Prevalence and treatment of panic disorder in bipolar disorder: systematic review and meta-analysis

Abstract

Question

Recent data suggest that anxiety disorders are as often comorbid with bipolar disorder (BD) as with unipolar depression. The literature on panic disorder (PD) comorbid with BD has been systematically reviewed and subject to meta-analysis.

Study selection and analysis

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines were thoroughly followed for literature search, selection and reporting of available evidence. The variance-stabilising Freeman-Tukey double arcsine transformation was used in the meta-analysis of prevalence estimates. Both fixed-effect and random-effects models with inverse variance method were applied to estimate summary effects for all combined studies. Heterogeneity was assessed and measured with Cochran’s Q and I² statistics.

Findings

Overall, 15 studies (n=3391) on cross-sectional prevalence and 25 independent lifetime studies (n=8226) were used to calculate pooled estimates. The overall random-effects point prevalence of PD in patients with BD, after exclusion of one potential outlier study, was 13.0% (95% CI 7.0% to 20.3%), and the overall random-effects lifetime estimate, after exclusion of one potential outlier study, was 15.5% (95% CI 11.6% to 19.9%). There were no differences in rates between BD-I and BD-II. Significant heterogeneity (I² >95%) was found in both estimates.

Conclusions

Estimates that can be drawn from published studies indicate that the prevalence of PD in patients with BD is higher than the prevalence in the general population. Comorbid PD is reportedly associated with increased risk of suicidal acts and a more severe course. There is no clear indication on how to treat comorbid PD in BD. Findings from the current meta-analysis confirm the highly prevalent comorbidity of PD with BD, implicating that in patients with BD, PD might run a more chronic course.

Background

Patients with bipolar disorder (BD) are exposed to psychiatric comorbidity, with longitudinal rates that can be higher than 50% and may reach even 70%. Psychiatric comorbidity is one of the major reasons BD is often as severe as schizophrenia.¹

Psychiatric comorbidity in BD goes often undetected and undertreated in the clinical setting.² This may depend on clinicians’ inclination to use a single, comprehensive primary diagnosis to deal with the patients, often neglecting the residuals symptomatology. Some symptoms—such as anxiety—may also hide within the multifaceted landscape of BD. Residual untreated symptoms are not unusual in BD with psychiatric comorbidity, and sometimes treatments aimed at the ‘primary’ condition might negatively impact on the course of the comorbid condition, worsening its course.

In patients with BD, anxiety is extremely common, and it can be expressed both as an isolated symptom and as a full-blown syndrome. Kraepelin included anxiety in the core features of BD, also considering it a typical feature of the different clinical subtypes of the major condition. However, it was not until the latest decades that the interest in the manifestations of anxiety in the course of BD peaked in the published literature.

In general, comorbid anxiety is related to worse outcome, may affect recovery, leading to longer time from index mood episode to full remission of symptoms, particularly in depression, may favour earlier relapse, and associates to lower quality of life.^{3 4} Its treatment is problematic.^5–8

Age at onset in patients with comorbid panic disorder (PD) was reported to be younger than in those without comorbidity.⁹ More depressive episodes, and possibly higher risk of suicide and suicide attempt, also were reported.^{10 11} Indeed, a negative impact of comorbidity of PD with the course and outcome of BD has been described.

Studies reported a wide variability of the cross-sectional prevalence of PD in BD, ranging from 2.3% to 62.5%, while the longitudinal prevalence ranged from 2.9% to 56.5%.^12–15

Overall, data on cross-sectional and longitudinal prevalence suggest that patients with BD share with patients with schizophrenia, or unipolar an equivalent rate of PD, which, nevertheless, is several times higher than the rate expected for the general population. However, detailed analysis is lacking. Precise information on comorbidity of PD with BD may serve the purpose of targeting those conditions that are known to affect treatment response and recovery, and that may increase the risk of suicidality and the chance of developing a substance use disorder. Moreover, the treatment of comorbid PD may reveal challenging, both because of the increased risk of adverse effects and the greater chance of medication-induced mood switch.

Objective

This study set out to systematically review the literature about BD comorbid with PD BD. Retrieved data were subject to meta-analysis to extract both cross-sectional (point) and longitudinal rates of comorbidity.

Study selection and analysis

The recommendations of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement were followed in performing the review.

Data on the comorbidity of PD with BD were extracted from published literature, which were retrieved from PubMed/MEDLINE, from inception until 1 August 2017, on the basis of a search code that was developed by two authors (KNF, JV).

