Papers

Commentary: Down's syndrome and parity

BMJ 1997; 314 doi: https://doi.org/10.1136/bmj.314.7082.721 (Published 08 March 1997) Cite this as: BMJ 1997;314:721
  1. R J Lilford, professor of health services researcha
  1. a Department of Public Health and Epidemiology University of Birmingham Birmingham B15 2TT

    Commentary

    Everyone knows that the risk of Down's syndrome rises with the mother's age. Parity also rises with age. Could some of the increased risk among older mothers be accounted for by their greater mean parity?

    Schimmel and colleagues found the expected sharp rise in risk of Down's syndrome with maternal age, but they also found an independent effect of parity—at a given age, risk rose with parity. They say that parity might be an additional factor to be taken into account in the calculation of the risk of Down's syndrome. Before accepting this conclusion, we need to ask ourselves whether the data may be confounded in some way.

    The data were derived from 37 110 births which occurred during 1981-9 at the Shaare Zedek Hospital. The authors defend the measurement of birth prevalence as a proxy for incidence (terminations and births) on the grounds that this institution serves an exclusively orthodox Jewish population who eschew prenatal diagnosis. But some women may have availed themselves of prenatal diagnostic services elsewhere. If this were true, and if they tended to be of lower parity (perhaps because they were also less averse to contraception), then this would bias the results towards greater risk at high parity. However, this effect would apply almost exclusively to older women, since only maternal age screening was available for Down's syndrome in Israel before 1990. The “parity effect” was indeed greatest among such women, but it was also present among younger women. Therefore, although this form of bias is supported by the data, it cannot fully explain the observed association.

    Maternal age was grouped in five year age bands. This can lead to “truncation,” a spurious or exaggerated association between two variables when they increase at different rates across the band range in the control variable.1 In this case, both risk of Down's and parity systematically favour the high end of each maternal age band, but Down's risk increases exponentially. Modelling would show how much of the additional risk attributed to parity could plausibly be ascribed to this problem. However, this bias would affect older mothers in particular, since the risk of Down's syndrome increases rapidly across the higher age bands. Again, therefore, the association among younger mothers cannot be attributed solely to this bias.

    The authors do not attempt to embed their results in a systematic review; they do not quote such a review, and they quote only one other paper, whose findings were negative.2 Castilla and Paz duplicated the Israeli analysis, and when five year age bands were used they also found an association between parity and risk of Down's syndrome—but this association was confined to the oldest age bands.3 Furthermore, mean maternal ages (within bands) were significantly higher for mothers of infants with Down's syndrome than for control mothers. Thus, truncation may have “explained” the findings, at least in part.

    Lastly, the Israeli paper showed a startlingly low prevalence of Down's syndrome among primiparous women—no cases in 4299 births to mothers aged 25-40. Associations based on atypically low risks in control populations call for wariness.

    If the association were true, what is the mechanism? It is hard to imagine that parity or high fertility predisposes to non-disjunction, and translocation carriers should be predisposed to high gravidity rather than parity (H Cuckle, personal communication). The most plausible hypothesis that I can think of is that parous women are less likely to abort not only normal fetuses but also those with aneupoidy. Before we investigate this further or advise women to take parity into account, corroboration of the main findings is required.

    References

    1. 1.
    2. 2.
    3. 3.
    View Abstract