I had a similar experience with a spike control appearing to disappear in plasma samples. Turned out that the spiked control (a peptide) was being degraded, presumably by enzymes in the plasma sample -- we published this observation. Article Intrinsic Peptidase Activity Causes a Sequential Multi-Step ...

Remember that plasma has many (MANY!!) enzymes. The lesson is that any spiked control may be subject to cleavage or other modification, changing to a new form. It's also possible that the spiked control molecule is becoming tightly bound up with other proteins (e.g. albumin) and being masked from detection.

Never assume stability of anything spiked into a blood sample -- better to prove it, and eliminate a significant potential problem in later analyses. In reality, all you really need to do is a simple control experiment: Spike your control into a fresh plasma sample. At regular intervals, withdraw an aliquot of the plasma and quench active enzyme activity (maybe boiling?), then perform MS analysis (sample sample as a function of time after spiking). Enzyme activity would be implicated if you observe a drop in your control over time. You might even see a rise in a the metabolic product, which can help confirm that it's actually an enzyme doing the damage -- as we did in our publication.

You might also try quenching by adding a general denaturant (SDS? urea?), which might also have the benefit of releasing your control molecule if it happened to be bound up in proteins like albumin --> in which case you might be able to define a better sample preparation method that would reliably release your molecule for subsequent detection.

DH in-source exchange could be an answer. Where are the four D atoms in the molecule?

C15H21N3O3S or C15H17D4N3O3S

Did you check for the mass of D3 or D2 analogs? They should be present if D4 is lost by deuterium-hydrogen exchange