Statistics from Altmetric.com
I write in response to the article by Rasinperä and colleagues (Gut 2004;53:1571–6) in which a DNA test was proposed for “adult-type hypolactasia”.
The ability to digest the milk sugar lactose as an adult (lactase persistence) is a variable genetic trait in human populations, lactase persistence being the most frequent phenotype in Northern Europe, while lactase non-persistence or “adult-type hypolactasia” is more frequent in most other populations.1 In sub-Saharan Africa for example, lactase persistence is found only at low frequency in the majority of populations that have been tested, but in some populations, particularly pastoralist groups, it is significantly more frequent.
A CT polymorphism located 13.9 kb upstream of exon 1 of the lactase gene (LCT) was previously shown in a Finnish population to be tightly associated with the lactase persistence phenotype2 and it is this change that is proposed as a DNA test for both Europeans and Africans. We agree that presence of a T at this polymorphic site is indeed a fairly good predictor of lactase persistence in Northern Europeans,3 and there is evidence that this nucleotide resides in a functional element.4,5 However, the presence of the alternative allele C at this site is not a good predictor of lactase non-persistence or “adult hypolactasia” in many non-Northern Europeans.6,7
I particularly draw readers’ attention to our recent study.6 We typed this polymorphism in 1671 individuals from seven African countries, which included 20 distinct cultural groups. In seven cases it was possible to match the groups tested with groups from the literature for whom phenotypic information was available. In five of these groups the published frequencies of lactase persistence were ⩾25%. We found the T allele in Cameroon but it was so rare elsewhere that it cannot explain the frequency of the lactase persistence phenotype throughout Africa and we devised a statistical test to show that these results were unlikely to have been obtained by chance.
Our ongoing results support this published information and we urge the community to refrain from using DNA tests on Africans and probably other non-Northern Europeans until an appropriate DNA change has been identified.
Dr Swallow raises a question about another DNA variant underlying adult-type hypolactasia in sub-Saharan populations and does not recommend analysis of the C/T−13910 variant as a genetic test in African and non-Northern European populations. Although the studies performed by us1–3 and others4,5 do not support the existence of another variant, we agree with Dr Swallow that well conducted studies are needed to confirm this.
The significance of the C/T−13910 variant underlying adult-type hypolactasia was questioned in the article by Mulcare.6 Their doubt is based on several assumptions that make it difficult to evaluate the significance of the findings. These assumptions can be listed as following:
It is not known whether or not the study subjects presented with adult-type hypolactasia. Thus there is a risk of wrong conclusions being drawn. It is well documented that the clinical diagnosis of adult-type hypolactasia is difficult to assess due to inaccurate diagnostic tests and variable, usually mild, symptoms.7,8,9,10 The diagnosis is usually based on indirect tests (lactose tolerance test or breath hydrogen test) whose specificity has been reported to range from 77% to 96% and sensitivity from 76% to 94%. In children in particular, the lactose tolerance test has not been ideal due to the high rate (up to 30%) of false positive results.11 The specificity of the breath hydrogen test varies between 89% and 100% and sensitivity from 69% to 100%.12 There is evidence that the breath hydrogen test may be an indicator of bacterial overgrowth rather than lactose malabsorption.13
Definition of ethnic origin was based on self definition and spoken language. As the authors themselves clarified, African populations have complex demographic histories. Many of the analysed groups were very small, and hence chance may have played a role. In contrast with the findings of Mulcare,6 our genotyping data in nomadic pastoralists Fulani-Sudanese were in agreement with the previously published figures of lactase persistence in this population.
There was no statistics shown against the C/T−13910 variant, only speculation presented in Mulcare’s paper.6
When conducting phenotype-genotype correlation studies in lactase persistence/non-persistence, detailed clinical studies are essential. The studies are difficult as it is unethical to take an intestinal biopsy from a healthy subject that would give the most reliable diagnosis. Measurement of lactase activity from hospitalised patients with a clinical indication for intestinal biopsy may reflect a disease in the gut and the result obtained may not correlate with the genotype. These uncertainties should be taken into account when interpreting the genotyping results in adult-type hypolactasia.
I write as the statistician on the paper by Mulcare and colleagues,1 which was criticised by Kolho and Järvelä above. I wish to correct two assertions made by Kolho and Järvelä. The first is their claim that with our statistical procedure “there is a risk of wrong conclusions being drawn” because “adult-type hypolactasia is difficult to assess due to inaccurate diagnostic tests”. This would indeed be true had we applied a “naïve” test (for example, a χ2 test) in which we had assumed the diagnoses of hypolactasia to be without error. In fact, not only did we assume that diagnoses occurred with error, but we did not even presume to know exactly what that level of error was. Instead, our uncertainty about the true level of error was modelled in a Bayesian framework and trained using available published data on comparisons of “true” diagnoses (for example, based on biopsy results) and “indirect” diagnoses (for example, based on breath hydrogen). To accomplish this, a novel statistical method was developed, which was described by Mulcare and colleagues.1 The fact that we incorporated these additional sources of error into our method means that the p values we obtained were not as low as they would have been had we applied a naïve test such as a χ2 test. Our remarkable finding was that, despite this, we still found significant departures in multiple sub-Saharan African populations. This led us to reject the null hypothesis that the presence of the C/T−13910 variant alone, even with diagnostic error, could explain the published data on lactase persistence in Africa. Other reasons must be sought to explain our results, one of which is the possibility that additional genetic variants influence lactase persistence.
The second assertion by Kolho and Järvelä was that there were “no statistics shown against the C/T−13910 variant, only speculation presented”. The meaning is unclear here. Certainly, statistics both in the sense of “data” and in the sense of “inference” were presented in our paper (see above). Our conclusions regarding the C/T−13910 variant were derived from carefully constructed statistical inference, and not mere “speculation”.
Conflict of interest: None declared.
Conflict of interest: None declared.
Conflict of interest: None declared.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.