Does the “Silver Bullet” Lose its Shine Over the Time? Assessment of Loss of Lithium Response in a Preliminary Sample of Bipolar Disorder Outpatients

2.1. Sample

A convenience sample of treatment-seeking patients admitted to two outpatient facilities in Italy (Genoa and Pisa) between April 2008 and June 2016 was retrospectively reviewed. Either self-referrals or those cases referred by their primary care physician were evaluated. Participants were adults aged 18-65 years, both sexes, with a primary diagnosis of BD either Type-I (BD-I), Type-II (BD-II), or not otherwise specified (BD-NOS) according to the Diagnostic and Statistical Manual for Mental Disorders, Fourth Edition (DSM-IV) criteria [40]. Eligible patients were those who failed to preserve achieved long-term lithium-based maintenance of BD. These later patients would include either i) cases who stopped lithium (any reason) before loss of maintenance (some of whom would then possibly had developed DIR); ii) cases who continued their lithium medication at the same, clinically valid, blood levels of the drug at the time they nonetheless lost maintenance (some of whom would have then eventually developed tolerance towards lithium). All eligible patients were then “exposed” to “restarted treatment with lithium” (at clinically valid doses equal or higher than the ones proven effective at the time of maintenance, still in the absence of hypothyroidism or other clinical confounding factors). Those patients who failed to respond at least once more toward restarted lithium were labeled as “cases” (either due to DIR or tolerance) in contrast to those who responded at least once again toward lithium following its discontinuation upon achieved maintenance of BD “controls”, actually being those subjects without the “condition at interest” for the present study: “refractoriness/loss of lithium response towards restarted treatment with lithium following loss of maintenance”. Controls were matched based only on the diagnostic Type of BD (diagnosis last vouched at the time of loss of long-term maintenance) in order to avoid “over-matching”.

Patients excluded in the present chart review were either: i) pregnant women or those at childbearing potential refusing to take a valid anticonception tool (as this cohort could not have been “exposed” to lithium; ii) patients with severe cognitive impairment; iii) patients with known history of poor treatment-adherence based on the clinician’/caregivers’ judgment; iv) patients who failed to undertake regular [41] lab screening of lithium blood levels, thyroid and kidney functioning; v) subjects refusing to provide a valid signed consent (actually, n=4, due to privacy concerns). Thus, a total of 37 out of 41 (90.24%) unique clinical records met our inclusion criteria.

2.2. Procedure

Relevant demographic, clinical, and treatment information (including lifetime non-pharmacological treatment self-reported by the patient) were recorded by a clinical investigator abstractor blind to the study hypothesis (PF) upon receipt of a general purpose signed waiver obtained from each participating subject. Data were then presented anonymously in an aggregated form using Microsoft Excel [42] to the principal investigator (MF) who reviewed the information relevant to the present research theme based on a number of pre-determined study variables, owing to critical methodological considerations for conducting retrospective studies [43, 44].

Specifically, the a priori postulated “study hypothesis” was presence of “cases” (of loss of lithium response upon re-attempted trial following loss of achieved maintenance of BD) either due to tolerance, DIR, or other/unknown causes. Lithium discontinuation or tolerance phenomena at the time of loss maintenance of BD were nonetheless assumed to potentially occur both among those subjects going to respond once again (“controls”) or the ones actually refractory toward re-introduced lithium upon its discontinuation (“cases”).

The main study “outcome” and “condition at interest” were the “response vs. refractoriness toward restarted lithium after loss lithium treatment-based long-term maintenance of BD” (thus, assessing the eventual differential clinical features associated with “refractoriness vs. response”).

Accounted, eventual, adjunctive treatments to lithium during either the maintenance phase, its subsequent loss, or the restarted trial attempt with lithium were: U.S. Food and Drug Administration (FDA)-approved drugs for “acute mania” (& adjunct) and/or “maintenance treatment (of BD)” (and adjunct), depression associated with BD-I (the list of allowed medications substantially stayed unvaried within the time of most remote vs. most updated clinical records within the timeframe of study retrospective follow-up) (www.fda.gov); off-label prescription of most recent second generation antipsychotics (SGAs), antidepressants, benzodiazepines; or their combination.

