Reader Comments

Post a new comment on this article

It is notoriously difficult to estimate human heritabilities

Posted by plosmedicine on 31 Mar 2009 at 00:18 GMT

Author: Ian Hastings
Position: Senior Lecturer
Institution: Liverpool School of Tropical Medicine
Submitted Date: December 14, 2007
Published Date: December 14, 2007
This comment was originally posted as a “Reader Response” on the publication date indicated above. All Reader Responses are now available as comments.

The analysis employed was a standard one used in quantitative genetics (e.g. [1,2]). It partitions the variation in the trait (malaria/fever incidence) to find the relative importance of genetics and environment. Critically, the genetic component is not measured directly but is inferred from the resemblance between relatives. Obviously relatives can resemble each other because of their similar genetic makeup, or because relatives tend to share the same environment. The analyst must identify and include all the common environmental factors shared between relatives--- it is then assumed that the remaining resemblance is purely caused by genetics. Critically, if some environmental factors shared between relatives are absent from the analysis then the effects of the missing factors will be erroneously attributed to genetics leading to substantial over-estimates of heritability. This requirement to correctly identify, quantify and include all common environmental factors is extremely difficult to achieve [1,2].

They attempted to achieve this by including an environmental factor ‘household’ in the analysis, noting that “A household typically comprises a group of 3-6 adjacent houses each occupied by one women and her children”. So the analysis is only valid if the individual house environments are identical within households. Differences between houses within the same household is obviously a cultural/environmental issue but seem certain to exist. For example, subtle economic disparities between houses may occur, maternal health awareness and malaria treatment practices may vary, some families may be HIV positive, and so on. More disturbingly, adjacent houses can vary substantially in mosquito biting rates, and hence malaria inoculation rates, (e.g. [3]), for example due to their differing proximity to local mosquito breeding sites. Consequently, families in the same household will differ from each other not just because of their genetics but also because of differences in house ‘environment’: these ‘house environment’ effects will have been erroneously attributed to genetic factors, over-inflating the heritability estimate by an unknown (and unknowable) amount. A previous analysis of heritability of human immunity to P. falciparum vaccination [4] was more explicit in this respect, the authors noting that “when a house effect was included in the analysis (model 2), most or all of the genetic variance could be explained by living in the same house” and that “this suggests that the genetic variation was highly confounded by the sharing of common environments by closely related relatives”.

Unfortunately confidence intervals for the estimates were not presented but can be approximated as ± 2 standard errors (obtained from Table 1) giving heritability estimates of approximately 25% (95% CI 3% to 45%) for Study 1 and 28% (95% CI 0% to 82%). The wide CI also suggest that statistical significance (i.e. that heritability > zero, indicating genetic effect are present) may be marginal in Study 1 and absent in Study 2.

In summary the heritability estimates are almost certainly overestimates, associated with very wide confidence intervals, and are likely to be of borderline statistical significance. This contrasts sharply with the bland assertion in their summary that genetic factors “accounted for around one quarter of the total variability in malaria incidence”; a line followed by the accompanying Synopsis. The problem is that such assertions easily enter the scientific consciousness as general ‘rules-of-thumb’ without being subjected to extended scientific scrutiny, hence the critique developed above.


1. Lynch M, Walsh B (1998) Genetics and analysis of quantitative traits. Sinauer.

2. Kruuk LEB (2004) Phil. Trans. Roy. Soc. Lond. B 359: 873-890.

3. Smith T et al (1995) Acta Trop 59: 1-18.

4. Stirnadel HA et al. (2000) Int. J. Epid. 29: 579-586.

Competing interests declared: I declare that I have no competing interests