DEPRESSION AND ANXIETY 32:527–538 (2015)
Review THE VALIDITY OF THE MOOD DISORDER QUESTIONNAIRE FOR SCREENING BIPOLAR DISORDER: A META-ANALYSIS Hee Ryung Wang, M.D., Ph.D.,1 Young Sup Woo, M.D., Ph.D.,1 Hyeong Sik Ahn, M.D., Ph.D.,2 Il Min Ahn,2,3 Hyun Jung Kim, M.P.H, Ph.D.,2 ∗ † and Won-Myong Bahk, M.D., Ph.D.1 ∗ †
We conducted a meta-analysis to review the diagnostic accuracy of the Mood Disorder Questionnaire (MDQ) among patients with mood disorders. We used a bivariate random effects model to calculate summary sensitivity and specificity. Twenty-one studies were included. At the standard or modified cutoff value of 7, summary sensitivity was .62 and summary specificity was .85. When we pooled 11 studies including both patients with bipolar disorder (BD) and those with unipolar depression, the summary sensitivity was .76 and summary specificity was .81. However, among the six studies that excluded patients with known BD, the summary sensitivity was significantly reduced to .37 and summary specificity was .88. There were no significant differences on the diagnostic accuracy of the MDQ between studies from Eastern and Western countries after adjusting for various clinical correlates. The overall diagnostic accuracy of the MDQ was relatively good. However, when the MDQ is applied among patients with depression without previous diagnoses of BD, its sensitivity was significantly reduced. This suggests that when the MDQ is applied among this population, its optimal cutoff value should be adjusted to enhance its sensitivity. Depression and Anxiety 32:527–538, C 2015 Wiley Periodicals, Inc. 2015. Key words: bipolar disorder; depression; mood disorders; assessment/diagnosis; measurement/psychometrics
INTRODUCTION
Bipolar disorder (BD) is a serious mental illness as-
1 Department
of Psychiatry, College of Medicine, The Catholic University of Korea, Seoul, Korea 2 Department of Preventive Medicine, College of Medicine, Korea University, Seoul, Korea 3 Department of Literary Arts, Brown University, Providence, Rhode Island † Both
authors contributed equally to this work.
sociated with significant morbidity and mortality; however, the early detection of BD can be challenging. Patients with BD are known to spend a much longer time in depressed phases than in manic or hypomanic phases.[1] Due to greater perceived distress from depressive symptoms, patients with BD have a greater tendency to seek help when depressed rather than when manic or hypomanic.[2] These tendencies may contribute to the misdiagnosis of BD as major depressive disorder (MDD). The misdiagnosis of BD as MDD and subsequent
∗ Correspondence
to: Won-Myong Bahk, Department of Psychiatry, Yeouido St. Mary’s Hospital, College of Medicine, The Catholic University of Korea, 10, 63(yuksam)-ro, Yeongdeungpo-gu, Seoul 150-713, Korea. E-mail: [email protected] Hyun Jung Kim, Department of Preventive Medicine, College of Medicine, Korea University 126-1, Anam Dong, Seong Buk Gu, Seoul 136-705, Korea. E-mail: [email protected]
C 2015 Wiley Periodicals, Inc.
Received for publication 3 January 2015; Revised 23 March 2015; Accepted 28 March 2015 DOI 10.1002/da.22374 Published online 22 May 2015 in Wiley Online Library (wileyonlinelibrary.com).
528
Wang et al.
mistreatment (e.g., antidepressant monotherapy) may lead to detrimental consequences such as mood destabilization, increased suicidality, or more aggravated mood symptoms.[3, 4] Thus, the early detection of BD among patients diagnosed as MDD is very important. Thus far, there have been several instruments developed to screen for BD among patients with depression.[5–8] One of the most commonly used is the Mood Disorder Questionnaire (MDQ) developed by Hirschfeld et al.[8] The MDQ is a self-administered screening instrument for BD composed of three parts. The first part consists of 13 yes-or-no statements to identify previous hypomanic or manic episodes; the second part consists of one yes-or-no question regarding the simultaneous occurrence of symptoms; and the third part comprises one question concerning the influence of the above symptoms. In the original study by Hirschfeld, the authors suggested a standard cutoff value of 7 (i.e., seven or more symptoms in the first part that occurred simultaneously, causing at least moderate impairment) as optimal.[8] According to previous studies, the MDQ seems to have relatively good sensitivity and specificity in detecting BD among psychiatric outpatients. However, it seems to have lower sensitivity in detecting BD in the general population. In addition, the MDQ has demonstrated higher sensitivity in screening for bipolar I disorder, compared to bipolar II disorder or BD not otherwise specified (NOS).[4, 8–14] The MDQ has been translated into many other languages, and the validity of these versions has been investigated among patients with mood disorders. Some studies reported that when the standard cutoff of 7 was used, the sensitivity was too low; thus, they suggested several modified cutoffs of 7 (including ignoring the responses in the second and third parts of the MDQ; ignoring the responses in only the third part of the MDQ; or requiring at least minor impairment in the third part) as an alternative optimal cutoff.[15–20] However, some studies have suggested other optimal cutoff values to compensate for poor screening accuracy caused by applying either the standard or modified cutoff of 7.[9–14, 21] Interestingly, there is a tendency for studies conducted in Eastern countries to demonstrate relatively lower optimal cutoff values than those conducted in Western countries. Thus, various culture-related factors may be related to these different optimal cutoff values.[4, 9, 11, 14] However, thus far, no meta-analyses have investigated the pooled screening accuracy of the MDQ or the appropriateness of the standard or modified cutoff value of 7 as an optimal cutoff. Furthermore, the impacts of various factors on screening accuracy have not been systematically analyzed. Thus, we performed a meta-analysis to systematically investigate the diagnostic accuracy of the MDQ among patients with mood disorders and examine the appropriateness of 7 as a cutoff value. In addition, the impacts of various correlates on the diagnostic accuracy were investigated. Depression and Anxiety
MATERIALS AND METHODS We conducted a comprehensive literature search using multiple databases. We followed the method of the Cochrane Diagnostic Test Accuracy Working Group for this meta-analysis.[22]
DATA AND LITERATURE SOURCES We searched MEDLINE, EMBASE, the Cochrane Library, and KoreaMed on September 10, 2014. In our literature search, there were no restrictions on publication year. We used the following keywords and MeSH terms for the MEDLINE search: BD, screening, sensitivity, specificity, MDQ. Please see the Index for the full search formula (see Supporting Information Table S1). Search strategies were tailored to the other databases. In addition, we manually searched the references of identified review articles.
STUDY SELECTION Two reviewers (HRW and WMB) independently assessed whether the studies met our selection criteria. The decision regarding study inclusion was done in two steps of screening. In the first step, the titles and abstracts were screened. The next step involved screening the full texts. Studies were included in this meta-analysis if (1) they provided data regarding the diagnostic accuracy of the MDQ for screening BD among patients with mood disorders; (2) the Diagnostic and Statistical Manual of Mental Disorders (DSM) or the International Classification of Diseases (ICD) was used as the standard reference diagnostic method; and (3) the studies were published in a peer-reviewed journal. The studies were excluded if (1) patients without mood disorders were included as study participants; (2) the studies were conducted in general population samples; or (3) the studies were written in languages other than English.
DATA EXTRACTION Two reviewers (HRW and WMB) used the predefined data extraction form to independently extract data from each study. Any disagreements between the two reviewers were resolved by discussion. The data extracted from each study were as follows: (1) diagnostic accuracy (true positive, true negative, false positive, false negative, sensitivity, specificity) of the MDQ for screening BD among patients with mood disorders at various cutoff values; (2) the continents in which each study was conducted (Eastern or Western countries); (3) population characteristics; and (4) definitions used for diagnosing BD.
ASSESSMENT OF METHODOLOGICAL QUALITY The quality of each study was assessed by two reviewers (HRW and WMB) using the Quality Assessment of Diagnostic Accuracy Studies2 (QUADAS 2) tool[23] as suggested by the Cochrane Diagnostic Test Accuracy Working Group.[24] Any disagreements were resolved with discussion between two reviewers.
