By using a 'targeted' invitation strategy we aimed to minimize the risk of selection bias caused by selective participation of, for example, women experiencing fertility problems.
Selective participation did occur to some degree, as participating mothers were younger, more often single and smokers, and newborns were more often boys.
The effect of selective participation at wave 1 on prevalence estimates (selection bias) was examined previously [ 13].
The study used a bivariate probit selection model to control for the potential bias due to selective participation [ 53].
Thus, selective participation could not be ruled out, and generalization of the results should be done with some caution, because selective participation could not be ruled out.
A major drawback of previous studies has been selective participation in the program, leading to biased estimates of program effects.
Whether there has been a selective participation based on headache status will become clearer when information about headache status among non-participants becomes available during 2009.
However, selective participation due to headache status seems less likely, because the prevalence of headache, migraine and TTH included were quite similar among participants and non-participants.
The biases introduced by selective participation in the first survey were uncertain, but we thought it likely that those troubled by headache would have been among the responders.
The invitation letter did not mention that a detailed headache interview would be performed, hence a selective participation of headache patients is less likely.
The procedure of following up both randomized and nonrandomized patients is recommended when there is concern about selective participation and reduced generalizability.
