Range of risk ratios across normal/abnormal karyotype, public/private clinic. in the detail that would be desired. Nonetheless, some suggestions emerge from the literature regarding specific agents that are worthy of further evaluation. Given the imprecision in exposure assessment which plagues most of this literature, the ability of negative studies to persuasively exonerate agents is quite limited. A summary of the evidence for several agents of particular interest is provided in Table IX. The recent studies suggesting positive associations of occupational exposure to metals (mercury and lead) to spontaneous abortion (Alcser et al., 1989; Lindbohm et al., 1991b; Cordier et al.. 1991] are of potential importance, especially given the evidence that lead [Lancranjan et al., 1973] and mercury [Lee and Dixon, 1973] may be disruptive to spermatogenesis. Although these agents are not thought to be mutagenic, developmental alterations through other processes such as epigenetic mechanisms may be involved. Mercury, in particular, has been found to be associated' with increased risk in two high-quality studies. .' Solvents -would be among the agents of greatest interest based on their wide- 'k R&S 143220 Paternal Exposure and Spontaneous Abortion TABLE IX. Summary of Strength of Evidence linking Selected Paternal Exposures to Spontaneous Abortion Agent Elevated RR Evidence from in >1 Study RR> 1.5 high quality studies Overall Lead Mercury Solvents Anesthetic gases Pesticides Phenoxy herbicides ionizing radiation Hydrocarbons No Yes Yes Yes Yes Yes No No Yes Yes Yes Yes Yes Mixed Yes Mixed Yes No Yes No No No No No Moderate Strong Moderate Strong Weak Weak Very weak Very weak 375 spread use and potential for mutagenicity, and there are sporadic indications of f increased risk associated with such exposures in the rubber and petrochemical indus tries. However, there has been little or no replication of the positive associations reported by Infante et al. [1976] for vinyl chloride, Sanotskii [1976] for chloroprene. or Taskinen et al. [1989] for a number of different solvents used in specific industrial settings. Known adverse effects of these agents have resulted in reductions in expo sure, so that replication is not feasible. Anesthetic gases have long been of interest as reproductive hazards given the link between exposure during pregnancy and risk of spontaneous abortion (Tannenbaum and Goldberg, 1985]. Studies of dentists and other health care workers con sistently suggest an association between male exposure and spontaneous abortion, but these are all based on mail surveys and many are from the same investigators. Refinements in exposure and outcome assessment are needed in future studies, but the reduction in anesthetic gas exposures in most settings may limit the informativeness of such efforts. Pesticides, especially Agent Orange, have been the focusof much effort but there g- are at most sporadic indications of positive associations. The diversity of agents and exposure circumstances that have been studied makes it particularly difficult to identify comparable studies, but there are few observations suggestive of an adverse effect of paternal pesticide exposure. Unless some means of overcoming the formidable chal lenges of exposure assessment can be identified (e.g, biological monitoring), addi tional simplistic efforts hold little promise. Exposure to Agent Orange, in particular, has been difficult to evaluate [Boyle et al., 1989; Ulienfeld and Gallo, 1989], Ionizing radiation and hydrocarbons have produced limited support for an as sociation with spontaneous abortion. However, the theoretical basis for interest in these highly mutagenic agents to which many people are exposed is strong enough to warrant further study, in spite of the limited epidemiologic support. Ionizing radiation, in particular, may be amenable to studies that take advantage of the unique individual monitoring programs of large numbers of workers in place at many industries. Methodological Issues The difficulties in conducting studies of paternal exposure and spontaneous abortion are substantial, including the challenge of isolating a paternal exposure from maternal influences, determining whether an