OBJECTIVE Nonadherence with scheduled saliva sampling, as encountered in ambulatory settings, can bias the estimation of salivary cortisol concentrations. This study is the first to estimate if such nonadherence is also associated with biased salivary testosterone concentration estimates. METHODS Using a standard ambulatory saliva-sampling protocol, we instructed pregnant women to collect saliva samples on two consecutive days at awakening, 1100h, 1500h, 2000h, and 2200h. We estimated testosterone concentrations in the saliva samples and participants' actual sampling times with an electronic medication event-monitoring system. We classified a saliva sample as adherent if it was sampled within a specific time window relative to its scheduled sampling time. We used a mixed-model analysis to distinguish between trait (number of adherent saliva samples per participant) and state (adherence status of a specific sample) adherence. RESULTS We included 60 pregnant women in this study. Seventy-five percent (448 of 600) of the scheduled samples indicated adherence with the sampling schedule. Participants' trait adherence was associated with their diurnal profiles of salivary testosterone estimates; that is, adherent participants had higher salivary testosterone estimates compared with nonadherent participants, F(1,58)=5.41, p=0.023, Cohen's d=0.67. The state adherence of a sample was associated with the salivary testosterone estimate of the related sample, F(1,469)=4.48, p=0.035, Cohen's d=0.20, with delayed sampling associated with lower salivary testosterone estimates. CONCLUSIONS The results suggest that common ambulatory nonadherence with scheduled saliva sampling is associated with biased salivary testosterone estimates. They will inform further studies estimating salivary testosterone with ambulatory saliva-sampling designs and highlight the relevance of strategies to improve or confirm adherence, beyond routinely used instructions.