Journal of Affective Disorders 136 (2012) 149–154

Contents lists available at SciVerse ScienceDirect

Journal of Affective Disorders

j ourna l homepage: www.e lsev ie r .com/ locate / jad

Brief communication

Episode cycles with increasing recurrences in first-episode bipolar-Idisorder patients

R.J. Baldessarini a,⁎, P. Salvatore a,b, H.-M.K. Khalsa a, H. Imaz-Etxeberria c,A. Gonzalez-Pinto c, M. Tohen d

a Department of Psychiatry, Harvard Medical School, International Consortium for Bipolar Disorder Research, McLean Division of Massachusetts General Hospital,Boston, MA, United Statesb Section of Psychiatry, Department of Neuroscience, University of Parma, Italyc Department of Psychiatry, Vittoria Medical Center, Vittoria, Spaind Department of Psychiatry, University of Texas Health Sciences Center, San Antonio, TX, United States

a r t i c l e i n f o

⁎ Corresponding author at: Mailman Research Center115 Mill Street, Belmont, MA 02478-9106, United Sta3203; fax: +1 617 855 3479.

E-mail address: rbaldessarini@mclean.harvard.edu

0165-0327/$ – see front matter © 2011 Elsevier B.V. Adoi:10.1016/j.jad.2011.08.037

a b s t r a c t

Article history:Received 30 July 2010Received in revised form 8 August 2011Accepted 30 August 2011Available online 23 September 2011

Background: Preliminary review of a century of studies of the course of manic-depressive syn-dromes produced 40 reports, of which approximately one-third report evidence of shorteningwellness intervals or cycle-lengths with more recurrences, and two-thirds did not.Methods:We evaluated inter-episode intervals (cycle-length) in 128 clinically-treated, DSM-IVbipolar-I disorder patients followed prospectively and systematically over 5.7 years, with 6.5episodes/person.Results: As expected, cycle-length varied inversely with total cycle-count/person; however,multivariate linear regression found only longer initial hospitalization and fewer total cyclesto be associated with cycle-length, whereas cycle-number (1, 2, 3, etc.), sex, intake-age, andfirst-episode polarity were not. Regression of within-subject cycle-length versus cycle-numberyielded individual slope-functions with pseudo-random distribution (28% fell within±1 month/cycle of the null [zero-slope]). Mean duration of early and late euthymic intervals(cycles 2 vs. 5) in patients with matched recurrence-counts was nearly identical.Conclusions: The course of bipolar-I disorder from onset was largely random or chaotic overnearly 6 years from onset. Only a minority of patients showed either cycle-acceleration orslowing, without changes in wellness intervals. The findings may be influenced by treat-ment-effects, but seem to indicate that most current bipolar-I disorder patients are unlikelyto show progressive shortening of recurrence-cycles.

© 2011 Elsevier B.V. All rights reserved.

Keywords:Bipolar disorderCycle-lengthFirst-episodesProgressionRecurrencesWellness-intervals

1. Introduction

It is evidently widely believed that bipolar I disorder is typ-ically a progressive condition, marked by shortening of well-ness intervals and of inter-episode cycles with the passage ofyears andwith rising episode-counts. This concept iswidely as-cribed by analogy to the process of induction and worsening of

312, McLean Hospital,tes. Tel.: +1 617 855

(R.J. Baldessarini).

ll rights reserved.

epileptic seizures in some species of laboratory animals, and isincluded in the most recent edition of the definitive workon the disorder (Goodwin and Jamison, 2007). However, ourclinical impression was that many bipolar I disorder patientsfollowed over years appeared to follow an erratic and unpre-dictable course. Our suspicion that progressive worsening orcycle-shortening may not be a typical feature of a majority ofbipolar I disorder patients was further stimulated by a transla-tion and review (Oepen et al., 2004) of a report in Germanby Slater (1938). He found that the origin of the concept of ac-celeration in manic-depressive illness from the initialanalyses by Emil Kraepelin in the 1890s may include a major

150 R.J. Baldessarini et al. / Journal of Affective Disorders 136 (2012) 149–154

sampling artifact. It is that there are fewer persons with higherrecurrence-counts, making sampling incomparable with morenumerous recurrences, with a need to adjust or control for ep-isode-count in assessments of cycle-progression (Oepen et al.,2004; Slater, 1938).