association reflects indirect exposure to the mother rather than a direct sperm-mediated process, heterogeneity in the causes t. 376 Savitz et al. of spontaneous abortion,.and the difficulty of accurately measuring the occurrence of spontaneous abortion. Assessment of spontaneous abortion, regardless of the exposure of interest, is challenging. Reports of clinically recognized spontaneous abortions are known to incompletely reflect the total number of conceptions that are lost [Wilcox et al., 1988] . Although valid results may still be obtained for the category of "clinically recognized spontaneous abortions," at the early gestational ages at which recognition is ambiguous the reporting may be both incomplete [Wiicox and Homey, 1984] and biased in relation to exposure status. The limited public perception of a paternal role in the etiology of spontaneous abortion makes differential reporting by exposure status somewhat less likely than for such highly publicized maternal exposures as anesthetic gases or use of video display terminals. Males are known to be particularly unreliable informants regarding reproductive history [Selevan, 1985]. Regarding spontaneous abortions, it is not only more likely that men would forget their wife's experience but even possible that they would be unaware that an early spontaneous abortion had occurred at all. Studies should rou tinely seek reproductive outcome information from the woman, whenever possible, contending with the logistical difficulties that can arise in locating and interviewing spouses no longer living with the male worker of interest. Research in which both the male (for exposure information) and female (for reproductive outcome information) are interviewed is needed in spite of the additional costs. Perhaps the most critical issue with regard to identification and classification of spontaneous abortions is the heterogeneity in the causes of spontaneous abortion. Isolation of the subset of all spontaneous abortions that are plausibly attributable to the paternal genome would markedly strengthen any true etiologic associations./One approach that has been taken in epidemiologic studies is to distinguish chromosomally normal and abnormal abortuses based on karyotyping the conceptus (Kline et al., 1989] , Some karyotypic abnormalities are attributable to the father, whereas others are clearly contributed by the mother, such that identification of the paternally derived subset of karyotypically abnormal abortuses could, in theory, be made. In practice, however, the logistical challenges of even obtaining information on the karyotype of the abortus are daunting, and the additional requirement to identify the source of the abnormality poses an additional challenge. For karyotypically normal abortuses, a potential role for paternal factors cannot be discounted, in that there may still be genetic determinants that do not act at the chromosomal level. It may be possible to identify more simplistic indicators of spontaneous abortions most likely to be influenced by paternal exposures by strati fying cases on gestational age to determine whether early or later losses are more affected, or to restrict the analysis to mothers at low risk (young mothers, nonsmok ers, etc.). Clinical and laboratory studies aimed at identifying the determining pa rental origin of spontaneous abortion would dramatically accelerate the development of this literature. Until recently, exposure classification has been rather poor, often based on the wife's report of her husband's job title or the male's employment in a broad industrial category. The limitations inherent in a job title as a marker of exposure are well known [Siemiatycki, 1991], with imperfect sensitivity and specificity as an indicator of actual exposure to any particular agent. The design of these studies would gener ally lead to a prediction of nondifferential misclassification, i.e., the uncertainty in R&S 143222 Paternal Exposure and Spontaneous Abortion 377 going from a job title to an exposure assignment would be similar for men who had fathered conceptions resulting in spontaneous abortion as in those who had fathered conceptions resulting in live bom children. Under those circumstances, the