Review Paper
Prevalence and Incidence Studies of Mood Disorders: A Systematic Review of the Literature
Paul Waraich, MHSc, MD1, Elliot M Goldner, MHSc, MD2, Julian M Somers, MSc, PhD1, Lorena Hsu, MSc3
| |
This is the second in a series of papers that present systematic reviews
of the prevalence and incidence of psychiatric disorders drawn from
studies
published in English literature in the years 1980 to 2000. The series discusses
the implications of these epidemiologic findings to mental health policy
and practice.
Objective: To present the results of a systematic review of literature
published between January 1, 1980, and December 31, 2000, that reports
findings on the prevalence and incidence of mood disorders in both general
population and primary care settings.
Method: We conducted a literature search of epidemiologic studies of mood
disorders, using Medline and HealthSTAR databases and canvassing English-language
publications. Eligible publications were restricted to studies that examined
subjects aged at least 15 years and over. We used a set of predetermined
inclusion and exclusion criteria to identify relevant studies. We extracted
and analyzed prevalence and incidence data for heterogeneity.
Results: Of general population studies, a total of 18 prevalence and 5
incidence studies met eligibility criteria. We found heterogeneity across
1-year and lifetime prevalence of major depressive disorder (MDD), dysthymic
disorder, and bipolar I disorder. The corresponding pooled rates for 1-year
prevalence were 4.1 per 100, 2.0 per 100, and 0.72 per 100, respectively.
For lifetime prevalence, the corresponding pooled rates were 6.7 per 100,
3.6 per 100, and 0.8 per 100, respectively. Significant variation was observed
among 1-year incidence rates of MDD, with a corresponding pooled rate of
2.9 per 100.
Conclusions: The prevalence of mood disorders reported in high-quality
studies is generally lower than rates commonly reported in the general
psychiatric literature. When controlled for common methodological confounds,
variation in prevalence rates persists across studies and deserves continued
study. Methodological variation among studies that have examined the prevalence
of depression in primary health care services is so large that comparative
analyses cannot be achieved.
(Can J Psychiatry 2004;49:124–138)
Click here for author affiliations.
Click here for research funding and support.
|
Clinical Implications
-
If health planners, clinicians, and researchers have been using commonly
reported prevalence rates to estimate the burden of disease as a result
of mood disorders, they may need to revise these estimates downward.
-
The prevalence of depression in primary care should be reexamined with
standardized methodologies.
-
The persistent variation in the incidence and prevalence of mood disorders
warrants further investigation.
Limitations
-
The best-estimate rates reported in this study should not be reported without
their confidence intervals, because significant heterogeneity exists among
studies.
-
This study reviews English-only reports between 1980 and 2000 and may have
missed some high-quality studies.
-
The preliminary analysis of variation in rates presented in this study
should be seen as hypothesis-generating. Further analysis must be cautious
of ecological fallacy.
|
Key Words: mood disorders, major depression, dysthymia, bipolar disorders, prevalence, incidence, systematic review
Résumé :Études de la prévalence et de l’incidence des troubles de l’humeur : un examen méthodique de la documentation
|
The diagnostic category of mood disorders is among the most prevalent of all mental health diagnoses (1–3). (For a definition of prevalence and incidence, refer to Goldner and colleagues; 4). Conducting a review of this category of mental disorders is challenging for several reasons. One challenge is that the various conditions that comprise this grouping of disorders can be quite disparate: some are transient, others are chronic or recurrent. Therefore, it is difficult to speak in generalities about the characteristics of mood disorders. Although most mood disorders result in some degree of temporary cognitive difficulty and mild deficits in attention or concentration, some forms of mood disorders are associated with profound cognitive and perceptual disturbance with delusional and hallucinatory features. Another set of challenges to reviewing the epidemiology of mood disorders involves the nosological frameworks and systems of psychiatric nomenclature currently in use (5,6). Mood disorders are constructs (that is, entities that cannot be directly measured or observed) developed through inference, hypothesis, deduction, and conjecture (1). Although mood disorders are described as taxonic (that is, defined in categorical terms) in current diagnostic nosological systems, some of their features may be more usefully conceptualized in dimensional terms (7). Given the uncertainties of our diagnostic constructs for mood disorders, previous reviews unsurprisingly report highly variable rates for all mood disorders (3,8–12) (Table 1). In addition, individuals may alternate between various mood disorders during their lifetime (13), making it difficult to determine the onset and cessation of disease episodes.