STATISTICAL ANALYSIS We calculated the summary sensitivity, summary specificity, and pooled diagnostic odds ratios (DORs) at various cutoff values. We conducted subgroup analyses to investigate the impact of the population characteristics on the diagnostic accuracy. All of the included studies were classified into two groups: studies conducted in Western countries and studies conducted in Eastern countries. We compared the diagnostic accuracy of the MDQ between Western and Eastern studies using an analysis of the hierarchical summary receiver operating characteristics (hsROC). To adjust for other confounding variables (such as the use of a broadened definition of the
Unclear 7
5
6
Unclear 7
6
Unclear 7
Modified 7
Modified 7
Standard 7
4
Modified 7
Modified 7
Modified 7a Standard 7
Chou[25]
Sasdelli[21]
Lee[12]
Bech[27]
Lin[11]
Haghighi[28]
Benazzi[15]
Cyprien[16]
de Dois[26]
Gan[9]
Gervasoni[17]
Kim[18]
Miller[19]
Standard 7
9
Tafalla[29]
Twiss[30]
Soares[32]
Optimal cutoff value
Study ID
No
No
No
Yes
Yes
No
No
No
No
No
No
No
No
Yes
No
No
The use of a broadened definition of DSM-IV
No
No
No
No
No
No
Yes
No
No
No
No
No
No
Yes
No
No
The use of final diagnosis by re-interviewing 1 year after initial screening
TABLE 1. Characteristics of the included studies
Outpatients with BD or MDD with current mood symptoms Outpatients with BD or MDD suffering from a current MDE Outpatients with BD or MDD
Outpatients with BD or MDD
Remitted outpatients with BD or MDD Inpatients with BD or MDD suffering from a current MDE Outpatients with MDD suffering from a current MDE Inpatients with BD or MDD or outpatients suffering from a current MDE Outpatients with BD or MDD with current mood symptoms Remitted outpatients with MDD
Outpatients with BD or MDD
Diagnosed with unipolar depression at least 3 months prior to the study, antidepressant use for at least 3 months Patients with a current or past diagnosis (preceding 6 months) of MDE Inpatients with BD or MDD or outpatients suffering from a current MDE Outpatients with BD or MDD with remission or partial remission of manic/depressive episode Outpatients with BD or MDD
Inclusion criteria of population
Not excluded
Not excluded
Not excluded
Not excluded
BD excluded
Not excluded
Not excluded
BD excluded
Not excluded
Not excluded
Not excluded
Not excluded
Not excluded
Not excluded
BD excluded
BD excluded
Bipolar disorder, excluded or not
Psychiatric outpatient Psychiatric outpatient Psychiatric outpatient Psychiatric inpatient Psychiatric outpatient Psychiatric inpatient + outpatient Psychiatric outpatient Psychiatric outpatient Psychiatric outpatient Psychiatric outpatient Psychiatric outpatient Psychiatric outpatient
Psychiatric inpatient + outpatient Psychiatric outpatient
Primary care setting
Psychiatric outpatient
Clinical setting
73
671
42
36
52
102
52
76
115
23
61
75
63
32
56
52
MDD (n)
(Continued)
54
232
81
36
59
44
70
11
80
75
102
95
59
81
11
7
Bipolar disorder (n)
Review: The Validity of the MDQ 529
Depression and Anxiety
Depression and Anxiety
2
8
5
Modified 7
Wang[14]
Weber[31]
Yang[10]
Yang[20] No
No
No
No
No
The use of a broadened definition of DSM-IV
No
No
No
No
No
The use of final diagnosis by re-interviewing 1 year after initial screening
Inpatients with MDD or outpatients suffering from a current MDE
Inpatients with BD or MDD or outpatients
Diagnosed with unipolar depression at least 3 months prior to the study, antidepressant use for at least 3 months Outpatients with BD or MDD
Outpatients with MDD
Inclusion criteria of population
BD excluded
Not excluded
Not excluded
BD excluded
BD excluded
Bipolar disorder, excluded or not
Psychiatric outpatient Psychiatric inpatient + outpatient Psychiatric inpatient + outpatient
Psychiatric outpatient Psychiatric outpatient
Clinical setting
1178
134
42
47
190
MDD (n)
309
284
54
6
60
Bipolar disorder (n)
MDD, major depressive disorder; BD, bipolar disorder; MDE, major depressive episode. “Unclear 7” means that there were no comments on whether the cutoff value of 7 is standard or modified. “Modified 7” means ignoring the responses in the second and third parts of the MDQ; ignoring the responses in the third part of the MDQ; or requiring at least minor impairment in the third part of the MDQ. “Standard 7” means seven or more endorsed symptoms in the first part that occurred simultaneously, causing at least moderate impairment. a Modified 7 means ignoring the responses only in the third part of the MDQ.
5
Optimal cutoff value
Waleeprakhon[13]
Study ID
TABLE 1. Continued
530 Wang et al.
Review: The Validity of the MDQ
531
Figure 1. Coupled forest plots and summary ROC plots. (A) Coupled forest plots at the standard or modified cutoff of 7 and the authors’ optimal cutoff. (B) Summary ROC plots of studies when applying the cutoff of 7 and the authors’ optimal cutoff. ROC, receiver operating characteristics; TP, true positive; FP, false positive; TN, true negative, FN, false negative. DSM-IV, the exclusion of patients with known BD, and the use of a standard or modified cutoff value of 7 or other optimal cutoff values), we performed subgroup analyses according to these variables. To identify the diagnostic accuracy according to different cutoff values within the same study, we calculated a linked ROC plot for direct comparison. We used a bivariate random effects model to calculate summary sensitivity, summary specificity, and pooled DORs using STATA. The hsROC curves and linked ROC plots for direct comparison were also calculated using STATA.
RESULTS IDENTIFICATION OF STUDIES
We identified 1,346 articles for possible inclusion. Of these, 1,276 articles were excluded because they did not meet our inclusion criteria at the first step of screening the title and abstract. For the remaining 70 articles, we screened the full manuscript. We excluded 49 of the 70 articles: 40 did not meet the inclusion criteria for study
population, four were poster abstracts, four did not provide information regarding the diagnostic accuracy of the MDQ, and one was a review article. Therefore, 21 studies were included in this meta-analysis. (See Supporting Information Fig. S1 for the study flow diagram.) STUDY CHARACTERISTICS AND PATIENT POPULATIONS
Seven of the studies excluded patients who were previously diagnosed as BD.[13, 14, 18, 20, 21, 25, 26] The remaining 14 studies included both patients with MDD and those with BD.[9–12, 15–17, 19, 27–32] Twenty of the 21 studies provided information regarding the diagnostic accuracy at the standard or modified cutoff value of 7. Of these 20 studies, 12 studies suggested the standard or modified cutoff value of 7 as their optimal cutoff, whereas the remaining eight studies provided diagnostic accuracy at this cutoff value as a reference used for comparison with other optimal cutoff values. Eleven studies were Depression and Anxiety
Wang et al.
532
QUALITY OF THE INCLUDED STUDIES
The quality of the included studies was assessed using QUADAS 2. We assessed both the risk of bias and applicability concerns. Overall, we consider the quality of each study to be heterogeneous. The results are presented in a QUADAS 2 table (Supporting Information Fig. S2). META-ANALYSIS OF THE SCREENING ACCURACY OF THE MDQ FOR BD
When applying the standard or modified cutoff of 7 (when possible, the standard cutoff of 7 took priority over the modified cutoff of 7), the summary sensitivity was .62 (95% CI = 0.51–0.72) and specificity was .85 (95% CI = 0.79–0.89)(Fig. 1A). This result indicated high heterogeneity. When applying the optimal cutoff values suggested by the authors in each study, the summary sensitivity was .78 (95% CI = 0.74–0.81) and summary specificity was .76 (95% CI = 0.71–0.81; pooled DOR = 11.23) (Fig. 1B). The heterogeneity remained, but was lower. META-ANALYSIS OF THE MDQ ACCORDING TO POPULATION CHARACTERISTICS
Figure 1. Continued
To determine the impact of several clinical correlates, we conducted subgroup analyses according to different population characteristics. First, we pooled 11 studies that included both patients with BD and those with MDD without using a broadened definition of the DSMIV for BD. The summary sensitivity was .76 (95% CI = 0.69–0.82) and specificity was .81 (95% CI = 0.75–0.86) at the cutoff of 7 (pooled DOR = 13.89)(Fig. 2A). Second, we pooled nine studies that excluded patients with known BD or used the broadened definition of the DSM-IV for BD. The summary sensitivity became significantly lower reaching .37 (95% CI = 0.29–0.45) but the summary specificity improved slightly to .88 (95% CI = 0.80–0.93; pooled DOR = 4.12)(Fig. 2B). Third, when we pooled the six studies that excluded patients with known BD, the summary sensitivity was .37 (95% CI = 0.22–0.54) and the summary specificity was .88 (95% CI = 0.79–0.94; pooled DOR = 4.46), indicating very high heterogeneity (Fig. 2C). Table 2 presents the summary sensitivity and specificity of the MDQ according to different inclusion/exclusion criteria of the included studies. META-ANALYSIS OF THE MDQ BETWEEN WESTERN AND EASTERN STUDIES
conducted in Western countries[15–17, 19, 21, 26, 27, 29–32] and the remaining 10 studies were performed in Eastern countries.[9–14, 18, 20, 25, 28] All 21 studies were based on the DSM-IV or DSM-IV-TR diagnostic criteria for diagnosing BD. However, three studies used a broadened definition of DSM-IV or DSM-IV-TR.[12, 18, 19, 33] The characteristics of the 21 included studies are presented in Table 1. Depression and Anxiety
To investigate the impact of cross-cultural differences between the West and East, we adjusted for the impact of other clinical correlates (including the use of different cutoff values, population characteristics, and the use of a broadened definition of the DSM-IV for BD). First, we compared the diagnostic accuracy between the six Western studies and one Eastern study that included both patients with BD and those with MDD without using the broadened definition of BD. This
Review: The Validity of the MDQ
533
Figure 2. Summary ROC plots according to various clinical correlates. (A) hsROC curve and summary point with 95% confidence region and 95% prediction region for MDQ in screening of bipolar disorder in 11 studies including patients with bipolar and unipolar depressive disorder, without using the broadened definition, using the standard or modified cutoff value of 7. (B) hsROC curve and summary point of nine studies that excluded known bipolar disorder or that used a broadened definition of bipolar disorder, using the standard or modified cutoff value of 7. (C) hsROC curve and summary point of 6 studies that excluded known bipolar disorder, using the standard or modified cutoff value of 7. (D) The comparison between Eastern and Western studies that included both BD and MDD, using the standard or modified cutoff of 7. ROC, receiver operating characteristics; hsROC, hierarchical summary receiver operating characteristics; MDQ, mood disorder questionnaire; BD, bipolar disorder; MDD, major depressive disorder. Depression and Anxiety
Wang et al.