Given uncertainty about the hypothesis of routinely pro-gressiveworseningwith rising recurrences of bipolar I disorder,we carried out a systematic review of findings on this topicreported over the past century (initially for manic-depressiveillness and later bipolar disorder) in a total of 40 studies. Thesummarized results of this analysis (Table 1) are strikingly in-consistent even in studies with adjustment or correction forthe sampling artifact noted by Slater (Oepen et al., 2004). Theremaining uncertainty pertaining to this important hypothesis,and its implications for long-term morbidity and clinical man-agement of bipolar disorder patients, especially from illness-onset, led us to study findings from the Harvard InternationalFirst-Episode Project, with efforts to control for potential sam-pling bias.

2. Methods

2.1. Subjects and clinical assessments

The study-protocolwas approved and reviewed annually bythe McLean Hospital IRB since 1989, and subjects providedwritten, informed consent for study-participation and for anon-ymous and aggregate reporting of information obtained. Clini-cal methods employed were detailed previously (Baldessariniet al., 2010; Khalsa et al., 2008; Salvatore et al., 2007, 2009;Tohen et al., 2000, 2003). Patients were evaluated initially atfirst-lifetime hospitalizations for a first episode of major affec-tive or psychotic illness, and followed prospectively and sys-tematically at intervals ranging from 1 week to 6 months overseveral years, with SCID-based assessments at intake and at24 months supporting final consensus diagnoses of DSM-IVtype I BPD. Two-year consensus diagnoses were employed to

Table 1Studies of cycle-shortening in manic-depressive patients.

Shortening Minor or a

Kraepelin (1921)a Swift (190Paskind (1930)a MacDonalAngst et al. (1968, 1973)b Pollock (1Taschev (1974)c Rennie (19Dunner et al. (1980)c LundquistZis et al. (1980)c Perris andAngst (1981)a Dunner etCutler and Post (1982)b Roy-ByrneMarneros et al., 1991b Tohen et aGitlin et al., 1995c Goldberg aVan Londen et al., 1998c Goldberg eKessing et al. (1998), Kessing andAndersen (1999), Kessing et al. (1999)b

Angst and

Kessing and Andersen (1999)b Cusin et alN=15 (37.5%) N=13 (32

The 40 studies reviewed include 15 (37.5%) with evidence of cycle-shortening withreported by Slater (1938)); 13 (32.5) with evidence of minor, ambiguous, or ageerror); and 12 (30.0%) finding little or no support for the hypothesis that cyclessampling error). Definitions: a. all patients included; b. probably avoided samplmethods unspecified. Most studies reported since the 1980s have considered bipol

minimize the instability of some initial DSM-IV diagnosesover time (Salvatore et al., 2009).

At follow-up assessments, interval weekly morbidity/well-ness status was reconstructed in life-charts, based on both in-person and by-telephone, semi-structured interviews of pa-tients, supplemented by interviews with relatives and treatingphysicians, as well as medical records. Patients were followedand treated by their individual clinicians based on prevailingcommunity standards. In this naturalistic, follow-up study,treatments were not controlled by experimental protocols orstandard requirements. For patientswhohad three ormore ep-isodes of illness to indicate possible trends, we estimated time(months) from the start of an index major episode to the startof the next (cycle-length) as the primary outcome measure.

