Biological Testing
Michael Gilding
Swinburne University of Technology
The discovery of blood groups in the early 20th century created the scope for biological paternity testing. In the 1970s, serological testing became much more sophisticated. Then, from the late 1980s DNA testing largely took its place. In fact, none of these techniques prove biological paternity per se. Rather, they establish biological nonpaternity, by ‘excluding’ men whose body substances – blood type, HLA antigens or DNA polymorphisms – are incompatible with biological paternity. In turn, a man is said to be the biological father on the basis of probability, drawing upon assumptions about the frequencies of biological markers in the general population. The probability of ‘exclusion’ using ABO blood types alone is about 17 per cent; for more sophisticated serological testing it is in excess of 95 per cent; and for DNA testing it is more than 99.99 per cent. DNA testing is not only more precise than serological techniques, but faster and cheaper (Richards, 2001: 676–7). In this context, a commercial paternity testing industry has emerged worldwide, mostly grounded in disputed paternity (Gilding, 2006: 91–3). During the same period, paternity testing has also become more widespread in the context of medical procedures, such as organ transplant and genetic screening (Lucassen and Parker, 2001: 1033).
Notwithstanding the proliferation of testing, there are no reliable population-based studies about the extent of paternal discrepancy. All test-based studies involve bias of some kind. On the one hand, there is bias towards high levels of paternal discrepancy in the context of disputed paternity by definition. O the other, there is bias towards low levels in many medical contexts, where participation is voluntary and presumably avoided by women fearful of disclosure.In other circumstances, bias is unknown, on account of lack of information
about the sample employed in the study.
Consider, for example, the studies employed by Bellis et al. in arriving at an
estimated paternal discrepancy rate of 3.7 per cent, grounded in ‘other testing’
– mostly for medical ends. The oldest study was published in 1957; the most
recent studies were published in 1999. Five studies were done in the US, four in
the UK, two each in France and Mexico, and one each in Canada, New
Zealand, Switzerland, and among the Yanomama tribe of southern Venezuela
and northern Brazil. In turn, these studies vary widely in their estimates of
paternal discrepancy: from 0.8 per cent in a 1994 Swiss sample, to 30.0 per cent
in a 1973 English sample. Given the variation, it is elementary to observe that
great care should be observed in applying a median to these studies.
The four British studies based upon ‘other testing’ – published in 1957,
1973, 1991 and 1999 – illustrate the point. The first, published in a medical
journal (Edwards, 1957), reported on an earlier study of heterospecific pregnancy based on ABO blood testing of 2578 families at a West London hospital
in 1949–50 (Johnstone, 1954). The study included all live-born infants during
the period of investigation, but no information was provided about the demographics of those who attended the hospital. Paternal discrepancy was observed
for 17 babies. Given ABO testing, Bellis et al. and Anderson agree that the
actual number of cases was almost 100, or 3.7 per cent (95% CI 3.0–4.4%).
Yet they disagree about the bias. Bellis et al. describe the bias as ‘unknown’, presumably guided by lack of information about the sample. Anderson describes paternity confidence as ‘relatively high’, presumably guided by the testing of all
live-born infants.
The 1973 study was more ambiguous than the 1957 study, and more influential. In 1991 it was described as the ‘most commonly quoted UK reference’
on the subject, routinely employed to justify estimates of 10–15 per cent in medical and genetics textbooks (MacIntyre and Sooman, 1991: 870). The actual
data were neither published nor refereed. What was published were passing
remarks made by Elliott Philipp, a medico, in the transcript of a symposium on
the ethics of artificial insemination by donor. Philipp described blood testing
200 to 300 women and their husbands in ‘a town in south-east England’. ‘From
our results,’ he stated, ‘we suddenly realized that 30% of the children could not
have been fathered by the men whose blood group we analysed’ (Wostenholme
and Fitzsimons, 1973: 63). Moreover, serological testing meant that the actual
paternal discrepancy rate was much greater than 30 per cent. A passage in the
transcript suggests that the families tested might have been ones where births
occurred outside of marriage, and another medico described the sample as
‘highly biased’ (p. 66). Again, Bellis et al. and Anderson classify the study differently. Bellis et al. observe a bias towards a lower ‘paternal discrepancy rate’
on account of ‘poor test sensitivity’ (2005: 751). Anderson includes the study
among those where paternity confidence is unknown, presumably because of
insufficient information about the sample.
The 1991 study was a letter to the editor of The Lancet, written in response to an article which suggested that high levels of paternal discrepancy had ‘taken on the character of urban folk tales – pieces of conventional wisdom which are widely believed but have little basis in fact’ (MacIntyre and Sooman, 1991: 869). Two medicos from the University of Edinburgh reported on a carrier testing programme for cystic fibrosis (CF) during pregnancy at the main Edinburgh maternity hospital, and for eight other DNA diagnostic laboratories around the UK (two each in London and Manchester, and one each in Belfast, Cardiff, East Anglia and Oxford). Of 521 families tested, there were seven results (1.4%; 95% CI 0.4–2.3%) incompatible with biological paternity. The medicos also reported another CF carrier testing trial, which provided an indication of bias arising from women avoiding the tests on account of paternity doubt. In this trial, women unsure of the father or whose partners were unavailable were asked not to participate. Of 1619 women, 17 (1.05%) disqualified themselves because they were unsure of the father, while another 31 (1.9%) did not participate for other reasons. The medicos concluded: ‘We believe that cited rates of non-paternity are somewhat exaggerated, and that the true rate lies closer to 1%’ (Brock and Shrimpton, 1991: 1151). Bellis et al. and Anderson agree that this study is biased towards those with high paternity confidence.
Finally, the 1999 study was an article in a medical journal directed towards the identification of genetic markers for multiple sclerosis (MS). The study involved a volunteer sample of 821 affected individuals from across the UK with living parents willing to take part. The mean age of onset was 36 years. By implication, the findings pertain to paternal discrepancy among those born in the 1970s and earlier (assuming adult participants only). Of the 821 individuals, 12 results (1.6%; 95% CI 0.7–2.5%) were incompatible with biological paternity (Chataway et al., 1999: 210). Again Bellis et al. and Anderson agree that this study is biased towards those with high paternity confidence.
The bottom line is that these studies are not of equal merit, and it is misleading to summarize their findings – in the manner of Bellis et al. and Anderson – through the median. The 1973 report does not include enough information to warrant consideration. The 1957 study is better, but based upon what are now rudimentary techniques and an unknown sample. The 1991 and 1999 studies are better again, on account of fuller information and better methods. It is revealing that Bellis et al. and Anderson disagree about the bias of the earlier studies, but agree on the later ones. Yet the 1999 study pertains to individuals born across decades, and offers no guide as to bias arising from a volunteer sample. The 1991 study, on the other hand, pertains to a birth cohort born in the early 1990s, and does provides an indication of the possible extent of bias on account of the sample. It thereby provides the most realistic, albeit conservative, estimate of paternal discrepancy based on testing procedures in the UK.
Comments