534
comparison did not reveal any significant difference, because the result for the one Eastern study was within the 95% confidence region of the six Western studies (Fig. 2D). Second, we compared the three Western studies and one Eastern study that included both patients with BD and those with MDD without using the broadened definition. This comparison also did not reveal any significant difference (Fig. 2D). DIRECT COMPARISON OF THE DIAGNOSTIC ACCURACY OF THE MDQ ACCORDING TO DIFFERENT CUTOFF VALUES WITHIN THE SAME STUDY
We compared the diagnostic accuracy of the MDQ according to different cutoff values within the same study. For this, we pooled six studies (including both patients with MDD and those with BD) that provided data for both the standard and modified cutoff of 7. Applying the modified cutoff of 7 improved the sensitivity to .74 (95% CI = 0.67–0.81) compared to .55 (95% CI = 0.34–0.74) at the standard cutoff of 7. The specificity decreased from .86 (95% CI = 0.75–0.93) when using the standard cutoff of 7 to .78 (95% CI = 0.69–0.85) when using the modified cutoff of 7 (Fig. 3A and B). We found two studies (including both patients with BD and those with MDD) that provided data for both the standard cutoff of 7 and other optimal cutoff values. When applying other optimal cutoff values compared to the standard cutoff of 7, the sensitivity improved considerably, however, the specificity was considerably reduced (Fig. 3C and D).
DISCUSSION In this meta-analysis, at the standard or modified cutoff of 7, the summary sensitivity was quite low and the heterogeneity was considerably high. The modified cutoff of 7 was initially suggested as an alternative cutoff to improve the sensitivity of the MDQ; however, in some studies where the modified cutoff of 7 resulted in low sensitivity, the authors compensated by lowering their optimal cutoff values. These trends were more prominent in studies from Eastern countries. In addition, we found that several clinical correlates had significant impacts on the diagnostic accuracy of the MDQ. First, when pooling studies excluding patients with known BD from the screened population, the summary sensitivity became significantly lower. This could be explained as follows: in studies where patients with known BD were initially excluded, the baseline prevalence of BD became lower than in studies that did not exclude patients with known BD. In addition, excluding patients with known BD might lead to a situation where patients who were relatively more difficult to diagnose with BD were more likely to be included in the screened population. These factors might lower the sensitivity of the MDQ. Depression and Anxiety
TABLE 2. The summary sensitivity and specificity of the MDQ according to different inclusion/exclusion criteria of included studies
Applying the standard or modified cutoff of 7 Applying the optimal cutoff values suggested by the studies Including subjects with both MDD and BD without using a broadened definition of the DSM-IV for BD Excluding subjects with known BD or using the broadened definition of the DSM-IV for BD Excluding subjects with known BD
Summary sensitivity
Summary specificity
Number of included studies
0.62 (95% CI = 0.51–0.72)
0.85 (95% CI = 0.79–0.89)
20
0.78 (95% CI = 0.74–0.81)
0.76 (95% CI = 0.71–0.81)
21
0.76 (95% CI = 0.69–0.82)
0.81 (95% CI = 0.75–0.86)
11
0.37 (95% CI = 0.29–0.45)
0.88 (95% CI = 0.80–0.93)
9
0.37 (95% CI = 0.22–0.54)
0.88 (95% CI = 0.79–0.94)
6
MDD, major depressive disorder; BD, bipolar disorder; DSM-IV, the Diagnostic and Statistical Manual of Mental Disorders, 4th Edition; CI, confidence interval.
Second, we found that when studies used a broadened definition of the DSM-IV for BD or when they used a final diagnosis elicited by re-interviewing 1 year later, the summary sensitivity became significantly lower. This may be explained as follows: the use of a broadened definition or final diagnosis 1 year later could both lead to a significant portion of patients who would be diagnosed as MDD in other studies to be diagnosed as BD in these studies. This, in turn, reduced the percentage of patients with MDQ positive in proportion to the total number of patients diagnosed with BD in these studies, thus leading to low sensitivity. In particular, when pooling only the studies that excluded patients with known BD from the screened population, the summary sensitivity was lowered to .37 at the standard or modified cutoff of 7. This suggests that applying the MDQ to screen for BD among populations with depression without previous diagnoses of BD using the standard or modified cutoff of 7 is not appropriate. On the contrary, when the 11 studies that
Review: The Validity of the MDQ
535
Figure 3. Linked ROC plots for direct comparison according to different cutoff values within the same study. (A) Coupled forest plots at the standard and modified cutoff of 7 (B) Linked ROC plots for direct comparison between the standard and modified cutoff of 7. (C) Coupled forest plots at the standard cutoff of 7 and other optimal cutoff values (D) Linked ROC plots for direct comparison between the standard cutoff of 7 and other optimal cutoff values. ROC, receiver operating characteristics.
included both patients with BD and those with MDD in the screened population were pooled, the summary sensitivity was quite satisfactory, at .76, with a standard or modified cutoff of 7. Taken together, these findings suggest that the cutoff values should be adjusted (especially lowered) to elicit better screening accuracy for BD when clinicians use the MDQ among populations
who have depression but who have not been previously diagnosed with BD. In addition, even though we could not calculate summary sensitivity and specificity by pooling only the studies that used a final diagnosis 1 year later due to a dearth of such studies, we found that in the existing studies, the sensitivity of the MDQ at a cutoff of 7 was very low.[9, 12] Depression and Anxiety
536
Wang et al.
Figure 3. Continued
Considering the fact that as time goes by, significant numbers of patients initially identified as having MDD experience a diagnostic conversion to BD,[34, 35] the true sensitivity of the MDQ might be even lower than expected in this meta-analysis. There have been reports that the optimal cutoff seemed to be lower in Eastern countries in comparison to their Western counterparts. However, after adjusting for various clinical correlates, these cross-cultural differences disappeared. This finding is dissimilar to previous observations or expectations that there might be crosscultural differences in response patterns toward screening instruments for mood disorders.[9, 36–38] Considering the findings of our meta-analysis, we expected that the tendency toward lower optimal cutoff values in previous Eastern studies might mainly stem from other factors, Depression and Anxiety
including population characteristics (such as relatively high proportions of bipolar II or NOS by excluding patients with known BD from the screened population), or the use of a broadened definition of BD, rather than cross-cultural differences. However, due to a dearth of studies that excluded patients with known BD, this finding only comes from pooling the studies that included both patients with BD and those with MDD. Thus, we cannot say that the diagnostic accuracy of the MDQ is identical in Eastern and Western countries among populations with depression without a previous diagnosis of BD. Further studies are needed to confirm this issue. In this meta-analysis, we compared the diagnostic accuracy according to different cutoff values within the same study. When moving from the standard cutoff of 7 to the modified cutoff of 7, the sensitivity increased.
Review: The Validity of the MDQ
Interestingly, the degree of increase in sensitivity was much greater in studies excluding patients with known BD than in studies including both patients with BD and those with MDD. This could be explained as follows: the studies initially excluding patients with known BD might include relatively more patients with depression, who were more difficult to diagnose with BD because of either their relatively weaker previous manic or hypomanic symptoms, or because of minor functional impairments. Thus, applying the strict standard cutoff of 7 might not detect these populations as MDQ positive, because they are more likely to respond negatively to the second or third part of the MDQ.[4, 15, 19] However, applying the modified cutoff of 7, on the contrary, by ignoring the responses in the second or third part, could more sensitively detect those populations as MDQ positive. Thus, this could lead to the discrepant findings between studies that included both BD and MDD and studies that excluded known BD. Taken together, these findings suggest that among patients with depression without a previous history of being diagnosed as BD, applying the modified cutoff of 7 (rather than the standard cutoff) can improve the diagnostic accuracy of the MDQ. When moving from the standard cutoff of 7 to other optimal cutoff values, the sensitivity became even higher; however, the specificity was sacrificed. Thus, these results indicate that clinicians should adjust the optimal cutoff value when applying the MDQ, considering both sensitivity and specificity, by taking into account the various population characteristics. There are some limitations in our meta-analysis. First, we excluded studies that were written in other languages besides English. Inclusion of those studies might have affected the results of our meta-analysis. Second, more than half of the included studies were conducted in university hospitals or tertiary care service settings. Thus, the populations included in this meta-analysis may not be representative of those who suffer from depression. Despite these limitations, our meta-analysis has merit: it revealed that various clinical correlates had significant influences on the diagnostic accuracy of the MDQ, and that when adjusting for these clinical correlates, there were no significant differences in diagnostic accuracy between Eastern and Western countries. This meta-analysis also indicated that, especially among patients with depression without previous diagnoses of BD, uniformly applying the standard or modified cutoff of 7 is not recommended; thus, further studies are needed to investigate the optimal cutoff values among these populations. In addition, further studies that investigate the optimal cutoff values of the MDQ incorporating long-term follow up of screened populations are needed, considering that a significant portion of patients initially diagnosed with MDD experience BD conversion during follow-up. Acknowledgments. There is nothing to declare. Conflict of interest. There is nothing to declare.
