Using Danish registry data, Kessing et al examined the relationship between lithium response and the timing of treatment (early v. delayed). ^{Reference Kessing, Vradi and Andersen1} Early treatment was associated with an increased probability of lithium response. This is a clinically important finding, given the increasing emphasis on early intervention in bipolar disorder. The results of the Kessing et al study are sobering. Only few patients, particularly among those for whom treatment was delayed, responded to lithium. Several factors may have contributed to the reported results.

The study did not - and possibly could not - control for the cycle shortening that is observed after successive episodes of bipolar disorder. Although the interpretation of such cycle shortening has been debated, ^{Reference Oepen, Baldessarini, Salvatore and Slater2} it is well established that early cycles are significantly longer than those occurring later; consequently, early in the course of illness one would expect longer spontaneous remissions regardless of treatment. This effect may be partially responsible for the greater treatment response in patients receiving early intervention in the Kessing et al study.

Naturalistic studies typically demonstrate full response in about 30% of participants ^{Reference Garnham, Munro, Slaney, MacDougall, Passmore and Duffy3} (that is, no recurrences, or the Kessing et al criterion, in treatment-adherent patients), which is markedly greater than the response rate observed by Kessing et al. This discrepancy could be related to age at first contact. The average age of participants whom Kessing et al reported as having received early and late treatment was 46.7 years and 49.1 years, respectively. The natural history of bipolar disorder includes an average age at onset in the second or third decade of life. The trajectory of the illness, where mania typically develops as the last stage, delays the diagnosis of bipolar disorder. Also, there is often a substantial delay in starting treatment even following the diagnosis of bipolar disorder. ^{Reference Ortiz, Bradler, Slaney, Garnham, Ruzickova and O'Donovan4,Reference Baldessarini, Tondo and Hennen5} These reports, in conjunction with the advanced age at index presentation, and high rates of antidepressant, antipsychotic and anticonvulsant use in the Kessing et al study suggest that participants may have been afflicted with bipolar disorder for some time before ‘first contact’. In a sample of 450 participants, Baldessarini et al reported a negative relationship between treatment latency and effect of treatment on time spent ill. ^{Reference Baldessarini, Tondo and Hennen5} If the aforementioned findings are generalisable to the Danish sample, the reduced overall treatment responses may be interpreted as a consequence of relatively advanced participant age.

Finally, Kessing et al analysed data collected since 1995. Is it possible that participants had received lithium during the years prior? This would further complicate the interpretations of sample responsiveness to lithium, regardless of early or late initiation. In conclusion, we suggest that the findings presented by Kessing et al are limited by the lack of control for inter-participant differences in the manifestation of the natural history of bipolar disorder. Such control may be difficult, or in some cases impossible, to achieve using registry-based observational data, but is nevertheless imperative to understanding the effects of early v. late treatment prophylaxis in relapsing-remitting illnesses such as bipolar disorder.