2.2. Data analyses

Data were analyzed initially by linear regression of estimat-ed cycle-length versus cycle-count, pooling data for all patients,regardless of matching of cycle-counts/patient. We also usedmultiple linear regressionmodeling to compute slope functions(beta [β]-coefficients) with their 95% confidence intervals (CI)for cycle-length (outcome) versus the following covariates:[a] sex, [b] intake age, [c] onset illness-polarity, [d] durationof initial hospitalization, and [e] total number of recurrences/patient. The multivariate modeling was carried out for cyclesinvolving major DSM-IV episodes only (mania, hypomania,mixed-states, major depression) and then with minor episodesincluded (major episodes plus dysthymia,mild or brief hypoma-nia, ormildmixed-states that did not fulfill all DSM-IV diagnosticcriteria, though considered clinically significant departures fromeuthymia), as detailed elsewhere (Baldessarini et al., 2010). Wealso carried out regression of cycle-length versus cycle-count foreach patient to estimate individual slope functions, which wereevaluated for frequency distribution, as illustrated with a histo-gram of proportion of patients versus slope (cycle-length/cycle-count). Potential changes in euthymic intervals were eval-uated as described below.

ge-limited Random course

7)a Lewis (1936)a

d (1918)a Slater (1938)b

931)a Bratfos and Haug (1968)c

42)a Winokur (1975)a

(1945)a Fukuda et al. (1993)c

d'Elia (1966)a Winokur et al. (1993)a

al. (1979)c Winokur et al. (1994)a

et al. (1985)b Goldberg et al. (1995a)a

l. (1990)a Haghighat (1996)c

nd Harrow (1994)a Turvey et al. (1999)c

t al. (1995b)a Solomon et al. (2000)a

Sellaro (2000)b,* Baldessarini (2011)b,*

. (2000)c

.5%) N=12 (30.0%)

rising recurrence-count (only 6 appear to have avoided the sampling error-related cycle-shortening (only 2 appear to have avoided Slater's samplingshortened with more recurrences (4 probably avoided or managed Slater'sing error reported by Slater (Oepen et al., 2004; Slater, 1938); c. analyticar-I disorder patients. * The present report.

151R.J. Baldessarini et al. / Journal of Affective Disorders 136 (2012) 149–154

3. Results

Patients meeting the stated inclusion criteria with multiplecycles of DSM-IV bipolar-I illness, and long-term follow-upfrom first episodes, totaled 128 adults (57.0% men, 43.0%women) of average age-at-intake of 30.1±11.3 years (median26 years, range 18–68 years). Initial DSM-IV illness episodesincluded mania (57.8%), mixed manic-depressive states(35.2%), or major depression (7.03%). Duration of follow-upobservations averaged 5.71±2.24 years (median=6.10[interquartile-range=3.90] years).

Initial simple regression of cycle-length versus cycle-countacross all subjects yielded amodest but significant correlation(r) of −0.123 (F=9.38, p=0.002), as expected, and with asubstantial negative slope of −3.05 (CI: –5.02 to −1.10)months/cycle, to suggest possible overall cycle-shortening.However, this method of analysis probably is inappropriate(Oepen et al., 2004; Slater, 1938) since sampling was unbal-anced for cycle-count, with progressively fewer patientsrepresented at higher numbers of recurrences (from 128 per-sons with 3 to only 10 with 6 recurrences).

To avoid analyses likely to be affected by inequitable sam-pling across cycle-counts, we initially employed multivariatelinear regression modeling (Table 2). Such modeling foundfew factors to be significantly and independently associatedwith cycle-length, whether based on recurrences of DSM-IVmajor episodes or of all episodes of illness. Factors relatedto greater cycle-length included fewer total episodes orcycles/patient, as anticipated by Slater (1938), and possiblyalso longer initial hospitalization. Factors not related tocycle-length included sex, intake age, and first-episode polar-ity. Notably, cycle number (first, second, third, etc.) was notrelated to cycle-length (Table 2).

We also determined slope-functions (cycle-length/cyclenumber) by bivariate linear regression of cycle-length versuscycle-number for each of the 128 patients (Fig. 1). This meth-od indicted a pseudo-random frequency distribution, withslope functions for nearly half of subjects falling below(53.9%) or above (46.1%) the null value of zero-slope. Howev-er, the slopes deviated from a normal Gaussian distribution,in that an excess of subjects fell very close to the null valueof zero (52% had slopes within ±1.0 month/cycle of zero).

