SEARCH

SEARCH BY CITATION

Keywords:

  • paternal age;
  • birth defects;
  • prevalence;
  • maternal age;
  • selection bias

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

BACKGROUND: Unlike maternal age, the effect of paternal age on birth defect prevalence has not been well examined. We used cases from the Texas birth defect registry, born during 1996–2002, to evaluate the association of paternal age with the prevalence of selected structural birth defects. METHODS: Poisson regression was used to calculate prevalence ratios (PRs) and 95% confidence intervals (CIs) associated with paternal age for each birth defect, adjusting for maternal age, race/ethnicity, and parity. RESULTS: Relative to fathers ages 25–29 years, fathers 20–24 years of age were more likely to have offspring with gastroschisis (PR 1.47, 95% CI: 1.12–1.94), and fathers 40+ years old were less likely to have offspring with trisomy 13 (PR 0.40, 95% CI: 0.16–0.96). No association was seen between paternal age and prevalence of anencephaly and encephalocele. A selection bias was observed for the other birth defects in which cases of younger fathers were more often excluded from study. CONCLUSIONS: In studies of birth defect risk and paternal age, the source of information may affect the validity of findings. Birth Defects Research (Part A) 2007. © 2006 Wiley-Liss, Inc.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

Strong associations have been found between maternal age and the prevalence of many birth defects, such as Down syndrome (Cuckle, 2005; Malini and Ramachandra, 2006; Newberger, 2000). The effect of paternal age on birth defect rates has not been as well studied, and the few studies that have been done have yielded inconsistent results. Theoretically, as paternal age increases, birth defect rates should also increase, based on Penrose's “copy-error” theory (1955). Penrose hypothesized that there is a higher rate of mutations in the male germ line than in the female germ line, because male germ cells are continuously replicating through spermatogenesis and thus have a greater chance to accumulate mutations with each new division. The older men get, the more divisions their spermatozoa undergo (Risch et al., 1987). If an increase in mutations is due to an increase in germ cell replications, there should be a linear increase in mutations and in birth defect rates in offspring as paternal age increases. A few studies have found a higher percentage of sperm with damaged DNA (double-strand breaks) in older men (Morris et al., 2002; Singh et al., 2003; Thacker, 2004), suggesting that advanced paternal age may result in an increase in birth defects.

Autosomal dominant disorders that require a single-gene mutation for expression are most commonly thought to be related to paternal age (Crow, 2000; Stene and Stene, 1977). Indeed, the only disorders that clearly appear to increase with advanced paternal age are de novo cases of Apert syndrome and achondroplasia (Glaser et al., 2003; Hurst and Ellegren, 2002; Risch et al., 1987; Singer et al., 1999; Thacker, 2004), which are both single-gene, autosomal dominant mutations. It is possible that these patterns might also hold true for more complex defects (Crow, 2000).

Previous studies have suggested that some groups of birth defects may be associated with advanced paternal age. Examples include ventricular and atrial septal defects, neural tube defects, trisomies 13, 18, and 21, craniosynostosis, cleft palate, and cleft lip (Balgir, 1984; Erickson and Bjerkedal, 1981; Glaser et al., 2003; Hook et al., 1981; Kazaura et al., 2004a; McIntosh et al., 1995; Olshan et al., 1994; Singer et al., 1999). An increased risk of gastroschisis and pyloric stenosis has been associated with younger maternal age in several studies (Baird et al., 1991; Haddow et al., 1993; Reefhuis and Honein, 2004; Schechter et al., 1997), but the paternal age effect for these birth defects has not been sufficiently investigated. In this study we examined the association between paternal age and selected structural birth defects in offspring.

METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

Data Collection

Birth defect data were obtained from the Texas Birth Defects Registry (TBDR) at the Texas Department of State Health Services (DSHS). Cases with the following types of birth defects were analyzed: ventricular septal defects, atrial septal defects, trisomies 13, 18, and 21, neural tube defects, pyloric stenosis, gastroschisis, craniosynostosis, cleft palate alone, and cleft lip (with or without cleft palate). Cases with more than one type of birth defect listed in the TBDR were counted in each relevant category. All children were delivered between January 1, 1996 and December 31, 2002. The TBDR uses active surveillance; its staff regularly visits health care facilities where affected children are delivered or treated, including hospitals, birthing centers, and midwives. To be included as a case in the TBDR, an infant or fetus must have a reported structural birth defect or developmental disability that is diagnosed either prenatally or within 1 year after delivery, and must have a maternal residence in an area covered by the registry at the time of delivery. During the years 1999–2002, affected births occurring to all mothers residing in Texas at time of delivery were included. During 1996–1998, only selected geographic regions (known as “public health regions”) were included in the surveillance system.

Date of case delivery was taken from the vital record (birth or fetal death certificate). If the delivery date was missing, then it was taken from the medical record. Date of father's birth was abstracted from the vital record and was subtracted from infant delivery date to calculate paternal age at the date of delivery. Denominator data consisted of all live births in Texas from 1996 to 2002, determined from birth certificate data with and without birth defects corresponding to the same time frame and in the same areas as the birth defect cases.

Analysis

Prevalence estimates were calculated for the selected birth defects for the following paternal age groups: less than 20, 20–24, 25–29, 30–34, 35–39, and 40 or over. Age-specific birth prevalence was calculated as the number of cases with a specific defect per paternal age group divided by the number of live births per age group (per 10,000). Using the paternal age group 25–29 years as a referent, prevalence ratios were calculated using Poisson regression to determine whether or not the birth defects were associated with paternal age. The 25 to 29-year-old paternal age group was chosen as the reference age category because it had the greatest number of fathers, and the choice of an age group in the middle of the possible paternal ages made it easier to interpret prevalence ratios for the very young as well as older paternal age groups.

Maternal age, parity, and maternal and paternal race/ethnicity were all examined as confounding factors in the Poisson regression analyses. We adjusted for maternal age as a continuous polynomial factor to minimize the possibility of residual confounding within broader categories of maternal age. We included a squared term because many of the associations between maternal age and birth defect prevalence appeared to be J-shaped or curved rather than linear. Because paternal age is highly correlated to maternal age, we adjusted for maternal age in every analysis. Maternal parity and maternal and paternal race/ethnicity, however, were removed from the Poisson regression analysis via backwards elimination if they were not significant (P ≤ .05), to maximize statistical precision (Hair et al., 1998). Trend tests were also performed for birth defects whose prevalences were associated with paternal age at a significance level of .10 or less, to see if a linear association existed. This was done in the Poisson regression by treating paternal age as a continuous variable instead of a categorical one. All analyses were conducted using SAS version 9 (SAS Institute, 2004).

A portion of the TBDR cases was missing paternal age information. To examine potential selection bias due to missing paternal age information, we calculated maternal age prevalence ratios for those birth defect cases with known paternal age and those with missing paternal age. The relation of maternal age with birth defect prevalence was compared between these two groups to ascertain the existence of selection bias due to missing values.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

Overall, higher crude prevalence estimates for the selected congenital malformations tended to occur among either the youngest or the oldest paternal age groups (Table 1). After adjustment for maternal age and other significant factors, older men seemed less likely to have offspring with pyloric stenosis; prevalence ratios for fathers 30 years of age and older ranged from 0.80 to 0.85. For trisomy 13, the oldest paternal age group also had a low prevalence ratio of 0.40, while the youngest paternal age group had a prevalence ratio of 2.16 (Table 1). The younger 20–24 paternal age group had significantly higher adjusted prevalence ratio values than other paternal ages for trisomy 21, cleft lip, and gastroschisis. Relative to fathers 25 to 29 years of age, fathers 20 to 24 years old were approximately 1.3 times more likely to have offspring with trisomy 21, 1.2 times more likely to have children with cleft lip, and 1.5 times more likely to have children with gastroschisis (Table 1). No significant paternal age effect was seen for any of the other birth defects.

