- Email: [email protected]
Antidepressant prescribing and suicide rate in Northern Ireland
European Psychiatry 18 (2003) 325–328 www.elsevier.com/locate/eurpsy
Original article
Antidepressant prescribing and suicide rate in Northern Ireland Christopher B. Kelly a,*, Tanzeel Ansari a, Thérèse Rafferty b, Mike Stevenson c a
Department of Mental Health, Queen’s University Belfast, Whitla Medical Building, 97 Lisburn Road, Belfast BT9 7BL, Northern Ireland, UK b Regional Prescribing Information Unit, Department of Therapeutics and Pharmacology, Queen’s University Belfast, 97 Lisburn Road, Belfast BT9 7BL, Northern Ireland, UK c Department of Epidemiology and Public Health, Queen’s University Belfast, Riddel Hall, Stranmillis Road, Belfast, UK Received 13 September 2002; received in revised form 20 February 2003; accepted 19 March 2003
Abstract Purpose. – Although antidepressants are the most commonly used treatment for depressive illness, there is uncertainty if their use is associated with a reduction in suicide rate. Antidepressant prescribing in Northern Ireland has increased over fivefold in the decade 1989–1999. The authors sought to explore whether this increase was associated with a reduction in suicide rate taking into account social and political factors thought also to have an influence on suicide. Materials and methods. – Factors that have been suggested to influence suicide were entered into a linear regression with frequency of suicide and undetermined deaths (referred to as suicide rate) as the dependent variable. The above factors were antidepressant prescribing, unemployment rate, household alcohol expenditure and persons charged with terrorist offences. The rise in younger suicides, in recent decades, suggests this analysis should be carried out separately for younger (less than 30 years) and older (30 years and above) suicides separately. The predictors in the two models are based on aggregate data for the total group. Result. – In the younger group there was no association between antidepressant prescribing and suicide. For the older group increased antidepressant prescribing was associated with a reduction in suicide rate over the 10 years of the study. Conclusion. – Increasing antidepressant prescribing appears to be an effective strategy for reducing suicide. This has been demonstrated in older individuals. © 2003 Éditions scientifiques et médicales Elsevier SAS. All rights reserved. Keywords: Suicide; Antidepressants; Unemployment
1. Introduction Suicide has been identified as a preventable cause of death and targeted for reduction in the UK and Europe. Psychological autopsy techniques have shown that a high percentage of suicides suffer from mental disorders and in particular depressive illness [5]. One prospective study has shown that increased detection and treatment of depressive illness was associated with a reduced suicide rate in a small rural area of Sweden [18]. Larger prospective studies of the effects of antidepressant treatment on suicide rate have proved impractical. In recent years educational programmes have raised awareness about the diagnosis and treatment of depressive illness. This, together with a range of new antidepressant agents marketed, has produced a marked increase in antidepressant prescribing in many countries. We hypothesised that * Corresponding author. E-mail address: [email protected] (C.B. Kelly). © 2003 Éditions scientifiques et médicales Elsevier SAS. All rights reserved. doi:10.1016/j.eurpsy.2003.03.005
the increase in antidepressant prescribing in Northern Ireland in the years 1989–1999 would be associated with a reduction in the suicide rate. In view of the higher reported frequency of depressive illness in older suicides and the worrying recent rise in younger suicides [2], data should be analysed separately for these groups. As several social factors have been suggested to alter suicide rate, these should be controlled for in any analysis. 2. Materials and methods 2.1. Data collection The number of recorded cases of both suicide and undetermined deaths were obtained for Northern Ireland in each of the years 1989–1999 inclusive. This information was released by the General Register Office in Belfast, where it is collated from Northern Ireland Coroners following inquests. In Northern Ireland, the Regional Prescribing Information Unit (RPIU), at Queen’s University Belfast, monitors pre-
326
C.B. Kelly et al. / European Psychiatry 18 (2003) 325–328
