The right answer is C.
It would be necessary to increase the size of the population as well as to vary it (to take a population of diffiffent families). To avoid skew errors as well as random errors.
Increasing the size of the study sample (without affecting the precision and validity of the measurements made on the subjects) makes it possible to limit the sampling fluctuations and therefore the random error.
The fact of not varying the population exposes us to a false estimate of the association between exposure and disease taking into account a possible intervention of a factor outside the causal chain .