537
REFERENCE 1. Judd LL, Akiskal HS, Schettler PJ, et al. The long-term natural history of the weekly symptomatic status of bipolar I disorder. Arch Gen Psychiatry 2002;59:530–537. 2. Ghaemi SN, Ko JY, Goodwin FK. “Cade’s disease” and beyond: misdiagnosis, antidepressant use, and a proposed definition for bipolar spectrum disorder. Can J Psychiatry 2002;47:125–134. 3. Ghaemi SN, Boiman EE, Goodwin FK. Diagnosing bipolar disorder and the effect of antidepressants: a naturalistic study. J Clin Psychiatry 2000;61:804–808. 4. Zimmerman M, Galione JN. Screening for bipolar disorder with the Mood Disorders Questionnaire: a review. Harv Rev Psychiatry 2011;19:219–228. 5. Angst J, Adolfsson R, Benazzi F, et al. The HCL-32: towards a self-assessment tool for hypomanic symptoms in outpatients. J Affect Disord 2005;88:217–233. 6. Bowden CL, Singh V, Thompson P, et al. Development of the bipolar inventory of symptoms scale. Acta Psychiatr Scand 2007;116:189–194. 7. Ghaemi SN, Miller CJ, Berv DA, et al. Sensitivity and specificity of a new bipolar spectrum diagnostic scale. J Affect Disord 2005;84:273–277. 8. Hirschfeld RM, Williams JB, Spitzer RL, et al. Development and validation of a screening instrument for bipolar spectrum disorder: the Mood Disorder Questionnaire. Am J Psychiatry 2000;157:1873–1875. 9. Gan Z, Han Z, Li K, et al. Validation of the Chinese version of the “Mood Disorder Questionnaire” for screening bipolar disorder among patients with a current depressive episode. BMC Psychiatry 2012;12:8. 10. Yang HC, Yuan CM, Liu TB, et al. Validity of the Chinese version Mood Disorder Questionnaire (MDQ) and the optimal cutoff screening bipolar disorders. Psychiatry Res. 2011;189:446–450. 11. Lin CJ, Shiah IS, Chu H, et al. Reliability and validity of the Chinese Version of the Mood Disorder Questionnaire. Arch Psychiatr Nurs 2011;25:53–62. 12. Lee D, Cha B, Park CS, et al. Usefulness of the combined application of the Mood Disorder Questionnaire and Bipolar Spectrum Diagnostic Scale in screening for bipolar disorder. Compr Psychiatry 2013;54:334–340. 13. Waleeprakhon P, Ittasakul P, Lotrakul M, et al. Development and validation of a screening instrument for bipolar spectrum disorder: The Mood Disorder Questionnaire Thai version. Neuropsychiatr Dis Treat 2014;10:1497–1502. 14. Wang YT, Yeh TL, Lee IH, et al. Screening for bipolar disorder in medicated patients treated for unipolar depression in a psychiatric outpatient clinic using the Mood Disorder Questionnaire. Int J Psychiatry Clin Pract 2009;13:117–121. 15. Benazzi F. Improving the Mood Disorder Questionnaire to detect bipolar II disorder. Can J Psychiatry 2003;48:770–771. 16. Cyprien F, Guillaume S, Jaussent I, et al. Impact of axis-I comorbidity and suicidal behavior disorders on sensitivity and specificity of the Mood Disorder Questionnaire in complex depressed inpatients. Compr Psychiatry 2014;55:876–882. 17. Gervasoni N, Weber Rouget B, Miguiez M, et al. Performance of the Mood Disorder Questionnaire (MDQ) according to bipolar subtype and symptom severity. Eur Psychiatry 2009;24:341–344. 18. Kim B, Wang HR, Son JI, et al. Bipolarity in depressive patients without histories of diagnosis of bipolar disorder and the use of the Mood Disorder Questionnaire for detecting bipolarity. Compr Psychiatry 2008;49:469–475. 19. Miller CJ, Klugman J, Berv DA, et al. Sensitivity and specificity of the Mood Disorder Questionnaire for detecting bipolar disorder. J Affect Disord 2004;81:167–171. Depression and Anxiety
538
Wang et al.
20. Yang HC, Liu TB, Rong H, et al. Evaluation of Mood Disorder Questionnaire (MDQ) in patients with mood disorders: a multicenter trial across China. PLoS One 2014;9:e91895. 21. Sasdelli A, Lia L, Luciano CC, et al. Screening for Bipolar Disorder Symptoms in Depressed Primary Care Attenders: Comparison between Mood Disorder Questionnaire and Hypomania Checklist (HCL-32). Psychiatry J 2013;2013:548349. 22. Leeflang MM, Deeks JJ, Gatsonis C, et al. Systematic reviews of diagnostic test accuracy. Ann Intern Med 2008;149:889–897. 23. Whiting PF, Rutjes AW, Westwood ME, et al. QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. Ann Intern Med 2011;155:529–536. 24. Deeks JJ. Assessing methodological quality. In: Deeks JJ, Bossuyt PM, Gatsonis C, editors. Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy Version 1.0.0. The Cochrane Collaboration; 2009. Available at: http://srdta.cochrane.org/. 25. Chou CC, Lee IH, Yeh TL, et al. Comparison of the validity of the Chinese versions of the Hypomania Symptom Checklist32 (HCL-32) and Mood Disorder Questionnaire (MDQ) for the detection of bipolar disorder in medicated patients with major depressive disorder. Int J Psychiatry Clin Pract 2012;16:132–137. 26. de Dios C, Ezquiaga E, Garcia A, et al. Usefulness of the Spanish version of the mood disorder questionnaire for screening bipolar disorder in routine clinical practice in outpatients with major depression. Clin Pract Epidemiol Ment Health 2008;4:14. 27. Bech P, Christensen EM, Vinberg M, et al. From items to syndromes in the Hypomania Checklist (HCL-32): psychometric validation and clinical validity analysis. J Affect Disord 2011;132:48– 54. 28. Haghighi M, Bajoghli H, Angst J, et al. The Farsi version of the Hypomania Check-List 32 (HCL-32): applicability and indication of a four-factorial solution. BMC Psychiatry 2011;11:14. 29. Tafalla M, Sanchez-Moreno J, Diez T, Vieta E. Screening for
Depression and Anxiety
30.
31.
32.
33.
34.
35.
36.
37.
38.
bipolar disorder in a Spanish sample of outpatients with current major depressive episode. J Affect Disord 2009;114:299–304. Twiss J, Jones S, Anderson I. Validation of the Mood Disorder Questionnaire for screening for bipolar disorder in a UK sample. J Affect Disord 2008;110:180–184. Weber Rouget B, Gervasoni N, Dubuis V, et al. Screening for bipolar disorders using a French version of the Mood Disorder Questionnaire (MDQ). J Affect Disord 2005;88:103–108. Soares OT, Moreno DH, Moura EC, et al. Reliability and validity of a Brazilian version of the Hypomania Checklist (HCL-32) compared to the Mood Disorder Questionnaire (MDQ). Rev Bras Psiquiatr 2010;32:416–423. Cerimele JM, Chwastiak LA, Dodson S, Katon WJ. The prevalence of bipolar disorder in primary care patients with depression or other psychiatric complaints: a systematic review. Psychosomatics 2013;54:515–524. Angst J, Sellaro R, Stassen HH, Gamma A. Diagnostic conversion from depression to bipolar disorders: results of a longterm prospective study of hospital admissions. J Affect Disord 2005;84:149–157. Gilman SE, Dupuy JM, Perlis RH. Risks for the transition from major depressive disorder to bipolar disorder in the National Epidemiologic Survey on Alcohol and Related Conditions. J Clin Psychiatry 2012;73:829–836. Chang SM, Hahm BJ, Lee JY, et al. Cross-national difference in the prevalence of depression caused by the diagnostic threshold. J Affect Disord 2008;106:159–167. Jang Y, Kim G, Chiriboga D. Acculturation and manifestation of depressive symptoms among Korean-American older adults. Aging Ment Health 2005;9:500–507. Lee JJ, Kim KW, Kim TH, et al. Cross-cultural considerations in administering the center for epidemiologic studies depression scale. Gerontology 2011;57:455–461.