Table 1. Number of Cases, Crude Prevalence Estimates, and Adjusted Prevalence Ratios for Paternal Age Groups Relative to the Reference Age Group of 25–29 Years for Selected Birth Defectsa
Birth defectPaternal age groupNumber of casesCrude prevalence per 10,000 births (95% confidence limits)Adjusted prevalence ratio (95% confidencelimits)
  • a

    Prevalence estimates and prevalence ratios are based on Texas births from 1996–2002. Prevalence estimates are per 10,000 live births. All adjusted prevalence ratios and 95% confidence limits for each paternal age group arc adjusted for maternal age plus other significant factors (listed below).

  • b

    Adjusted for maternal age and parity.

  • c

    Adjusted for maternal age, paternal race/ethnicity, and parity.

  • d

    Adjusted for maternal age and maternal race/ethnicity.

  • e

    Adjusted for maternal age, paternal race/ethnicity, and maternal race/ethnicity.

  • f

    Adjusted for maternal age and paternal race/ethnicity.

  • g

    Adjusted for maternal age, maternal race/ethnicity, and parity.

  • h

    Adjusted for maternal age, paternal race/ethnicity, maternal race/ethnicity, and parity.

Anencephalyb<20151.44 (0.81–2.38)0.54 (0.27–1.03)
20–24561.50 (1.13–1.95)0.67 (0.45–0.98)
25–29821.73 (1.38–2.15)1.00 (Reference)
30–34551.32 (1.00–1.72)0.94 (0.65–1.36)
35–39240.99 (0.64–1.47)0.75 (0.46–1.67)
40+161.21 (0.69–1.96)0.90 (0.46–1.67)
Spina bifidac<20282.69 (1.79–3.89)0.69 (0.38–1.20)
20–241193.19 (2.61–3.76)0.87 (0.63–1.19)
25–291633.45 (2.92–3.98)1.00 (Reference)
30–341323.18 (2.63–3.72)0.94 (0.70–1.26)
35–39783.22 (2.55–4.02)0.93 (0.64–1.34)
40+443.32 (2.41–4.45)0.88 (0.53–1.39)
Encephaloceled<2080.77 (0.33–1.51)0.60 (0.24–1.41)
20–24240.64 (0.41–0.96)0.69 (0.40–1.20)
25–29330.70 (0.48–0.98)1.00 (Reference)
30–34260.63 (0.41–0.92)1.04 (0.62–1.73)
35–39170.70 (0.41–1.12)1.19 (0.62–2.22)
40+100.75 (0.36–1.39)1.27 (0.55–2.71)
Ventricular septal defecte<2039137.58 (33.85–41.30)0.97 (0.86–1.10)
20–24136236.48 (34.54–38.42)0.97 (0.91–1.05)
25–29178137.68 (35.93–39.43)1.00 (Reference)
30–34159138.29 (36.41–40.17)0.97 (0.90–1.03)
35–39100141.34 (38.78–43.90)0.94 (0.87–1.03)
40+66450.03 (46.23–53.84)1.02 (0.92–1.13)
Atrial septal defectsf<2037736.23 (32.58–39.89)1.04 (0.90–1.20)
20–4131035.09 (33.19–36.99)1.03 (0.95–1.24)
25–29165234.95 (33.27–36.64)1.00 (Reference)
30–34145334.97 (33.17–36.76)0.92 (0.85–1.00)
35–3995639.48 (36.98–41.98)0.92 (0.83–1.01)
40+65549.35 (45.57–53.13)1.01 (0.89–1.13)
Cleft palate alone (without cleft lip)d<20504.81 (3.57–6.34)0.87 (0.62–1.22)
20–241814.85 (4.14–5.55)0.91 (0.75–1.11)
25–292505.29 (4.63–5.94)1.00 (Reference)
30–342215.32 (4.62–6.02)0.98 (0.82–1.17)
35–391265.20 (4.29–6.11)0.92 (0.73–1.15)
40+926.93 (5.59–8.50)1.23 (0.94–1.61)