The following terms were used to scan PubMed database: ‘Bipolar’ (all fields) OR ‘manic’ (all fields) OR ‘mania’ (all fields) OR ‘manic depression’ (all fields) OR ‘manic depressive’ (all fields) AND ‘panic’ (all fields).

Online details the combination of search terms that were used in the PubMed/MEDLINE interrogation, as well as all other search details.

Only studies in the English language were included. The search was enriched by a thorough scan of the reference lists of relevant books and reviews.^13–17

A first reviewer (KNF) conducted a screen of titles and abstracts of the extracted list of references for inclusion, with a validation check by a second reviewer (JV). In case of uncertainty on inclusion, the full-text article was accessed. Discrepancies were solved through discussion, until the achievement of a consensus.

Criteria for study selection

• Studies that included primary data concerning the comorbidity of PD in adult (over 18 years old) patients with BD: essentially, the number of patients with confirmed diagnosis of BD and with a comorbid diagnosis of PD.

• Studies published in the English language.

Data abstraction and quality assessment

A previously pilot-tested standardised coding system was used by two authors (JV, KNF) for data extraction. The following information was derived from the articles: authors’ name, publication year, location, sample size, criteria for diagnosis, procedure for diagnosis (either clinical decision or diagnosis based on standardised or semistandardised interview), number of cases with BD, number of cases with PD and number of cases with any other diagnostic group when used as comparison. For each study, one reviewer (KNF) abstracted the relevant data and a second reviewer (JV) verified the extraction for completeness and accuracy. Discrepancies in extraction/scoring were solved through discussion.

Findings

Eventually, 16 studies with cross-sectional data and 26 with longitudinal data were included in the analysis. Overall, nine studies with data on BD-I were entered in the related sensitivity analysis, while four studies with data on BD-II were entered in the sensitivity analysis on BD-II.

The PRISMA flow chart is shown in the online. The detailed results are also shown in the online supplementary appendix. The included studies with cross-sectional estimates are reported in table 1, while the details on included studies with lifetime estimates are reported in table 2.

Table 1.

Characteristics of studies included in the meta-analysis concerning the cross-sectional prevalence of panic disorder in patients with bipolar disorder