scribing practices throughout Northern Ireland, using data supplied from the Central Services Agency (CSA). When scripts are taken to the pharmacy, they are coded by the pharmacist and submitted monthly for payment to the CSA. The data is felt to be an accurate reflection of actual drug use. The RPIU was able to supply details of all antidepressants prescribed annually from 1989 to 1999. Drugs were classified according to the World Health Organisation Anatomical Chemical Classification System. To monitor the utilisation of drugs, the Defined Daily Dose (DDD) System is an internationally accepted technique [13]. It is defined as the assumed average dose per day for a drug used for its main indication in adults. The DDD is a technical unit of measurement and does not reflect the prescribed or actual daily dose. Nevertheless the DDD method has stood the test of time and remains the best method available for its purpose. A variety of social factors such as unemployment and alcohol use have been associated with suicide rate in some [10], but not all studies [15]. It was felt that such factors should be included in any analysis of suicide trends. Percentage unemployment figures were obtained, for the relevant years, from the Department of Economic Development for Northern Ireland. Household alcohol expenditure was obtained from the Northern Ireland Family Expenditure Survey for the years 1989–1998 only. Due to the changing security situation in Northern Ireland over the years 1989–1999, which is not present in other European countries, an indicator of terrorist activity (persons charged with terrorist offences, security situation statistics from the Royal Ulster Constabulary) was also included in the analysis. 2.2. Statistical analysis Separate analyses were carried out for younger suicides (and undetermined deaths) below 30 years of age and for those 30 years and above. These figures were adjusted for the population at risk, using yearly census estimates for each age group. Predictors entered into a linear regression model were antidepressant prescribing (log DDD in thousands), unemployment rate, household alcohol expenditure and persons
charged with terrorist offences. Frequency of suicide and undetermined deaths (referred to as suicide rate) was the dependent variable. Recent reports suggest that undetermined or open deaths should be included in suicide research studies for greater accuracy [11]. A problem of analysis with opportunistic data-sets such as these is that one may merely report on parallel, but unconnected trends over time. To adjust for this, all variables that had a statistically significant linear time component had that component removed. This was necessary for both suicide age groups and all explanatory variables except terrorist offences. Statistical significance was set at P < 0.05 (twotailed). Non-parametric statistics were not used for the primary analysis as they would not have allowed inclusion of the above social factors and removal of linear time components. 3. Results Data for the years 1989–1999 is shown in Table 1. For the younger group (below 30 years) the overall significance of the model was P = 0.03 (F = 5.48, adjusted R2 = 0.47). There was an inverse association between the frequency of persons charged with terrorist offences and suicide rate (t = –2.99, P = 0.02). There was a trend for a positive relationship between unemployment and suicide but this did not reach statistical significance (t = 1.96, P = 0.09). There was no statistically significant association between suicide and either alcohol expenditure or antidepressant prescribing for this age group. For the older age group (30 years and above) there was an inverse relationship between antidepressant prescribing and suicide (t = –2.90, P = 0.02). Unemployment also showed a negative association with suicide (t = –3.61, P = 0.007). Other factors entered were not statistically significant. The overall significance of the model for the older age group was P = 0.02 (F = 6.89, adjusted R2 = 0.54). If the most stringent Bonferroni correction was used, given that four variables were examined, there would be a
Table 1 Antidepressant prescribing and demographic data for Northern Ireland population (1989–1999) Year
Population 0–29 (years)
Population 30+ (years)
S&U 0–29 (years)
S&U 30+ (years)
1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999
778 690 773 632 772 690 776 741 777 885 775 794 770 014 769 137 766 715 761 696 752 617
811 745 821 963 834 605 847 856 860 389 872 118 884 866 899 940 913 576 926 920 939 207
39 38 40 38 59 59 46 44 50 54 59
95 130 109 90 92 92 101 99 88 96 95
S & U, suicides and unexplained deaths.
Unemployment Alcohol Persons rate (%) expenditure charged with (Pounds Sterling/ terrorist week) offences 13.98 5.97 433 12.74 8.13 383 12.85 7.36 404 13.70 9.18 418 13.64 9.85 372 12.58 9.42 349 11.23 12.64 440 10.75 9.21 595 8.11 10.49 405 7.35 12.00 459 6.52 – 294
Antidepressant prescribing (DDD’s) 4962 5678 6408 7669 9158 11 214 13 478 15 664 21 958 24 680 28 182
C.B. Kelly et al. / European Psychiatry 18 (2003) 325–328
trend toward statistical significance for the overall model in both age groups (P = 0.12, under 30 years; P = 0.08, 30 years and above).
4. Discussion This study demonstrates a statistically significant association between increased antidepressant prescribing and fall in suicide rate in the population over 30 years of age. Unemployment was also inversely associated with suicide in this age group. Suicide in the population younger than 30 years was significantly and negatively related to the number of persons charged with terrorist offences. It is perhaps surprising that there is relatively little published evidence on the efficacy of antidepressants in preventing suicide. Published antidepressant trials (either of acute treatment or of continuation/maintenance) are usually either too small in number or too brief to give a meaningful assessment of effect on suicide rate. One previously published epidemiological study has shown an association between greater prescribing of antidepressant drugs and reduced suicide rate in Sweden between 1990 and 1996. These findings were extended to other Scandinavian countries [8]. However, this study used multiple simple non-parametric correlations without controlling for social factors and compensating for multiple testing. In Hungary there has also been a decline in suicide rate (by over 30% between 1984 and 1998) in association with a substantial rise in antidepressant prescribing [17]. One other epidemiological study in Italy (1988–1996) showed no statistically significant relationship between antidepressant use and suicide rate [1]. There are three possible reasons for the discrepancy between the above national studies. Firstly it may be easier to demonstrate associations in countries where the suicide rate is relatively high (Scandinavia and Hungary) than those with a traditionally low rate (Italy). Secondly there is considerable evidence for under-prescribing of antidepressants to those with depressive illness. This is also true with respect to suicide. It is estimated that at least 50% of suicides have suffered from a depressive disorder [7] yet toxicological analysis for the presence of antidepressants in consecutive series of completed suicides show low rates between 8% and 15% [16]. With such a low base of antidepressant use, it is probably that effects on suicide rate would only be seen after a considerable increase in prescribing. It is notable that this study and the reports from Sweden and Hungary showing a relationship between increased antidepressant prescribing and reduction in suicide occurred during periods when antidepressant prescription rose by over fivefold, while in the negative report from Italy change in prescribing was at a much lower level. Third it is possible that trends in social factors may also influence suicide rate. A variety of factors such as unemployment [14] and alcohol use [6] have been implicated. In looking at the effects of antidepressant prescribing on suicide rate, it is important to control the above
327
factors. This has not been done in the epidemiological studies mentioned earlier. Although antidepressants are the most commonly used treatment for depressive illness there is better evidence for the prophylactic effect of lithium on suicide and suicidal behaviour [19]. One prospective study, on the island of Gotland, has shown an effect of increased detection and treatment of depressive illness on suicide rate [18]. However, the population involved was small and the resources and training intensive. Our current study confirms the finding of earlier Scandinavian and Hungarian studies and expands them by controlling for social factors and time trends. Taken together these studies suggest that significantly increasing antidepressant prescribing is associated with a lower suicide rate and may be a realistic method of helping achieve targets for suicide reduction. In this study there was no association between antidepressant prescribing and younger suicides. In the absence of knowledge of prescribing volume between the two age groups, this association should be considered tentative as it is not fully clear if it is related to inefficacy or to insufficient treatment. 4.1. Data selection The study relied on available population based data-sets, which limited the total information for analysis. Age and sex-group specific data were not available from the prescription unit. Also no information on the prevalence of psychopathological diagnoses in the community, over the time period of the study, could be obtained. Our original hypothesis involved the effects of antidepressant prescribing on suicide rate in different age groups. We were unable to break down the social factors investigated and prescribing information by age, nonetheless many of the social factors, e.g. family alcohol expenditure, terrorist offences, have impact regardless of age across a community. The above data refinements would have aided the analysis if available. The analysis of suicides was split into those under 30 years and 30 years and above based on two important findings. Firstly suicide in Britain and Ireland has risen steeply in the 15–24-year-old group in the last 20 years. Trends for the 25–34 year age group have also been upward but less marked [2]. For this reason separation at 30 years was chosen to capture the group of suicides which have shown the most recent change, including the youngest half of the 25– 34 group. The separation at 30 years is a compromise to avoid missing relevant changes in the 25–34 group, while not making the cut off age too old. Secondly psychological autopsy data suggest that diagnoses of affective disorders in suicides rise with age [3,4]. Prescription of antidepressants would most likely influence suicide in groups with higher rates of affective illness. Alcohol abuse is thought to have a significant influence in between one fifth to one half of suicides [6]. We choose to use family alcohol expenditure as a marker of population alcohol use. Other markers are available such as drink-driving con-
328
C.B. Kelly et al. / European Psychiatry 18 (2003) 325–328
victions, alcohol related psychiatric/medical admissions but because of significant changes in policing policy and policies related to alcohol related admissions, over the time of the study, these proxy measures were avoided. It is known that alcohol consumption in a population effects the prevalence of alcohol abuse (the Lederer effect). Using expenditure on alcohol as a marker of abuse would seem plausible, therefore, given the limitations of other data sources. Northern Ireland is virtually unique in Western Europe in having suffered considerable civil disturbance over the past 30 years. It is known that when the “troubles” began in Northern Ireland the suicide rate declined [12]. It was suggested that this reflected greater inter-dependence within religious and political groups. We felt such a known influence on suicide rate could not be ignored and should be incorporated into the study. During the years of the study the nature of the civil disturbance has altered and incidents of shooting or bombing would not have reflected the level of background terrorist activity. Due to the nature of the legal system, convictions do not necessarily reflect the time point of the related incident and were avoided. The security statistic of persons charged with terrorist offences offered the best and most current indicator of violent disturbance related to the “troubles”. It is interesting that the suicide rate for those under 30 years was inversely related to the number of persons charged with terrorist offences. This parallels Lyon’s findings in Belfast in 1972 and may suggest that the continuing fall in terrorist offences may tend to elevate suicide rate in those under 30 years, at least in the short term. Unemployment was positively associated with suicide rate in the younger group, at trend level, but significantly and inversely associated for those over 30 years. The relationship between suicide and unemployment is complex, with a full discussion beyond the scope of this paper. However, studies showing positive, negative and no associations have been reported [14]. Gender is an important factor in suicide. Suicide rates are higher among men yet depression is more common among women. Women are more apt to seek help and receive a prescription for an antidepressant. Unfortunately because the prescription data were not sex-specific we could not investigate the influence of gender further in this study. 5. Conclusion Increased antidepressant prescribing at significant levels (fivefold from a low baseline) is associated with reduced suicide rate in those over 30 years. Although there is no association between antidepressant prescribing and suicide for those under 30 years, the ecological nature of the analysis does not allow a definitive conclusion that antidepressant prescribing has no effect on suicide rate in younger people. However, given the psychopathological differences between older and younger suicides, found in psychological autopsy studies, it is possible that different approaches to suicide reduction may be required for separate age groups.
It is important to separate the efficacy of antidepressants as potential agents to reduce suicide with the difficulties in recognition of depressive illness and prescription of antidepressants at service level (efficiency). The above findings suggest antidepressants are efficacious at reducing suicide, difficulties in effectively prescribing them may require service delivery changes [9].
References [1]
[2]
[3]
[4]
[5]
[6] [7]
[8] [9]
[10] [11] [12] [13] [14]
[15]
[16]
[17] [18]
[19]
Barbui C, Campomani A, D’Avanzo B, Negri E, Gorattini S. Antidepressant drug use in Italy since the introduction of the SSRI’s: national trends, regional differences and impact on suicide rates. Soc Psychiatry Psychiatr Epidemiol 1999;34:152–6. Cantor C. Suicide in the western world. In: Hawton K, et al., editors. International handbook of suicide and attempted suicide. Chichester: Wiley; 2000. p. 9–28. Carlson GA, Rich CL, Grayson P, Fowler RC. Secular trends in psychiatric diagnoses of suicide victims. J Affect Disord 1991;21: 127–32. Cornwell Y, Duberstein PR, Cox C, Herrman JH, Forbes NT, Caine ED. Relationship of age and axis I diagnoses in victims of completed suicide: a psychological autopsy study. Am J Psychiatry 1996;153:1001–8. Foster T, Gillespie K, McClelland R, Patterson C. Risk factors for suicide independent of DSM-III-R Axis I disorder. Case control psychological autopsy study in Northern Ireland. Br J Psychiatry 1999; 175:175–80. Foster T. Dying for a drink. Br Med J 2001;323:817–8. Isacsson G, Bergman U, Rich CL. Epidemiological data suggests antidepressants reduce suicide risk among depressives. J Affect Disord 1996;41:1–8. Isacsson G. Suicide prevention—a medical breakthrough? Acta Psychiatr Scand 2000;102:113–7. Katon W, van Korff M, Lin E, et al. Stepped collaborative care for primary care patients with persistent symptoms of depression. Arch Gen Psychiatry 1999;56:1109–15. Kupinski J, Tiller JWG, Burrows GD, Hallenstein H. Youth suicide in Victoria: a retrospective study. Med J Aust 1994;160:113–6. Linsley KR, Schapira K, Kelly TP. Open verdict v suicide— importance to research. Br J Psychiatry 2001;178:465–8. Lyons HA. Depressive illness and aggression in Belfast. Br Med J 1972;1:342–4. Oydvin K, Kristinnson A. Guidelines for DDD. World Health Organisation Collaborating Centre for Drug Statistics Methodology; 1991. Platt S, Hawton K. Suicidal behaviour and the labour market. In: Hawton K, et al., editors. International handbook of suicide and attempted suicide. Chichester: Wiley; 2000. p. 309–84. Platt S, Micciolo R, Tansella M. Suicide and unemployment in Italy: description, analysis and interpretation of recent trends. Soc Sci Med 1992;34:1191–201. Rich CL, Isacsson G. Suicide and antidepressants in South Alabama: evidence for improved treatment of depression. J Affect Disord 1997; 45:135–43. Rihmer Z. Can better recognition and treatment of depression reduce suicide rates? A brief review. Eur Psychiatry 2001;16:406–9. Rutz W, Van Knorring L, Walinder J. Long-term effect of an educational programme for general practitioners given by the Swedish Committee for the Prevention and Treatment of Depression. Acta Psychiatr Scand 1992;85:83–8. Tondo L, Baldessarini RJ, Hernen J, Flavis G, Silvetti F, Tohen M. Lithium treatment and risk of suicidal behaviour in bipolar disorder patients. J Clin Psychiatry 1998;59:405–14.
Original article
Antidepressant prescribing and suicide rate in Northern Ireland Christopher B. Kelly a,*, Tanzeel Ansari a, Thérèse Rafferty b, Mike Stevenson c a
Department of Mental Health, Queen’s University Belfast, Whitla Medical Building, 97 Lisburn Road, Belfast BT9 7BL, Northern Ireland, UK b Regional Prescribing Information Unit, Department of Therapeutics and Pharmacology, Queen’s University Belfast, 97 Lisburn Road, Belfast BT9 7BL, Northern Ireland, UK c Department of Epidemiology and Public Health, Queen’s University Belfast, Riddel Hall, Stranmillis Road, Belfast, UK Received 13 September 2002; received in revised form 20 February 2003; accepted 19 March 2003
Abstract Purpose. – Although antidepressants are the most commonly used treatment for depressive illness, there is uncertainty if their use is associated with a reduction in suicide rate. Antidepressant prescribing in Northern Ireland has increased over fivefold in the decade 1989–1999. The authors sought to explore whether this increase was associated with a reduction in suicide rate taking into account social and political factors thought also to have an influence on suicide. Materials and methods. – Factors that have been suggested to influence suicide were entered into a linear regression with frequency of suicide and undetermined deaths (referred to as suicide rate) as the dependent variable. The above factors were antidepressant prescribing, unemployment rate, household alcohol expenditure and persons charged with terrorist offences. The rise in younger suicides, in recent decades, suggests this analysis should be carried out separately for younger (less than 30 years) and older (30 years and above) suicides separately. The predictors in the two models are based on aggregate data for the total group. Result. – In the younger group there was no association between antidepressant prescribing and suicide. For the older group increased antidepressant prescribing was associated with a reduction in suicide rate over the 10 years of the study. Conclusion. – Increasing antidepressant prescribing appears to be an effective strategy for reducing suicide. This has been demonstrated in older individuals. © 2003 Éditions scientifiques et médicales Elsevier SAS. All rights reserved. Keywords: Suicide; Antidepressants; Unemployment
1. Introduction Suicide has been identified as a preventable cause of death and targeted for reduction in the UK and Europe. Psychological autopsy techniques have shown that a high percentage of suicides suffer from mental disorders and in particular depressive illness [5]. One prospective study has shown that increased detection and treatment of depressive illness was associated with a reduced suicide rate in a small rural area of Sweden [18]. Larger prospective studies of the effects of antidepressant treatment on suicide rate have proved impractical. In recent years educational programmes have raised awareness about the diagnosis and treatment of depressive illness. This, together with a range of new antidepressant agents marketed, has produced a marked increase in antidepressant prescribing in many countries. We hypothesised that * Corresponding author. E-mail address: [email protected] (C.B. Kelly). © 2003 Éditions scientifiques et médicales Elsevier SAS. All rights reserved. doi:10.1016/j.eurpsy.2003.03.005
the increase in antidepressant prescribing in Northern Ireland in the years 1989–1999 would be associated with a reduction in the suicide rate. In view of the higher reported frequency of depressive illness in older suicides and the worrying recent rise in younger suicides [2], data should be analysed separately for these groups. As several social factors have been suggested to alter suicide rate, these should be controlled for in any analysis. 2. Materials and methods 2.1. Data collection The number of recorded cases of both suicide and undetermined deaths were obtained for Northern Ireland in each of the years 1989–1999 inclusive. This information was released by the General Register Office in Belfast, where it is collated from Northern Ireland Coroners following inquests. In Northern Ireland, the Regional Prescribing Information Unit (RPIU), at Queen’s University Belfast, monitors pre-
326
C.B. Kelly et al. / European Psychiatry 18 (2003) 325–328
scribing practices throughout Northern Ireland, using data supplied from the Central Services Agency (CSA). When scripts are taken to the pharmacy, they are coded by the pharmacist and submitted monthly for payment to the CSA. The data is felt to be an accurate reflection of actual drug use. The RPIU was able to supply details of all antidepressants prescribed annually from 1989 to 1999. Drugs were classified according to the World Health Organisation Anatomical Chemical Classification System. To monitor the utilisation of drugs, the Defined Daily Dose (DDD) System is an internationally accepted technique [13]. It is defined as the assumed average dose per day for a drug used for its main indication in adults. The DDD is a technical unit of measurement and does not reflect the prescribed or actual daily dose. Nevertheless the DDD method has stood the test of time and remains the best method available for its purpose. A variety of social factors such as unemployment and alcohol use have been associated with suicide rate in some [10], but not all studies [15]. It was felt that such factors should be included in any analysis of suicide trends. Percentage unemployment figures were obtained, for the relevant years, from the Department of Economic Development for Northern Ireland. Household alcohol expenditure was obtained from the Northern Ireland Family Expenditure Survey for the years 1989–1998 only. Due to the changing security situation in Northern Ireland over the years 1989–1999, which is not present in other European countries, an indicator of terrorist activity (persons charged with terrorist offences, security situation statistics from the Royal Ulster Constabulary) was also included in the analysis. 2.2. Statistical analysis Separate analyses were carried out for younger suicides (and undetermined deaths) below 30 years of age and for those 30 years and above. These figures were adjusted for the population at risk, using yearly census estimates for each age group. Predictors entered into a linear regression model were antidepressant prescribing (log DDD in thousands), unemployment rate, household alcohol expenditure and persons
charged with terrorist offences. Frequency of suicide and undetermined deaths (referred to as suicide rate) was the dependent variable. Recent reports suggest that undetermined or open deaths should be included in suicide research studies for greater accuracy [11]. A problem of analysis with opportunistic data-sets such as these is that one may merely report on parallel, but unconnected trends over time. To adjust for this, all variables that had a statistically significant linear time component had that component removed. This was necessary for both suicide age groups and all explanatory variables except terrorist offences. Statistical significance was set at P < 0.05 (twotailed). Non-parametric statistics were not used for the primary analysis as they would not have allowed inclusion of the above social factors and removal of linear time components. 3. Results Data for the years 1989–1999 is shown in Table 1. For the younger group (below 30 years) the overall significance of the model was P = 0.03 (F = 5.48, adjusted R2 = 0.47). There was an inverse association between the frequency of persons charged with terrorist offences and suicide rate (t = –2.99, P = 0.02). There was a trend for a positive relationship between unemployment and suicide but this did not reach statistical significance (t = 1.96, P = 0.09). There was no statistically significant association between suicide and either alcohol expenditure or antidepressant prescribing for this age group. For the older age group (30 years and above) there was an inverse relationship between antidepressant prescribing and suicide (t = –2.90, P = 0.02). Unemployment also showed a negative association with suicide (t = –3.61, P = 0.007). Other factors entered were not statistically significant. The overall significance of the model for the older age group was P = 0.02 (F = 6.89, adjusted R2 = 0.54). If the most stringent Bonferroni correction was used, given that four variables were examined, there would be a
Table 1 Antidepressant prescribing and demographic data for Northern Ireland population (1989–1999) Year
Population 0–29 (years)
Population 30+ (years)
S&U 0–29 (years)
S&U 30+ (years)
1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999
778 690 773 632 772 690 776 741 777 885 775 794 770 014 769 137 766 715 761 696 752 617
811 745 821 963 834 605 847 856 860 389 872 118 884 866 899 940 913 576 926 920 939 207
39 38 40 38 59 59 46 44 50 54 59
95 130 109 90 92 92 101 99 88 96 95
S & U, suicides and unexplained deaths.
