Is the observed association between increasing paternal age and delayed conception an artefact?

Markku Sallmén1,3 and Ritva Luukkonen2

1 Division of Epidemiology and Social Sciences and 2 Division of Biostatistics, Department of Epidemiology and Biostatistics, Finnish Institute of Occupational Health, Helsinki, Finland

Dear Sir,

The assessment of factors influencing human fertility is an important field of research (Carlsen et al.1992Go; Bonde, 1999Go). The effects of age, in particular, are of interest in assessing the prognosis for infertile couples. Also, age should be adequately taken into account in any studies aiming to find out the risk factors for reduced fertility. We have read with interest the recent paper by Ford and co-workers (2000) in this journal (Ford et al.2000Go). They conducted a cross-sectional study with retrospective data on time to conception (TTC), and observed a strong association between increasing paternal age and delayed conception. In that study, all the mothers and their partners living in the area administered by the Avon Health Authority and with expected delivery date within a 21 month calendar period were invited to enrol in the study. Information on TTC and on related factors were collected by self-administered questionnaires from each mother and her partner at 18 weeks gestation. Time-to-conception data were divided into four bands, <=6 months, >6 but <=12 months, >12 months but <=3 years and >3 years. Maternal and paternal ages were defined for the time the pregnancy was conceived. We consider that this definition of age will cause a positive bias for increasing age of both parents because couples have just aged during their time to conception according to the length of TTC.

We have conducted altogether 11 simulations to find out the biasing potential of `defining age at the time of conception' under the following assumptions: (i) Age has no effect on fertility; (ii) Only couples who achieved a pregnancy within 10 years are included; (iii) Fecundabilities vary among the couples according to beta distribution (mean fecundability 0.28, SD 0.18) to get a TTC distribution close to that observed in the study by Ford et al., 2000. In that study, 74, 14, 8.5 and 3.5% of the couples achieved a pregnancy in the consecutive time-to-conception bands; (iv) We defined father's age at the onset of TTC so that the age distribution at the time of conception mimics the age distribution in the study by Ford et al., 2000; (v) To define the age at the time of conception, we added the duration of TTC in months and subtracted one month because we supposed that TTC = 1 means that the pregnancy begun during the first menstrual cycle. For the men we additionally added 3 months (estimated time of spermatogenesis) for each TTC. (vi) Women who conceived after 49 years of age were excluded; (vii) To assess the impact of the findings related to age differences between the spouses, on average we selected a female partner 3 years younger than the male partner. We assumed that the age difference between the man and his partner are normally distributed (SE 1.9, 1.9, 2.0, 2.0, and 2.2 in consecutive age groups respectively). The entire age distribution of both parents is available from the first author upon request; (viii) We also assumed that maternal age has no impact on fertility because female age was adjusted in the original paper. We used logistic regression to define the odds ratios of paternal age at the time of conception and of age difference for achieving conception within 6 or 12 months.

Increasing paternal age at the time of conception was strongly related to prolonged TTC in our simulation data (Table IGo). The reason for this finding is that some of the men in age groups <=24 up to 39 years will impregnate their wives only after reaching the lower limit of the next age group. The proportion of men sharing this characteristic will depend on the length of TTC (i.e. on the fecundity or the fertility potential of the couple). Therefore, the association of age and fertility was stronger among those achieving conception after 12 months than among those who achieved conception after 6 months. Also, the observed association was strongly related to the relative size of the consecutive age groups. Therefore, the odds ratios (OR) were very low in the relatively small age groups 35–39 and >=40 years. In addition, no men in the oldest age group can progress to an older age group during the time to conception. By contrast, the reference age group (<=24 years) will lose predominantly subfertile couples during their TTC. In terms of selection, there is a strong selection towards subfertile couples occurring in the oldest age groups if age is defined at the time of conception. All in all, the associations found in our simulations were stronger than those observed in the real data. Our findings suggest that within the age range studied, increasing paternal age was not related to reduced fertility in the study by Ford et al., 2000.


View this table:
[in this window]
[in a new window]
 
Table I. The association of paternal age on the odds ratios (OR) for achieving conception within 6 or 12 months in fictitious populations under the assumption that paternal age has no effect on fertility at the studied age interval (11 simulations)
 
The findings of age differences between the parents may well be explained by the same age definition error. In our simulation populations, the index families `women whose partners were >=5 years older than themselves' had a median OR of 0.80 (range 0.72, 0.88) and 0.67 (range 0.60, 0.82) at the time of conception for conceptions within 6 and 12 months respectively, as compared with `partner at most 2 years older'. The obvious reason for this finding is that only a few men in the index group can be <=24 years of age, which seemingly are the most fertile men.