Cleft lip (with or without cleft palate)g<2010410.00 (8.07–11.92)1.13 (0.86–1.47)
20–2440110.74 (9.69–11.79)1.18 (1.00–1.38)
25–294429.35 (8.48–10.22)1.00 (Reference)
30–343518.45 (7.56–9.33)0.88 (0.75–1.02)
35–392359.70 (8.46–10.95)0.97 (0.80–1.17)
40+1249.34 (7.70–10.99)0.91 (0.71–1.16)
Pyloric stenosise<2023622.68 (19.79–25.58)1.12 (0.96–1.30)
20–2475620.25 (18.81–21.69)1.03 (0.94–1.13)
25–2988818.79 (17.55–20.02)1.00 (Reference)
30–3459214.25 (13.10–15.39)0.80 (0.73–0.88)
35–3935314.58 (13.06–16.10)0.85 (0.76–0.96)
40+18213.71 (11.72–15.71)0.84 (0.72–0.98)
Craniosynostosisc<20222.11 (1.33–3.20)0.71 (0.44–1.13)
20–24912.44 (1.96–2.99)0.78 (0.60–1.02)
25–291613.41 (2.88–3.93)1.00 (Reference)
30–341684.04 (3.43–4.65)1.06 (0.85–1.30)
35–391235.08 (4.18–5.98)1.20 (0.94–1.54)
40+715.35 (4.18–6.75)1.19 (0.87–1.62)
Gastroschisisg<2011310.86 (8.86–12.86)1.42 (0.98–2.05)
20–242446.54 (5.72–7.36)1.47 (1.12–1.94)
25–291122.37 (1.93–2.81)1.00 (Reference)
30–34571.37 (1.04–1.78)0.98 (0.67–1.42)
35–39190.78 (0.47–1.23)0.84 (0.45–1.47)
40+50.38 (0.22–0.88)0.49 (0.15–1.24)
Trisomy 21 (Down syndrome)h<20656.25 (4.82–7.96)1.25 (0.93–1.65)
20–242546.80 (5.97–7.64)1.28 (1.08–1.51)
25–293186.73 (5.99–7.47)1.00 (Reference)
30–3443210.40 (9.42–11.38)1.03 (0.90–1.18)
35–3938615.94 (14.35–17.53)0.94 (0.81–1.10)
40+41531.27 (28.26–34.28)1.05 (0.88–1.24)
Trisomy 13 (Patau syndrome)b<20111.06 (0.53–1.89)2.16 (0.78–5.67)
20–24220.59 (0.37–0.89)1.14 (0.56–2.26)
25–29280.59 (0.39–0.86)1.00 (Reference)
30–34370.89 (0.63–1.23)1.07 (0.60–1.92)
35–39331.36 (0.94–1.91)0.99 (0.52–1.93)
40+120.90 (0.47–1.58)0.40 (0.16–0.96)
Trisomy 18 (Edwards syndrome)b<2090.86 (0.40–1.64)0.96 (0.45–1.93)
20–24320.86 (0.59–1.21)0.91 (0.59–1.39)
25–29531.12 (0.84–1.47)1.00 (Reference)
30–34561.35 (1.02–1.75)0.80 (0.57–1.15)
35–39642.64 (2.04–3.37)0.91 (0.62–1.32)
40+644.82 (3.71–6.16)0.87 (0.57–1.33)

Tests for trend showed that younger paternal age was associated with higher prevalence for pyloric stenosis (P = .002), cleft lip (P = .043), gastroschisis (P = .004), and trisomy 13 (P = .068). Even though the prevalence ratios were highest for the two youngest age groups, no linear trend was noted between young paternal age and increased prevalence of trisomy 21.

The number of cases available for study ranged from 192 (encephalocele) to 8,113 (ventricular septal defect) (Table 2). Even though the number of cases was sufficiently large to be able to generate valid prevalence estimates for most birth defects, a large percentage of cases in the TBDR were missing paternal age data. Overall, 18.6% of the TBDR cases were missing paternal age information, and the percentage of records missing paternal age ranged from 11.2% (craniosynostosis) to 55.0% (anencephaly).

