select search filters
roundups & rapid reactions
before the headlines
Fiona fox's blog

expert reaction to smoking in boys and fatness in their future sons

A new study published in the European Journal of Human Genetics investigated links between smoking in boys and men, and BMI and fat mass of their children. A press briefing was held to present the findings of the study.


Prof Wolf Reik, Head of the Epigenetics and Chromatin Programme, Babraham Institute, said:

“The paper makes a fascinating observation which further supports the idea that transgenerational epigenetic effects can occur in humans. These are also increasingly observed in animal models, showing that while there is a mechanism for ‘erasure’ of epigenetic information between generations, this can fail on occasion. The inheritance effects reported here (and those in some animal models) occur with transmission from father to son which means that the information needs to be carried in the sperm. Since the sperm is quite small there are some candidate molecules which one could suspect could carry the information. These include methylation (chemical modification) of DNA, some histone (chromatin) marks, or for example small RNAs. It will be exciting to examine these candidates in animal models, and eventually in humans as well.”


Prof Sir Richard Peto, Professor of Medical Statistics & Epidemiology, University of Oxford, said:

“The main claim in this report is unjustified. This new study does not provide good evidence that fathers who started to smoke before age 11 have fatter teenage children as a result. Indeed, on the contrary, it provides reasonably good evidence that paternal smoking is of little or no relevance to teenage body mass in the next generation. For, among the children who were weighed in their teenage years there were about 40 times as many with fathers who had started to smoke at age 15+ as there were with fathers who had started before age 11, and this report found no material effect of fathers having started to smoke at age 15+ on the body-mass index (BMI) of their teenage sons or daughters.

“The false-positive result in this report arose mainly because the statistical analyses put unduly data-dependent emphasis on what could well (if appropriate allowance is made for multiple hypothesis testing) have been chance findings in the small subgroup of a few dozen teenagers whose fathers had started to smoke at a very early age, while largely ignoring what happened in the far larger numbers of teenage sons and daughters whose fathers had started to smoke at age 15+.”


Prof Tim Spector, Professor of Genetic Epidemiology, King’s College London, said:

“Transgenerational effects of environment have been clearly shown in rodents but not in humans. This is a rare study showing transgenerational effects in humans. The data are persuasive but not yet definitive as we need to confirm the same smoking related epigenetics changes in the kids’ DNA. We urgently need more studies like this.”


Dr Graham Burdge, Reader in Human Nutrition, University of Southampton, said:

“The findings of this study may potentially provide new insights into factors that may influence development of obesity in childhood.

“However, the findings only show associations and cannot be interpreted as indicating that paternal smoking at an early age causes obesity in their sons.

“Although the authors have been careful to control their data for major confounding influences, they did not take into account parental BMI, which is a major factor in childhood obesity, or the level of exercise the children undertook.  These are important omissions that could potentially confound the findings of the study. 

“The small number of fathers in the early smoking category (less than 1% of the whole subject group) can give rise to results that are not representative of the general population and are open to analytical errors. It is unclear whether there were sufficient father-son pairs in which the father started smoking young to allow firm conclusions to be drawn.

“The association between fathers and sons, but not fathers and daughters, is intriguing, but has been shown previously.  Weight gain and waist circumference during childhood has been related to parental weight such that fathers’ weight was associated with that of their sons, but not daughters, and vice versa.

“The authors only tested one gene out of the 35 or more genes that have been linked to obesity.  So, these findings cannot rule out the possibility that inheritance of genes is involved.

“This study adds to the list of possible indicators of obesity risk in children and adolescents, but does not increase knowledge about the underlying processes.  A possible alternative interpretation is that that exposure of parent and child to a common environment has an effect on the child’s BMI that might be influenced by the lifestyle of the father.”


Prof Patrick Wolfe, Professor of Statistics, UCL, said:

“I just don’t think this is very compelling science.  The authors tested various combinations of parental smoking relative to children’s BMI in different age groups (7, 9, 11, 13, 15, and 17 years).  Of these tests, they refer in the title and abstract to just one particular combination which they found to be significant: adjusted mean difference of BMI in sons aged 13, 15 and 17 from fathers that started smoking from age under 11 (not mothers’ smoking; not daughters, and not for sons aged 7, 9, and 11).  Thus, because of multiple comparisons, the true significance level of this finding is likely to be substantially less than the authors report – the authors tested numerous combinations of things, as outlined above, but didn’t adjust their measures of significance for the fact that they tested the same data so many times.

“The result that the authors emphasise comes from comparing just 0.16% of the sample against the remaining 99.84%.  I see three main limitations with the study’s conclusions:

“1)  In the study, the authors assessed effects of both mothers and fathers, but in the title and abstract of the paper, they only report their mean BMI results about fathers – they did find an association between fathers smoking from before age 11 and sons’ BMI, but no evidence of an effect with mothers. The authors attribute this to the “very few” (1% of smoking mothers) mothers who reported starting smoking before age 11; but the men who reported starting smoking before age 11 are themselves only 3% of smoking fathers.

“2) The authors tested for an effect of higher mean BMI in daughters as well as sons, and found none (cf. Figs 1a and 1b).  They then conclude that they “cannot rule out a smaller effect in daughters.”  It is important to remember that no study of this type can ever rule out any effect – we cannot prove a negative!  Similarly the press release says “The effect, although present, was not seen to the same degree in girls.”  There is no evidence of any BMI effect in girls, according to the authors’ own analysis (see Fig 1b – the confidence intervals include 0, which means no effect can be concluded).

“3) Likewise, the authors looked for effects across a range of children’s ages, and again emphasised certain ages (13, 15, 17) over others (7, 9, 11).

“Together, these three points increase the chance of getting a false positive result.  The results need to be considered in an overall context, including the fact that the authors tested for a number of other associations and did not find evidence for them.  That is because, for a reported significance level to be valid, a single hypothesis must be formulated before a single experiment is performed – rather than multiple hypotheses all simultaneously tested (mothers, fathers, sons, daughters, and different children’s age ranges).

“Another cause for concern is that the authors do not state whether the parental-smoking-age cutoff of 11 years old was chosen before looking at the data.  Thus we do not know how sensitive the authors’ reported results are to this cutoff.

“Even if the results can be considered as meaningful, this study cannot rule out hidden associations because one cannot control for all confounding factors: for example, the authors report on the tiny fraction of the sample comprising sons whose fathers started smoking before age 11; who is to say that these sons themselves might not have started smoking prior to age 11, thus potentially influencing their weight in one way or another?”


‘Prepubertal start of father’s smoking and increased body fat in his sons: further characterisation of paternal transgenerational responses’ by Kate Northstone et al. published in the European Journal of Human Genetics on Wednesday 2 April 2014.

in this section

filter RoundUps by year

search by tag