Which Test to use for difference between two time points AND correcting for confounders?


I am a Phd student currently working on biomarkers during pregnancy and am unsure which test to use.
I want to assess whether there are significant differences in my biomarker at the beginning of pregnancy compared to a time point later on.
From experience and also a Pearson's correlation analysis I conducted, I know that there are several potential confounders for my biomarker due to changes during pregnancy (f.i. renal function changes in pregnancy).

When not correcting for confounders (I am using SPSS) I know to use a simple paired samples T test (which was significant: p<0.001)

However, I think it would be very sensible to correct for at least 4 or 5 variables that change during pregnancy. Which test should I use?

--> I tried univariate ANCOVA (biomarker as dependent variable, time points as fixed factors, covariates as covariates), however, I am unsure about the validity of the results as my parameters are dependent and one of the prerequisites of ANOVA is to have independent variables.
--> I also tried a repeated measure ANCOVA, however, I do not have any groups to compare just one variable (continuous) and 2 time points...

Is there any other way I missed? Is any of the options above ok to use?

Would appreciate any help very much!
2 time points of 60 people. The study question is how the biomarker behaves during pregnancy/does the biomarker rise/fall during pregnancy?


Less is more. Stay pure. Stay poor.
Is the biomarker bound by zero or does it have a limit of detection?

Well I believe the literature supports including the baseline value in the model and the post value as the dependent value. You can include variables to control for that you think may impact the DV (e.g., gestational diabete status, etc.), but I am unsure if you would deem these confounders, since what relationship are they confounding? There would need to be a systematic effect modification between the starting value and other variable or the variable would have to be a common cause of both variables and not Markovian independent for them to have an association with both terms. Last comment would be with 60 subjects you could likely only be able too adjust for a couple of variables before sparsity becomes a concern and these variables would have to be fairly balance between the groups for the SEs not to become too large.