resulting risk estimates are diluted, understating the magnitude of any edologic association [Copeland et al., 1977]. This is the principal reason that "negative" studies should not be viewed as persuasive evidence that the agents of interest do not produce increased risks;' Instead, it would be safe to infer that the job categories do not have a widespread, extremely strong adverse influence on spontaneous abortion. Confounding is always a concern, with speculation limited only by the limited knowledge of risk factors for spontaneous abortion. Paternal occupation is undoubt edly correlated with such lifestyle factors as smoking [Sterling and Weinkam, 1976; Siemiatycki et al., 1988], yet there is little evidence to suggest that paternal lifestyle factors such as smoking [Beckman and Nordstrom, 1982; Taskinen et al., 1989] or alcohol use [Halmesmaki et al., 1989; Parazzini et al., 1990] influence risk of spontaneous abortion. Confounding by other workplace agents is always possible, given that jobs with an environmental hazard rarely involve only one such agent. At present, however, with no paternal exposure that constitutes a proven etiologic agent, such concerns are highly speculative. Even among maternal factors, the only proven determinants are history of spontaneous abortion (which may reflect persistent ex posure or inherent characteristics of the woman), advanced maternal age, and pos sibly cigarette smoking or alcohol use (Kline et al.. 1989]. Since paternal occupation is a determinant of social class, and social class is closely related to maternal smoking during pregnancy [Williamson et al., 1989], maternal smoking should be considered as a potential confounding factor, yet only about half of the studies we reviewed did so (Appendix table). A number of study design issues need to be considered in evaluating this literature and making recommendations for future research. Some studies have com pared preemployment and postemployment outcomes in the same individuals [Townsend et al., 1982; Restrepo et al., 1990]. Although such a design offers the advantage of controlling for any intrinsic factors to the couple, the potential for confounding by maternal and paternal age is substantial. Induced abortions have the potential for biasing comparisons of spontaneous abortion risks in comparisons of populations with differing induced abortion occur rence. Specifically, induced abottions remove pregnancies from being at risk of spontaneous abortion, and failure to consider them as "pregnancies at risk" artifi cially inflates the risk of spontaneous abortions relative to what would be found in a life table analysis [Olsen, 1984]. Alternatively, if induced abortions are counted as pregnancies at risk without noting that they ended prior to the end of the risk period for spontaneous abortion, the risk of spontaneous abortion will be artificially reduced relative to what would be found in a life table analysis [Modvig et al., 1990],4foe ideal approach is to determine precise timing of all induced and spontaneous abortions and to conduct an analysis of spontaneous abortion incidence by week of gestation. However, the accuracy of reported induced abortions (especially by males) and the precision of the timing of induced and spontaneous abortions makes this strategy difficult to implement in practice. The size and resulting statistical power of these studies should also be consid ered. Starting with a large work force, the number of informative events (pregnancies among spouses) is limited by the proportion of men who are married and have 378 Savitz et al. reproductive experiences in the time period of employment, often resulting in highly imprecise risk estimates. Many of the most interesting groups, for example, workers exposed to DBCP, are too small to generate reliable risk estimates. Accepting the inherent constraints in addressing the topic of paternal exposure and spontaneous abortion, there are a number of opportunities for advancing knowl edge through improved study methods. Opportunistic studies in which paternal ex posures are included in studies focused on other issues (e.g., maternal exposures) can still add valuable information. In the context of a well-designed study