The lifetime risk of suicide in schizophrenia: a reexamination
Palmer BA, Pankratz VS and Bostwick JM;
Commented by , 24 Apr 2005
Aim of the study
One of the most frequently quoted epidemiological numbers in the field of schizophrenia is an approximately 10% lifetime prevalence of suicide. This textbook wisdom has usually been based on a meta-analysis published in 1977 (ref. 1) and two studies of chronic schizophrenics (ref. 2; ref. 3).
According to the authors these publications may, however, have overestimated the true suicide risk due to methodological limitations. The previous analyses based their estimation on so-called proportionate mortality (PM). This means that they analysed the percentage of the dead who died by suicide.
Such an approach can result in biased estimates by several factors: When the studies included in the previous meta-analysis had started, many patients had already survived their early years of the illness when the suicide risk seems to be especially high.
On the other hand, in many studies the patients had not been followed-up for a very long time, but with increasing age deaths due to medical reasons become more frequent than suicides. PM, however, assumes a constant rate of suicide over a life-time. Since this is probably wrong, PM could have overestimated suicide risk. The aim of the current study, therefore, was to establish a new methodology for extrapolating lifetime suicide prevalence estimates.
Methods
The authors searched MEDLINE (1966-present) for studies that observed cohorts of patients with schizophrenia and included those studies that a) had a follow-up of at least 2 years, b) with a drop-out rate below 10%, and c) that reported suicides.
Two authors independently extracted data on sample size, number of deaths, number of suicides, percentage of follow-up, and the diagnostic criteria used. Analyses were performed using two different samples - 32 studies of patients enrolled at various illness points (25,578 subjects) and 29 studies of patients at first admission or at illness onset illness onset or first admission (22,598 subjects).
Regression models considering both - proportionate mortality (PM, the percentage of the dead who died by suicide) and case fatality (CF - the percentage of the total sample who died by suicide) were used to estimate suicide risk.
Results
Compared to the previous studies the estimated lifetime suicide prevalence was much lower: 5.6% (95% confidence interval 3.7%-8.5%) in patients observed from first admission or from illness onset and 1.8% (95% confidence interval, 1.4%-2.3%) in mixed samples. The diagnostic criteria used did not have a major impact on the results.
Dr Leucht's comments
This study is a good example how difficult it is to interprete the results of epidemiological studies, and how unclear many results that we consider to be "proven medical facts" are. The authors explain quite clearly that proportionate estimates (PM) of lifetime suicide prevalence are accurate only when either all subjects are followed up until death or if the suicide rate is constant over the lifetime.
The first condition was not met by almost any study included in the previous analyses. The second assumption is probably not correct. It is more likely that patients with schizophrenia commit suicide more often in the first years after the onset of the illness than after having lived for decades with the disease.
This hypothesis probably also explains why the suicide rate in the sample of first admission patients was higher than in the mixed sample. But there are also a number of limitations of the authors’ approach. As in any meta-analysis the results are based on composite data from collections of studies, individual patient data on length of follow-up etc. were not available.
A second one is that many studies had short follow-up periods so that all data beyond these periods are mathematical extrapolations. Therefore, the ideal study would follow up a large sample of patients with schizophrenia from the onset of the disease until their death and use a survival analysis as a statistical approach.
References
1. Caldwell CB and Gottesman II. Schizophrenics kill themselves too: a review of risk factors for suicide. Schizophrenia Bulletin 1990; 16; 571-588
2. Bleuler M. The Schizophrenic Disorders: Long Term Patient and Family Studies. Clemens SM, trans. New Haven, Conn: Yale University Press; 1978
3. Tsuang MT. Suicide in schizophrenics, manics, depressives, and surgical controls: a comparison with general population suicide mortality. Archives of General Psychiatry 1978; 35; 153-155