Review THE VALIDITY OF THE MOOD DISORDER QUESTIONNAIRE FOR SCREENING BIPOLAR DISORDER: A META-ANALYSIS Hee Ryung Wang, M.D., Ph.D.,1 Young Sup Woo, M.D., Ph.D.,1 Hyeong Sik Ahn, M.D., Ph.D.,2 Il Min Ahn,2,3 Hyun Jung Kim, M.P.H, Ph.D.,2 ∗ † and Won-Myong Bahk, M.D., Ph.D.1 ∗ †
We conducted a meta-analysis to review the diagnostic accuracy of the Mood Disorder Questionnaire (MDQ) among patients with mood disorders. We used a bivariate random effects model to calculate summary sensitivity and specificity. Twenty-one studies were included. At the standard or modified cutoff value of 7, summary sensitivity was .62 and summary specificity was .85. When we pooled 11 studies including both patients with bipolar disorder (BD) and those with unipolar depression, the summary sensitivity was .76 and summary specificity was .81. However, among the six studies that excluded patients with known BD, the summary sensitivity was significantly reduced to .37 and summary specificity was .88. There were no significant differences on the diagnostic accuracy of the MDQ between studies from Eastern and Western countries after adjusting for various clinical correlates. The overall diagnostic accuracy of the MDQ was relatively good. However, when the MDQ is applied among patients with depression without previous diagnoses of BD, its sensitivity was significantly reduced. This suggests that when the MDQ is applied among this population, its optimal cutoff value should be adjusted to enhance its sensitivity. Depression and Anxiety 32:527–538, C 2015 Wiley Periodicals, Inc. 2015. Key words: bipolar disorder; depression; mood disorders; assessment/diagnosis; measurement/psychometrics
INTRODUCTION
Bipolar disorder (BD) is a serious mental illness as-
1 Department
of Psychiatry, College of Medicine, The Catholic University of Korea, Seoul, Korea 2 Department of Preventive Medicine, College of Medicine, Korea University, Seoul, Korea 3 Department of Literary Arts, Brown University, Providence, Rhode Island † Both
authors contributed equally to this work.
sociated with significant morbidity and mortality; however, the early detection of BD can be challenging. Patients with BD are known to spend a much longer time in depressed phases than in manic or hypomanic phases.[1] Due to greater perceived distress from depressive symptoms, patients with BD have a greater tendency to seek help when depressed rather than when manic or hypomanic.[2] These tendencies may contribute to the misdiagnosis of BD as major depressive disorder (MDD). The misdiagnosis of BD as MDD and subsequent
∗ Correspondence
to: Won-Myong Bahk, Department of Psychiatry, Yeouido St. Mary’s Hospital, College of Medicine, The Catholic University of Korea, 10, 63(yuksam)-ro, Yeongdeungpo-gu, Seoul 150-713, Korea. E-mail: [email protected] Hyun Jung Kim, Department of Preventive Medicine, College of Medicine, Korea University 126-1, Anam Dong, Seong Buk Gu, Seoul 136-705, Korea. E-mail: [email protected]
C 2015 Wiley Periodicals, Inc.
Received for publication 3 January 2015; Revised 23 March 2015; Accepted 28 March 2015 DOI 10.1002/da.22374 Published online 22 May 2015 in Wiley Online Library (wileyonlinelibrary.com).
528
Wang et al.
mistreatment (e.g., antidepressant monotherapy) may lead to detrimental consequences such as mood destabilization, increased suicidality, or more aggravated mood symptoms.[3, 4] Thus, the early detection of BD among patients diagnosed as MDD is very important. Thus far, there have been several instruments developed to screen for BD among patients with depression.[5–8] One of the most commonly used is the Mood Disorder Questionnaire (MDQ) developed by Hirschfeld et al.[8] The MDQ is a self-administered screening instrument for BD composed of three parts. The first part consists of 13 yes-or-no statements to identify previous hypomanic or manic episodes; the second part consists of one yes-or-no question regarding the simultaneous occurrence of symptoms; and the third part comprises one question concerning the influence of the above symptoms. In the original study by Hirschfeld, the authors suggested a standard cutoff value of 7 (i.e., seven or more symptoms in the first part that occurred simultaneously, causing at least moderate impairment) as optimal.[8] According to previous studies, the MDQ seems to have relatively good sensitivity and specificity in detecting BD among psychiatric outpatients. However, it seems to have lower sensitivity in detecting BD in the general population. In addition, the MDQ has demonstrated higher sensitivity in screening for bipolar I disorder, compared to bipolar II disorder or BD not otherwise specified (NOS).[4, 8–14] The MDQ has been translated into many other languages, and the validity of these versions has been investigated among patients with mood disorders. Some studies reported that when the standard cutoff of 7 was used, the sensitivity was too low; thus, they suggested several modified cutoffs of 7 (including ignoring the responses in the second and third parts of the MDQ; ignoring the responses in only the third part of the MDQ; or requiring at least minor impairment in the third part) as an alternative optimal cutoff.[15–20] However, some studies have suggested other optimal cutoff values to compensate for poor screening accuracy caused by applying either the standard or modified cutoff of 7.[9–14, 21] Interestingly, there is a tendency for studies conducted in Eastern countries to demonstrate relatively lower optimal cutoff values than those conducted in Western countries. Thus, various culture-related factors may be related to these different optimal cutoff values.[4, 9, 11, 14] However, thus far, no meta-analyses have investigated the pooled screening accuracy of the MDQ or the appropriateness of the standard or modified cutoff value of 7 as an optimal cutoff. Furthermore, the impacts of various factors on screening accuracy have not been systematically analyzed. Thus, we performed a meta-analysis to systematically investigate the diagnostic accuracy of the MDQ among patients with mood disorders and examine the appropriateness of 7 as a cutoff value. In addition, the impacts of various correlates on the diagnostic accuracy were investigated. Depression and Anxiety
MATERIALS AND METHODS We conducted a comprehensive literature search using multiple databases. We followed the method of the Cochrane Diagnostic Test Accuracy Working Group for this meta-analysis.[22]
DATA AND LITERATURE SOURCES We searched MEDLINE, EMBASE, the Cochrane Library, and KoreaMed on September 10, 2014. In our literature search, there were no restrictions on publication year. We used the following keywords and MeSH terms for the MEDLINE search: BD, screening, sensitivity, specificity, MDQ. Please see the Index for the full search formula (see Supporting Information Table S1). Search strategies were tailored to the other databases. In addition, we manually searched the references of identified review articles.
STUDY SELECTION Two reviewers (HRW and WMB) independently assessed whether the studies met our selection criteria. The decision regarding study inclusion was done in two steps of screening. In the first step, the titles and abstracts were screened. The next step involved screening the full texts. Studies were included in this meta-analysis if (1) they provided data regarding the diagnostic accuracy of the MDQ for screening BD among patients with mood disorders; (2) the Diagnostic and Statistical Manual of Mental Disorders (DSM) or the International Classification of Diseases (ICD) was used as the standard reference diagnostic method; and (3) the studies were published in a peer-reviewed journal. The studies were excluded if (1) patients without mood disorders were included as study participants; (2) the studies were conducted in general population samples; or (3) the studies were written in languages other than English.
DATA EXTRACTION Two reviewers (HRW and WMB) used the predefined data extraction form to independently extract data from each study. Any disagreements between the two reviewers were resolved by discussion. The data extracted from each study were as follows: (1) diagnostic accuracy (true positive, true negative, false positive, false negative, sensitivity, specificity) of the MDQ for screening BD among patients with mood disorders at various cutoff values; (2) the continents in which each study was conducted (Eastern or Western countries); (3) population characteristics; and (4) definitions used for diagnosing BD.
ASSESSMENT OF METHODOLOGICAL QUALITY The quality of each study was assessed by two reviewers (HRW and WMB) using the Quality Assessment of Diagnostic Accuracy Studies2 (QUADAS 2) tool[23] as suggested by the Cochrane Diagnostic Test Accuracy Working Group.[24] Any disagreements were resolved with discussion between two reviewers.