of spontaneous abortion (including registry-based studies), addition of a modest series of questions about the father's exposures to potentially harmful agents through habits (tobacco, alcohol, illicit drugs), at home (hobbies, home repairs), and at work (querying work activities and locations as well as specific agents) would be contributory. A+tnore ambitious step would be the detailed ascertainment of occupational exposure histories as developed by Siemiatycki-and colleagues for a community-based study of occu pational carcinogens (Siemiatycki, 1991]. (Their two-stage approach involves an ini tial interview followed by an interview with an industrial hygienist focusing in detail on activities and agents associated with the individual's work setting. In addition to opportunistic studies, sufficient evidence has accrued to justify development of targeted studies of paternal exposures. Populations of men with potentially harmful exposures could be studied in retrospective cohort designs with attention to accurate classification of exposure presence and timing based on industryspecific job-exposure matrices. In addition to industrial exposures, men exposed to medications or men with identified sperm characteristics could be studied to assess the reproductive outcomes in their spouses. Carefully designed case-control studies in the community could obtain data on lifestyle factors as well as retrospective reports of occupational exposures. In parallel with the evolving epidemiologic evidence, the need for additional laboratory studies should be emphasized. Insofar as the toxicologists can pinpoint exposures or suggest subsets of paternally derived spontaneous abortions, the epide miologic literature would benefit markedly. In the absence of a clearly demonstrated example of paternally caused spontaneous abortions in humans, skepticism about the potential for this process to operate will remain. However, the absence of such clear-cut evidence is not a result of a series of carefully designed negative studies. In contrast, the present literature consists of studies of improving quality that contain a number of suggestively positive results. In combination with laboratory evidence and public concern, this research avenue deserves to be pursued to evaluate thoroughly the possibility that paternal exposures constitute an important, preventable cause of spontaneous abortions. REFERENCES Alcser KH, Brix KA, Fine LJ, Kalienbach LR, Wolfe RA (1989): Occupational mercury exposure and male reproductive health. Am i Ind Med 15:517-529. Aschengrau A, Monson RR (1989): Paternal military service in Vietnam and risk of spontaneous abor tion. J Occup Med 31:619-623. Beckman L, Nordstrom S (1982): Occupational and environmental risks in and around a smelter in northern Sweden. Hereditas 97:1--7. BEIR (1990): "Health Effects of Exposure to Low Levels of Ionizing Radiation: BEIR V." Committee on the Biological Effects of Ionizing Radiations, National Research Council. Washington, DC: National Academy Press. R&S 143224 ;a & t Paternal Exposure and Spontaneous Abortion 379 Boue J, Booe A, Lazar P (1973): Retrospective and prospective epidemiological studies of 1300 kary otyped spontaneous human abortions. Teratology 12:11-26. -Boyfe<CA,-rDccotifl6- P,-0'-Brien..TR (1989): Long-term health, consequences of military service in ^ietnam. Epidenbol*Rev:-l 1:1-27" Brodsky JB. Cohen EN. Whitcher C. Brown BW, Wu ML (1983): .Occupational exposure to mercury in dentistry and pregnancy outcome. JADA 11:779-780. Cinnelli D, Hofherr L, Tomsic J, Morgan RW (1981): "A Case-Control Study of the Relationship Between Exposure to 2,4-0 and Spontaneous Abortions in Humans." Final report to the National Forest Products Association and U.S. Department of Agriculture--Forest Service. Menlo Park, CA: SRI International, August 14, 1981. Centers for Disease Control Vietnam Experience Study (CDC) (1988): Health status of Vietnam veterans, III. Reproductive outcomes and child health. JAMA 239:2713-2719. Cohen EN. for the Ad Hoc Committee on the Effect of Trace Anesthetics on the Health of Operating Room Personnel, American Society of Anesthesiologists (1974): Occupational disease among operating room personnel: A national