There are some earlier population-based studies focusing on the effects of parental age on fecundability (Spira et al., 1985Go; Olsen, 1990Go) or in which the crude association of age and fertility can be observed (Curtis et al., 1997Go). In two studies, parental age was defined for the time of onset of TTC (Spira et al., 1985Go; Curtis et al., 1997Go). No decrease in fertility with increasing paternal age was observed in these studies. By contrast, men in their 30's seemed to be more fertile than the men in their 20's. These findings seem to contradict the observations by Ford et al. 2000. This apparent discrepancy may well be explained by the difference in age definition between studies. In one study, the age of the parents was defined at the time of pregnancy (Olsen, 1990Go). In that study, a slight tendency towards reduced fertility in older men and a very strong effect among the older women was observed. The OR of `delayed conception >1 year' ranged from 2–20 in the 5 year age groups as compared with reference group (15–19 years) after adjustment of male age. At least a part of the finding may be explained by the definition of age in that study as well.

The findings of two recent studies have shown increasing fecundity with increasing age (Jensen et al.2000Go; ,Juul et al.2000Go). The reasons for these unexpected findings are for example, the tendency of older couples to give up more easily trying to conceive a child than the younger ones, higher success rates of younger couples in the medical treatment of infertility, and possible age related differences in the use of effective contraceptive methods (Jensen et al.2000Go). Based on Monte Carlo simulations, the other study (Juul et al.2000Go) showed that the inverse trend in fecundability may be observed if the individual fecundity is assumed to decrease with increasing age.

In general, it is almost impossible to study the effect of age on fecundability retrospectively (Olsen et al., 1998Go; ,Jensen et al.2000Go). Moreover, if TTC is categorized already in the questionnaire, the definition of age at the onset of TTC is difficult for the subfertile couples in particular. Occasionally age distribution is truncated at the time of pregnancy. Then there will be an excess of short TTCs of older parents (right truncation) and long TTCs of younger parents (left truncation) if age is defined for the beginning of TTC. This type of data characteristic will further complicate the problem (Sallmén, 2000Go).

The method of measuring time to pregnancy is vulnerable to biases (Weinberg et al., 1994Go). Analytic thinking through causal diagrams (Greenland et al., 1999Go) may be helpful in discriminating between apparent and causal effects in studies of fertility (Sallmén, 2000Go). The findings of our simulations suggest that the observed decreasing fertility with increasing male age (Ford et al., 2000Go) may well be explained by erroneous definition of age. No conclusion on the effect of age on fertility should be made on the basis of studies defining age at the time of conception. Thus far, there is no evidence for reduced fertility with increasing paternal age at least within the age range 20–40 years.

Notes

3 To whom correspondence should be addressed at: Department of Epidemiology and Biostatistics, Finnish Institute of Occupational Health, FIN-00250, Helsinki, Finland E-mail: Markku.Sallmen{at}occuphealth.fi Back

References

Bonde, J.P. (1999) Environmental fertility research at the turn of the century. Scand. J. Work Environ. Health, 25 (special issue), 529–536.[ISI][Medline]

Carlsen, E., Giwercman, A., Keiding, N. and Skaggebæk, N.E. (1992) Evidence for decreasing quality of semen during past 50 years. BMJ., 305, 609–613.[ISI][Medline]

Curtis, K.M., Savitz, D.A. and Arbuckle, T.E. (1997) Effects of cigarette smoking, caffeine consumption, and alcohol intake on fecundability. Am. J. Epidemiol., 146, 32–41.[Abstract]

Ford, W.C.L., North, K., Taylor, H. et al. (2000) Increasing paternal age is associated with delayed conception in a large population of fertile couples: evidence for declining fecundity in older men. Hum. Reprod., 15, 1703–1708.[Abstract/Free Full Text]

Greenland, S., Pearl, J. and Robins, J.M. (1999) Causal diagrams for epidemiologic research. Epidemiology, 10, 37–48.[ISI][Medline]

Jensen, T.K., Scheike, T., Keiding, N. et al. (2000) Selection bias in determining the dependence of waiting time to pregnancy. Am. J. Epidemiol., 152, 565–572.[Abstract/Free Full Text]

Juul, S.,.Keiding, N. and Tvede, M. on behalf of the European Infertility and Subfecundity Study Group. (2000) Retrospectively sampled time-to-pregnancy data may make age-decreasing fecundity look increasing. Epidemiology, 11, 717–719.[ISI][Medline]

Olsen, J. (1990) Subfecundity according to the age of the mother and the father. Dan. Med. Bull., 37, 281–282.[ISI][Medline]

Olsen, J., Juul, S.and Basso, O. (1998) Measuring time to pregnancy. Methodological issues to consider. Hum. Reprod., 13, 1751–1753.[Free Full Text]

Sallmén, M. (2000) Fertility among workers exposed to solvents or lead. Academic dissertation. People and Work. Research reports 37. Finnish Institute of Occupational Health, Helsinki, Finland.

Spira, N., Spira, A. and Schwartz, D. (1985) Fertility of couples following cessation of contraception. J. Biosoc. Sci., 17, 281–290.[ISI][Medline]

Weinberg, C.R., Baird, D.D. and Wilcox, A.J. (1994) Sources of bias in studies of time to pregnancy. Stat. Med., 13, 671–681.[ISI][Medline]