Since cycle-length can be affected by either episode-duration or the length of wellness intervals, and given theevidence of a lack of systematic change in cycle-length, wetested further for specific changes in the duration of euthy-mic intervals. We restricted sampling to patients with six re-currences (5 cycles) from illness-onset, and evaluated the

Table 2Multiple linear regression model of factors associated with greater cycle-length in

Factor β-Coefficient [95% CI]

Longer initial hospitalization +0.308 [+0.09 to +0.5Fewer cycles/patient −2.83 [−4.90 to −0.75Cycle number −1.52 [−3.92 to +0.88

The model shown included cycles between either major or minor illness-episodes;unrelated to cycle-length (β-coefficient=−0.392 [CI: −0.943 to +0.155], t=associated with cycle-length (β=2.91 [CI: −0.484 to +6.30]; t=1.60, p=0.09).sex, [b] age-at-intake, and [c] polarity of first-lifetime episode (manic, mixed, or de

duration of their earliest and latest periods of emotional sta-bility. The euthymic periods sampled averaged from cycles2 and 5; their respective durations averaged (±SD) 401±361 versus 391±396 days, respectively, indicating a veryminor decrement of 2.5% that was nonsignificant (byANOVA, F [df=1; 172]=0.030, p=0.863).

4. Discussion

The present findings are instructive, in that they verify thepredictions of Slater (1938), dating from observations reportedby Kraepelin (Oepen et al., 2004), that sampling is biasedwhenall patient-subjects with recurrent manic-depressive disordersare pooled and analyzed for association of episode-numberand inter-episode cycle-length. This bias evidently reflectsa tendency for fewer patients (different persons) to remainin an initial sample as the total recurrence or cycle countrises, indicating a need to adjust or control for episode- orcycle-number (first, second, third, etc.). We attempted toavoid biasing effects of higher episode-counts by use of multi-variate regression modeling. This approach indicated a strongassociation of cycle-length with total cycle-count (shorter av-erage cycle-length with rising number of recurrences, possiblyreflecting sampling error), but not for cycle-number. These re-sults were similar whether DSM-IV major episodes or theseplus sub-syndromal morbid states were included in the defini-tion of recurrences (Table 2). We also found that individualslope functions (cycle-length/cycle-number) followed a pseu-do-random distribution, with an excess of values close to thenull of zero (slope neither increasing nor decreasing as cycle-count increased; Fig. 1).

Potential limitations of the present findings include lack ofcontrol over treatment, which may well modify the courseof bipolar I disorder, and consideration of only the initial peri-od of risk (up to approximately six years from onset with hos-pitalization) of the patients studied—both possibly limitinggeneralizability of the findings. Although long-term mood-stabilizing treatmentsmaymodify the course of bipolar disor-ders, to what extent such effects may be reflected in reducedrecurrence rates, episode-severity, episode-duration, or lon-ger euthymic intervals remains uncertain (Baldessarini andTarazi, 2005). Even if treatment introduces confounds, largesamples of bipolar I disorder patients followed for years with-out treatment are highly unlikely to be found. Although cyclesinclude both episode-duration and euthymic intervals, sepa-rate analyses indicated a lack of effect of rising episode-countson euthymic intervals, and it follows that episode duration(cycle-length minus euthymic interval) also was not affected.

first-episode bipolar I disorder patients.

t-value p-value

3] 2.74 0.006] 2.68 0.008] 1.24 0.22

with only major episodes considered cycle-number (first, second, etc.) was1.41, p=0.16), and duration of first-lifetime hospitalization only weaklyOther factors not associated with cycle length in either model included: [a]pressive).