Table 2. Total Number of Specific Birth Defect Cases, Number and Percentage of Cases Missing Paternal Age, and Total Number Used in the Analysisa
Birth defectTotal (n)Missing paternal age (%)Total used in analysis (n)
  • a

    Texas births, 1996–2002.

Anencephaly551303 (55.0)248
Spina bifida753189 (25.1)564
Encephalocele19274 (38.5)118
Ventricular septal defect81131323 (16.3)6790
Atrial septal defect76731270 (16.6)6403
Cleft palate alone (without cleft lip)1118198 (17.7)920
Cleft lip (with or without cleft palate)2063406 (19.7)1657
Pyloric stenosis3597590 (16.4)3007
Craniosynostosis71680 (11.2)636
Gastroschisis746196 (26.3)550
Trisomy 21 (Down syndrome)2335465 (19.9)1870
Trisomy 13 (Patau syndrome)22582 (36.4)143
Trisomy 18 (Edwards syndrome)416138 (33.2)278

Table 3 shows maternal age prevalence ratios stratified by whether paternal age was known or was missing. The results show that the maternal age relation with prevalence differed between the two groups. Although a few birth defects (anencephaly, encephalocele, gastroschisis, and trisomy 13) exhibited the same general maternal age trend whether or not paternal age was known, differences in maternal age trends were seen for the majority of birth defects, including spina bifida, ventricular septal defect, atrial septal defect, cleft palate, cleft lip, pyloric stenosis, craniosynostosis, trisomy 21, and trisomy 18.

Table 3. Adjusted Prevalence Ratios for Maternal Age Groups Relative to the Reference Age Group of 25–29 Years for Selected Birth Defectsa
Birth defectMaternal age groupNumber of cases, paternal age knownPrevalence ratio (95% confidence limits), paternal age knownNumber of cases, paternal age missingPrevalence ratio (95% confidence limits), paternal age missing
  • a

    Separate prevalence ratios were calculated for cases with paternal age known and missing. Prevalence ratios are based on Texas births from 1996 to 2002. The prevalence ratios and 95% confidence limits for each paternal age group are shown. Prevalence ratios for all birth defects were adjusted for maternal race/ethnicity and parity.