Study	Location	Diagnostic group	Criteria for diagnosis	Procedure for diagnosis	n	% Male	Age	Prevalence, n (%)	Comments
Bellani et al (2012)	University of Texas Health Science Center at San Antonio, USA	BD	DSM-IV	SCID	205	60 (29.3%)	36.6±11.5	60 (29.3)	Rates of depressive episodes are similar in the two study groups.
		MDD			105	31 (29.5%)	38.0±13.1	10 (9.8)
Boylan et al (2004)	McMaster Regional Mood Disorders Program (Hamilton, Ontario, Canada)	BD	DSM-IV	SCID	138	31.9%	~41	37 (26.8)	Outpatient sample, including 70.3% BD-I and 29.7% rapid cycling.
Ciapparelli et al (2007)	Department of Psychiatry, Neurobiology, Pharmacology and Biotechnology, University of Pisa, Italy	Schiz	DSM-IV	SCID	23	19 (82.6%)	36.5±8.8	24 (24.5)	BD with psychotic features.
		Schizoaff			19	14 (73.7%)	34.9±10.6
		BD			56	46 (83.6%)	35.8±12.1
Coryell et al (2009)	Multisite  (USA)	BD-I (mania)	DSM-IV	SCID	92	63	36.5±13.3	1 (1.1)
		BD-I (depression)			168	59	36.6±14.4	9 (5.4)
		BD-I (cycling)			167	75	35.9±11.9	7 (4.2)
Dilsaver and Chen (2003)	Harris County Psychiatric Hospital, Houston, Texas, USA	BD-I (mania)	DSM-III-R	SCID	19	47.4%	30.9±8.8	1 (4.0)	In the sample of patients with pure mania, >90% had psychotic features. In the sample with depressive mania, >90% had psychotic features.
		BD-I (depressive mania)			25	56.0%	34.7±8.9	16 (84.2)
Dilsaver et al (1997)	Harris County Psychiatric Center, University of Texas, Houston, USA	Bipolar depression	DSM-III-R	SCID	53			33 (62.3)
		Pure mania			32			1 (2.3)
		Depressive mania			44			20 (62.5)
Okan Ibiloglu and Caykoylu (2011)	Psychiatry Outpatient Clinics, Ataturk Training and Teaching Hospital, Ankara, Turkey	BD-I	DSM-IV	SCID	50	17 (34.0%)	37.8±9.51	30 (60.0)
		BD-II			46	12 (26.1%)		22 (47.8)
Pini et al (2003)	Department of Psychiatry, Neurobiology, Pharmacology and Biotechnology, University of Pisa, Italy	BD-I	DSM-III-R	SCID	151	43.7%		35 (23.2)
Strakowski et al (1992)	Psychotic Disorders Program, McLean Hospital and the Consolidated Department of Psychiatry, Harvard University Medical School	BD-I (manic or mixed)	DSM-III-R	SCID	41	39.0%	29.5±12.7 (manic)	2 (4.9)	Sample of inpatients with first episode of mania or mixed status.
							40.4±11.7 (mixed)
Vieta et al (2001)	BDs Programme of the Hospital Clinic and University of Barcelona	BD-I	DSM-III-R	SCID	129	14 (35%)	39.7±10.96 (with psychiatric comorbidity)	3 (2.3)	Sample of patients in remission, all with BD-I.
						39 (39%)	41.51±15.21 (without)
Cosoff and Hafner (1998)	Adelaide, Australia	BD	DSM-III-R	SCID	20	60%	34.8±10 (men)	3 (15.0)
		Schiz			60		34.9±9.6 (women)	3 (5.0)
		Schizoaff			20
Dell’Osso et al (2011)	University Department of Psychiatry of Milan, Italy	BD	DSM-IV	SCID	508	224 (44.1%)	>40	32 (6.3)	56.7% BD-I, 76.1% without any substance or alcohol abuse.
Mantere et al (2006)	Mood Disorders Research Unit, NPHI, Helsinki, Finland	BD	DSM-IV	SCID	191 (90: BD-I; 101: BD-II)	47.1%	37.7±12.2	46 (24.1)	Acute-phase BD, 47.1% BD-I, inpatients or outpatients. 32.5% with rapid cycling. 16.2% with psychotic symptoms.
		MDD			269	26.8%	NA	45 (16.7)
McElroy et al (2001)	Multisite (USA and the Netherlands)	BD	DSM-IV	SCID	288	44.0%	42.8±11.3	27 (9.4)	STANLEY  Foundation data.
Simon et al (2004)	Multisite (USA)	BD-I	DSM-IV	SCID	360	40.6%	41.7±12.8	33 (9.2)	First 500 patients of STEP-BD.
		BD-II			115			5 (4.4)
Otto et al (2006)	Multisite (USA)	BD	DSM-IV	SCID	918	41.0%	40.6±12.7	78 (8.5)	First 1000 patients of STEP-BD. >75% BD-I, 50% in recovery, ~25% depressed.
Tamam and Ozpoyraz (2002)	Cukurova University, Adana, Turkey	BD-I	DSM-IV	SCID	70	29 (41.6%)	33.4±10.3	4 (5.7)	Included only patients with BD-I.
Zutshi et al (2006)	National Institute of Mental Health and Neurosciences, Bangalore, India	BD	DSM-IV	SCID	80	57 (71.3%)	30.06±7.77	4 (5.0)	Patients with BD in remission; the controls were relatives of neurological patients.
		Controls			50	38 (76.0%)	31.44±7.85	0 (0)

BD, bipolar disorder; DSM, Diagnostic Statistic Manual; MDD, major depressive disorder; Schiz, schizophrenia; NA, not available; Schizoaff, schizo-affective disorder; SCID, Structured Clinical Interview for DSM; STEP-BD, Systematic Treatment Enhancement Program for BD.

Table 2.

Characteristics of studies included in meta-analysis concerning the lifetime prevalence of panic disorder in patients with bipolar disorder (see online