study. Anesthesiology 41:321-340. Cohen EN, Brown BW, Bruce DL, Cascorbi HF, Corbett TH, Jones TW, Whitcher CE (1975): A survey of anesthetic, health hazards among demists. JADA 90:1291-1296. Cohen EN, Brown BW, Wu ML, Whitcher CE, Brodsky JB, Gift HC, Greenfield W, Jones TW, Driscoll EJ (1980): Occupational disease in dentistry and chronic exposure to trace anesthetic gases. JADA 101:21-31. Copeland KT, Checkoway H, McMichael AJ. Holbrook RH (1977): Bias due to misclassification in the estimation of relative risk. Am J Epidemiol 105:488-495. Cotdier S. Deplan F, Mandereau L. Hemon D (1991): Paternal exposure to mercury and spontaneous abortions. Sri Ind Med 48:375-381. - Daniell W, Vaughan TL (1988): Paternal employment in solvent related occupations and advene preg nancy outcomes. Br J Ind Med 45:193-197, Erickson JD, Mulinare J, McClain PW. Fitch TG. James LM, McClearn AB, Adams MJ Jr. (1984): Vietnam veterans' risks for fathering babies with birth defects. JAMA 252:903-912. Eskenazi B, Fenster L, Hudes M, Wyrobek AJ. Katz DF. Gerson J, Rempel DM (1991): A study of the effect of perchloroethylene exposure on the reproductive outcomes of wives of dry-cleaning workers. Am J Ind Med 20:593-600. Furuhjelm M, Jonson B, Lagergren C-G (1962): The quality of human semen in spontaneous abortion. IntJ Feitil 7:17-21. Gardner MJ, Snee MP, Hall AJ, Powell CA. Downes S. Terrell JD (1990): Results of case-control study of leukaemia and lymphoma among young people near Sellafield nuclear plant in West Cumbria. Br Med J 300:423-434. Goldsmith JR, Potashnik G, Israeli R (1984): Reproductive outcomes in families of DBCP-exposed men. Arch Envir Health 39:85-89. Green S, Auletta A, Fabricant J. Kapp R, Manandhar M. Sheu C. SpringerJ, Whitfield B (1985): Current status of bioassays in genetic toxicology--the dominant lethal assay: A report of the U.S. Envi ronmental Protection Agency Genc-Tox Program. Mutat Res 154:49-67. Guirguis SS. Pelmear PL, Roy ML. Wong L (1990): Health effects associated with exposure to anaes thetic gases in Ontario hospital personnel. Br J Ind Med 47:490-497. Halmesmaki E, Valimaki M, Roine R, Ylikahri R, Ylikorkala O (1989): Maternal and paternal alcohol consumption and miscarriage. Br J Obstet Gynecol 96:188-191. Hamill PW, Steinbergcr E, Levine RJ, Rodriguez-Rigau LI, Lemeshow S, Avrunin JS (1982): The epidemiologic assessment of male reproductive hazard from occupational exposure to TDA and DNT. J Occup Med 24:985-994. Hemminki K, Kyyronen P. Hiemi M-L, Koskinen K. Sallmen M, Vainio H (1983): Spontaneous abor tions in an industrialized community in Finland. Am J Publ Health 73:32-37. Infante PF, Wagoner JK. McMichael AJ, Waxweiier RJ. Falk H (1976): Genetic risks of vinyl chloride. Lancet 1:734. Joel CA (1986): New etiologic aspects of habitual abortion and infertility, with special reference to the male factor. Fertil Steril 17:374-380. Khartazi M. Potashnik G, Goldsmith JR (1980): Reproductive effects of dibromochloropropane. Israel J Med Sci 16:403-406. Kline J, Stein Z. Hatch M. Hutzler M. Lai A. Pantel-Silverman J (1982): "Surveillance of Parental R&S 143225 380 Savitz et si. Employment and Spontaneous Abortion." Final Report to the National Institute for Occupational \ Safety and Health. Contract No. 210815002. Cincinnati. OH: National Institute for Occupation^; Safety and Health. Kline J, Stein Z, Susscr M (1989): "Conception to Birth. Epidemiology of Prenatal Development." New York: Oxford University Press. Knill-Jones RP. Newman BJ, Spence AA(1975): Anaesthetic practice and pregnancy. Lancet 2:807-809. Knishkowy B, Baker EL (1986): Transmission of occupational disease to family members. Am J Ind Med 9:545-550. Lancranjan I, Popescu HI. Gavanescu O. Klepsch I, Serbanescu M (1975): Reproductive ability of workmen occupationally exposed to lead. Arch Environ Health 30:396--401. Lauwerys R. Siddons M, Misson CB. Borlee I. Bouckaen A. Lechat MF. Temmerman PD (1981): Anaesthetic health hazards among Belgian nurses and physicians. Int Arch Occup Environ Health 48:195-203. Lee IP, Dixon RL (1975): Effects of mercury on spermatogenesis studied by velocity sedimentation cell separation and serial mating. J Pharmacol Exp Ther 194:171-181. Levine RJ, Mathew RM, Chenault