Anencephaly<20421.40 (0.98–1.97)381.64 (1.09–2.45)
20–24761.12 (0.84–1.48)521.00 (0.71–1.43)
25–29731.00 (Reference)541.00 (Reference)
30–34450.78 (0.56–1.08)330.80 (0.54–1.19)
35–3980.28 (0.14–0.50)231.17 (0.73–1.82)
40+40.60 (0.22–1.31)40.95 (0.32–2.17)
Spina bifida<20710.83 (0.60–1.14)382.18 (1.40–3.42)
20–241450.86 (0.67–1.10)561.69 (1.15–2.53)
25–291641.00 (Reference)311.00 (Reference)
30–341221.03 (0.79–1.33)180.82 (0.48–1.36)
35–39480.90 (0.62–1.28)141.42 (0.79–2.47)
40+141.33 (0.69–2.32)73.54 (1.59–7.08)
Encephalocele<20161.22 (0.44–3.14)111.58 (0.82–3.05)
20–24411.54 (0.75–3.31)171.21 (0.69–2.14)
25–29271.00 (Reference)131.00 (Reference)
30–34180.89 (0.35–2.17)80.88 (0.43–1.70)
35–39141.56 (0.55–4.04)31.49 (0.68–3.07)
40+21.11 (0.05–6.13)11.31 (0.17–4.80)
Ventricular septal defect<208100.78 (0.71–0.85)3602.41 (1.93–3.01)
20–2416970.89 (0.83–0.96)4231.52 (1.23–1.87)
25–2917731.00 (Reference)2551.00 (Reference)
30–3414771.18 (1.10–1.27)1350.76 (0.57–0.99)
35–397801.44 (1.31–1.58)961.19 (0.87–1.62)
40+2532.36 (2.04–2.73)392.34 (1.45–3.60)
Atrial septal defect<207870.81 (0.73–0.90)3753.24 (2.64–3.99)
20–2415390.86 (0.79–0.94)4241.88 (1.55–2.28)
25–2916901.00 (Reference)2111.00 (Reference)
30–3413751.13 (1.04–1.24)1340.88 (0.68–1.13)
35–397681.45 (1.30–1.60)901.28 (0.95–1.70)
40+2442.32 (1.96–2.73)261.73 (1.04–2.72)
Cleft palate alone (without cleft lip)<201190.88 (0.72–1.07)602.73 (1.73–4.37)
20–242190.83 (0.71–0.98)561.36 (0.88–2.13)
25–292651.00 (Reference)381.00 (Reference)
30–341970.99 (0.84–1.17)150.56 (0.28–1.02)
35–39971.10 (0.88–1.35)171.42 (0.75–2.57)
40+231.33 (0.89–1.92)10.42 (0.02–2.11)
Cleft lip (with or without cleft palate)<202050.85 (0.73–0.99)1093.41 (2.51–4.66)
20–244741.00 (0.89–1.12)1201.78 (1.34–2.40)
25–294731.00 (Reference)671.00 (Reference)
30–343100.88 (0.77–1.00)410.82 (0.56–1.18)
35–391580.99 (0.84–1.17)361.50 (1.00–2.21)
40+371.14 (0.84–1.53)101.88 (0.93–3.45)
Pyloric stenosis<204791.04 (0.92–1.18)2164.09 (2.99–5.66)
20–248761.02 (0.92–1.12)2022.04 (1.51–2.79)
25–298381.00 (Reference)931.00 (Reference)
30–345500.90 (0.80–1.00)440.66 (0.42–1.02)
35–392220.82 (0.70–0.96)290.98 (0.57–1.61)
40+420.80 (0.56–1.09)30.51 (0.08–1.64)
Craniosynostosis<20480.48 (0.37–0.63)202.89 (1.01–8.92)
20–241300.71 (0.58–0.86)282.08 (0.82–5.89)
25–291771.00 (Reference)131.00 (Reference)
30–341711.32 (1.10–1.58)121.28 (0.39–4.11)
35–39861.54 (1.23–1.92)51.22 (0.21–5.01)
40+242.27 (1.54–3.22)11.29 (0.01–11.63)
Gastroschisis<202275.70 (4.06–8.16)9010.05 (6.36–16.61)
20–242283.33 (2.40–4.70)603.56 (2.26–5.85)
25–29631.00 (Reference)161.00 (Reference)
30–34190.43 (0.22–0.77)100.88 (0.44–1.71)
35–39110.60 (0.26–1.21)20.41 (0.09–1.23)
40+20.59 (0.06–2.27)11.11 (0.11–4.51)
Trisomy 21 (Down syndrome)<201340.80 (0.65–0.97)682.50 (1.60–3.93)
20–242850.90 (0.77–1.05)861.69 (1.12–2.60)
25–293021.00 (Reference)471.00 (Reference)
30–343891.79 (1.55–2.07)491.49 (0.93–2.40)
35–394534.69 (4.08–5.41)1037.04 (4.71–10.74)
40+30715.74 (13.44–18.42)6722.20 (14.17–35.13)
Trisomy 13 (Patau syndrome)<20181.41 (0.80–2.45)91.21 (0.46–3.03)
20–24240.92 (0.55–1.52)181.30 (0.61–2.86)
25–29261.00 (Reference)131.00 (Reference)
30–34371.90 (1.21–3.01)80.86 (0.32–2.15)
35–39262.88 (1.75–4.75)204.89 (2.33–10.68)
40+126.43 (3.33–11.89)56.20 (1.81–17.74)
Trisomy 18 (Edwards syndrome)<20220.94 (0.48–1.76)91.27 (0.46–3.36)
20–24360.76 (0.44–1.31)151.10 (0.47–2.62)
25–29471.00 (Reference)131.00 (Reference)
30–34531.50 (0.92–2.46)141.49 (0.63–3.58)
35–39714.37 (2.76–7.03)368.56 (4.27–18.59)
40+4914.71 (8.78–24.66)2529.70 (13.92–67.13)

