Pages within the "support" section:

10. Effects on the next generation

Some assessments involve health effects on the next generation (i.e. on the offspring of the population experiencing the dietary change). There are two alternative ways to represent such effects when using the Qalibra framework and software, as described below. Both of them involve additional complications for the user when preparing model inputs and require additional care when interpreting results.

One method is to consider effects on offspring and their mothers together in the calculation of net health impact. In this approach, the offspring are not included as separate individuals in the calculation; instead, the effect on them is included in the calculation for their mothers. This requires the assessor to specify p effect for females, as the probability of giving birth during the current year to a child that has the effect [2] , which will be a function of the mother’s current age. If the effect is gender-specific (e.g. an effect on sperm counts of male offspring) then p effect needs to be the probability of giving birth to a child of the relevant gender, that has the effect. When using this approach for next-generation effects, other parameters in the calculation (CA, LE, disease weights, etc.) should in principle be set to the values appropriate to the attributes of the offspring at the age when the effect occurs (e.g. zero for effects in newborn infants, puberty for reproductive effects, etc.). However, it was not feasible to provide for this in the current version of the Qalibra software. Instead, Qalibra uses values based on the attributes of the mother for all parameters except current age CA, which is set to zero automatically when the user indicates (by ticking a check box in the inputs) that the effect is on offspring. Consequently, other parameters will be set to the values appropriate for individuals of age zero, but with other attributes equal to those of their mothers. This means, for example, that LE (life expectancy) will be set to the female value for all newborns, which will normally be higher than the value that should be used for male newborns, leading to some over-estimation of the health impact. It also means health impacts will be over-estimated for effects that actually begin after age zero (e.g. at puberty). This and any other consequences of this approach need to be borne in mind when interpreting the results: if there is concern that the result will be materially misleading (e.g. enough to change the balance of risk and benefit), then it would be advisable to try using the alternative approach for representing next-generation effects (see next paragraph), which allows all parameters to be based on the attributes of the offspring. Note that, when using the above method, treating mothers and their offspring together, the total number of individuals represented in the overall assessment will be greater than the number of individuals specified as input (the current generation), due to addition of offspring generated in modelling the next-generation effect. Also, when examining variation between individuals in contribution to the population’s net health impact, the contribution of mother and offspring will be shown together, although their relative importance may be seen by examining the breakdown of effects produced by Qalibra, or by rerunning the assessment without the next-generation effect. Significant additional care is therefore required when interpreting the results of next-generation assessments using this method.

The alternative method for representing next-generation effects avoids the difficulties of the first method, but presents a different challenge. This method requires the user to include an appropriate proportion of newborn children in the population for the assessment. This allows parameters including CA and LE for offspring to be set according to their age and other attributes at the point of disease onset. It also means that these individuals will be represented separately in the assessment results, simplifying interpretation. The challenge for this method relates to the inputs for intake or exposure to the substances causing the next-generation effects, as these should be intakes of the mother and not of the newborn individuals themselves. These intakes could be based on a dietary survey data for women of child-bearing age, however it may be necessary to include a substantial number of newborns in the population to provide a sufficient sample to adequately represent variation in the intakes of mothers, especially if the effect is associated with unusual or extreme intakes. This in turn will either require that the overall population is very large (in order to include an adequate number of newborns), leading to long model run times, or that newborns are over-represented in the calculations but that other age groups are scaled up to compensate (using the scaling factor, I sf ) [3] .

[2] The possibility of twins, triplets etc. may be incorporated by specifying p effect not as a probability of giving birth, but as the average number of offspring born with the effect during the current year for a woman of the relevant age.

[3] If scaling factors are used, care should be taken to ensure that the number of individuals modelled explicitly (before scaling up) is sufficient to represent adequately the variability of intakes amongst individuals of similar attributes.