STATISTICAL ANALYSIS We calculated the summary sensitivity, summary specificity, and pooled diagnostic odds ratios (DORs) at various cutoff values. We conducted subgroup analyses to investigate the impact of the population characteristics on the diagnostic accuracy. All of the included studies were classified into two groups: studies conducted in Western countries and studies conducted in Eastern countries. We compared the diagnostic accuracy of the MDQ between Western and Eastern studies using an analysis of the hierarchical summary receiver operating characteristics (hsROC). To adjust for other confounding variables (such as the use of a broadened definition of the
Unclear 7
5
6
Unclear 7
6
Unclear 7
Modified 7
Modified 7
Standard 7
4
Modified 7
Modified 7
Modified 7a Standard 7
Chou[25]
Sasdelli[21]
Lee[12]
Bech[27]
Lin[11]
Haghighi[28]
Benazzi[15]
Cyprien[16]
de Dois[26]
Gan[9]
Gervasoni[17]
Kim[18]
Miller[19]
Standard 7
9
Tafalla[29]
Twiss[30]
Soares[32]
Optimal cutoff value
Study ID
No
No
No
Yes
Yes
No
No
No
No
No
No
No
No
Yes
No
No
The use of a broadened definition of DSM-IV
No
No
No
No
No
No
Yes
No
No
No
No
No
No
Yes
No
No
The use of final diagnosis by re-interviewing 1 year after initial screening
TABLE 1. Characteristics of the included studies
Outpatients with BD or MDD with current mood symptoms Outpatients with BD or MDD suffering from a current MDE Outpatients with BD or MDD
Outpatients with BD or MDD
Remitted outpatients with BD or MDD Inpatients with BD or MDD suffering from a current MDE Outpatients with MDD suffering from a current MDE Inpatients with BD or MDD or outpatients suffering from a current MDE Outpatients with BD or MDD with current mood symptoms Remitted outpatients with MDD
Outpatients with BD or MDD
Diagnosed with unipolar depression at least 3 months prior to the study, antidepressant use for at least 3 months Patients with a current or past diagnosis (preceding 6 months) of MDE Inpatients with BD or MDD or outpatients suffering from a current MDE Outpatients with BD or MDD with remission or partial remission of manic/depressive episode Outpatients with BD or MDD
Inclusion criteria of population
Not excluded
Not excluded
Not excluded
Not excluded
BD excluded
Not excluded
Not excluded
BD excluded
Not excluded
Not excluded
Not excluded
Not excluded
Not excluded
Not excluded
BD excluded
BD excluded
Bipolar disorder, excluded or not
Psychiatric outpatient Psychiatric outpatient Psychiatric outpatient Psychiatric inpatient Psychiatric outpatient Psychiatric inpatient + outpatient Psychiatric outpatient Psychiatric outpatient Psychiatric outpatient Psychiatric outpatient Psychiatric outpatient Psychiatric outpatient
Psychiatric inpatient + outpatient Psychiatric outpatient
Primary care setting
Psychiatric outpatient
Clinical setting
73
671
42
36
52
102
52
76
115
23
61
75
63
32
56
52
MDD (n)
(Continued)
54
232
81
36
59
44
70
11
80
75
102
95
59
81
11
7
Bipolar disorder (n)
Review: The Validity of the MDQ 529
Depression and Anxiety
Depression and Anxiety
2
8
5
Modified 7
Wang[14]
Weber[31]
Yang[10]
Yang[20] No
No
No
No
No
The use of a broadened definition of DSM-IV
No
No
No
No
No
The use of final diagnosis by re-interviewing 1 year after initial screening
Inpatients with MDD or outpatients suffering from a current MDE
Inpatients with BD or MDD or outpatients
Diagnosed with unipolar depression at least 3 months prior to the study, antidepressant use for at least 3 months Outpatients with BD or MDD
Outpatients with MDD
Inclusion criteria of population
BD excluded
Not excluded
Not excluded
BD excluded
BD excluded
Bipolar disorder, excluded or not
Psychiatric outpatient Psychiatric inpatient + outpatient Psychiatric inpatient + outpatient
Psychiatric outpatient Psychiatric outpatient
Clinical setting
1178
134
42
47
190
MDD (n)
309
284
54
6
60
Bipolar disorder (n)
MDD, major depressive disorder; BD, bipolar disorder; MDE, major depressive episode. “Unclear 7” means that there were no comments on whether the cutoff value of 7 is standard or modified. “Modified 7” means ignoring the responses in the second and third parts of the MDQ; ignoring the responses in the third part of the MDQ; or requiring at least minor impairment in the third part of the MDQ. “Standard 7” means seven or more endorsed symptoms in the first part that occurred simultaneously, causing at least moderate impairment. a Modified 7 means ignoring the responses only in the third part of the MDQ.
5
Optimal cutoff value
Waleeprakhon[13]
Study ID
TABLE 1. Continued
530 Wang et al.
Review: The Validity of the MDQ
531
Figure 1. Coupled forest plots and summary ROC plots. (A) Coupled forest plots at the standard or modified cutoff of 7 and the authors’ optimal cutoff. (B) Summary ROC plots of studies when applying the cutoff of 7 and the authors’ optimal cutoff. ROC, receiver operating characteristics; TP, true positive; FP, false positive; TN, true negative, FN, false negative. DSM-IV, the exclusion of patients with known BD, and the use of a standard or modified cutoff value of 7 or other optimal cutoff values), we performed subgroup analyses according to these variables. To identify the diagnostic accuracy according to different cutoff values within the same study, we calculated a linked ROC plot for direct comparison. We used a bivariate random effects model to calculate summary sensitivity, summary specificity, and pooled DORs using STATA. The hsROC curves and linked ROC plots for direct comparison were also calculated using STATA.
RESULTS IDENTIFICATION OF STUDIES
We identified 1,346 articles for possible inclusion. Of these, 1,276 articles were excluded because they did not meet our inclusion criteria at the first step of screening the title and abstract. For the remaining 70 articles, we screened the full manuscript. We excluded 49 of the 70 articles: 40 did not meet the inclusion criteria for study
population, four were poster abstracts, four did not provide information regarding the diagnostic accuracy of the MDQ, and one was a review article. Therefore, 21 studies were included in this meta-analysis. (See Supporting Information Fig. S1 for the study flow diagram.) STUDY CHARACTERISTICS AND PATIENT POPULATIONS
Seven of the studies excluded patients who were previously diagnosed as BD.[13, 14, 18, 20, 21, 25, 26] The remaining 14 studies included both patients with MDD and those with BD.[9–12, 15–17, 19, 27–32] Twenty of the 21 studies provided information regarding the diagnostic accuracy at the standard or modified cutoff value of 7. Of these 20 studies, 12 studies suggested the standard or modified cutoff value of 7 as their optimal cutoff, whereas the remaining eight studies provided diagnostic accuracy at this cutoff value as a reference used for comparison with other optimal cutoff values. Eleven studies were Depression and Anxiety
Wang et al.
532
QUALITY OF THE INCLUDED STUDIES
The quality of the included studies was assessed using QUADAS 2. We assessed both the risk of bias and applicability concerns. Overall, we consider the quality of each study to be heterogeneous. The results are presented in a QUADAS 2 table (Supporting Information Fig. S2). META-ANALYSIS OF THE SCREENING ACCURACY OF THE MDQ FOR BD
When applying the standard or modified cutoff of 7 (when possible, the standard cutoff of 7 took priority over the modified cutoff of 7), the summary sensitivity was .62 (95% CI = 0.51–0.72) and specificity was .85 (95% CI = 0.79–0.89)(Fig. 1A). This result indicated high heterogeneity. When applying the optimal cutoff values suggested by the authors in each study, the summary sensitivity was .78 (95% CI = 0.74–0.81) and summary specificity was .76 (95% CI = 0.71–0.81; pooled DOR = 11.23) (Fig. 1B). The heterogeneity remained, but was lower. META-ANALYSIS OF THE MDQ ACCORDING TO POPULATION CHARACTERISTICS
Figure 1. Continued
To determine the impact of several clinical correlates, we conducted subgroup analyses according to different population characteristics. First, we pooled 11 studies that included both patients with BD and those with MDD without using a broadened definition of the DSMIV for BD. The summary sensitivity was .76 (95% CI = 0.69–0.82) and specificity was .81 (95% CI = 0.75–0.86) at the cutoff of 7 (pooled DOR = 13.89)(Fig. 2A). Second, we pooled nine studies that excluded patients with known BD or used the broadened definition of the DSM-IV for BD. The summary sensitivity became significantly lower reaching .37 (95% CI = 0.29–0.45) but the summary specificity improved slightly to .88 (95% CI = 0.80–0.93; pooled DOR = 4.12)(Fig. 2B). Third, when we pooled the six studies that excluded patients with known BD, the summary sensitivity was .37 (95% CI = 0.22–0.54) and the summary specificity was .88 (95% CI = 0.79–0.94; pooled DOR = 4.46), indicating very high heterogeneity (Fig. 2C). Table 2 presents the summary sensitivity and specificity of the MDQ according to different inclusion/exclusion criteria of the included studies. META-ANALYSIS OF THE MDQ BETWEEN WESTERN AND EASTERN STUDIES
conducted in Western countries[15–17, 19, 21, 26, 27, 29–32] and the remaining 10 studies were performed in Eastern countries.[9–14, 18, 20, 25, 28] All 21 studies were based on the DSM-IV or DSM-IV-TR diagnostic criteria for diagnosing BD. However, three studies used a broadened definition of DSM-IV or DSM-IV-TR.[12, 18, 19, 33] The characteristics of the 21 included studies are presented in Table 1. Depression and Anxiety
To investigate the impact of cross-cultural differences between the West and East, we adjusted for the impact of other clinical correlates (including the use of different cutoff values, population characteristics, and the use of a broadened definition of the DSM-IV for BD). First, we compared the diagnostic accuracy between the six Western studies and one Eastern study that included both patients with BD and those with MDD without using the broadened definition of BD. This
Review: The Validity of the MDQ
533
Figure 2. Summary ROC plots according to various clinical correlates. (A) hsROC curve and summary point with 95% confidence region and 95% prediction region for MDQ in screening of bipolar disorder in 11 studies including patients with bipolar and unipolar depressive disorder, without using the broadened definition, using the standard or modified cutoff value of 7. (B) hsROC curve and summary point of nine studies that excluded known bipolar disorder or that used a broadened definition of bipolar disorder, using the standard or modified cutoff value of 7. (C) hsROC curve and summary point of 6 studies that excluded known bipolar disorder, using the standard or modified cutoff value of 7. (D) The comparison between Eastern and Western studies that included both BD and MDD, using the standard or modified cutoff of 7. ROC, receiver operating characteristics; hsROC, hierarchical summary receiver operating characteristics; MDQ, mood disorder questionnaire; BD, bipolar disorder; MDD, major depressive disorder. Depression and Anxiety
Wang et al.
534
comparison did not reveal any significant difference, because the result for the one Eastern study was within the 95% confidence region of the six Western studies (Fig. 2D). Second, we compared the three Western studies and one Eastern study that included both patients with BD and those with MDD without using the broadened definition. This comparison also did not reveal any significant difference (Fig. 2D). DIRECT COMPARISON OF THE DIAGNOSTIC ACCURACY OF THE MDQ ACCORDING TO DIFFERENT CUTOFF VALUES WITHIN THE SAME STUDY
We compared the diagnostic accuracy of the MDQ according to different cutoff values within the same study. For this, we pooled six studies (including both patients with MDD and those with BD) that provided data for both the standard and modified cutoff of 7. Applying the modified cutoff of 7 improved the sensitivity to .74 (95% CI = 0.67–0.81) compared to .55 (95% CI = 0.34–0.74) at the standard cutoff of 7. The specificity decreased from .86 (95% CI = 0.75–0.93) when using the standard cutoff of 7 to .78 (95% CI = 0.69–0.85) when using the modified cutoff of 7 (Fig. 3A and B). We found two studies (including both patients with BD and those with MDD) that provided data for both the standard cutoff of 7 and other optimal cutoff values. When applying other optimal cutoff values compared to the standard cutoff of 7, the sensitivity improved considerably, however, the specificity was considerably reduced (Fig. 3C and D).
DISCUSSION In this meta-analysis, at the standard or modified cutoff of 7, the summary sensitivity was quite low and the heterogeneity was considerably high. The modified cutoff of 7 was initially suggested as an alternative cutoff to improve the sensitivity of the MDQ; however, in some studies where the modified cutoff of 7 resulted in low sensitivity, the authors compensated by lowering their optimal cutoff values. These trends were more prominent in studies from Eastern countries. In addition, we found that several clinical correlates had significant impacts on the diagnostic accuracy of the MDQ. First, when pooling studies excluding patients with known BD from the screened population, the summary sensitivity became significantly lower. This could be explained as follows: in studies where patients with known BD were initially excluded, the baseline prevalence of BD became lower than in studies that did not exclude patients with known BD. In addition, excluding patients with known BD might lead to a situation where patients who were relatively more difficult to diagnose with BD were more likely to be included in the screened population. These factors might lower the sensitivity of the MDQ. Depression and Anxiety
TABLE 2. The summary sensitivity and specificity of the MDQ according to different inclusion/exclusion criteria of included studies
Applying the standard or modified cutoff of 7 Applying the optimal cutoff values suggested by the studies Including subjects with both MDD and BD without using a broadened definition of the DSM-IV for BD Excluding subjects with known BD or using the broadened definition of the DSM-IV for BD Excluding subjects with known BD
Summary sensitivity
Summary specificity
Number of included studies
0.62 (95% CI = 0.51–0.72)
0.85 (95% CI = 0.79–0.89)
20
0.78 (95% CI = 0.74–0.81)
0.76 (95% CI = 0.71–0.81)
21
0.76 (95% CI = 0.69–0.82)
0.81 (95% CI = 0.75–0.86)
11
0.37 (95% CI = 0.29–0.45)
0.88 (95% CI = 0.80–0.93)
9
0.37 (95% CI = 0.22–0.54)
0.88 (95% CI = 0.79–0.94)
6
MDD, major depressive disorder; BD, bipolar disorder; DSM-IV, the Diagnostic and Statistical Manual of Mental Disorders, 4th Edition; CI, confidence interval.
Second, we found that when studies used a broadened definition of the DSM-IV for BD or when they used a final diagnosis elicited by re-interviewing 1 year later, the summary sensitivity became significantly lower. This may be explained as follows: the use of a broadened definition or final diagnosis 1 year later could both lead to a significant portion of patients who would be diagnosed as MDD in other studies to be diagnosed as BD in these studies. This, in turn, reduced the percentage of patients with MDQ positive in proportion to the total number of patients diagnosed with BD in these studies, thus leading to low sensitivity. In particular, when pooling only the studies that excluded patients with known BD from the screened population, the summary sensitivity was lowered to .37 at the standard or modified cutoff of 7. This suggests that applying the MDQ to screen for BD among populations with depression without previous diagnoses of BD using the standard or modified cutoff of 7 is not appropriate. On the contrary, when the 11 studies that
Review: The Validity of the MDQ
535
Figure 3. Linked ROC plots for direct comparison according to different cutoff values within the same study. (A) Coupled forest plots at the standard and modified cutoff of 7 (B) Linked ROC plots for direct comparison between the standard and modified cutoff of 7. (C) Coupled forest plots at the standard cutoff of 7 and other optimal cutoff values (D) Linked ROC plots for direct comparison between the standard cutoff of 7 and other optimal cutoff values. ROC, receiver operating characteristics.
included both patients with BD and those with MDD in the screened population were pooled, the summary sensitivity was quite satisfactory, at .76, with a standard or modified cutoff of 7. Taken together, these findings suggest that the cutoff values should be adjusted (especially lowered) to elicit better screening accuracy for BD when clinicians use the MDQ among populations
who have depression but who have not been previously diagnosed with BD. In addition, even though we could not calculate summary sensitivity and specificity by pooling only the studies that used a final diagnosis 1 year later due to a dearth of such studies, we found that in the existing studies, the sensitivity of the MDQ at a cutoff of 7 was very low.[9, 12] Depression and Anxiety
536
Wang et al.
Figure 3. Continued
Considering the fact that as time goes by, significant numbers of patients initially identified as having MDD experience a diagnostic conversion to BD,[34, 35] the true sensitivity of the MDQ might be even lower than expected in this meta-analysis. There have been reports that the optimal cutoff seemed to be lower in Eastern countries in comparison to their Western counterparts. However, after adjusting for various clinical correlates, these cross-cultural differences disappeared. This finding is dissimilar to previous observations or expectations that there might be crosscultural differences in response patterns toward screening instruments for mood disorders.[9, 36–38] Considering the findings of our meta-analysis, we expected that the tendency toward lower optimal cutoff values in previous Eastern studies might mainly stem from other factors, Depression and Anxiety
including population characteristics (such as relatively high proportions of bipolar II or NOS by excluding patients with known BD from the screened population), or the use of a broadened definition of BD, rather than cross-cultural differences. However, due to a dearth of studies that excluded patients with known BD, this finding only comes from pooling the studies that included both patients with BD and those with MDD. Thus, we cannot say that the diagnostic accuracy of the MDQ is identical in Eastern and Western countries among populations with depression without a previous diagnosis of BD. Further studies are needed to confirm this issue. In this meta-analysis, we compared the diagnostic accuracy according to different cutoff values within the same study. When moving from the standard cutoff of 7 to the modified cutoff of 7, the sensitivity increased.
Review: The Validity of the MDQ
Interestingly, the degree of increase in sensitivity was much greater in studies excluding patients with known BD than in studies including both patients with BD and those with MDD. This could be explained as follows: the studies initially excluding patients with known BD might include relatively more patients with depression, who were more difficult to diagnose with BD because of either their relatively weaker previous manic or hypomanic symptoms, or because of minor functional impairments. Thus, applying the strict standard cutoff of 7 might not detect these populations as MDQ positive, because they are more likely to respond negatively to the second or third part of the MDQ.[4, 15, 19] However, applying the modified cutoff of 7, on the contrary, by ignoring the responses in the second or third part, could more sensitively detect those populations as MDQ positive. Thus, this could lead to the discrepant findings between studies that included both BD and MDD and studies that excluded known BD. Taken together, these findings suggest that among patients with depression without a previous history of being diagnosed as BD, applying the modified cutoff of 7 (rather than the standard cutoff) can improve the diagnostic accuracy of the MDQ. When moving from the standard cutoff of 7 to other optimal cutoff values, the sensitivity became even higher; however, the specificity was sacrificed. Thus, these results indicate that clinicians should adjust the optimal cutoff value when applying the MDQ, considering both sensitivity and specificity, by taking into account the various population characteristics. There are some limitations in our meta-analysis. First, we excluded studies that were written in other languages besides English. Inclusion of those studies might have affected the results of our meta-analysis. Second, more than half of the included studies were conducted in university hospitals or tertiary care service settings. Thus, the populations included in this meta-analysis may not be representative of those who suffer from depression. Despite these limitations, our meta-analysis has merit: it revealed that various clinical correlates had significant influences on the diagnostic accuracy of the MDQ, and that when adjusting for these clinical correlates, there were no significant differences in diagnostic accuracy between Eastern and Western countries. This meta-analysis also indicated that, especially among patients with depression without previous diagnoses of BD, uniformly applying the standard or modified cutoff of 7 is not recommended; thus, further studies are needed to investigate the optimal cutoff values among these populations. In addition, further studies that investigate the optimal cutoff values of the MDQ incorporating long-term follow up of screened populations are needed, considering that a significant portion of patients initially diagnosed with MDD experience BD conversion during follow-up. Acknowledgments. There is nothing to declare. Conflict of interest. There is nothing to declare.