Given this situation, the birth defects with dissimilar maternal age trends differed almost solely in the prevalence ratios between the two youngest maternal age groups. For these age groups, birth defect cases where paternal age was missing tended to have significantly higher prevalence ratios than did cases where paternal age was known. For example, ventricular septal defect cases who were missing paternal age information had prevalence ratios of 2.41 (95% CI: 1.93–3.01) and 1.52 (95% CI: 1.23–1.87) for maternal ages <20 and 20–24, respectively. For ventricular septal defect cases where paternal age was known, the prevalence ratios for these same two maternal age groups were 0.78 (95% CI: 0.71–0.85) and 0.89 (95% CI: 0.83–0.96) (Table 3). Comparing the number of cases and prevalence ratios for each of the maternal age groups where paternal age was known versus unknown, it appears that a higher percentage of younger-aged mothers (25 years of age or less) were excluded from the paternal age regression analyses than were mothers of other ages.

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

The relatively large number of cases for most of the different birth defect types was one of the strengths of this study, as it should have yielded adequate statistical power to detect a difference in prevalence ratios. Also, the use of Poisson regression allowed for adjustment of multiple potential confounders and for consideration of maternal age as a polynomial function. Adjusting for maternal age as a continuous polynomial factor helped to minimize residual confounding and made it possible to model J-shaped or curved as well as linear associations. Even though we modeled maternal age as a continuous polynomial factor in this analysis, we also looked at the results of adjusting for maternal age in other ways (as a linear factor as well as a categorical factor). We noticed that although the same general trends were seen, the magnitude of the association between young paternal age and birth defect prevalence varied depending upon the way maternal age was modeled. With a linear maternal age model, young paternal age was found to be more strongly associated with more birth defects, whereas with a categorical maternal age model, young paternal age was less strongly associated with birth defects. Prevalence ratio estimates were in the middle of these two extremes when maternal age was modeled as a continuous polynomial factor. Paternal age studies that vary in their modeling of maternal age may arrive at different results and conclusions; some consensus should probably be reached on the best method for maternal age adjustment in these types of analyses.

The main weakness of this study was the relatively large percentage of cases that were excluded from the analyses due to missing paternal age. The percentage of missing birth defect cases in this study was higher than for most studies thus far (Kazaura et al., 2004a; Lian et al., 1986; McIntosh et al., 1995; Olshan et al., 1994). For some birth defects, the percentage of records missing paternal age was extremely high, up to 55% (for anencephaly). The birth defects with the highest percentage of records missing paternal age data tended to be more severe disorders that often result in fetal death, such as neural tube defects and trisomies 13 and 18. This may reflect a greater difficulty in linking birth defects registry records with fetal death records than with birth certificates. It is also likely that paternal information is less complete on fetal death records than on birth certificates.

The analysis of maternal age effects showed that only four birth defects—anencephaly, encephalocele, gastroschisis, and trisomy 13—had similar maternal age trends regardless of whether or not paternal age was known. Because differences in maternal age trends were observed for all the rest of the birth defect types (spina bifida, ventricular septal defect, atrial septal defect, cleft palate, cleft lip, pyloric stenosis, craniosynostosis, trisomy 21, and trisomy 18), it is likely that for these latter defects, there is a large potential for selection bias and results may not be generalizable to the cases missing paternal age information.

Because more young mothers were missing paternal age information and thus were more often excluded from paternal age analyses than were older mothers, and because maternal and paternal age are highly correlated, this would suggest that more young case fathers (<25) were also excluded than were fathers of other ages. It has also been reported that fathers with no age listed on their children's birth certificates are much younger than fathers for whom age is reported (Landry and Forrest, 1995). This could have led to an underascertainment of risk for younger fathers, in the same way that many of the maternal age prevalence ratios for cases where paternal age was known seem to have underascertained the risk for younger mothers. On the other hand, because selection bias can be in either direction, it is also possible that the large amount of selection bias could have led to an overascertainment of risk for younger fathers.