Study	Location	Diagnostic group	Criteria for diagnosis	Procedure for diagnosis	n	Age	Male (%)	Prevalence, n (%)	Comments
Altshuler et al (2010)	Stanley Foundation Bipolar Treatment Outcome Network	BD-I BD-II	DSM-IV	SCID	572 139	Male: 43.5±12.4 Female: 40.7±10.9	34 (4.78%)	96 (13.5)	>80% BD-I, STANLEY  Foundation data
Azorin et al (2009)	Sainte Marguerite Hospital, University of Marseilles	BD-I	DSM-IV	SCID	1090	43±14		56 (5.1)	Hospitalised patients with acute mania
Chen and Dilsaver (1995); Robins and Regier (1991)	Harris County Psychiatric Centre (HCPC), University of Texas, Houston	BD-I	DSM-III	SCID	168		9 (17.3%)	35 (20.8)	Epidemiological catchment area
		MDD			557		17 (8.3%)	56 (10.0)
		General population			17 143			138 (0.8)
Coryell et al (2009)	Massachusetts General Hospital in Boston, New York State Psychiatric Institute and Columbia-Presbyterian Hospital in New York City, Rush Presbyterian Hospital in Chicago, Washington University in St Louis, University of Iowa Hospitals and Clinics in Iowa City	BD-I	DSM-IV	SADS	290	Mania: 36.5±13.3 Depression: 36.6±14.4 Cycling: 35.9±11.9		17 (4.0)
		BD-II			143
Craig et al (2002)	Suffolk County Mental Health Project (12 inpatient facilities of Suffolk County, New York) Public sector outpatient clinic for the destitute in Starr County National Epidemiologic Survey on Alcohol and Related Conditions	BD with psychosis	DSM-III-R	SCID, Alcohol Use Disorder and Associated Disabilities Interview Schedule-DSM-IV Version	138	15–60 years		4 (2.9)	BD with psychosis
		MDD with psychosis			87			6 (5.2)
		Schiz/schizoaff			225			9 (4.0)
Dilsaver et al (2008)	Suffolk County Mental Health Project (12 inpatient facilities of Suffolk County, New York)	BD	DSM-IV	SCID	69	BD: 34.9±11.8 Unipolar: 36.9±12.8	BD: 36.2% Unipolar: 31.2%	39 (56.5)	Data from Latinos
		MDD			118			27 (22.9)
Goldstein and Levitt (2008)	Public sector outpatient clinic for the destitute in Starr County	BD	DSM-IV	SCID	1411	Male: 35.3±16.7 Female: 40±16.2	432 (74.6%)	853 (60.4)	2001–2002 National Epidemiologic Survey on Alcohol and Related Conditions
Henry et al (2003)	Charles Perrens Hospital, Bordeaux, Chenevier Hospital, Creteil	BD	DSM-IV	French version of DIGS	318	53.3±15.1	41%	52 (16.4)	75% BD-I, half psychotic currently or longitudinal
Kawa et al (2005)	South Island Bipolar Study in New Zealand (Otago, Southland, and Canterbury regions of New Zealand)	BD	DSM-IV	DIGS	211	Male: 43 Female: 44	17 (15.0%)	41 (19.4)
Kessler et al (1997)	National Comorbidity Survey of the USA	BD	DSM-III-R	CIDI	29	15–54	49.1%	8 (33.1)	BD-I National Comorbidity Survey
Levander et al (2007)	Stanley Foundation Bipolar Network (University of California Los Angeles; University of Texas Southwestern, Dallas; University of Cincinnati; NIMH, Bethesda; University Medical Centre, Utrecht; University of Munich	BD-I	DSM-IV	SCID	276	AUD: 41.6±11.2 No AUD: 41.7±11.2	28 (8.0%)	54 (19.5)	STANLEY Foundation data, 2/3 with alcohol use
		BD-II			71			18 (25.4)
		Schizoaff			3			0 (0)
