Saliva testing as a means to monitor therapeutic lithium levels in patients with psychiatric disorders: Identification of clinical and environmental covariates, and their incorporation into a prediction model

Abstract

Objective: The narrow therapeutic window of lithium medications necessitates frequent serum monitoring, which can be expensive and inconvenient for the patient. Compared to blood, saliva collection is easier, non-invasive, requires less processing, and can be done without the need for trained personnel. This study investigated the utility of longitudinal salivary lithium level monitoring.

Methods: We measured salivary lithium levels using ICP-OES in n = 169 passive drool samples, collected both as single observations and longitudinally for up to 18 months, from a multi-center cohort of n = 75 patients with bipolar disorder or other psychiatric conditions.

Results: Saliva and serum lithium levels were highly correlated. Adjustment for daily lithium dose, diabetes, and smoking improved this relationship (r = 0.77). Using the adjusted intersubject equation and a patient's salivary lithium value, we observed a strong correlation between the predicted vs. observed serum lithium levels (r = 0.70). Most patients had highly stable saliva/serum ratios across multiple visits, with longitudinal variability significantly greater with age. Use of the intrasubject saliva/serum ratio from a single prior observation had similar predictive power to the use of the adjusted intersubject equation. However, the use of the mean intrasubject ratio from three prior observations could robustly predict serum lithium levels (predicted vs. observed r = 0.90).

Conclusions: These findings strongly suggest that saliva could be used for lithium monitoring, and open the door for the development and implementation of a point-of-care salivary lithium device for use at home or the clinic. We propose that the use of saliva will dramatically improve treatment opportunities for patients with mood disorders.

Keywords: biofluid; bipolar disorder; blood; lithium; peripheral; saliva; therapeutic drug monitoring.

FIGURE 1

Correlations between salivary and serum lithium levels, unadjusted (A), and adjusted for relevant covariables, (B). Unadjusted correlation was determined by Spearman correlation analysis (r = 0.74; p < 0.0001). Adjusted correlation (Spearman rho = 0.77, p < 0.0001) was determined by multivariate linear regression analysis including adjustment for daily lithium dose, type 2 diabetes, and smoking

FIGURE 2

Relationship between salivary and serum lithium levels in the Determination cohort (A) and the relationship between predicted and observed serum lithium levels in the Validation cohort (B). The correlation between the salivary and serum lithium values in panel A were adjusted for daily lithium dose, diabetes status, and recruitment site (Pearson r = 0.77, p < 0.0001). The correlation between the observed and predicted serum lithium levels using the validation cohort equation from 2A was significant at a Pearson r = 0.70, p < 0.0001. Predicted serum lithium levels that differed from observed levels by more than 0.2 mmol/L are highlighted in grey in panel B

FIGURE 3

Comparisons of predicted vs. actual lithium levels, based on a single prior within‐subject saliva/serum lithium ratio (A) and an average of three prior within‐subject saliva/serum ratios (B). In panel A, the predicted versus observed serum lithium levels were significantly correlated (Pearson r = 0.71; p < 0.0001), however, this correlation improved dramatically when an average of three ratios were used to predict the 4th appointment (r = 0.90; p < 0.0001). Overall, patients 55 years of age and above showed the most variability in saliva/serum lithium ratio levels (black data points in A, B). Patients were included in Figure 3A if they attended two or more appointments, and Figure 3B if they attended four or more appointments

FIGURE 4

Saliva/serum ratios for individual patients across multiple visits in patients younger than 55 years (A) and equal or greater than 55 years (B). Patients' first study appointment is represented by month 1, followed by every subsequent appointment, spaced by months since the first appointment. For clarity, only data from patients attending visits spanning 10 months or greater from their initial appointment are shown. Each color represents an individual patient. Saliva/serum ratios for all patients attending at least two appointments are shown in Table S1