To the best of our knowledge, no studies thus far have examined selection bias for birth defect cases in this way, that is, by comparing differences in maternal age effect among cases in which paternal age was available and those in which it was missing. This is most likely because other studies either had much lower percentages of cases missing paternal age information than did ours (Dzurova and Pikhart, 2005; Erickson and Bjerkedal, 1981; McIntosh et al., 1995; Olshan et al., 1994; Roecker and Huether, 1983), only mentioned the percentage of records missing paternal age for the whole birth registry instead of for the birth defect cases (Kazaura et al., 2004a; Polednak, 1976), or were case-control studies that gave no information about the number of cases missing paternal age (Balgir, 1984; Singer et al., 1999; Zhan et al., 1991). Unlike our study, which used both live births and fetal deaths, most studies providing information about missing paternal age used only live births in their analyses (Dzurova and Pikhart, 2005; McIntosh et al., 1995; Olshan et al., 1994; Roecker and Huether, 1983). Because fetal death records are generally missing more paternal information than are live birth records, the use of fetal deaths probably contributed to the high percentages of missing paternal age information. Studies based on data from the Medical Birth Registry of Norway were the only other studies that included both live births and fetal deaths (Erickson and Bjerkedal, 1981; Kazaura et al., 2004a, b). Kazaura et al. (2004b) found that 20% of gastroschisis cases in Norway were missing paternal age information, which is comparable to that found in our study (26% for gastroschisis).

Studies that include only live births will have more complete paternal information and selection bias will seemingly be minimized. However, it is unclear how the omission or exclusion of fetal deaths or the incomplete ascertainment of cases would bias study findings. It is interesting to note that in our study the birth defects with similar maternal age prevalence ratios, whether or not paternal age was known, were not the birth defects with the lowest percentages of missing paternal age information. In fact, three of the four defects with similar maternal age effects were severe disorders, and all four birth defects had high percentages of cases with unknown paternal age (the lowest was gastroschisis, with 26% of cases missing father's age). This seems to suggest that selection bias is not necessarily tied to the percent of cases missing age information, and could mean that even though only a small percentage of cases may be missing paternal information in a study, significant selection bias might exist nonetheless. These findings may be relevant to future studies using birth defect registry data that include fetal deaths.

Because anencephaly, encephalocele, gastroschisis, and trisomy 13 were the only birth defects in this study that seemed unaffected by selection bias, these four defects are the only ones with results we can interpret with some validity. Our results showed that younger paternal age groups (less than 25 years of age) were associated with a higher risk of gastroschisis and trisomy 13 than were more advanced paternal ages. The comparable study by Kazaura et al. (2004b) found that younger fathers were at higher risk of having a child with gastroschisis, after adjusting for maternal age. As with most trisomy disorders, the prevalence of trisomy 13 is known to significantly increase with increasing maternal age (Hassold and Jacobs, 1984; Jacobs and Hassold, 1995; Risch et al., 1986). Therefore, the decreased prevalence of trisomy 13 observed with advanced paternal age in this study was unexpected. The results observed for trisomy 13 in this study should be interpreted with caution because of the small sample sizes for some of the age categories. Even though the association between paternal age and trisomy disorders has been examined in other studies, no study has investigated paternal age and risk of trisomy 13 alone.

In our study, no significant paternal age association was found for anencephaly or encephalocele. Strassburg et al. (1983) also found no significant paternal age effect for anencephaly after controlling for maternal age. However, some studies have reported that very young fathers (less than 20 years of age) had an increased risk of having children with neural tube defects (Kazaura et al., 2004a; McIntosh et al., 1995).

In conclusion, we were unable to confirm the relation between paternal age and risk for the majority of birth defects, including spina bifida, ventricular septal defect, atrial septal defect, cleft palate, pyloric stenosis, craniosynostosis, trisomy 21, and trisomy 18, due the observed bias in which cases of young paternal age were likely selectively excluded. We recommend that future studies carefully examine the source of paternal information and potential biases related to missing information.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES