Sixteen studies published between 1968 and 2012 met the inclusion criteria (supplementary
), a surprisingly small number. Eleven were studies of clinical samples (6 of first-episode cases and 5 of mixed first- and non-first-episode cases) and 5 of general population samples (). Fourteen studies reported a positive association between adult life events and onset of psychotic disorder or occurrence of subclinical experiences. Within the clinical studies, cases with psychosis were over 2 times6
to 8 times7
more likely to report life events compared with controls in the period leading up to onset. In the general population studies, those with psychotic experiences (vs those without) were between 2 times8
and 7 times9
more likely to report recent life events.
Studies Investigating the Associations Between Adult Life Events and Psychosis in First-Episode Samples
Studies Investigating the Associations Between Adult Life Events and Psychosis Experiences in General Population Samples
Studies Investigating the Associations Between Adult Life Events and Psychosis in Mixed Onset and Relapse Samples
The picture is the same when only those studies6,10–15
that received a quality score of 10 or above (n
= 7) are considered, ie, 6 studies6,11–15
reported some evidence that the number and/or severity of events was associated with around a 3- to 5-fold increased risk of psychosis.
The majority of these more robust studies found life events were elevated prior to onset of psychosis, with the time period under consideration ranging between 3 months12
and 3.6 years.15
The seminal article of Brown and Birley,12
eg, found life events were increased in the 3-week period pre-onset of psychotic symptoms. The sample, however, was small (n
= 50, 13 of whom were non-first-episode), and no subsequent studies have reported similar findings. Later studies suggest that life events may exert their influence over a longer period. For example, 2 studies11,13
of disorder found life events were around 2–3 times higher in cases compared with controls across a 1-year period. Further, a general population study15
of 1722 young adults found that exposure to life events over the previous 3 years was associated with an increased risk of psychotic experiences.
Severity and Type
Some studies assessed other contextual elements of events, such as their severity and type. In a study of 97 cases (35 first episode) and 207 controls, Bebbington et al14
found that moderate and severe life events (vs mild) were higher in the 3-month period pre-onset in patients with schizophrenia (ie, 52% cases with moderate/severe events vs 10% controls).
More specifically, using a sample of 41 first-onset patients, Raune et al6
found that intrusive events, such as a physical assault or invasive operation, were more likely to be associated with an increased risk of psychosis and were most common in the 3 months pre-onset (ie, 34% cases vs 3% controls). However, for these analyses, the control sample was taken from 2 studies16,17
conducted 20 years previously.
Others have found evidence for specific effects for certain types of events. Dohrenwend et al,13
eg, in a sample of 66 schizophrenia cases (21 first onset) and 197 population-based controls, found physical illness and injury (akin to intrusive events) were around 2 times more common in cases than controls.
Independence of Events
One way of clarifying the causal relationship between events and onset is to distinguish events that are independent of emerging symptoms, eg, death of a close relative, from those which may be influenced by mental state, eg, interpersonal conflict.
Five of the more robust studies6,11–14
distinguished between possibly dependent and independent events. For example, Brown and Birley12
found 46% of cases were exposed to recent independent events compared with 14% of controls. More recent research suggests similar conclusions.6,11,14
Raune et al,6
eg, found that almost all cases (95%) experienced a life event 1 year prior to the development of symptoms, and that in 76%, these event(s) were independent. These authors also found that cases were 2 times more likely to report independent life events in the 3 months pre-onset than controls (34% cases vs 14% controls).
In contrast, Dohrenwend et al13
did not find any increase in independent events prior to onset. However, they did find a higher number of “non-fateful” events (a similar concept to dependent events, ie, events that are influenced by prior mental state and personality characteristics, such as relationship difficulties) in the year prior to onset.
Additionally, we carried out a meta-analysis of a subset of 13 studies in which the number exposed and not exposed to life events had been reported7–9,10,11–15,18–20
for more detail).
The meta-analysis yielded an overall weighted OR of 3.19 (95% CI 2.15–4.75), which suggests that individuals with psychotic disorder/experiences are roughly 3 times more likely than controls to be exposed to recent life events (). The OR from the clinical samples6–8,10–14,18,19
are higher than the general population studies,9,15,20
but this was not statistically significant. There is substantial heterogeneity between studies (Higgins’ I
= 87.27% [95% CI 70.34%–96.36%]). The heterogeneity was not removed by meta-regression using any of the 4 possible moderators (year of publication, life events period, quality score, and type of sample, ie, clinical or general population), possibly due to rather restricted variability on all of them.
Forest plot for the meta-analysis examining the overall association between recent life events and psychosis.
Across all studies, there were a number of common methodological issues that merit specific consideration. First, the majority of studies were cross sectional, introducing potential recall bias and limiting inferences concerning direction of causation. In relation to psychotic disorder, it is difficult to envisage longitudinal studies being feasible, given the low incidence of disorders. Consequently, efforts to minimize recall bias and carefully date exposure to events and onset of disorder are essential but were rarely made.
Further, not all the studies in this review included a comparison group, and, of the 10 that did,6,7,10–14,18,19,22
not all drew controls from the same populations as cases. Bias in selection of comparison group(s), therefore, cannot be excluded.
Within the clinical studies, causal interpretations are limited by the small number of first-episode only samples and by the failure of papers based on mixed samples to report findings specifically for first-episode cases. This noted, both types of study in the main reported positive associations. It is clear that more first-episode studies are needed, which utilize appropriate control groups and objective ratings of the impact of events.
Differences in life events measurement make comparisons between studies difficult and this may account for some of the variations in findings. Instruments to assess life events generally fall into 2 categories: checklist or semi-structured interview. The Life Events and Difficulties Schedule23
(LEDS) is considered the gold standard as it takes account of context, eg, timing, severity, and independence of events. It is, however, time consuming to administer and rate. Of the studies reviewed, 4 used checklists8,9,20,24
, 2 used a checklist that was interviewer administered10,18
, and 10 used semi-structured interviews,6,7,11–15,19,21,22
of which 46,11,12,14
used the LEDS.23
Of these latter studies, 3 studies6,12,14
found positive associations between recent life events and psychosis onset.
Where the severity of life events was evaluated, this was mainly determined using objective criteria and not based on subjective appraisals. This is because subjective perceptions of severity may be affected by mood and mental state, which would then risk confusing exposure and outcome and make it impossible to distinguish cause and effect.
Although most of the assessments of life events used in the reviewed studies do enquire about positive events, only one of the studies drew a distinction between positive (desirable) and negative (undesirable) events in their analyses,22
finding associations only for negative events and psychosis. No study specifically discussed the valence of events in relation to psychosis onset.
Finally, adjustment for potential confounders was inconsistent. Where adjustments were made, the majority controlled for age, gender, and ethnicity, with some controlling for a wider range of factors, such as urbanicity, education, IQ, substance use, comorbid neurosis,8
and past year cannabis use.9
No study adjusted for childhood adversity.