Table 1 Summary of approaches used by current studies of precision psychiatry compared with the R-LiNK study (see text for details)

Studies of lithium response biomarkers are largely based on secondary analyses of efficacy RCTs, or convenience samples. The lack of clinical representativeness and/or small sample sizes may have contributed to biases in findings in biomarker studies (Carvalho et al. 2016; Hoertel et al. 2013)

To increase the translational potential, the R-LiNK study employs a pragmatic design that reflects clinical reality. Each centre will recruit 20–30 patients who have agreed to initiate a trial of lithium treatment (on the recommendation of their treating clinician). Exclusion criteria are minimized to enhance the generalizability and external validity of study findings

The definition and measurement of lithium response varies between publications. Some studies focus only on ‘good response’ subgroups and compare this group to the rest of the population; others identify several response categories or compare a range of categorical and continuous measures of response. Many studies assess lithium response using cross- sectional retrospective assessments rather than prospective monitoring. Also, only few studies follow guidelines on differentiating non-response from non-adherence (Howes et al. 2017)

R-LiNK will follow participants prospectively for 2 years after initiation of lithium and will systematically assess clinical symptoms, illness activity and medication adherence over time. An independent panel of experts will examine all this longitudinal data to classify the clinical response of each participant according to response categories that have been agreed a priori (with the aim of reducing phenotypic misclassifications)

Lithium adherence and risk of sub-optimal adherence will be monitored. A brief intervention may be offered to maintain engagement with treatment

Samples vary in homogeneity or heterogeneity, in the reliability of diagnosis and the range of BD subtypes included

Measures of illness activity may vary significantly across studies: some use retrospective clinical reports, others use established observer ratings, others combine observer and objective ratings. The type of ratings selected, and the weightings given to individual symptoms of BD or to illness dimensions in the scales selected may affect the identification of clinical predictor variables or influence the concordance between clinical and biological variables

Few biomarker studies include patient-related outcomes (PRO)

Many components of the methodology adversely affect the potential signal-to-noise ratio (South et al. 2017)

R-LiNK will recruit individuals with symptoms that meet internationally accepted diagnostic criteria for BD-I and includes methodological strategies that try to increase the ‘signal’ and reduce the ‘noise’ (Scott and Etain 2018)

Symptom measures have been selected based on (i) good psychometric, item response theory (IRT) and clinimetric properties; (ii) the weighting given to symptoms that may be particularly sensitive to change early during lithium treatment (e.g. activity, energy and mood); (iii) a balanced assessment of key symptom dimensions

We will use electronic self-monitoring of core BD symptoms. This ecological momentary assessment (EMA) approach will include daily ratings of a unique subset of selected symptoms of BD plus additional PRO items that can be used to formulate individualized ratings of personal recovery (which also can be compared with other R-LiNK response categories)

Many studies focus on a single biomarker or select markers from one specific domain (e.g. focusing only on fMRI, omics, etc.). Recent research indicates that prediction may be enhanced by using combinations of factors, rather than trying to identify single or unidimensional biomarkers of outcome or treatment response (Trivedi et al. 2016)

It is unlikely that there is a single biomarker for lithium response (or non-response or tolerability), so R-LiNK employs an integrative science approach to try to identify combinations of clinical, functional, structural, molecular and metabolic biomarkers (called biosignatures) that may be included in a composite prediction tool. In addition, R-LiNK will examine clinical and biological moderators and mediators of lithium response