McElroy et al (2001b)	Stanley Foundation Bipolar Treatment Outcome Network	BD	DSM-IV	SCID-P and clinician-administered and self-rated questionnaires	288	42.8±11.3		58 (20.1)	STANLEY Foundation data
Mula et al (2008)	Multicentre Italian study University Hospital Outpatient Psychiatric Department of the University of Pisa and from three collaborating community mental health centre outpatient clinics	BD-I	DSM-IV	SCID-I, SCID-P	70	41.3±11.6	43%	50 (27.6)
		BD-II			51	48.1±11.7	43%
		MDD			60	49.2±11	30%
Nakagawa et al (2008)	Stanley Foundation Bipolar Network (University of California Los Angeles; University of Texas Southwestern, Dallas; University of Cincinnati; NIMH, Bethesda; University Medical Centre, Utrecht; University of Munich	BD	DSM-III-R	SCID	116	18–73 years	34.5%	37 (31.9)	BD depression
Pini et al (1997)	Stanley Foundation Bipolar Treatment Outcome Network	BD	DSM-III-R	SCID-P and clinician-administered and self-rated questionnaires	24	37.9±12.0	11 (45.4%)	9 (36.8)	BD depression
		MDD			38			15 (24.0)
		Dysthymia			25
Rihmer et al (2001)		BD-I	DSM-III-R	Hungarian version of the DIS	95	18–64 years		7 (7.4)	Patients with BD-I and BD-II, Hungarian epidemiological study
		BD-II			24			3 (12.5)
		MDD			443			55 (12.4)
Schaffer et al (2006)	Canadian Community Health Survey: Mental Health and Well-Being	BD	DSM-IV	WMH-CIDI	852	37.3±13.7	60 (15.5%)	164 (19.3)	Canadian Community Health Survey: Mental Health and Well-Being
Simon et al (2004)	STEP-BD (multicentre project funded by the NIMH, USA)	BD-I	DSM-IV	Mini International Neuropsychiatric Interview (MINI Plus V.5.0)	360	41.7±12.8	40.6%	66 (18.3)	STEP-BD data
		BD-II			115			16 (13.9)
Slama et al (2004)	Two university-affiliated hospitals (in Paris and Bordeaux, France)	BD	DSM-IV	Semistructured diagnostic interviews (the DIGS and the Family Interview for Genetic Studies)	307	44.2±14.7	124 (40.4%)	25 (7.9)	Patients in remission
Szadoczky et al (1998)		BD	DSM-III-R	Hungarian version of the DIS	149		44%	16 (10.6)	Epidemiological study
		MDD			443			55 (12.4)
Tamam and Ozpoyraz (2002)	Balcali Hospital, Cukurova University, Adana, Turkey	BD-I	DSM-IV	SCID, Hamilton Depression Scale, Bech-Rafaelsen Mania Scale	70	33.4±10.3	29 (41.6%)	7 (10.0)	BD-I
Yerevanian et al (2001)	Mood disorders clinic of a major university centre Bipolar Clinic of the Mood Disorders Programme, Clarke Institute of Psychiatry, University of Toronto	BD-I: 8	DSM-III	SCID-P SADS-Longitudinal Version	8	40.4±15.1	5 (62.5%)	2 (5.7)	Most patients with BD-II
		BD-II: 27			27	39.5±11.8	13 (51.9%)
		MDD			98			18 (18.4)
Young et al (1993)	Psychotic disorders unit, McLean Hospital, Belmont, Massachusetts, USA	BD	NA	SCID-I	81	37.6	32 (39.5%)	26 (32.1)
Young et al (2013)	National Institute of Mental Health and Neurosciences, Bangalore, India	BD	DSM-IV-TR	SCID-CV	304	36.92±13.55	49.34%	21 (6.9)
Zutshi et al (2006)	Hospitals of Taiwan	BD	DSM-IV	Chinese versions of CIDI and SCID	80	30.06±7.77	57 (71%)	6 (7.5)	BD patients in remission
		Controls			50			0 (0.0)
Tsai et al (2012)		BD-I	DSM-IV	Hungarian version of the DIS	306	37.07±12.3	147 (48.03%)	36 (12.6)