0

5

10

15

20

25

30

Pro

po

rtio

n o

f C

ases

(%

)

(120) (105) (90) (74) (59) (44) (29) (14) 1 17 32 47 62 77 92 108 123 138 153 168

Slope (Cycle-length vs. count, weeks)

Normal

Observed

0–100 –50 –25 +25 +50 +75–75 +100 +125

gniwolsnoitarelecca

Fig. 1. Histogram of distribution (% of cases) of slopes (based on linear regression of cycle-length in weeks versus cycle-count or number [first, second, third, etc.])for 128 individual bipolar-I disorder patients with ≥3 recurrent episodes of different polarities (5.5 episodes/patient, total of 1132 episodes and 577 cycles, overan average of 5.7 years). In most cases the relationship of cycle-length and cycle-count was chaotic, whether based on major or minor recurrences. Nearly half ofcases (45.1%) tended to accelerate (shorter cycles with increasing recurrence-count) and a similar proportion (54.7%) tended toward slowing (longer cycles). Forcomparison, a computed normal Gaussian distribution of the same data is shown in the bell-shaped curve (dotted line). Compared to the normal distribution,there was a several-fold excess of cases with little or no tendency toward altered cycle-length (28.1% of patients were within ±1 month of the null or modalvalue of zero slope).

152 R.J. Baldessarini et al. / Journal of Affective Disorders 136 (2012) 149–154

It is also important to emphasize that previous studies(Table 1), employing a variety of data-acquisition and analyticmethods, have yielded strikingly inconsistent findings aboutthe relationship of cycle-count to cycle-length, even with ef-forts to address the sampling artifact pointed out by Slater(1938; Oepen et al., 2004). It may be that the apparently near-ly randomor even chaotic course of bipolar I disorders thatweobserved (Fig. 1) may contribute to the varied outcomes ofprevious studies (Table 1).

Although the present findings did not support the hypothe-sis that increasing episode-count would be associatedwith shortening inter-episode cycle-length, some caveats arewarranted. These include the following points. First, not all bi-polar I disorder patients show a lack of progression (cycle-shortening) over time or recurrence-count, as indicated by in-consistent previous findings (Table 1) and that only a minorityof the current patients showed evidence of cycle-shortening(and others, cycle-lengthening; Fig. 1). Second, even withoutprogressive worsening of the illness-course, increased totalmorbidity over years of illness is likely to have an increasinglysevere clinical and personal impact on bipolar I patents, proba-bly contributing to substance-abuse, disability, and mortality

(Baethge et al., 2008; Huxley and Baldessarini, 2007; Tondoet al., 2007; Wingo et al., 2010).

In conclusion, we did not find support for predominant,progressiveworsening of illness-course in the early years of bi-polar I disorder, as there was no evidence of decreasing time ineuthymia, rising episode-duration, or cycle-shortening overlonger follow-up with more recurrences. The findings suggesta typically more nearly random or even chaotic course of bipo-lar I disorder over years of illness. Nevertheless, it remains ofinterest to study patients with apparently progressive illnesses(who may represent as many as 40% of first-episode, DSM-IVbipolar I disorder patients; Fig. 1) as a potential subgroup-of-special-interest. Such patients should be evaluated for possiblebiological, clinical, or treatment-response differences fromother bipolar I disorder patients.

Role of funding source

Supported in part by a grant from the Bruce J. Anderson Foundation andthe McLean Private Donors Bipolar Disorder Research Fund (to RJB), by agrant from NARSAD (to PS), and by the Atlas Foundation (to MT). Thesefunding sources had no influence on the design, analysis, or reporting ofthis study.

153R.J. Baldessarini et al. / Journal of Affective Disorders 136 (2012) 149–154

Conflict of interestDr. Baldessarini has no recent financial relationships with industrial or-

ganizations, is not a member of corporate speakers' panels, and neither henor any immediate family member holds equity relationships with pharma-ceutical or biotechnology corporations. Dr. Tohen was employed by Eli Lillyto 2008, has recently received honoraria from or consulted for Astra Zeneca,BristolMyers-Squibb, and GlaxoSmithKline, Johnson & Johnson, Sepracor,and Wyeth Corporations; his spouse is a current employee and minor stock-holder at Eli Lilly. Other coauthors have no relevant disclosures bearing onpotential conflicts of interest regarding this study.

AcknowledgmentsSupported in part by a grant from the Bruce J. Anderson Foundation and

the McLean Private Donors Bipolar Disorder Research Fund (to RJB), by agrant from NARSAD (to PS), and by the Atlas Foundation (to MT).