CB, Brown MH. Hum ME, Bentley KS. Mohr KL. Working PK (1990): Differences in the quality of semen in outdoor workers during summer and winter. N Engl J Med 323:12-16. Liiienfeld DE. Gallo MA (1989): 2,4-D, 2.4,5-T, and 2,3,7,8-TCDD: An overview. Epidemiol Rev 11:28-58. Lindbohm M-L, Hcmminki K, Bonhomme MG. Anttila A. Rantala K, Heikkila P. Rosenberg MJ (1991a): Effects of paternal occupational exposure on spontaneous abortions. Am J Pub Health 81:1029-1033. Lindbohm M-L, Sallmen M, Anttila A. Taskinen H, Hcmminki K (1991b): Paternal occupational lead exposure and spontaneous abortion. Scand J Work Environ Health 17:95-103. Lindbohm M-L, Hcmminki K, Hyyronen P (1984): Parental occupational exposure and spontaneous abortions in Finland. Am J Epidemiol 120:370-378. McDonald AD, McDonald JC, Armstrong B, Cherry N, Delorme C, Nolin AD, Robert D (1987): Occupation and pregnancy outcome. Br J Ind Med 44:521-526. McDonald AD, McDonald JC. Armstrong B. Cherry NM. Nolin AD. Robert D (1989): Father's occu pation and pregnancy outcome. Br J Ind Med 46:329-333. Modvig J. Schmidt L. Damsgaard MT (1990): Measurement of the total risk of spontaneous abortion: The virtue of conditional risk estimation. Am J Epidemiol 132:1021-1038. Morgan RW, Kheifcis L. Obrinsky DL, Whorton MD. Foliart DE (1984): Fetal loss and work in a waste water treatment plant. Am J Publ Health 74:499-501. Nordstrom S. Birke E. Gustavsson L (1983): Reproductive hazards among workers at high voltage substations. Bioclcctromagnciics 4:91-101. Olsen J (1984): Calculating risk ratios for spontaneous abortions: The problem of induced abortions. Int J Epidemiol 13:347-50. Oishan AF, Faustman E (1993): Male-mediated developmental toxicity. Ann Rev Public Health, 14: 159-181. Parazzini F, Bocciolone L, La Vecchia C, Negri E. Fedele L (1990): Maternal and paternal moderate daily alcohol consumption and unexplained miscarriages. Br J Obstet Gynaecol 97:618-622. Restrepo M, Munoz N, Day NE, Parra JE. de Romero L, Nguyen-Dinh X (1990): Prevalence of adverse reproductive outcomes in a population occupationally exposed to pesticides in Colombia. Scand J Work Environ Health 16:232-238. Roan CC. Matanoski GE. Mcllnay CQ, Olds KL. Pylant F, Trout JR, Wheeler P, Morgan DP (1984): Spontaneous abortions, stillbirths, and birth defects in families of agricultural pilots. Arch Env Health 39:56-60. Rosenberg PH, Vanttinen H (1978): Occupational hazards to reproduction and health in anaesthetists and paediatricians. Acta Anaesth Scand 22:202-207. Rothman KJ (1986): "Modern Epidemiology." Boston: Little, Brown and Company. Rupa DS, Reddy PP, Reddi OS (1991): Reproductive performance in population exposed to pesticides in cotton fields in India. Environ Res 55:123-128. Russell LB, Shelby MD (1985): Tests for heritable genetic damage and for evidence of gonadal exposure in mammals. Mutat Res 154:69-84. R&S 143226 Paternal Exposure and Spontaneous Abortion 381 Russell WL, Hunsicker PR (1988): Dominant mutagenic effect of ENU on fust-generation litter size in mice following treatment of spermatogonial stem cells. Environ Mol Mutag 1 l(Suppl 11):90. Sanotskii IV (1976): Aspects of the toxicology of chloroptene: Immediate and long-term effects. Environ Health Ferspect 17:83-93. Savitz DA, Chen i (1990): Parental occupation and childhood cancer Review of epidemiologic studies. Environ Health Perspect 88:323-37. Savitz DA. Harley B, Krekel S, Marshall J. Bondy J. Orleans M (1984): Survey of reproductive hazards among Oil, Chemical, and Atomic Workers exposed to halogenated hydrocarbons. Am J Ind Med 6:233-264. Selby PB. Russell WL (1983): First-generation litter size reduction following irradiation of spermatogo nial stem cells in mice and its use in risk estimation. Environ Mutag 7:431-69. Selevan SG (1985): Design of pregnancy outcome studies of industrial exposure. In Hemminki K. Sorsa M, Vainio H (ed): "Occupational Hazards and Reproduction." Washington, DC: Hemisphere Publishing Company. Siemiatycki J (1991): "Risk Factors for Cancer in the Workplace." Boca Raton, FL: CRC Press. Siemiatydd J, Wacholder S, Dewar R. Wald L, Bigin D. Richardson L, Rosenman K, Girin M (1988): Smoking and degree of occupational exposure: Are internal analyses