|
Table 1 Variation of rates of mood disorders in selected review studies
|
|
Disorder
|
Review study
|
Prevalence rate (per 100)
|
Incidence
|
|
|
|
Point
|
1-year
|
Lifetime
|
|
|
Depression
|
Bland (3)
|
—
|
0.6 to 7.0
|
0.9 to 16.1
|
—
|
|
|
Wittchen and others (8)
|
1.5 to 4.0
|
2.6 to 9.8
|
4.4 to 18
|
—
|
|
|
Boyd and Weissman (9)
|
—
|
—
|
—
|
104 to 519 per 100 000
|
|
Depression in primary care
|
Lemelin and others (10)
l Self-report
l Interview- based
|
4.8 to 8.6
9.0 to 30.0
|
—
—
|
—
—
|
—
—
|
|
Dysthymia
|
Wittchen and others (8)
|
1.2 to 3.9
|
2.3 to 4.6
|
3.1 to 3.9
|
—
|
|
|
Bland (3)
|
—
|
—
|
4.1 to 8.6
|
—
|
|
Bipolar I disorder
|
Wittchen and others (8)
|
0.1 to 2.3
|
1.0 to 1.7
|
0.6 to 3.3
|
—
|
|
|
Weissman and others (11)
|
—
|
0.3 to 1.5
|
—
|
|
|
Bebbington and Ramana (12)
|
—
|
—
|
2.6 to 20.8 per 100 000 yearly
|
This review addresses some of the challenges described above by presenting the results of a systematic review of literature published between January 1, 1980, and December 31, 2000, that reports findings of the prevalence and incidence of mood disorders in both general population and primary care settings. We examine the variability of the rates of disorders across studies and attempt to determine how much of this variance can be accounted for by various factors, using hetero- geneity analysis. This technique should decrease the variance associated with heterogeneous study methods and diagnostic constructs and subsequently provide a more robust estimate of the variation in rates of disorders.
We focus our discussion on the mood disorders that have been most extensively examined in epidemiologic studies: major depressive disorder (MDD), dysthymia, and bipolar I disorder (BD I). MDD has been found to be one of the most prevalent mood disorders in previous reviews (Table 1). It is important not only because of its prevalence but also because of its associated disability. The Global Burden of Disease study found that the disability associated with MDD is surpassed only by that associated with ischemic heart disease in industrialized countries (14). Previous reports have also highlighted recent trends toward an increasing prevalence of depression. This may mean that depression will be even more of a burden on society in future years (14,15). We selected the epidemiology of MDD in primary care as an additional topic of interest for this review. Several jurisdictions are currently attempting to address care gaps in the management of depression in primary care, such as the low detection rate by primary care physicians (16). Establishing robust estimates of the prevalence of depression in primary care will inform these policy initiatives.
Dysthymia is characterized by chronic low-grade depressive symptoms and can be particularly disabling when it is comorbid with other mood disorders, such as MDD. About 70% of those with dysthymia may eventually go on to develop MDD (13). Relative to other mood disorders, research on the epidemiology of dysthymia has been scarce (3); however, several recent studies have examined this disorder in more detail (17–19). BD I, also known as manic-depressive disorder, may be the most disabling of mood disorders (14). About 10% of people with MDD eventually develop BD, which makes studying the epidemiology of this disorder difficult (13).
Methods
The methods employed in this review have been presented in more detail elsewhere (4). We searched the Medline and HealthSTAR databases for relevant studies, using the key indexing terms epidemiology, prevalence, and incidence, combined with the search terms mental disorders, mood disorders, depressive disorders, major depression, dysthymia, and bipolar disorders. We limited the search to English-language studies published between January 1, 1980, and December 31, 2000. Reference lists of relevant primary and review articles identified were also searched.
General population prevalence and incidence studies were eligible for inclusion if they were community surveys using probability sampling techniques. For primary care studies, those that randomly or consecutively sampled a population of primary care attendees for a given period were included. Eligible publications were restricted to studies having sample sizes of 450 or over. We chose this number as the lower limit of sample size based on preliminary calculations (using the formula in Kelsey and colleagues; 20, p 282) that demonstrated adequate error rates for a range of expected prevalence rates. Studies were also eligible for inclusion if they examined age ranges covering the adult population. Only studies using operationalized diagnostic criteria and case identification based on either standardized instruments or clinician diagnosis were included. Prevalence data, including overall, sex-specific, and age-specific rates, were extracted from eligible studies.
