Stephen Senn's first point is a valid one and we are grateful for the opportunity to clarify the situation. The 2002 paper1 should have stated the relationships between probands. This information is not only relevant to the statistical analysis as he points out but also to the issue of whether there might have been selection for a genetic variant carried by survivors of the grandparental generation that predisposes to diabetes in the probands generation. The relationship was checked regarding the 19 diabetic probands and they all had different paternal grandfathers. The nondiabetic probands could have been related, falsely narrowing the confidence intervals, but sensitivity analyses by randomly keeping just one cousin (where cousin relationship existed) does not change results significantly.

With respect to Senn's point about mass significance, we did not screen a lot of variables describing the same factor, we only ruled out confounding between ancestors. There were few hypotheses tested. The type of exposure (food supply) and the timing of exposure (slow growth period) had been selected on the basis of earlier work.2 Only two outcomes were tested, cardiovascular and diabetic deaths, both prior hypotheses.

Senn's third point relates to the control of age. At first glance, an outcome variable was used as predictor, but the outcome was not death but cause of death. As we knew that the paternal grandparents ‘starving’ during the slow growth period, when they were around 10–11 years old, was followed by a long life of the probands, we were careful not to overstate the probands cardiovascular and diabetes mortality, causes of death that have become so common during aging.

It is early days in the study of male-line transgenerational responses in humans, but some coherence is emerging with respect to observed outcomes. They have features of the Metabolic Syndrome, which some regard as a ‘maladaptation’ to a modern world.3 In addition to the cardiovascular and diabetic risks observed in the Överkalix cohort, the contemporary ALSPAC cohort has found an association between mid-childhood paternal smoking and a raised body mass index in sons4 and in Taiwan paternal betel nut (Areca catechu) chewing has been linked to early onset of the metabolic syndrome in the offspring.5 Still our findings were both significant and nonsignificant and more human replicating data are necessary.

Finally, we should emphazise that we do not claim that these transgenerational responses are mediated by epigenetic inheritance (ie where the DNA sequence itself is unchanged), only that this mechanism is a good candidate.