An essential referenced summary of the definitions and outcomes adopted for the present study is provided in Appendix A.

Finally, upon conception of the study and generation of study’ hypotheses, a clinical scan of the research question and hypothesis was made seeking out clinical expertise with an independent co-author who assisted in the clinical interpretation of the results (DDB) and proof-reading alongside with additional co-authors (MS, AV, BS, FI, AC, GP, ADB).

2.3. Data Analysis

Following dummy coding of the variables at interest (pre-determined) as “yes” =1 or “no” =0, the distribution of data and the homogeneity of variance were assessed using the Kolmogorov-Smirnov and Levene tests respectively. Parametric analyses for demographic and clinical characteristics of the groups were performed using the independent-sample t-Student test for continuous variables (Mann-Whitney-U (Wilcoxon rank-sum) for non-parametric distributions) and the Pearson’s χ2 analysis (chi-square goodness-of-fit test) for categorical variables. IBM SPSS Statistics™ v.22 for Windows [45] was used for statistical analysis, with α=.05, two-tailed.

3.1. Overall Clinical Features of the Whole Sample

Median age of reviewed patients was 52 years old (IQR, or inter-quartile range=19), overall median duration in years (and IQRs) of phases I-, II-, III- and IV- were as follows, respectively: 8.5(6.5), 4(4.25), 5.5(1.75), 1.15(1.43). Median duration of tapering-off phase, in days, was 1(.75). All patients (F:M=17:20 - 46% of the total) had a diagnosis of BD-I.

Please refer to Appendix A for additional coding.

None of the patients included in the presented chart-review over a cohort study developed tolerance towards lithium within the retrospectively assessed period of time. Only 4 cases developed loss of lithium response/lithium-refractoriness upon its discontinuation (10.8% of the total) in contrast to 33 controls who actually lost maintenance of BD due to discontinuation occurring for a variety of reasons, yet going to respond once more to re-instituted lithium.

Overall, the included cases and controls were retrospectively assessed for an average period of 8.1 months after loss of lithium-based maintenance of BD and included only BD-I patients.

A quantitative synthesis of the clinical and demographic features of the patients included in the present study is presented in Table 1 and Fig. (1) discriminating between cases and controls.

3.2. Comparison of Cases vs. Controls

No statistically significant difference between cases and controls was observed with respect to a number of demographic and clinical features but for time spent before first trial ever with lithium in life (8.5 vs. 3 years; U=24.5, Z=-2.048, p=.041) and length of lithium discontinuation until new therapeutic attempt (5.5 vs. 2 years; U=8, Z=-2.927, p=.003) between cases vs. controls respectively. Tapering off of lithium was significantly faster among cases vs. controls (1 vs. 7 days; U=22, Z=-2.187), though both subgroups had worrisome high rates of poor adherence overall.

4.1. Main Findings of the Study: Clinical Features Associated to DIR vs. Preserved Response

In our convenience sample, loss of lithium response upon loss of maintenance occurred just in 4 out 41 patients (roughly 11% of the total). None of the reviewed cases developed tolerance towards lithium. Yet, the infrequency of the later phenomenon and the unrepresentativeness of the sample size would nonetheless require additional research to assess whether the prophylactic effect of lithium would be transient or persistent over the time, though reappearance of affective recurrences after years of successful treatment of BD could not be excluded by supposed persistence of the prophylactic effects of lithium anyway [7].

There was no statistical difference in the otherwise clinically valid (all in the 0.4-1.1mmol/L range) [46, 47] average serum level of lithium at the time of last reliable assessment prior loss of maintenance between cases or controls. However, stating the recall bias and self-report assessment nature of the present study, the related results must be interpreted with caution, as the actual time between loss of response towards lithium and the measurement of the last lithium level could not be reliably determined consistently in the preliminary sample in point. Similar considerations would apply to the statistical difference seen about phase I or ADHD comorbidity (as outlined in Table 1). Neither, a statistical difference was documented about the most recent levels of triiodothyronine (T3), though the actual validity of such clinical routine test has been questioned [48, 49].