References

Angst, J., 1981. Course of affective disorders. In: van Praag, H.M., Lader, M.,Rafaelson, O.J., Sachar, E.J. (Eds.), Handbook of Biological Psychiatry.Part IV. Brain Mechanisms and Abnormal Behavioral Chemistry. MarcelDekker, New York, pp. 225–242.

Angst, J., Sellaro, R., 2000. Historical perspectives and natural history of bipo-lar disorder. Biological Psychiatry 48, 445–457.

Angst, J., Grof, P., Hippius Hpoeldinger,W.,Weis, P., 1968. La psychosemaniaco-depressive est-elle periodique ou intermittente. In: de Ajuriaguerra, J.(Ed.), Cycles Biologiques et Psychiatrie. George & Cie, Geneva, pp. 339–351.

Angst, J., Baastrup, P., Grof, P., Hippius, H., Poeldinger, W., Weis, P., 1973.Course of monopolar depression and bipolar psychoses. Psychiatria,Neurologia, Neurochirurgia 76, 489–500.

Baethge, C., Hennen, J., Khalsa, H.-M.K., Salvatore, P., Tohen, M., Baldessarini,R.J., 2008. Sequencing of substance use and mood changes in 166 first-episode bipolar I disorder patients. Bipolar Disorders 10, 738–741.

Baldessarini, R.J., Tarazi, F.I., 2005. Drugs for psychosis and mania. Chapt 18,In: Brunton, L.L., Lazo, J.S., Parker, K.L. (Eds.), Goodman and Gilman'sGoodman and Gilman's The Pharmacological Basis of Therapeutics, elev-enth edition. McGraw-Hill Press, New York, pp. 461–500.

Baldessarini, R.J., Salvatore, P., Khalsa, H.M., Gebre-Medhin, P., Imaz, H.,González-Pinto, A., Perez, J., Cruz, N., Maggini, C., Tohen, M., 2010. Mor-bidity in 303 first-episode bipolar I disorder patients. Bipolar Disorders12, 264–270.

Bratfos, O., Haug, J.O., 1968. Course of manic-depressive psychosis. Follow-up investigation of 215 patients. Acta Psychiatrica Scandinavica 44,89–112.

Cusin, C., Serretti, A., Lattuada, E., Mandelli, L., Smeraldi, E., 2000. Impact ofclinical variables on illness time course in mood disorders. PsychiatryResearch 97, 217–227.

Cutler, N.R., Post, R.M., 1982. Life-course of illness in untreated manic-depressive patients. Comprehensive Psychiatry 23, 101–115.

Dunner, D.L., Murphy, D., Stallone, F., Fieve, R.R., 1979. Episode frequencyprior to lithium treatment in bipolar manic-depressive patients. Com-prehensive Psychiatry 20, 511–515.

Dunner, D.L., Murphy, D., Stallone, F., Fieve, R.R., 1980. Cyclicity of affectivedisorders and its modifications by drugs: affective episode frequencyand lithium therapy. Psychopharmacology Bulletin 16, 49–50.

Fukuda, K., Etoh, T., Iwadate, T., Atsushi, I., 1993. Course and prognosis ofmanic-depressive psychosis: quantitative analysis of episodes and inter-vals. The Tohoku Journal of Experimental Medicine 139, 299–307.

Gitlin, M.J., Swendsen, J., Heller, T.L., Hammen, C., 1995. Relapse and impair-ment in bipolar disorder. The American Journal of Psychiatry 152,1635–1640.

Goldberg, J.F., Harrow, M., 1994. Kindling in bipolar disorders: longitudinalfollow-up study. Biological Psychiatry 35, 70–72.

Goldberg, J.F., Harrow, M., Grossman, L.S., 1995a. Course and outcome in bi-polar affective disorder: longitudinal follow-up study. The AmericanJournal of Psychiatry 152, 379–384.