Unemployment Alcohol Persons rate (%) expenditure charged with (Pounds Sterling/ terrorist week) offences 13.98 5.97 433 12.74 8.13 383 12.85 7.36 404 13.70 9.18 418 13.64 9.85 372 12.58 9.42 349 11.23 12.64 440 10.75 9.21 595 8.11 10.49 405 7.35 12.00 459 6.52 – 294
Antidepressant prescribing (DDD’s) 4962 5678 6408 7669 9158 11 214 13 478 15 664 21 958 24 680 28 182
C.B. Kelly et al. / European Psychiatry 18 (2003) 325–328
trend toward statistical significance for the overall model in both age groups (P = 0.12, under 30 years; P = 0.08, 30 years and above).
4. Discussion This study demonstrates a statistically significant association between increased antidepressant prescribing and fall in suicide rate in the population over 30 years of age. Unemployment was also inversely associated with suicide in this age group. Suicide in the population younger than 30 years was significantly and negatively related to the number of persons charged with terrorist offences. It is perhaps surprising that there is relatively little published evidence on the efficacy of antidepressants in preventing suicide. Published antidepressant trials (either of acute treatment or of continuation/maintenance) are usually either too small in number or too brief to give a meaningful assessment of effect on suicide rate. One previously published epidemiological study has shown an association between greater prescribing of antidepressant drugs and reduced suicide rate in Sweden between 1990 and 1996. These findings were extended to other Scandinavian countries [8]. However, this study used multiple simple non-parametric correlations without controlling for social factors and compensating for multiple testing. In Hungary there has also been a decline in suicide rate (by over 30% between 1984 and 1998) in association with a substantial rise in antidepressant prescribing [17]. One other epidemiological study in Italy (1988–1996) showed no statistically significant relationship between antidepressant use and suicide rate [1]. There are three possible reasons for the discrepancy between the above national studies. Firstly it may be easier to demonstrate associations in countries where the suicide rate is relatively high (Scandinavia and Hungary) than those with a traditionally low rate (Italy). Secondly there is considerable evidence for under-prescribing of antidepressants to those with depressive illness. This is also true with respect to suicide. It is estimated that at least 50% of suicides have suffered from a depressive disorder [7] yet toxicological analysis for the presence of antidepressants in consecutive series of completed suicides show low rates between 8% and 15% [16]. With such a low base of antidepressant use, it is probably that effects on suicide rate would only be seen after a considerable increase in prescribing. It is notable that this study and the reports from Sweden and Hungary showing a relationship between increased antidepressant prescribing and reduction in suicide occurred during periods when antidepressant prescription rose by over fivefold, while in the negative report from Italy change in prescribing was at a much lower level. Third it is possible that trends in social factors may also influence suicide rate. A variety of factors such as unemployment [14] and alcohol use [6] have been implicated. In looking at the effects of antidepressant prescribing on suicide rate, it is important to control the above
327
factors. This has not been done in the epidemiological studies mentioned earlier. Although antidepressants are the most commonly used treatment for depressive illness there is better evidence for the prophylactic effect of lithium on suicide and suicidal behaviour [19]. One prospective study, on the island of Gotland, has shown an effect of increased detection and treatment of depressive illness on suicide rate [18]. However, the population involved was small and the resources and training intensive. Our current study confirms the finding of earlier Scandinavian and Hungarian studies and expands them by controlling for social factors and time trends. Taken together these studies suggest that significantly increasing antidepressant prescribing is associated with a lower suicide rate and may be a realistic method of helping achieve targets for suicide reduction. In this study there was no association between antidepressant prescribing and younger suicides. In the absence of knowledge of prescribing volume between the two age groups, this association should be considered tentative as it is not fully clear if it is related to inefficacy or to insufficient treatment. 4.1. Data selection The study relied on available population based data-sets, which limited the total information for analysis. Age and sex-group specific data were not available from the prescription unit. Also no information on the prevalence of psychopathological diagnoses in the community, over the time period of the study, could be obtained. Our original hypothesis involved the effects of antidepressant prescribing on suicide rate in different age groups. We were unable to break down the social factors investigated and prescribing information by age, nonetheless many of the social factors, e.g. family alcohol expenditure, terrorist offences, have impact regardless of age across a community. The above data refinements would have aided the analysis if available. The analysis of suicides was split into those under 30 years and 30 years and