We conducted qualitative analyses of variables related to methodology to summarize and elucidate any observed differences between rates. As well, we pooled each set of rates according to a Bayesian approach to metaanalysis, using the Fastpro software program. Readers interested in a more detailed discussion of this approach should refer to Eddy and colleagues (21). We analyzed each of the pooled rates for heterogeneity, using chi-square tests according to Fleiss’ method (22).
Results
Description of Studies
General Population Studies. From the citations and abstracts generated from the initial electronic search, we identified 38 prevalence and 12 incidence studies potentially meeting our inclusion criteria, in addition to 8 review papers. The full texts of these articles were retrieved. All reference lists of identified studies and reviews were searched, generating an additional 39 prevalence and 15 incidence studies, for which full-text articles were obtained.
Of the 77 prevalence studies for which full-text articles were reviewed, 42 studies were excluded: 30 studies did not meet eligibility criteria, and 12 presented duplicate data. Thus, 35 prevalence papers of mood disorders met eligibility criteria for the review (17–19,23–54); 18 unique primary investigations of prevalence were included. The reasons for excluding prevalence studies fulfilling all but 1 of the inclusion criteria are documented in Table 2 (55–66). Of the 27 incidence studies identified, 22 were excluded: 16 did not meet inclusion criteria, and 6 were based on duplicate data. Therefore, 5 general population incidence studies of mood disorders were included (17,44,46,67,68). Five of the excluded studies almost met inclusion criteria, and reasons for their exclusion are presented in Table 2 (61,69–72).
|
Table 2 Mood disorder studies excluded at article review stage
|
|
Study reference
|
Reason for exclusion
|
Comments
|
|
General population studies
|
|
|
|
Prevalence studies
|
|
|
|
Mumford and others (66)
|
Does not report prevalence rates
|
Rates cannot be determined from data presented
|
|
Angst (55)
|
Does not meet age criteria
|
Assesses age range of 22 to 35 years
|
|
Takeuchi and others (57)
|
Population too specific
|
Looks at population of Chinese Americans
|
|
Angst and Merikangas (56)
|
Does not meet age criteria
|
Only looks at ages 18 to 19 years
|
|
Angst (58)
|
Does not meet age criteria
|
Prospective 10-year cohort study; presents rates applicable to age 20 to
30 years
|
|
Rumble and others (59)
|
Diagnoses based on ICD-8 criteria
|
Community sample
|
|
Fournier and Kovess (64)
|
Not clear whether operationalized criteria used
|
Comparison of mail and telephone surveys
|
|
Stefansson and others (63)
|
Does not include entire adult age group
|
Presents rates applicable to age 55 to 57 years
|
|
Surtees and Sashidharan (60)
|
Limited to population of women
|
Comparison of 2 community samples
|
|
Dilling and Weyerer (62)
|
Uses ICD-8 diagnostic criteria
|
Community sample
|
|
Halldin (65)
|
Uses ICD-8 diagnostic criteria
|
Community sample
|
|
Shen and others (61)
|
Diagnostic criteria used unclear
|
Large community sample
|
|
Incidence studies
|
|
|
|
Murphy and others (69)
|
Does not use operationalized criteria
|
Follow-up of cohort from 1970 to 1992
|
|
Eaton and others (70)
|
Based on duplicate data
|
Looks at Baltimore site of ECA study with 12-year follow-up
|
|
Horwath and others (71)
|
Based on duplicate data
|
Looks at 4 sites of ECA study with 1-year follow-up
|
|
Rorsman and others (72)
|
Does not use operationalized criteria
|
Study carried out between 1957 and 1972
|
|
Shen and others (61)
|
Diagnostic criteria used unclear
|
Large community sample
|
|
Primary care studies
|
|
|
|
Prevalence studies
|
|
|
|
Zinbarg and others (89)
|
Inadequate sample size
|
Examines both primary care and outpatient psychiatry populations
|
|
Barrett and others (83)
|
Does not use operationalized criteria
|
Presents rates based on screening scales
|
|
Kessler and others (84)
|
Rates not representative of primary care population
|
Denominator used in rate is total household population
|
|
Schulberg and others (85)
|
Rate may not be representative
|
Overall response rate < 50%
|
|
Parker and others (85)
|
Rate may not be representative
|
1st-stage response rate not assessed; very low 2nd-stage response rate
|
|
Zung and others (87)
|
Does not use operationalized criteria
|
Presents rates based on screening scales
|
|
Nielsen and Williams (88)
|
Diagnoses based on Feighner (1972) criteria
|
Data collected in 1971
|
Primary Care Studies. For MDD in primary care, 9 prevalence and 2 incidence studies as well as 4 review papers were identified and retrieved. An additional 34 prevalence studies and 1 incidence study resulted from the reference lists of identified articles. Of the 43 prevalence studies identified, 33 were excluded as they did not fulfill inclusion criteria. Thus, 10 primary care prevalence studies met eligibility criteria for the review (73–82). Excluded studies almost meeting eligibility criteria are presented in Table 2 (83–89). Of the 3 incidence studies identified, 2 met eligibility criteria (90,91).