AUD, alcohol use disorder; BD, bipolar disorder; CIDI, Composite International Diagnostic Interview; DIGS, Diagnostic Interview for Genetic Studies; DIS, Diagnostic Interview Schedule; DSM, Diagnostic Statistic Manual; DSM-III-R, Diagnostic and Statistical Manual of Mental Disorders, Third Edition, Revised; DSM-IV-TR, Diagnostic and Statistical Manual of Mental Disorders-Fourth Edition (Text Revision); MDD, major depressive disorder; NA, not available; NIMH, National Institute of Mental Health; SADS, Schedule for Affective Disorders and Schizofrenia; Schiz, schizophrenia; Schizoaff, schizo-affective disorder; SCID, Structured Clinical Interview for DSM; SCID-I, Structured Clinical Interview for DSM Axis I Disorders; SCID-CV, Structured Clinic Interview for DSM, Clinician Version; SCID-P; Structured Clinic Interview for DSM, Patient Version; STEP-BD, Systematic Treatment Enhancement Program for BD; WMH, WHO World Mental Health (WMH) Survey Initiative.

Cross-sectional (point) prevalence of comorbid PD

The overall cross-sectional estimate of PD in patients diagnosed with BD was 11.5% (95% CI 10.4% to 12.6%) in the fixed-effect model, and it was 15.1% (95% CI 7.9% to 24.0%) in the random-effects model (figure 1). Across studies there was a variation in the cross-sectional prevalence of PD in BD, a likely reflection of the sociodemographics and clinical characteristics of the samples (figure 1). No relevant publication bias emerged from the funnel plot and the Egger’s or Begg and Mazumdar rank correlation test. Heterogeneity was substantial: I²=95.9% (95% CI 94.5% to 97.0%), estimated between-study variance=0.15 (figure 1 and table 3). However, the Baujat plot suggested that the studies with the greatest contribution to the overall heterogeneity had a small to moderate influence on the result but one. The standardised residuals plot and the radial plot suggested just one sample was potential outlier.

Figure 1.

Cross-sectional prevalence of panic disorder (PD) in patients diagnosed with bipolar disorder.

Table 3.

Effect sizes in the meta-analysis of studies on panic disorder in patients with bipolar disorder

	k	n	Prevalence, n (%)	95% CI (%)	Q	P value	I2, n (%)	95% CI (%)
Cross-sectional studies
FE model	15	3391	11.5	10.4 to 12.6
RE model	15	3391	15.1	7.9 to 23.9	343.5	<0.001	95.9	94.5 to 97.0
RE model without outliers	14	3295	13.0	7.0 to 20.3	253.9	<0.001	94.9	92.9 to 96.3
RE in subgroup analysis 1	7	1228	12.3	1.2 to 31.3	120.9	<0.001	95.0	92.0 to 96.9
FE in subgroup analysis 2	2	161	12.9	8.0 to 18.7
Subgroup analysis 1: BD-I samples only. Subgroup analysis 2: BD-II samples only; since there were two studies only, the FE model was reported.
Longitudinal studies
FE model	25	8226	15.5	14.7 to 16.3
RE model	25	8226	16.8	12.2 to 22.0	546.5	<0.001	95.6	94.5 to 96.5
RE model without outliers	24	8157	15.5	11.6 to 19.9	491.7	<0.001	95.3	94.0 to 96.3
RE in subgroup analysis 1	9	2852	14.0	8.0 to 20.9	137.7	<0.001	94.2	91.0 to 96.3
RE in subgroup analysis 2	4	217	14.8	1.7 to 35.8	11.1	0.011	73.1	24.2 to 90.4
Subgroup analysis 1: BD-I samples only. Subgroup analysis 2: BD-II samples only.

BD, bipolar disorder; FE, fixed-effect model; k, number of included studies; n, number of patients in the included studies; RE, random-effects model with empirical Bayes estimator.

Indeed, this sample, with patients with both BD-I and BD-II, had one of the highest cross-sectional estimated prevalence of PD.

After exclusion of this outlier, the overall fixed-effect cross-sectional estimate of PD in patients diagnosed with BD went to a small decrease, being now 10.6% (9.5%–11.7%), while the random-effects estimate was 13.0% (7.0%–20.3%) (table 3).

Subgroup analyses, with inclusion/exclusion of studies according to the nature of the sample, did not reveal relevant changes in the estimates of the random-effects model, nor a substantial attenuation of heterogeneity (see table 3 for details).

No difference in cross-sectional prevalence rates was found between BD-I and BD-II.

The meta-regression of data without the outlier on age and gender ratio showed that neither age (coefficient=0.011; z=−0.48; P=0.64; I²=94.6%) nor gender ratio (coefficient=−0.150; z=−0.94; P=0.36; I²=94.6%) was related to the cross-sectional prevalence estimate of PD in BD.

Longitudinal prevalence of comorbid PD

There were 26 studies detailing data on lifetime prevalence of PD in patients diagnosed with BD (figure 2), yielding 25 rates, some of them further grouped into BD-I and BD-II estimates (table 2). The longitudinal estimate of PD in patients diagnosed with BD was 15.5% (95% CI 14.7% to 16.3%) in the fixed-effect model, and it was 16.8% (95% CI 12.2% to 22.0%) in the random-effects model. Again, the composition of the sample by sociodemographic (male:female ratio, age range) and clinical (BD subtypes, phase of the disorder at assessment) variables conditioned a wide variability of the longitudinal prevalence of PD across studies (figure 3).

Figure 2.

Lifetime prevalence of panic disorder (PD) in patients diagnosed with bipolar disorder.

Figure 3.

Graphical representation of the sample size of studies reporting cross-sectional and lifetime estimates of panic disorder in patients diagnosed with bipolar disorder.

