This material was presented in the 25th European Congress of Psychiatry which was organised by the European Psychiatry Association in Florence, Italy from 1-4th April 2017. I had access to the material as I attended the full Congress. Please note that this material remains the intellectual property of the European Psychiatry Association and its authors and any citation should include a full reference to it. For more information on the EPA and how to become a member, please visit: http://www.europsy.net/

1. Giovanni de Girolamo*, Jessica Dagani*, Giulia Signorini*, Olav
Nielssen**, Matthew Large***
• Age of Onset of bipolar disorder and the
interval between its onset and the initial
management: a meta-analysis
• *IRCCS Fatebenefratelli, Brescia (Italy)
• **St Vincents Hospital, Sydney, University of Sydney, University of New South
Wales, NSW, Australia
• ***The School of Psychiatry, University of New South Wales, NSW, Australia

3. There are several, heterogeneous, criteria to define the Age Of Onset:
1. The age when the first morpho-functional pathological processes
which lead to a clear disorder (e.g. cellular or receptor alterations)
appear
 The age when the first signs of a disorder appear
(e.g. cognitive dysfunctions)
3. The age when the first symptoms of a disorder appear
( e.g. depressed mood)
4. The age when the first overt episode of a disorder appears
(e.g. first psychotic episode)
5. The age of the first contact with any health service
6. The age of the first pharmacological treatment
7. The age of the first hospitalization
WHAT IS THE AGE OF ONSET?

4. 1. It allows to study the precursors and prodromes of a
disorder
2. It allows to study the disorder lifetime risk, which is the
estimated proportion of the population who will have the
disorder by the end of their lives.
3. It allows to study comorbidity in a diachronic sense
(primary disorder, e.g. Depression  self-medication with
alcohol or substances  alcohol/substance abuse as
secondary disorder)
4. It allows to plan primary and secondary prevention, and
early interventions.
WHY IT IS IMPORTANT TO STUDY THE AGE OF ONSET?

5. 64 weeks of
duration of
untreated
psychosis

7. BACKGROUND
• Currently, there are no widely agreed definitions of
the onset and initial management of BD, and the
effect of treatment delay on the outcome of BD is
unknown.
• The onset of BD is difficult to define prospectively
because early symptoms can include both
depression and attenuated mania, which can be
difficult to distinguish from normal fluctuations in
mood.

8. OBJECTIVES
• To carry out a meta-analysis of the period of time between
the AOO and the age of initial management of BD, referred
to here as the interval.
• The term management is used rather than treatment,
because the contact with services may have resulted in a
range of interventions that often did not include specific
treatment for BD.
• The aims of the study were to calculate a pooled estimate of
this interval and to explore possible reasons for the
variation in this interval between-studies.

9. METHODS
Studies were included if they:
• reported on cohorts of patients with BD, including BD I, BD II and BD NOS, and
• provided data on the mean AOO of BD (defined as either the AOO of symptoms,
the onset of mood syndromes or an episode or illness onset) and
• provided data on the mean age of first management of BD (defined as either age
at first treatment, first diagnosis or first hospitalization).
Studies were excluded if they:
• reported mixed samples of bipolar and schizophrenia spectrum disorders, or did
not separate those disorders;
• reported the duration between the onset of BD and the initial management,
without data regarding sample size, mean age and standard deviation for both
measures;
• were studies of affective psychosis;
• reported on age-specific populations, such older (>64) or younger (<14) patients.