Prevalence Studies. We present findings for the 24 papers reporting overall or sex-specific 1-year and lifetime prevalence rates and age-specific lifetime prevalence rates for mood disorders in general, for MDD, for dysthymic disorder, and for BD I (Table 3 to 5, Figure 1). The results of 2 studies reporting only 1-year age-specific rates are not reported here (35,44). The results of studies reporting only data for point prevalence or 6-month prevalence (23,30,38–40,45,46) or for mood disorder categories other than those stated above (23,42,92) are also not presented.
|
Table 3 One-year prevalence rates of mood disorders
|
|
Study authors
|
Study site
|
Total number of subjects (n)
|
Response rate
(%)
|
Case-finding method
|
Prevalence rate (per 100 persons)
|
|
|
|
|
|
|
Any mood disorder
|
Major depressive disorder
|
Dysthymic disorder
|
Bipolar I disorder
|
|
Henderson and others (32)
|
Australia - national
|
10 600
|
78.0
|
Census; CIDI-A/ICD-10; lay interviewers
|
—
|
5.1
|
1.1
|
—
|
|
Bijl and others (18)
|
Netherlands - national
|
7146
|
69.7
|
Census; CIDI/DSM-III-R; lay interviewers; algorithm diagnosis
|
7.6
|
5.8
|
2.3
|
1.1
|
|
Pakriev and others (17)
|
Udmurt Republic - region of Udmurtia (rural areas)
|
855
|
85.9
|
Census; CIDI/ICD-10 & DSM-III-R; diagnosis made by clinician
|
—
|
22.5c
15.4b
|
4.1c
2.6b
|
0.1c
0.2b
|
|
Szadoczky and others (17)
|
Hungary - national
|
2953
|
85.0
|
Census (GP lists for 5 different areas); DIS/DSM-III-R; lay interviewers;
algorithm diagnosis
|
—
|
7.1
|
—
|
0.9
|
|
Offord and others (19)
|
Canada - province of Ontario
|
8116
|
76.5
|
Census; UM-CIDI/DSM-III-R; lay interviewers; algorithm diagnosis
|
4.5
|
4.1
|
0.8
|
0.6
|
|
Kessler and others (36)
|
USA (NCS) - national
|
8098
|
82.4
|
Census; UM-CIDI/DSM-III-R; lay interviewers; algorithm diagnosis
|
11.3
|
10.3
|
2.5
|
1.3
|
|
Bourdon and others (26)
|
USA (ECA) - 5 sites, mainly urban
|
20291
|
68 to 80
|
Census; DIS/DSM-III; lay interviewers; algorithm diagnosis
|
6.3
|
3.5
|
3.3
|
0.6
|
|
Faravelli and others (54)
|
Italy - 3 local health units of Florence
|
1000
|
100.0
|
Census (GP lists for 3 local health units); structured interview including
items from SADS-L; DSM-III & DSM-III-R; interviews by GPs
|
—
—
|
6.3a
6.2b
|
3.0a
2.6b
|
1.3a
1.5b
|
|
Oakley-
Browne and others (43)
|
New Zealand - area of Christchurch, mostly urban
|
1498
|
70.0
|
Census; DIS/DSM-III; lay interviewers; algorithm diagnosis
|
10.4
|
6.7
|
—
|
0.2
|
|
Hwu and others (34)
|
Taiwan -
metropolitan
Taipei
small towns
rural villages
|
11 004
5 005
—
3 004
2 995
|
95.0
—
—
—
—
|
Census; DIS-CM/DSM-III; lay interviewers; method of diagnosis unclear
|
—
|
—
0.64
—
1.1
0.81
|
—
—
—
—
—
|
—
1.2
—
0.3
1.0
|
|
Bland and others (24)
|
Canada - metropolitan Edmonton
|
3 258
|
71.6
|
Census; DIS-DSM-III; lay interviewers; algorithm diagnosis
|
6.8
|
4.6
|
—
|
0.2
|
|
|
|
|
|
Best-estimate
(95%CI)
|
7.5
(5.7 to 9.7)
|
4.1
(2.4 to 6.2)
|
2.0
(1.3 to 2.8)
|
0.72
(0.5 to 1.0)
|
|
— = Not reported; aDSM-III criteria; bDSM-III-R criteria; cICD-10 criteria.