Some asymmetry, suggesting publication bias, emerged from the funnel plot, but the Egger’s or Begg and Mazumdar rank correlation test did not reveal statistically significant bias. Heterogeneity was substantial: I²=95.6% (95% CI 94.5% to 96.5%). The Baujat plot suggested that two studies among those with the greatest contribution to the overall heterogeneity also had the greatest influence on the result. However, the radial plot and the standardised residuals plot suggested just one potential outlier, and a different one.

After exclusion of one potential outlier study, the longitudinal estimate of PD in patients diagnosed with BD was 15.5% (95% CI 11.6% to 19.9%) in the random-effects model. No difference in longitudinal prevalence rates was found between BD-I and BD-II (table 3).

Heterogeneity remained substantial in analyses by subgroup, suggesting the sample diagnosis or phase of the disorder was not the main reason explaining it (table 3). Meta-regression showed no relationship of estimates with age (coefficient=0.005; z=0.44; P=0.66; I²=95.9%), gender ratio (coefficient=−0.029; z=−0.38; P=0.70; I²=92.8%) or the diagnostic procedure (Structured Clinical Interview for Diagnostic Statistic Manual (DSM) vs any other: coefficient=0.039; z=0.34; P=0.73; I²=95.5%).

Comparison of BD with other diagnoses

Table 3 summarises the details on the prevalence rates of PD in patients with BD. Cross-sectional pooled rates indicate that 13% of patients with BD suffer from PD. This estimate is roughly similar to that reported for patients with schizophrenia or unipolar depression, and it is several times higher than the rate expected for the general population. The pooled longitudinal rate for PD in patients with BD is 15.5%, which is similar to that reported for patients with unipolar depression but several times higher in comparison with that reported in the control population.

PD probably presents with a chronic rather than episodic course in patients with BD and unipolar depression, as suggested by the negligible difference between cross-sectional and longitudinal prevalence.

The prevalence of PD in people without comorbid disorder was reported in just one study, and it was 0% (out of 50 subjects), way much lower than the prevalence found in patients with BD in the same study (5%) or the estimates found in the present meta-analysis.

Cross-sectional estimates of PD in patients with BD were confronted with those observed in patients with major depressive disorder (MDD) in two studies only. A higher cross-sectional prevalence of PD in patients with BD was found than in patients with MDD: test for subgroup differences (based on random-effects model): Q=7.98, df=1, P=0.0047. However, in these two studies the cross-sectional estimates of PD in patients with BD were higher than the pooled estimates in the overall meta-analysis on cross-sectional data. Moreover, heterogeneity was substantial: I²=86.4% (66.9%–94.4%); Q=22.0, df=3, P<0.0001.

Six studies (published in seven papers) reported longitudinal estimates of PD in patients with BD and with MDD. No statistically significant difference was detected in the longitudinal prevalence of PD in patients with BD or MDD: test for subgroup differences (based on random-effects model): Q=0.05, df=1, P=0.82. Heterogeneity was substantial: I²=88.2% (81.3%–92.6%); Q=93.3, df=11, P<0.0001.

Conclusions and clinical implications

There is a consensus in the literature of higher rate of psychiatric comorbidity in patients with BD, but with some inconclusiveness about the rates of specific comorbid disorders. Several methodological issues contribute to these inconsistencies. Studies differ in the characteristics of the population under study, by gender, age and composition of samples in terms of BD subtypes or phase of the disorder at the time of assessment. The methods of assessment also influence the chance of correct detection of the cases. Trained lay interviewers are usually employed in epidemiological studies, while clinical studies more often rely on highly experienced researchers. Thus, reliability is often higher in clinical studies, at the expense of the inclusion of more severe cases. Conversely, epidemiological studies, which apply structured interviews, may yield artificially inflated rates because of the multiple allocation of the same symptom.¹⁶

The current paper analysed 15 studies with data on cross-sectional (point) prevalence of PD in BD (n=3391). It also analysed 25 studies with longitudinal data (n=8226). Cross-sectional prevalence of PD ranged in these studies from 2.3% to 62.5%, while longitudinal prevalence ranged from 2.9% to 56.5%. The analysis returned a random-effects point prevalence (after exclusion of one potential outlier study) of 13.0% (95% CI 7.0% to 20.3%), and a random-effects lifetime estimate (after exclusion of one potential outlier study) of 15.5% (95% CI 11.6% to 19.9%).

Significant heterogeneity was found in both cross-sectional and longitudinal meta-analysis (94.9% and 95.3%, respectively). In a previous meta-analysis, lifetime comorbidity of PD in patients with BD was equal to 16.8% (95% CI 13.7 to 20.1),¹⁸ a similar estimate to ours.