537
REFERENCE 1. Judd LL, Akiskal HS, Schettler PJ, et al. The long-term natural history of the weekly symptomatic status of bipolar I disorder. Arch Gen Psychiatry 2002;59:530–537. 2. Ghaemi SN, Ko JY, Goodwin FK. “Cade’s disease” and beyond: misdiagnosis, antidepressant use, and a proposed definition for bipolar spectrum disorder. Can J Psychiatry 2002;47:125–134. 3. Ghaemi SN, Boiman EE, Goodwin FK. Diagnosing bipolar disorder and the effect of antidepressants: a naturalistic study. J Clin Psychiatry 2000;61:804–808. 4. Zimmerman M, Galione JN. Screening for bipolar disorder with the Mood Disorders Questionnaire: a review. Harv Rev Psychiatry 2011;19:219–228. 5. Angst J, Adolfsson R, Benazzi F, et al. The HCL-32: towards a self-assessment tool for hypomanic symptoms in outpatients. J Affect Disord 2005;88:217–233. 6. Bowden CL, Singh V, Thompson P, et al. Development of the bipolar inventory of symptoms scale. Acta Psychiatr Scand 2007;116:189–194. 7. Ghaemi SN, Miller CJ, Berv DA, et al. Sensitivity and specificity of a new bipolar spectrum diagnostic scale. J Affect Disord 2005;84:273–277. 8. Hirschfeld RM, Williams JB, Spitzer RL, et al. Development and validation of a screening instrument for bipolar spectrum disorder: the Mood Disorder Questionnaire. Am J Psychiatry 2000;157:1873–1875. 9. Gan Z, Han Z, Li K, et al. Validation of the Chinese version of the “Mood Disorder Questionnaire” for screening bipolar disorder among patients with a current depressive episode. BMC Psychiatry 2012;12:8. 10. Yang HC, Yuan CM, Liu TB, et al. Validity of the Chinese version Mood Disorder Questionnaire (MDQ) and the optimal cutoff screening bipolar disorders. Psychiatry Res. 2011;189:446–450. 11. Lin CJ, Shiah IS, Chu H, et al. Reliability and validity of the Chinese Version of the Mood Disorder Questionnaire. Arch Psychiatr Nurs 2011;25:53–62. 12. Lee D, Cha B, Park CS, et al. Usefulness of the combined application of the Mood Disorder Questionnaire and Bipolar Spectrum Diagnostic Scale in screening for bipolar disorder. Compr Psychiatry 2013;54:334–340. 13. Waleeprakhon P, Ittasakul P, Lotrakul M, et al. Development and validation of a screening instrument for bipolar spectrum disorder: The Mood Disorder Questionnaire Thai version. Neuropsychiatr Dis Treat 2014;10:1497–1502. 14. Wang YT, Yeh TL, Lee IH, et al. Screening for bipolar disorder in medicated patients treated for unipolar depression in a psychiatric outpatient clinic using the Mood Disorder Questionnaire. Int J Psychiatry Clin Pract 2009;13:117–121. 15. Benazzi F. Improving the Mood Disorder Questionnaire to detect bipolar II disorder. Can J Psychiatry 2003;48:770–771. 16. Cyprien F, Guillaume S, Jaussent I, et al. Impact of axis-I comorbidity and suicidal behavior disorders on sensitivity and specificity of the Mood Disorder Questionnaire in complex depressed inpatients. Compr Psychiatry 2014;55:876–882. 17. Gervasoni N, Weber Rouget B, Miguiez M, et al. Performance of the Mood Disorder Questionnaire (MDQ) according to bipolar subtype and symptom severity. Eur Psychiatry 2009;24:341–344. 18. Kim B, Wang HR, Son JI, et al. Bipolarity in depressive patients without histories of diagnosis of bipolar disorder and the use of the Mood Disorder Questionnaire for detecting bipolarity. Compr Psychiatry 2008;49:469–475. 19. Miller CJ, Klugman J, Berv DA, et al. Sensitivity and specificity of the Mood Disorder Questionnaire for detecting bipolar disorder. J Affect Disord 2004;81:167–171. Depression and Anxiety
538
Wang et al.
20. Yang HC, Liu TB, Rong H, et al. Evaluation of Mood Disorder Questionnaire (MDQ) in patients with mood disorders: a multicenter trial across China. PLoS One 2014;9:e91895. 21. Sasdelli A, Lia L, Luciano CC, et al. Screening for Bipolar Disorder Symptoms in Depressed Primary Care Attenders: Comparison between Mood Disorder Questionnaire and Hypomania Checklist (HCL-32). Psychiatry J 2013;2013:548349. 22. Leeflang MM, Deeks JJ, Gatsonis C, et al. Systematic reviews of diagnostic test accuracy. Ann Intern Med 2008;149:889–897. 23. Whiting PF, Rutjes AW, Westwood ME, et al. QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. Ann Intern Med 2011;155:529–536. 24. Deeks JJ. Assessing methodological quality. In: Deeks JJ, Bossuyt PM, Gatsonis C, editors. Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy Version 1.0.0. The Cochrane Collaboration; 2009. Available at: http://srdta.cochrane.org/. 25. Chou CC, Lee IH, Yeh TL, et al. Comparison of the validity of the Chinese versions of the Hypomania Symptom Checklist32 (HCL-32) and Mood Disorder Questionnaire (MDQ) for the detection of bipolar disorder in medicated patients with major depressive disorder. Int J Psychiatry Clin Pract 2012;16:132–137. 26. de Dios C, Ezquiaga E, Garcia A, et al. Usefulness of the Spanish version of the mood disorder questionnaire for screening bipolar disorder in routine clinical practice in outpatients with major depression. Clin Pract Epidemiol Ment Health 2008;4:14. 27. Bech P, Christensen EM, Vinberg M, et al. From items to syndromes in the Hypomania Checklist (HCL-32): psychometric validation and clinical validity analysis. J Affect Disord 2011;132:48– 54. 28. Haghighi M, Bajoghli H, Angst J, et al. The Farsi version of the Hypomania Check-List 32 (HCL-32): applicability and indication of a four-factorial solution. BMC Psychiatry 2011;11:14. 29. Tafalla M, Sanchez-Moreno J, Diez T, Vieta E. Screening for
Depression and Anxiety
30.
31.
32.
33.
34.
35.
36.
37.
38.
bipolar disorder in a Spanish sample of outpatients with current major depressive episode. J Affect Disord 2009;114:299–304. Twiss J, Jones S, Anderson I. Validation of the Mood Disorder Questionnaire for screening for bipolar disorder in a UK sample. J Affect Disord 2008;110:180–184. Weber Rouget B, Gervasoni N, Dubuis V, et al. Screening for bipolar disorders using a French version of the Mood Disorder Questionnaire (MDQ). J Affect Disord 2005;88:103–108. Soares OT, Moreno DH, Moura EC, et al. Reliability and validity of a Brazilian version of the Hypomania Checklist (HCL-32) compared to the Mood Disorder Questionnaire (MDQ). Rev Bras Psiquiatr 2010;32:416–423. Cerimele JM, Chwastiak LA, Dodson S, Katon WJ. The prevalence of bipolar disorder in primary care patients with depression or other psychiatric complaints: a systematic review. Psychosomatics 2013;54:515–524. Angst J, Sellaro R, Stassen HH, Gamma A. Diagnostic conversion from depression to bipolar disorders: results of a longterm prospective study of hospital admissions. J Affect Disord 2005;84:149–157. Gilman SE, Dupuy JM, Perlis RH. Risks for the transition from major depressive disorder to bipolar disorder in the National Epidemiologic Survey on Alcohol and Related Conditions. J Clin Psychiatry 2012;73:829–836. Chang SM, Hahm BJ, Lee JY, et al. Cross-national difference in the prevalence of depression caused by the diagnostic threshold. J Affect Disord 2008;106:159–167. Jang Y, Kim G, Chiriboga D. Acculturation and manifestation of depressive symptoms among Korean-American older adults. Aging Ment Health 2005;9:500–507. Lee JJ, Kim KW, Kim TH, et al. Cross-cultural considerations in administering the center for epidemiologic studies depression scale. Gerontology 2011;57:455–461.