CIDI-A = Composite International Diagnostic Interview ; CIDI/ICD = Composite
International Diagnostic Interview/International Classification of Diseases;
DIS = Diagnostic Interview Schedule; DISCM = Diagnostic Interview Schedule
Chinese Modified; UM-CIDI = University of Michigan Composite International
Diagnostic Interview.
|
Because the diagnostic nomenclature for mood disorders has undergone revisions from DSM-III to DSM-IV criteria, it was necessary to make assumptions with respect to the classification of various terms in these studies to be in accordance with DSM-IV criteria. These terms, along with the DSM-IV categories to which they are assumed to be equivalent, are in Table 6.
|
Table 4 Lifetime prevalence rates of mood disorders
|
|
Study authors
|
Study site
|
Total number of subjects (n)
|
Response rate
(%)
|
Case-finding method
|
Prevalence rate (per 100 persons)
|
|
|
|
|
|
|
Any mood disorder
|
Major depressive disorder
|
Dysthymic disorder
|
Bipolar I disorder
|
|
Murphy and others (42)
|
Atlantic Canada - region of Stirling County
|
1 396
|
86.0
|
Census; DIS/DSM-III; lay interviewers; method of diagnosis unclear
|
—
|
7.9
|
—
|
—
|
|
Bijl and others (18)
|
Netherlands - national
|
7 146
|
69.7
|
Census; CIDI/DSM-III-R; lay interviewers; algorithm diagnosis
|
19.0
|
15.4
|
6.3
|
1.8
|
|
Szadoczky and others (48)
|
Hungary - national
|
2 953
|
85.0
|
Census (GP lists for 5 different areas); DIS/DSM-III-R; lay interviewers;
algorithm diagnosis
|
24.2
|
15.1
|
4.5
|
1.5
|
|
Fournier and others (31)
|
Canada - Montreal
|
893
|
63.6
|
Telephone survey; CIDIS/DSM-III-R; lay interviewers; algorithm diagnosis
|
31.4
|
29.6
|
14.0
|
—
|
|
Carta and others (28)
|
Italy - Cagliari and Scano Montiferro
|
480
|
87.0
|
Census; CIDI/DSM-III-R; physician interviewers
|
—
|
13.3
|
4.1
|
—
|
|
Kessler and others (36)
|
USA (NCS) - national
|
8 098
|
82.4
|
Census; UM-CIDI/DSM-III-R; lay interviewers; algorithm diagnosis
|
19.3
|
17.1
|
6.4
|
1.6
|
|
Chen and others (29)
|
Hong Kong - national
|
7 229
|
77.8
|
Census; DIS-III-CM/DSM-III; lay interviewers; algorithm diagnosis
|
—
|
1.9a
|
2.0a
|
0.15a
|
|
Wacker and others (49)
|
Switzerland - city of Basel
|
470
|
52.2
|
Census; CIDI/DSM-III-R & ICD-10; interviewers with training in psychiatry
or psychology
|
19.4b
25.7c
|
15.7b
22.8c
|
7.2b
7.0c
|
—
|
|
Bourdon and others (26)
|
USA (ECA) - 5 sites, mainly urban
|
20 291
|
68-80
|
Census; DIS/DSM-III; lay interviewers; algorithm diagnosis
|
8.3
|
5.9
|
3.3
|
0.8
|
|
Wittchen and others (53)
|
Germany - former West Germany
|
483
|
73.5
|
Census; DIS/DSM-III; clinical interview and diagnosis
|
12.9
|
9.0
|
4.0
|
0.24
|
|
Oakley-Browne and others (43)
|
New Zealand - area of Christchurch, mostly urban
|
1 498
|
