Research report
The empirical redefinition of the psychometric criteria for remission in bipolar disorder

https://doi.org/10.1016/j.jad.2007.06.011Get rights and content

Abstract

Background

Current definitions of remission for mania and bipolar depression are convention-rather than empirically-based, and their clinical salience is unclear, as few studies have attempted to calibrate them against objective clinical criteria. This study aimed to determine equivalence scores on two widely used clinical rating scales, the Young Mania Rating Scale (YMRS) and Montgomery-Åsberg Depression Rating Scale (MADRS), that corresponded with an objective global clinical measure of remission in bipolar disorder patients.

Methods

Data from four pharmacological randomised controlled trials in bipolar I disorder were analysed. Two trials were conducted for bipolar depression (N = 410 and 833), and two for manic or mixed episodes (N = 136 and 110). In this study, a Clinical Global Impression-Bipolar Version (CGI-BP) severity score of 1 (normal, not at all ill) was used as the primary comparative measure of remission. The mean total YMRS and MADRS scores in the mania and depression studies, respectively, that corresponded with a CGI-BP severity score of 1 were determined.

Results

The mean YMRS score that corresponded with a CGI-BP severity score of 1 was < 4 in both trials (2.6 and 3.0, respectively), and the mean corresponding MADRS score was < 5 (4.1 and 4.6, respectively).

Limitations

Utilising a psychometric definition of remission.

Conclusions

This study suggests that a cut-off score of < 5 on the MADRS and < 4 on the YMRS approximates a CGI-BP definition of complete remission. Although lower than conventional cut-off scores, these perhaps better represent clinical reality and patient expectations. In the context of clinical trials, study end-points may be more difficult to reach with lower cut-offs, but the outcomes achieved are more likely to be clinically meaningful.

Introduction

In bipolar disorder, response and remission are pertinent concepts that encapsulate the purpose and utility of any intervention, and the ability to translate research findings into clinical practice often hinges on the meaningfulness of their definitions. Whilst response is a conventional and useful outcome measure in clinical trials, its arbitrary definition, usually set as a 30–50% score reduction on standard psychometric scales for partial response and > 50% score reduction for full response, is difficult to conceptualize in clinical terms. Furthermore, the presence of residual affective symptoms is associated with a greater risk of relapse (Keller et al., 1992) and poorer functional outcomes (Tohen et al., 1990, Perlis et al., 2006, Tohen et al., 2006), thus arguing for remission as a more desirable treatment target.

The text revised version of the Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM IV-TR) defines full remission of bipolar episodes as the absence of significant signs and symptoms for at least two months (APA, 2000). In research, operationalized criteria for remission are largely based on expert clinical consensus and practice norms or convention. For mania, endpoint scores of ≤ 12 on the Young Mania Rating Scale (YMRS) are often used to define remission (Tohen et al., 2000a). For bipolar depression, remission is traditionally defined as scores of ≤ 7 on the Hamilton Depression rating Scale (HAM-D17) (Nierenberg and DeCecco, 2001), or ≤ 12 on the Montgomery-Åsberg Depression Rating Scale (MADRS) (Thase et al., 2006), in accordance with unipolar depression criteria. The inadequacies of these conventional definitions have been suggested by studies in major depression, and lower cut-off scores have been deemed more valid (Zimmerman et al., 2004c). Zimmerman et al. (2004b), in attempting to link objective criteria for remission to cut-off scores on the MADRS in unipolar depression, found that a score of < 4 described an absence of clinically significant features of depression, while a score of < 9 described a softer definition of remission. An attempt to examine MADRS scores in healthy control populations yielded comparable results, with a mean score of 4 (SD 5.8) (Zimmerman et al., 2004a). A higher remission threshold on the MADRS of < 10 was proposed by Hawley et al. (2002), using a different analytical approach that included a CGI score of 2, or borderline ill, as an anchor. More stringent definitions of remission that incorporate multiple symptomatic and global measures of illness have been attempted. In a study of mania (Chengappa et al., 2003), the remission criteria required concurrent scores of ≤ 7 on YMRS, ≤ 7 on HAM-D21 and ≤ 2 on the severity scale of the Clinical Global Impressions-Bipolar Version (CGI-BP).

Even with such refined definitions, it remains unclear whether these arbitrary criteria represent clinical remission with sufficient reliability to serve as acceptable proxy markers for research. A further complexity exists for bipolar disorder research, in that the conventional scales measure a specific affective pole and do not accommodate situations of polarity switch. As the CGI-BP (Spearing et al., 1997), like its parent, the CGI, is a non-symptom-based instrument that provides global measures of disease severity and improvement, it may be a better candidate than symptom-based scales to capture clinical remission. In particular, the score of 1 on its severity scale (CGI-S), defined as “normal, not at all ill”, appears to be a succinct approximation of the essence of clinical remission. A CGI-S score of 2, formally defined as “borderline mentally ill”, is a softer approximation of remission.

The aim of this study was to determine scores on the YMRS and MADRS that corresponded with an objective clinical definition of remission, the CGI-S score of 1, in order to empirically derive clinically meaningful cut-off scores for episodic remission in bipolar disorder.

Section snippets

Methods

This study was based on data generated by four clinical trials of pharmacotherapy in bipolar I disorder, which recruited large samples over multiple sites. Two of these trials were conducted with patients meeting the DSM-IV criteria for major depression using the Structured Clinical Interview for DSM-IV (SCID) (bipolar depression trial A and bipolar depression trial B), and two trials investigated patients in mania or mixed episode, also defined by DSM-IV criteria using the SCID (bipolar mania

Results

The sample sizes and visit structure of the four analysed trials are shown in Table 1. Table 2 shows a categorisation of the pooled CGI-S scores, with corresponding MADRS and YMRS total scores. The mean YMRS score that corresponded with a CGI-BP severity score of 1, a global assessment of “normal, not at all ill”, was below 4 for both mania trials, (2.6 and 3.0 respectively). The mean corresponding MADRS score for both depression trials was below 5 (4.1 and 4.6 respectively). The mean YMRS

Discussion

This study suggests that cut-off values of < 5 on the MADRS total score and < 4 on the YMRS total score are appropriate to define full remission from acute episodes of bipolar disorder. Using a softer definition of remission or remission with mild symptoms, cut-off scores < 7 on the YMRS score and < 10 on the MADRS are supported. These scores are lower than those conventionally established, which may demand that trial findings utilising conventional scales be reinterpreted in a slightly different

Role of funding source

No specific funding was received for this study.

Conflict of interest

Dr Tohen and Ms Wang are employees of Lilly.

Professor Berk has received funding for research from the Stanley Medical Research Institute, Medical Benefits Fund of Australia Limited, The National Health and Medical Research Council, Beyond Blue, and pharmaceutical companies (Bristol-Myers Squibb, Eli Lilly, GlaxoSmithKline, Organon, Novartis, Mayne Pharma). He has served as a consultant to pharmaceutical companies (Astra Zeneca, Bristol Myers Squibb, Eli Lilly, GlaxoSmithKline, Janssen Cilag,

Acknowledgment

We thank Eli Lilly for providing the data on which this study has been performed, and for statistical support.

References (20)

There are more references available in the full text version of this article.

Cited by (74)

View all citing articles on Scopus
View full text