in cohort studies likely to be confounded by smoking status? Am J Ind Med 13:59-69. Smith AH, Fisher DO. Pearce N, Chapman CJ (1982): Congenital defects and miscarriages among New Zealand 2,4,5-T sprayers. Arch Environ Health 37:197-200. Stachei B. Dougherty RC, Lahl U. Schlosser M. Zcschmar B (1989): Toxic environmental chemicals in human semen: Analytical method and case studies. Andrologia 21:282-91. Stellman SD, Stellman JM. Sommer JF (1988): Health and reproductive outcomes among American Legionnaires in relation to combat and herbicide exposure in Vietnam. Environ Res 47:150-174. Sterling TD, Arundel A (1986): Review of recent Vietnamese studies on the carcinogenic and teratogenic effects of phenoxy herbicide exposure. Int J Health Serv 16:265-278. Sterling TD, Weinkam JJ (1976): Smoking characteristics by type of employment. J Occup Med 18: 743-754. Suskind RR, Hertzberg VS (1984): Human health eflccts of 2.4,5-T and its toxic contaminants. JAMA 251:2372-2380. Tannenbaum TN, Goldberg RJ (1985): Exposure to anesthetic gases and reproductive outcome. J Occup Med 27:659-668. Taskinen H, Anttiia M Lindbohm M-L. Sallmcn M. Hemminki K (1989): Spontaneous abortions and congenital malformations among the wives of men occupationally exposed to organic solvents. Scand J Work Environ Health 15:345-352. Tomlin PJ (1979): Health problems of anaesthetists and their families in the West Midlands. Br Med J 1:779-784. Townsend JC. Bodner KM, Van Peenen PFD, Olson RD, Cook RR (1982): Survey of reproductive events of wives of employees exposed to chlorinated dioxins. Am J Epidemiol 115:695-713. U.S. Department of Health and Human Services. Centers for Disease Control (1989): "Health Status of Vietnam Veterans, Vol. V, Reproductive Outcomes and Child Health." Atlanta, Georgia. Whoiton D, Krauss RM, Marshall S (1977): Infertility in male pesticide workers. Lancet 2:1259-1261. Wilcox AJ, Homey LF (1984): Accuracy of spontaneous abortion recall. Am J Epidemiol 120:727-733. Wilcox AJ, Weinberg CR, O'Connor JF, Baitd DD. Schlatterer JP, Canfield RE, Armstrong EG, Nisula BC (1988): Incidence of early loss of pregnancy. N Engl J Med 319:189-194. Williamson DF, Sctdula MK. Kendrick JS. Binkin NJ (1989): Comparing the prevalence of smoking in pregnant and nonpregnant women. 1985 to 1986. JAMA 261:70-74. Wyrobek AJ, Gordon LA, Burkhart JG, Francis MW, Kapp RW, Jr., Letz G. Mailing HV. Topham JC. Whorton MD (1983a): An evaluation ot the mouse sperm morphology test and other sperm tests in nonhuman animals. A report of the U.S. Environmental Protection Agency Gcne-Tox Program. Mutat Res 115:1-72. Wyrobek AJ. Gordon LA, Burkhart JG. Francis MW. Kapp RW. Jr., LetzG. Mailing HV, Topham JC. Whorton MD (1983b): An evaluation of human sperm as indicators of chemically induced alter ations of spermatogenic function. A report of the U.S. Environmental Protection Agency GeneTox Program. Mutat Res 115:73-148. Yazigi RA, Odem RR. Polakoski KL (1991): Demonstration of specific binding of cocaine to human spermatozoa. JAMA 266:1956-1959. APPENDIX. Paternal Occtipaftonal Exposure and Spontaneous Abortion: Study Type, Confounders, Case Definitions Reference Cohen el al. [1974] Bouc et al. [1975] Cohen et al. [I975J K nilMopes et al. (19751 Jnfanteel aL [1976] Sanotskii [19761 Rosenberg & Vanttinen Tomlin |!979| Cohen et a!. |1980| Kharra/i cl al |I9A0| CarmcIM cl at. | 19ft11 Lauwerys et al. 11981] Beckman & Nordstrom (1982] Hamillet tl. [1982] Kline et al. [1982] Smith et al. 11982) Townsend et al. [1982] Canet al. [19831' Hemmlnki et al. (1983) Nordstrom el al. 11983) Lindbohm et al. |I984| Study location United States France United Slates United Kingdom United States Soviei Union Finland Uniicd Kingdom Uniled States Israel United Slates Belgium Sweden United Stales United States Study type Retrospeciivc cohort Cross-sec! tonal Retrospective cohort Retrospeciivc cohort Retrospective Retrospective cohort Retrospective cohort Retrospct'ine cohort Retrosp<M i\c cohort RclroKpciiite cohori Nested i .i't lonind Rcirospetmu Retrospects c cohort Retrospective cohort Case control New Zealand United Slates Vietnam Finland Sweden Finland Retrospeciivc cohort Retrospective cohort Retrospeciivc cohort Retrospective