Goldberg, J.F., Harrow, M., Grossman, L.S., 1995b. Recurrent affective syn-dromes in bipolar and unipolar mood disorders at follow-up. The BritishJournal of Psychiatry 166, 382–385.

Goodwin, F.K., Jamison, K.R., 2007. Manic-Depressive Illness, second edition.Oxford Press, New York.

Haghighat, R., 1996. Lifelong development of risk of recurrence in depressivedisorders. Journal of Affective Disorders 41, 141–147.

Huxley, N., Baldessarini, R.J., 2007. Disability and its treatment in bipolar dis-order patients. Bipolar Disorders 9, 183–196.

Kessing, L.V., Andersen, P.K., 1999. Effect of episodes on recurrence in affec-tive disorder: case register study. Journal of Affective Disorders 53,225–231.

Kessing, L.V., Andersen, P.K., Mortensen, P.B., Bolwig, T.G., 1998. Recurrencein affective disorder: case register study. The British Journal of Psychia-try 172, 23–28.

Kessing, L.V., Olsen, E.W., Andersen, P.K., 1999. Recurrence in affective disor-der: analyses with frailty models. American Journal of Epidemiology149, 404–411.

Khalsa, H.M.K., Salvatore, P., Hennen, J., Tohen, M., Baldessarini, R.J., 2008.Suicidal events and accidents in 216 first-episode bipolar-I disorder pa-tients: predictive factors. Journal of Affective Disorders 106, 179–184.

Kraepelin, E., 1921. Manic-Depressive Insanity and Paranoia. E & S Living-stone, Edinburgh, pp. 137–138.

Lewis, A., 1936. Manic-depressive psychosis. The Journal of Mental Science82, 488–558.

Lundquist, G., 1945. Prognosis and course in manic-depressive psychoses:follow-up study of 319 first-admissions. Acta Psychiatrica et Neurologica20 (Suppl. 35), 56–68.

Macdonald, J.B., 1918. Prognosis in manic-depressive insanity. The Journal ofNervous and Mental Disease 47, 20–30.

Marneros, A., Deister, A., Rohde, A., 1991. Affective, schizoaffective and schizo-phrenic psychoses. A comparative long-term study. Monographien ausdem Gesamtgebiete der Psychiatrie 65, 1–454.

Oepen, G., Salvatore, P., Baldessarini, R.J., 2004. On the periodicity of manic-depressive insanity by Eliot Slater (1938): translation and commentary.Journal of Affective Disorders 78, 1–9.

Paskind, H.A., 1930. Manic-depressive psychosis in private practice: lengthof the attack and length of the interval. Archives of Neurology and Psy-chiatry 23, 789–794.

Perris, C., D'Elia, G., 1966. Study of bipolar (manic-depressive) and unipolarrecurrent depressive psychoses. Acta Psychiatrica Scandinavica 42(Suppl 194), 153–171.

Pollock, H.M., 1931. Recurrence of attacks in manic-depressive psychoses.The American Journal of Psychiatry 11, 567–574.

Rennie, T.A.C., 1942. Prognosis in manic-depressive psychoses. The AmericanJournal of Psychiatry 98, 801–814.

Roy-Byrne, R., Post, R.M., Uhde, T.W., Porcu, T., Davis, D., 1985. Longitudinalcourse of recurrent affective illness: life-chart data from research pa-tients at the NIMH. Acta Psychiatrica Scandinavica 71 (Suppl. 317),1–34.

Salvatore, P., Khalsa, H.M.K., Hennen, J., Tohen, M., Yurgelun-Todd, D., Casolari,F., De Panfilis, C., Maggini, C., Baldessarini, R.J., 2007. Psychopathology fac-tors in first-episode affective and nonaffective psychotic disorders. Journalof Psychiatric Research 41, 724–736.

Salvatore, P., Baldessarini, R.J., Tohen, M., Khalsa, H.M.K., Perez Sanchez-Toledo, J., Zarate Jr., C.A., Vieta, E., Maggini, C., 2009. McLean-Harvard In-ternational First-Episode Project: two-year stability of DSM-IV diagno-ses in 500 first-episode psychotic disorder patients. The Journal ofClinical Psychiatry 70, 458–466.