11. Study author and
country
Location and Setting Sample
Akiskal et al. (2003) 28 Multicentre Study, France 196 (144 f, 52 m) patients with BD II
Ancin et al. (2010)29 Inpatients and outpatients from Madrid, Spain 143 (84 f, 59 m) patients with BD I (115) or BD II (28)
Carballo et al. (2008) 30 Inpatients and outpatients of two university hospitals in Pennsylvania and
New York State, USA
168 (85 f, 83 m) patients with BD
Cha et al. (2009)31 Inpatients at Seoul National University Hospital, Korea. 258 (140 f, 118 m) patients with BD I
Col etal. (2014)32 Outpatients from Ankara, Turkey 78 (50 f, 28 m) patients with BD I
Dell'Osso et al. (1993)33 Inpatients from a binational study from Pisa, Italy and Tennessee, USA. 51 female patients with Biploar Mania
Dienes et al. (2006) 34 Outpatients recruited in Los Angeles, USA 58 (29 f, 29 m) patients with BD I
Drancourt et al. (2013) 17 Outpatients recruited from university clinics in France 401 (292 f, 209 m) patients with BP I (418) or BP II (83)
Goldberg et al. (2002)22 Outpatients from a specialist Bipolar Clinic, New York, USA 56 (21 f, 35 m) patients with BD I (46) or BD II (7)
Gutierrez-Rojas et al. (2008)35 Outpatients from Jaen, Spain 108 (75 f, 33 m) patients with BD
Haro et al. (2006)36 Inpatients from a multicentre European study. 3,536 (1953 f, 1584 m) patients with Bipolar Mania
Lee & Dunner (2008)37 Outpatients from a university Clinic in Seattle, USA 44(26 f, 18 m) patients in a depressed phase of BD I (12), BD II (16) and BD
NOS (16)
Lex et al. (2011)38 Inpatients and out patients from Vienna and Villach, Austria 26 (11 f, 15 m) patients with BD I.
Lieberman et al. (2010)39 Volunteers recruited Washington, USA 282 (221 f, 71 m) patients with BD I (87), BD II (117) and BD NOS (78)
Loranger et al. (1978) 40 Inpatients from New York, USA 200 (100 f, 100 m) patients with Bipolar Mania
Marangell et al. (2008) 41 Trial participants from Texas, USA 9 (7 f, 2 m) patients with Rapid Cycling BD
Mazzarini et al. (2010)42 Outpatients from Barcelona, Spain 164 (92 f, 72 m) patients with BD II
Meeks (1999)43 Outpatients from Kentucky, USA 86 (64 f, 22 m) patients with BD I (74) or BD II (12)
Nivoli et al. (2014) 44 Outpatients from Barcelona, Spain 593 (326 f, 267 m) Patients with BD I (332), BD II (120) and BD NOS (18)
Okan Ibiloglu et al. (2011) 45 Outpatients from Ankara, Turkey 96 (67 f, 29 m) patients with Bipolar I (50) or Bipolar II (46)
Paholpak et al. (2014)46 Outpatient and inpatients from a multicentre study, Thailand 424 (258 f, 166 m) Patients with BD I (64) or BD II (146)
Parker et al. (2013)47 Outpatients recruited from a specialist clinic, NSW, Australia 210 (133 f, 177 m) patients with BD I (98) or BD II (534)
Perugi et al. (2000) 48 Inpatients from Pisa, Italy 320 (179 f, 141 m) patients with BD I
Schoyen et al. (2011) 49
Inpatients and out patients, Norway 482 (268 f, 214 m) patients with BD I (315), BD II (140) and BD NOS (27)
Serretti et al. (2002)50 Inpatients from Milan, Italy 1,004 (606 f, 398 m) patients with BD I (863) or BD II (141)
Skjelstad et al. (2011)51 Outpatients from Southern Norway 15 (11 f, 4 m) patients with BD II
Slama et al. (2004)52 Inpatients and outpatients from two Paris and Bordeaux, France 307 (183 f, 124 m) patients with BD I (226) or BD II (81)

13. Table 2. Main result and sensitivity analysis of meta analysis of the interval between the age at onset of bipolar affective disorder and initial management
N Samples Difference in
means in years
Standardized
difference in
means
Standard error Variance Lower Limit Upper Limit Z-Value P-Value I-square Q-value P-value of Q
Main result
All studies 50 5.81 .53
.04 .002 .45 .62 12.2 <.0001 92.6 672 <.0001
Sensitivity analysis of onset definitions
First symptoms 16 5.43 .40 .05 .003 .29 .51 7.34 <.0001 87.4 14.2 <.001
First episode 22 6.79 .72 .08 .007 .56 .88 8.82 <.0001 94.4
Illness onset 13 4.16 .35 .07 .005 .21 .48 5.02 <.0001 73.4
Sensitivity analysis of initial management definitions
First diagnosed 13 10.03 1.09 .16 .03 .77 1.42 6.58 <.0001 93.4 15.8 <.0001
First treated 25 3.87 .391 .07 .004 .26 .52 5.97 <.0001 93.0
First hospitalized 13 5.54 .44 .06 .003 .33 .55 7.74 <.0001 87.3
Sensitivity analysis of method of diagnosis
Clinical diagnosis 11 5.47 .49 .07 .005 .34 .63 6.58 <.0001 94.7 .37 .54
Used a structured
interview
40 5.90 .54 .06 .003 .44 .65 9.82 <.0001 93.0
Sensitivity analysis of chronology method
Unspecified
method
39 4.72 .42 .04 .002 .34 .49 10.55 <.0001 88.8 13.9 <.0001
Chronology
method
12 7.96 .89 .12 .02 .65 1.13 7.35 <.0001 91.7
Sensitivity analysis of overall reporting strength
Less strong
reporting
36 5.74 .48 .05 .002 .39 .57 10.71 <.0001 93.3 2.11 0.14
Stronger
reporting
15 5.97 .61 .08 .006 .46 .76 7.83 <.0001 88.1

14. RESULTS
• The pooled interval between the onset of BD and initial
management was almost six years. There was very high
between sample heterogeneity. There was some evidence
of publication bias towards samples that reported a longer
interval (Eggers Regression, intercept= 3.18, t-value =
4.84, df=49, P-value <.0001).
• Duval and Tweedie's trim and fill analysis (6 studies
trimmed to the right of mean and correction of point
estimate to .46 95% CI .37, to .55) suggested publication
bias had inflated the pooled effect size by 16%.