70.0
|
Census; DIS/DSM-III; lay interviewers; algorithm diagnosis
|
14.7
|
12.6
|
6.4
|
0.7
|
|
Hwu and others (34)
|
Taiwan
Metropolitan
Taipei
Small towns
Rural villages
|
11 004
5 005
3 004
2 995
|
95.0
|
Census; DIS-CM/DSM-III; lay interviewers; method of diagnosis unclear
|
—
—
—
—
—
|
—
0.88
—
1.7
0.97
|
—
0.92
—
1.5
0.94
|
—
1.6
—
0.7
1.0
|
|
Bland and others (93)
|
Canada - metropolitan Edmonton
|
3 258
|
71.6
|
Census; DIS/DSM-III; lay interviewers; algorithm diagnosis
|
10.2
|
8.6
|
3.7
|
0.6
|
|
Lee and others (37)
|
Korea - Dong, Seoul (urban) and Eub, Myeon (rural)
|
5 100
|
81.8
|
Census; DIS/DSM-III; lay interviewers; algorithm diagnosis
|
5.4
|
3.4
|
2.2
|
0.42
|
|
Canino and others (27)
|
Puerto Rico - entire island nation
|
1 513
|
91
|
Census; DIS/DSM-III; lay interviewers; algorithm diagnosis
|
7.9
|
4.6
|
4.7
|
0.5
|
|
|
|
|
|
Best-estimate
(95%CI)
|
14.1
(10.2 to 18.7)
|
6.7
(4.2 to 10.1)
|
3.6
(2.5 to
5.0)
|
0.82
(0.56 to 1.1)
|
|
aOverall rate calculated from raw data (only sex and age-specific rates
reported); bDSM-III-R criteria; cICD-10 criteria; CIDI-S = CIDI simplified
|
All the studies presented are community surveys using samples ranging in size from approximately 500 (49) to 20 000 (26) subjects. For each study, the percentage confidence interval (CI) width or error rate for estimated prevalence at a 95% confidence level may be calculated using the formula provided by Kelsey and colleagues (20, p 282). Most studies used either the Diagnostic Interview Schedule (DIS) or the Composite International Diagnostic Interview (CIDI) administered by trained lay interviewers and applied algorithms to derive diagnoses.
|
Table 5 Sex-specific 1-year and lifetime prevalence rates of mood disorders
|
|
Study authors
|
Study site
|
Prevalence rate (per 100 persons)
|
|
|
|
Any mood disorder
|
Major depressive
disorder
|
Dysthymic
disorder
|
Bipolar I
disorder
|
|
|
|
Male
|
Female
|
Male
|
Female
|
Male
|
Female
|
Male
|
Female
|
|
|
|
|
|
|
|
|
|
|
|
|
1-year prevalence
|
|
|
|
|
|
|
|
|
|
Henderson and others (32)
|
Australia
|
—
|
—
|
3.4
|
6.8
|
1.0
|
1.3
|
—
|
—
|
|
Bijl and others (18)
|
Netherlands
|
5.7
|
9.7
|
4.1
|
7.5
|
1.4
|
3.2
|
1.1
|
1.1
|
|
Pakriev and others (17)
|
Udmurt Republic
|
—
|
—
|
13.1
|
29.6
|
2.7
|
5.1
|
—
|
—
|
|
Szadoczky and others (48)
|
Hungary
|
—
|
—
|
4.7
|
9.0
|
—
|
—
|
1.0
|
0.9
|
|
Offord and others (19)
|
Ontario, Canada
|
3.2
|
5.9
|
2.8
|
5.4
|
—a
|
0.8
|
—a
|
0.6
|
|
Kessler and others (36)
|
USA (NCS)
|
8.5
|
14.1
|
7.7
|
12.9
|
2.1
|
3.0
|
1.4
|
1.3
|
|
Faravelli and others (54)
|
Florence, Italy
|
—
|
—
|
3.5
|
8.8
|
2.2
|
3.7
|
0.7
|
1.9
|
|
Weissman and others (51)
|
USA (ECA)
|
—
|
—
|
—
|
—
|
—
|
—
|
0.9
|
1.1
|
|
|
Best-estimate
(95%CI)
|
|