Slater, E., 1938. Zur Periodik des manisch-depressiven Irreseins. Zeitschriftfür die Gesamte Neurologie und Psychiatrie 162, 794–801.

Solomon, D.A., Keller, M.B., Leon, A.C., Mueller, T.I., Lavori, P.W., Shea, M.T.,Coryell, W., Warshaw, M., Turvey, C., Maser, J.D., Endicott, J., 2000. Mul-tiple recurrences of major depressive disorder. The American Journal ofPsychiatry 157, 229–233.

Swift, H.M., 1907. Prognosis of recurrent insanity of the manic-depressivetype. American Journal of Insanity 64, 311–326.

Taschev, T., 1974. Course and prognosis of depression on the basis of 652 pa-tients deceased. In: Angst, J. (Ed.), Classification and Prediction of Out-come of Depression. FK Schattauer Verlag, New York, pp. 156–172.

Tohen, M., Waternaux, C.M., Tsuang, M.T., 1990. Outcome in mania: 4-yearprospective follow-up of 75 patients utilizing survival analysis. Archivesof General Psychiatry 47, 1106–1111.

Tohen, M., Hennen, J., Zarate Jr., C.A., Baldessarini, R.J., Strakowski, S.M., Stoll,A.L., Faedda, G.L., Suppes, T., Gebre-Medhin, P., Cohen, B.M., 2000. TheMcLean First Episode Project: two-year syndromal and functional recov-ery in 219 cases of major affective disorders with psychotic features. TheAmerican Journal of Psychiatry 157, 220–228.

Tohen, M., Zarate Jr., C.A., Hennen, J., Khalsa, H.M., Strakowski, S.M., Gebre-Medhin, P., Salvatore, P., Baldessarini, R.J., 2003. The McLean-HarvardFirst-Episode Mania Study: prediction of recovery and first recurrence.The American Journal of Psychiatry 160, 2099–2107.

Tondo, L., Lepri, B., Baldessarini, R.J., 2007. Risks of suicidal ideation, attemptsand suicides among 2826 men and women with types I and II bipolar,and recurrent major depressive disorders. Acta Psychiatrica Scandina-vica 116, 419–428.

Turvey, C.L., Coryell, W.H., Solomon, D.A., Leon, A.C., Endicott, J., Keller, M.B.,Akiskal, H., 1999. Long-term prognosis of bipolar-I disorder. Acta Psy-chiatrica Scandinavica 99, 110–119.

Van Londen, L., Molenaar, R.P.G., Goekoop, J.G., Zwinderman, A.H., Rooijmans,H.G.M., 1998. Three- to five-year prospective follow-up of outcome inmajor depression. Psychological Medicine 28, 731–735.

154 R.J. Baldessarini et al. / Journal of Affective Disorders 136 (2012) 149–154

Wingo, A.P., Baldessarini, R.J., Holtheimer, P.E., Harvey, P.D., 2010. Factors as-sociated with functional recovery in bipolar disorder patients. BipolarDisorders 177, 131–134.

Winokur, G., 1975. The Iowa 500: heterogeneity and course in manic-depressive (bipolar) illness. Comprehensive Psychiatry 16, 125–131.

Winokur, G., Coryell, W., Keller, M., Endicott, J., Akiskal, H., 1993. Prospectivefollow-up of patients with bipolar and primary unipolar affective disor-der. Archives of General Psychiatry 50, 457–465.

Winokur, G., Coryell, W., Akiskal, H.S., Endicott, J., Keller, M., Mueller, T.,1994. Manic-depressive (bipolar) disorder: course in light of a prospec-tive ten-year follow-up of 131 patients. Acta Psychiatrica Scandinavica89, 102–110.

Zis, A.P., Grof, P., Webster, M.A., Goodwin, F.K., 1980. Cyclicity of affectivedisorders and its modifications by drugs: prediction of relapse in recur-rent affective disorder. Psychopharmacology Bulletin 16, 47–49.