15. N Samples Difference in
means in years
P-Value
All studies 50 5.8 <.0001
Sensitivity analysis of onset definitions
First symptoms 16 5.4 <.0001
First episode 22 6.8 <.0001
Illness onset 13 4.2 <.0001
Sensitivity analysis of initial management definitions
First diagnosed 13 10.0 <.0001
First treated 25 3.9 <.0001
First hospitalized 13 5.5 <.0001

16. RESULTS
• Samples reporting the AOO defined by the AOO of
an episode of BD reported a longer interval
between onset and initial management of BD than
studies reporting first symptoms.
• Studies that reported the onset of symptoms had a
longer interval between onset and management
than studies that used less-well specified
definitions of illness onset. Post hoc testing found
significant differences between all three groups.

17. N
Samples
Difference in
means in
years
P-Value
Main result
All studies 50 5.8 <.0001
Sensitivity analysis of method of diagnosis
Clinical diagnosis 11 5.5 <.0001
Used a structured
interview
40 5.9 <.0001
Sensitivity analysis of chronology method
Unspecified method 39 4.7 <.0001
Chronology method 12 7.9 <.0001

18. Table 3. Meta-regression of continuous variables potentially associated with between sample heterogeneity
Point estimate Standard Error Lower limit Upper limit Z-value P-value
Mean age at onset -.05 .006 -.060 -.036 -7.92 <.0001
Mean age at initial
management
-.004 .008 -.020 .01 -.50 .61
Proportion Bipolar
1#
-.005 .001 -.007 -.002 -4.27 <.0001
Proportion Male -.001 .002 -.005 .003 -.46 .64
Proportion with
substance use†
.004 .002 -.001 .009 1.53 .12
Year of publication .02 .005 .01 .03 3.84 .0001
† 33 samples, #=48 samples

19. RESULTS
• Meta-regression found that samples with an earlier mean
AOO had a longer interval between onset and management.
Samples with a later mean age at initial management did
not have a longer interval.
• Studies with a higher proportion of BD I subjects tended to
have a shorter interval between the onset of BD and
management while more recently published studies
reported longer interval.
• The proportion of males and the proportion of patients with
substance or alcohol use disorder did not contribute to
between-study heterogeneity. .

20. Table 4. Multiple Meta-regression of moderator variables associated with between sample heterogeneity
Covariate Coefficient Standard
Error
lower 95%
CI
upper 95%
CI
Z-value P-value
Used first episode as measure of
onset
0.03 0.07 -0.10 0.17 0.47 0.64
Used first diagnosed as measure of
initial management
0.17 0.09 -0.004 0.38 1.92 0.06
Chronology method 0.26 0.08 0.11 0.41 3.34
0.0008
Mean age at onset -0.03 0.01 -0.04 -0.02 -4.56
<0.0001
Proportion Bipolar I -0.003 0.001 -0.005 -0.001 -3.64
0.0003
Year of publication 0.01 0.004 0.001 0.02 2.14
0.03
Intercept -15.54 7.96 -31.2 0.07 -1.95 0.05

21. RESULTS
• Multiple meta-regression (Table 4) model suggested that
independent contributions to between-study heterogeneity
(resulting in a longer period between onset and
management) were made by samples that used a
chronology method, used age at first diagnosis as the
indicator of management, had a younger mean AOO, had
fewer BD I patients or were from more recently published
studies.
• This model explained 63% of the observed between study
variance.

22. • Our main finding was a pooled interval of almost
six years between the onset of BD and its initial
management.
• The marked variation in this interval was partly
explained by:
– the way in which onset was defined,
– how the chronology of onset was assessed,
– the average AOO,
– the proportion of patients with BD I and
– when the study was published.
CONCLUSIONS

23. CONCLUSIONS
• Our finding that the interval between onset and
management was longer in studies that defined onset using
the age at first episode rather than the onset of symptoms
was surprising, because logically the emergence of
symptoms precedes the development of a full syndrome.
• A possible explanation is that the primary researchers
sometimes defined the onset of BD symptoms by the onset
of manic symptoms occurring in an initial manic episode,
which emerged later that symptoms of depression.
Moreover, clinicians may not recognize that an episode of
depression is part of bipolar disorder.

24. CLINICAL IMPLICATIONS
• The main clinical implication is that even in research centres, there is an
unacceptably long interval between the emergence of symptoms and
the identification and treatment of BD. The findings of this study
identify an opportunity for intervention in the early stages of illness,
which may in turn postpone or prevent the progression of BD.
• Recent clinical research to develop a staging model of BD that takes
into consideration not only the history of the disorder but also the
heterogeneous manifestation of the condition, may improve our ability
to detect BD earlier and also help develop guidelines for early
intervention, including stage-of illness specific interventions.