Too fast tapering of lithium among cases (of whom, one with ascertained attention deficit hyperactivity disorder) and the median duration of phases -II and -IV were the only statistically different outcomes observed between cases and controls. Both outcomes have major clinical implications. In our sample, the exposure to antidepressant (adjunctive) therapy across varying phases was not associated with differential outcomes after lithium trial re-attempt.

In this regard, although adjunctive antidepressant treatment for established mood stabilizers has been excluded to be associated with increased risk for treatment-emergent switch (or antidepressant-efficacy) in BD [50], any conclusive statement about the actual clinical predictive value of lifetime antidepressant exposure (especially during phases -I or -III) cannot be made based on the preliminary results from the present study, thus urging for caution and parsimony in prescription. Moreover, rates of reported loss of maintenance due to poor adherence were strikingly high even among those patients originally considered as adherent ones, albeit actually taped off their lithium much faster than controls who still allowed for few more days on average. This finding prompts for vigilance and for delivery of enduring psychoeducation even after achievement of maintenance of BD, striving at enhancing the tolerability and long-term management of lithium side-effect profile. In this regard, it is worth notice that lithium was reported less frequently associated to polypharmacy in BD compared to other medication, whereas polypharmacy is a known contributor for reduced compliance overall [51]. This later evidence further promotes preservation of lithium whenever possible and also earlier prescription of lithium in those BD-I patients who would be considered optimal candidates - indeed, in our sample the average amount of time spent in phase I was much higher among cases vs. controls [52].

According to Kraepelin’s approach, mood disorders are best studied following a longitudinal prospective approach. Yet, in this case, we could rely only on a retrospective approach, which further undermines the validity and generalizability of our preliminary results based on a small sample size (actually precluding any multivariate analysis or control for any factor/covariate, or any direct comparison of random-selected cases vs. controls sharing common features). While this later approach owed to feasibility considerations, and would be preferred to readily assess relatively infrequent outcomes [53], a number of potential biases should nonetheless be considered prior the critical interpretation of the results coming from the present pilot case-control over a cohort study. Among others, major limitations of the present study are due to recall and measurement biases that may have occurred, especially stating the lack of any clinician-assisted retrospective life-chart measure or external validator of the diagnosis (including biomarker genetic or other neurobiological validators), rating of mania/depression/global assessment of BD or objective assessment of lithium adherence other than regular clinical lab testing, and lack of axis-II personality disorder assessment (or demographic record about marital status or education), despite the unquestionable usefulness of graphical charting of mood disorders [54, 55]. Also, the inclusion of a convenience BD-I only sample precluded the opportunity to readily assess the incidence of new cases of lithium refractoriness, or to compute overall rates of lithium-discontinuation. In addition, failure to achieve euthymia upon an average follow-up of 8.1 months actually may represent failure of acute treatment rather than failure to achieve maintenance, further soliciting for more accurate prospective studies. Similarly, since the sample at study included only treatment-seeking outpatients rather than inpatient admitted to a lithium clinic. While this option gathered some preliminary insight about the outpatient setting, a Berkson’s bias (namely the systematic exclusion of most severe cases: indeed most patients self-reported to be substantially stable within phase II trial thus probably representing a less severe subset of BD-I cases) may have likewise limited the generalizability of the present results as the unrepresentativeness of the control cases subset precluded any statistically meaningful conclusive comparison and the lack of BD-II cases would have precluded additional stratification/correction for multiple comparisons of results based on some otherwise clinically suggestive predictors of lithium poor adherence and/or refractoriness rather than optimal response, namely, differential sequence of lifetime mood episodes/predominant polarity, current of lifetime confounding factors (e.g. polypharmacy), presence of mixed/mood-incongruent psychotic features, cyclothymia and/or lifetime rapid-cycling course, family history for BD, pyknic body build, and/or presence of psychotic rather than anxious/obsessive traits [30, 47, 56-64].

