Circadian variation in attempted suicide by deliberate self poisoning

BMJ 1994; 309 doi: (Published 24 September 1994) Cite this as: BMJ 1994;309:774
  1. R Manfredini,
  2. M Gallerani,
  3. S Caracciolo,
  4. A Tomelli,
  5. G Calo,
  6. C Fersini
  1. Institute of Internal Medicine, University of Ferrara Medical School, I- 44100 Ferrara, Italy Accident and Emergency Department, St Anna Hospital, Ferrara Department of Psychology and Institute of Pharmacology, University of Ferrara Medical School, Ferrara Mental Health Service, Operative Unit of World Health Organisation Parasuicide Multicentre Study, Ferrara
  1. Correspondence to Dr Manfredini.
  • Accepted 10 March 1994

Circadian patterns have been shown for many physiological variables and several medical diseases.1 Suicidal behaviours usually occur with symptoms of depression, although only a few people who attempt suicide fully satisfy diagnostic criteria for having an affective disorder. Clinical and experimental evidence suggests that affective disorders are related to disturbances in the phase or amplitude of biological rhythms - for example, sleep-wake cycles, environmental dark-light cycles, and hormonal rhythms. We determined whether a specific temporal risk exists in the occurrence of attempted suicide by deliberate self poisoning.

Patients, methods, and results

We studied prospectively all patients who attempted suicide by deliberate self poisoning and were admitted to the accident and emergency department of St Anna Hospital, Ferrara, Italy, from 1 January 1989 to 31 December 1991. This hospital alone deals with medical emergencies occuring in the city and suburban area of Ferrara (about 150 000 residents).

Attempted suicide by deliberate self poisoning was diagnosed from the clinical history (from relatives, witnesses, or the patients themselves), physical examination, and response to the benzodiazepine flumazenil. In most cases the presumptive diagnosis was then confirmed by specific toxicological assay. All patients were interviewed by a consultant psychiatrist, who recorded demographic characteristics, the nature of the drugs or substances ingested, and the time of ingestion.

Seventy one patients were men (mean age 41 (SD 18) and 141 women (mean age 39 (17)). Ninety eight patients had taken benzodiazepines, 10 antidepressants, six neuroleptics, four barbiturates, 10 sodium hypochlorite and 29 other drugs; 32 had taken a mixture of drugs and data were not available for 22.

The time of self poisoning was categorised into 24 increments of one hour, 0600 to 0659 being classed as 6 am (table). Data in the group as a whole and by sex were analysed by cosinor analysis, in which the cosine curve best fitting the data was determined by multiple linear regression.2 For rhythms detected by rejection of the zero amplitude hypothesis (at P<0.05), the cosinor procedure yields (a) the rhythm adjusted mean (or mesor) which measures the extent of the rhythm; (b) the amplitude (the distance from the mesor to the peak or through of the cosine curve best fitting the data); and (c) the acrophase (peak time of the cosine curve best fitting the data). Each of the values has an estimate of variance.

Distribution of attempts at suicide by time

View this table:

A significant circadian rhythm was found for the whole population (P=0.007, mesor 8.83) and for men (P=0.007, mesor 2.86) and women (P=0.026, mesor 5.87). The acrophases were respectively 1745 (95% confidence interval to 1432 to 2258), 1837 (1537 to 2148), and 1716 (1303 to 2129).


The sociodemographic data on our sample were consistent with published data. In particular, women attempting suicide outnumbered men by two to one. This might indicate women's greater demonstrative tendency as men outnumber women by two to one in studies of completed suicide.

Our results suggest that the risk of attempting suicide by self poisoning is greatest in the early evening, which is in agreement with previous studies.3 The risk of suicide is greatest during the late morning and early afternoon.4 As suicide and attempted suicide have different psychological causes and epidemiological risk factors, these data support the hypothesis that chronobiologically determined changes influence suicidal behaviour. The acrophases of important biological variables such as temperature, heart rate and cortisol secretion vary in people with affective disorders compared with control subjects and may affect behaviour in addition to diurnal changes in mood and social activities. Increased adrenergic activity and lowered serotoninergic activity5 in the afternoon might play a part in mood changes.

It could also be argued, however, that people who attempt suicide by minor self poisoning often do not want to die and their cry for help is more likely to be heard in the hours of the day when social and psychiatric services are available.

In conclusion, a specific temporal risk, defined as the point of confluence of many factors (hormonal, environmental and behavioural), probably plays a part in suicidal behaviour. Treatment of depressive disorders might therefore be improved by aiming for peak drug concentrations at vulnerable times.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
View Abstract