Retrospective cohort Retrospective cohort Confounders addressed Malernat age. smoking None Maternal age. smoking, pregnancy history Maternal age, smoking, birth order; palernal age Previous SAbh, palernal age NA1' Malernat nge Nutiu M^lcruul age. smoking, previous SAh M.itcrn.il age Maternal age. ciparcitc. marijuana smoke, paternal age. cdoealmn. occupation mantal stains, previous SAh Smoking. drug condompi ion Maternal age. smoking, occupation, gravidity, paternal age. alcohol Paternal alcohol, smoking, race, marital status Maternal age. previous SAb. alcohol, smoking, palernal unemployment, ethnicity, public/private care Maternal age, smoking, ethnicity Maternal age. birth control method, medical Ion during pregnancy, smoking, alcohol None Season, maternal occupation Maternal age, smoking, alcohol, medical ions, gravidity, chemical exposures Maternal age, piece of residence, parity, marital states Miscarriage case definition GA* <20 weeks GA <12 weeks and abnormal karyotype GA <20 weeks Mean GA * 1 i weeks (excludes slillbirths) SAbs and stillbirths NA Clinically recognized NA GA <20 weeks NA Esdudcs slillbirrhs, heavy tme periods Excludes stillbirths Excludes stillbirths Couples with hiMory of SAb GA <28 weeks with known karyotype Excludes stillbirths GA <20 weeks Escludes stillbirths Hospitalized SAbs GA <28 weeks Excludes stillbirths Hospilalized SAbs GA <28 weeks Baseline risk of spontaneous abortion ) 12.5 NA1 9.0 10.9 sn f NA 132 9K 67 ft ft 1 1 O'1 ft 1 74 14 5 NA ion 11.9 59 9.8 95 7.2' Morgan cl al [ 19X4 | k.unvUl |I9H4| S.A.f/ .-i .,1 | I9X4| Suskmd A; 1 ten/berg [I9M4] Jlrudsky a of | I9K5J Uamell A Vaughan | I9HK| United Slates United States United Stales United States United States Untied States Su llm.m ..I .it 11`JMKJ Cjh |rwi| United St.lies United States Mim'kmI | 19Xl)| WJienei.m \ Mniisnn JI9M9] MJ>un.ikK|.il |I9K9| i.i-kiik'ii el .d |19H9| CiLjirgms cr al ]1990| Rest repo el al. | I99t)| Cordier ct al. 1199 11 l mdhohm el al 1199 la] 1 injhohm el at 1199 Ih] Hup-ieial |1991 | United Stales United States Canada Finland OnUrm. Canada Colombia France Finland Finland India '(ic',t;ilion;il age. l'Spnnianec)us abortion. `Not available. ''Iru-lmles exposed population. 'filed in Sterling and Arundel (19861 'Cited m McDonald ct al. [I987j. Retrospective cohort Retrospective cohort Retrospective cohort Retrospective cohort Retrospective cohort Retrospective cohort Retrospective cohort Retrospective cohort RelrospLCike co hi <n Case komrol Rein ^|k.\.lise cohort Nested ease con l ml Retrospective cohort Retrospective cohort Retrospective cohort Retrospective" cohort Nested case control Retrospective cohort Maternal face. age. cdiiLaiioti. smoking. alu>hnl Maternal age, paternal age, education Maternal age, paternal age r.u.c. marital -slaiiis. None Maternal age, smoking Maternal age. martial siatus. employ inent status, census tract, median income, gravidity. previous SAbs, lime since last pregnancy, paternal age Maternal age, smoking, paternal age. combat exposure Paternal age, taic. enlistment status. education. military specialty, smoking akoluil, drug use. marital status Matornjl age. smoking, ah ohol ir^,- i-uiptoyment. previous ,S \b. cdiuaiion. palcru.il age Maternal age ran; rmplosmciil smoking parity, prior pregn mo losses p.ivuuni ">nr.e paternal age, rati.', education, lunhpl.uc Maternal age gr.mdriv pro urns S\h .-iluikity. education ' moVing. .Hiup.iinin k- liol Maternal age date n| (.oiucpiion Maternal age. smoking. hnlli oulci. pioious SAh None Maternal age, gravidity, smoking, alcohol, paternal age, length or recall Maternal age, socioeconomic status, occupational hazards Maternal age. smoking, febrile diseases, paternal age, smoking, previous SAb Paternal alcohol, smoking, drug use. nutritional status, toxic exposure* SAbs and stillbirths $Abs and stillbirths SAhs and stillbirth* Excludes stillbirths GA <20 weeks Prior losses noted on birth certificate Includes stillbirths Excludes stillbirths SAhs arnl stillbirths G-\ 2Huick\ UA ,'h uuk\ C tuusally rciogru/ed (liA * h -28 weeks) CiA <20 weeks GA <20 weeks GA <28 weeks Clinically recognized Clinically recognized GA <20 weeks K .1 ft 2 92 fI 9 8 .1 18.6 58 87 V .1 NA 104' KK 5.1 1.9 9.6 89 NA 11 4