Furthermore, the eventual existence of a “latent time period” related to loss of lithium response [27] would likewise deserve additional investigation, as the present study had not enough power to assess this otherwise clinically suggestive issue, otherwise not necessarily pertaining just to overt BD-spectrum disorders [65]. Finally, from an epidemiological and neurobiological standpoint, due to the intrinsic limitation of retrospective assessment it cannot be excluded “reverse causality” of BD-treatment trajectory interplay, as in some cases the disease at issue could cause the outcome itself [66, 67]. Thus, in the absence of ad-hoc, controlled prospective studies carried across varying clinical populations, no reliable causal inference could be made about “timing”, “strength of relationship” (in terms of relative risk), “dose-dependent” relation, or “biological plausibility” about a plethora of pleomorphic phenomena eventually associated to lithium refractoriness in BD. We therefore propose to re-conceptualize lithium discontinuation- “induced” refractoriness (DIR) into a more prudent (yet still tentative) evidence-base sound “post-discontinuation” (lithium) refractoriness”, in line with the findings from a large, prospective naturalistic study by Coryell et al., (1998) which documented no evidence for reduced lithium prophylactic value after recurrence of mania (or depression) at the time of resumed treatment with lithium [68]. In addition, even in the presence of refractoriness, inability to make (or exclude) any conclusive causal nexus between lithium discontinuation and subsequent loss of response would not per se exclude the chance of “pseudo”-refractoriness at least in some instances.

While it must be further emphasized that the power of the present pilot retrospective study was insufficient to reliably detect rare and complex outcomes as loss of lithium response over the time may be, it is nonetheless important to remark the value of lithium, both as a clinical tool and as a research paradigm for a better understanding of “manic-depressive illness”. The present study included BD-I patients who may have been exposed to the SGA class over the past decade or so, while previous (post-hoc) prospective reports could not do so owing to publication bias for the SGA compounds, which have nonetheless been speculated to potentially contribute themselves towards mood stabilization of BD [69]. Yet, further, large-sampled and rigorous, ad-hoc designed, controlled prospective studies are warranted aiming at shedding light about the actual occurrence of “loss of shine of the silver bullet over the time”. Current tentative evidence nonetheless urges for caution in discontinuing lithium as “valued goods should be handled with care”, especially owing to a long-term perspective avoiding too fast withdrawal since this later may otherwise trigger unfavorable epigenetic adaptations [70-72] resulting in neuroprogression of BD and increased risk for the associated much burdensome outcomes of uncontrolled BD, with a special emphasis towards the risk for suicide and other major personal, familial and social issues related to BD progression and failure to ensure maintenance, which would go beyond the single individual to encompass lithium response across generations sharing common endo-phenotypes [73, 74], predicting different trajectories towards lithium response [75-78].

While there is no consensus about the ideal duration of the maintenance treatment of BD, considering the risk for recurrence, established international guidelines recommend to continue as long as possible [79-81]. In some cases (patient’ decision, safety concerns, drug intoxication, planned pregnancy) tapering off of lithium under clinician’ guidance despite achieved maintenance of BD, which should be done gradually in order to minimize the risk of relapse, ideally over three months or longer in case of lithium treatment [82-84] as manic or hypomanic rebound would otherwise occur in up to 79% of the abrupt withdrawal cases [85].

Please refer to Appendix A for additional coding.

Albeit with these later reservations, intrinsic limitations of the present study, and need for the identification of definite biomarkers of BD progression [86], prudence in discontinuing lithium treatment upon achieved maintenance of BD would therefore represent a primary goal both for the clinician, the patient and his/her caregivers striving at enhancing safety beyond the sole efficacy, at least in the absence of more conclusive results coming from the above aimed rigorous, replication studies.