Have you considered multilevel modeling as a way of doing this analysis?

This paper sets out the pros and cons of multilevel versus SEM

http://eprints.lse.ac.uk/52203/1/Steele_multilevel_models_longitudinal_2008.pdf

The ML approach is very easy to implement and works well where there is missing data so that occasions are seen as nested in children and there are some occasions where the child was not measured ( as long as the value of the response variable  is not the cause of the missingness this is OK). The approach handles time varying and time-invariant predictors and time (or Age) itself can be a CONTINUOUS predictor which can be very important in growth analysis of children as categorical or clinic time is too coarse when children are undergoing fairly rapid development. (Clinic time - 1,2,3,4 is rather coarse and is akin to measurement error in the predictor variable leading to attenuated effects.) You may prefer SEM when the response has not been directly measured ; that is you have several measures of the outcome.

This site has lots of material on multilevel analysis of repeated measures:

http://www.bristol.ac.uk/cmm/software/mlwin/mlwin-resources.html#multipleresp

This Manual on RGate has two chapters on the  modeling of the development of children. 

https://www.researchgate.net/publication/268331676_Vol2Traingmanual2014withminorcorrections

Data Vol2Traingmanual2014withminorcorrections

I would check out Patrick Curran's work. He is head of the psychometric lab at UNC and has done a lot of work using multilevel and SEM for modeling growth in children. His web address is:http://www.unc.edu/~curran/manuscripts.htm

Thanks for all your comments and answers.I will try to work on it.

Do your 'repeated risk factors' change with time? Either way, you could look at them as time varying or time-invariant predictors. 

You could model children growth (CG) using a growth curve model by specifying slope and intercept: 

CG1 = fCG_alpha +  D1, 

CG2 = fCG_alpha + 1 fCG_beta + D2, 

CG3 = fCG_alpha + 2 fCG_beta + D3, 

CG4 = fCG_alpha + 3 fCG_beta + D4, 

CG5 = fCG_alpha + 4 fCG_beta + D5, 

CG6 = fCG_alpha + 5 fCG_beta + D6;

Of course, if you feel that growth is not linear, you could change the coefficients to reflect quadratic etc. relations. 

Then you could relate your risk factors to fCG_beta, adjusting for fCG_alpha (your intercept, or baseline value). 

ie. 

fCG_beta = * fCG_alpha + *age + *BMI + *fTimeVaryingRiskFactor_beta; 

in which case, fTimevaryingRiskFactor would make it a parallel growth curve model. 

You should be careful with optimization problems and missing data though.