There is some evidence that patients with pure mania quite never report panic, which, conversely, is much more prevalent in patients with mixed states or depression (up to 80%, depending on the sample), and often the picture is very complex.¹⁹ Patients regularly admitted to hospital are likely to display higher rates of psychiatric comorbidity, including PD, than those treated primarily in the community. As a consequence, clinical samples including hospitalised patients with BD will include a higher proportion of patients with psychiatric comorbidity.

As far as PD is concerned, in the general population the 12-month prevalence of PD was found to vary from 0.8% to 2.8%,^20–25 while the lifetime prevalence was reported to range from 1.4% to 5.1%.21 23 24 26–28 Thus the reported rates for patients with BD are approximately more than 7-fold higher for cross-sectional prevalence, and more than 4.5-fold higher for longitudinal prevalence, than in the general population. It is worth noticing that the longitudinal rates are approximately double of the cross-sectional rates in the general population (although some overlap was also reported), while in patients with BD these rates are very close. It can be advanced that in patients with BD, PD is rather exclusively chronic and not episodic, and probably related to temperament,²⁹ which does not seem to be the case in the general population.

In the current analysis, both point and longitudinal estimates of PD in BD were related to significant heterogeneity, which probably reflects heterogeneity in sample composition, differences in methodology and medical settings, as well as geographical differences. No relationship of the prevalence rates with age or gender ratio was found, which is consistent with the incidence and prevalence of PD in the general population.

This widely accepted worst outcome of BD when comorbid with PD may depend on PD being often complicated by symptoms of depression,^30–32 which may trigger a depressive phase in the course of the BD or agitation.³³

A higher risk of suicide and suicide attempt was reported in patients with BD comorbid with PD. The frequent association of PD with major depression or substance use and related disorders might explain this finding. As a matter of fact, the evidence in so far is inconclusive. Some studies reported an increased risk of suicide attempts and self-harm in BD comorbid with PD, while other studies failed to find such an association.¹⁰

Overall, the treatment of BD is complex, while the treatment of many of its facets remains poorly researched.^5–8 While in general antidepressants are the treatment of choice for PD, this might not be the case in patients with BD (at least not in monotherapy) because of the risk of inducing or exacerbating mania. Patients with BD comorbid with PD are ideal candidates to the use of mood stabilisers, since there is evidence that lithium, lamotrigine and valproic acid/divalproex sodium may alleviate symptoms of anxiety both alone and in combination with antidepressants or second-generation antipsychotics, especially in mixed and rapid cycling patients.^34–39 However, the impact of PD on mood stabilisation and the load of caregiver burden is unclear,⁴⁰ and hard data are not available for the treatment of PD in the frame of BD, neither concerning pharmacotherapy nor psychotherapy.^{41 42} The true risk for committing suicide under antiepileptic drugs is still a matter of debate,^43–45 while psychosocial interventions do not seem to attenuate it.^46–49

It is worth noticing that most of the papers included in this systematic review were not identified through the scanning of the search engine (PubMed/MEDLINE), but instead they were derived from reference lists of review papers and books. There is no clear explanation for this, but it is possible that the MEDLINE algorithm was unable to locate the information when the search is aimed at data that were not clearly related to the main objective of the study. Overall, accuracy and completeness of review and meta-analytic studies cannot be assured when the data are retrieved from secondary sources, since there always might be an additional secondary source unknown to the authors.

An additional problem hampers the validity of the findings that can be derived from systematic reviews and meta-analysis studies that are based on psychiatric comorbidity in BD. Standard criteria assume that a comorbid anxiety disorder should be diagnosed in BD only if one can ascertain the independence of the symptoms of the comorbid disorder from those of the main disorder. However, it is probable that most studies reviewed here had used the very simple approach of merely adding up the symptoms. As a matter of fact, no DSM edition made explicit the requirement for independency of symptoms and diagnosis; thus, it cannot be excluded PD symptoms at least partially overlap with some core features of BD (eg, irritability, excitement).

In conclusion, this meta-analysis proved that there is high rate of comorbidity between BD and PD, with prevalence rates that are several times higher than in the general population, although with wide variation across studies. PD probably runs a more chronic course in patients with BD than in the general population. These findings highlight the importance of early detection and treatment of PD in BD to prevent chronic outcome, lessen symptom severity, increase chance of remission from a manic or depressive episode, and reduce the risk of suicide and self-harm.