above based on two important findings. Firstly suicide in Britain and Ireland has risen steeply in the 15–24-year-old group in the last 20 years. Trends for the 25–34 year age group have also been upward but less marked [2]. For this reason separation at 30 years was chosen to capture the group of suicides which have shown the most recent change, including the youngest half of the 25– 34 group. The separation at 30 years is a compromise to avoid missing relevant changes in the 25–34 group, while not making the cut off age too old. Secondly psychological autopsy data suggest that diagnoses of affective disorders in suicides rise with age [3,4]. Prescription of antidepressants would most likely influence suicide in groups with higher rates of affective illness. Alcohol abuse is thought to have a significant influence in between one fifth to one half of suicides [6]. We choose to use family alcohol expenditure as a marker of population alcohol use. Other markers are available such as drink-driving con-
328
C.B. Kelly et al. / European Psychiatry 18 (2003) 325–328
victions, alcohol related psychiatric/medical admissions but because of significant changes in policing policy and policies related to alcohol related admissions, over the time of the study, these proxy measures were avoided. It is known that alcohol consumption in a population effects the prevalence of alcohol abuse (the Lederer effect). Using expenditure on alcohol as a marker of abuse would seem plausible, therefore, given the limitations of other data sources. Northern Ireland is virtually unique in Western Europe in having suffered considerable civil disturbance over the past 30 years. It is known that when the “troubles” began in Northern Ireland the suicide rate declined [12]. It was suggested that this reflected greater inter-dependence within religious and political groups. We felt such a known influence on suicide rate could not be ignored and should be incorporated into the study. During the years of the study the nature of the civil disturbance has altered and incidents of shooting or bombing would not have reflected the level of background terrorist activity. Due to the nature of the legal system, convictions do not necessarily reflect the time point of the related incident and were avoided. The security statistic of persons charged with terrorist offences offered the best and most current indicator of violent disturbance related to the “troubles”. It is interesting that the suicide rate for those under 30 years was inversely related to the number of persons charged with terrorist offences. This parallels Lyon’s findings in Belfast in 1972 and may suggest that the continuing fall in terrorist offences may tend to elevate suicide rate in those under 30 years, at least in the short term. Unemployment was positively associated with suicide rate in the younger group, at trend level, but significantly and inversely associated for those over 30 years. The relationship between suicide and unemployment is complex, with a full discussion beyond the scope of this paper. However, studies showing positive, negative and no associations have been reported [14]. Gender is an important factor in suicide. Suicide rates are higher among men yet depression is more common among women. Women are more apt to seek help and receive a prescription for an antidepressant. Unfortunately because the prescription data were not sex-specific we could not investigate the influence of gender further in this study. 5. Conclusion Increased antidepressant prescribing at significant levels (fivefold from a low baseline) is associated with reduced suicide rate in those over 30 years. Although there is no association between antidepressant prescribing and suicide for those under 30 years, the ecological nature of the analysis does not allow a definitive conclusion that antidepressant prescribing has no effect on suicide rate in younger people. However, given the psychopathological differences between older and younger suicides, found in psychological autopsy studies, it is possible that different approaches to suicide reduction may be required for separate age groups.
It is important to separate the efficacy of antidepressants as potential agents to reduce suicide with the difficulties in recognition of depressive illness and prescription of antidepressants at service level (efficiency). The above findings suggest antidepressants are efficacious at reducing suicide, difficulties in effectively prescribing them may require service delivery changes [9].
References [1]
[2]
[3]
[4]
[5]
[6] [7]
[8] [9]
[10] [11] [12] [13] [14]
[15]
[16]
[17] [18]
[19]
Barbui C, Campomani A, D’Avanzo B, Negri E, Gorattini S. Antidepressant drug use in Italy since the introduction of the SSRI’s: national trends, regional differences and impact on suicide rates. Soc Psychiatry Psychiatr Epidemiol 1999;34:152–6. Cantor C. Suicide in the western world. In: Hawton K, et al., editors. International handbook of suicide and attempted suicide. Chichester: Wiley; 2000. p. 9–28. Carlson GA, Rich CL, Grayson P, Fowler RC. Secular trends in psychiatric diagnoses of suicide victims. J Affect Disord 1991;21: 127–32. Cornwell Y, Duberstein PR, Cox C, Herrman JH, Forbes NT, Caine ED. Relationship of age and axis I diagnoses in victims of completed suicide: a psychological autopsy study. Am J Psychiatry 1996;153:1001–8. Foster T, Gillespie K, McClelland R, Patterson C. Risk factors for suicide independent of DSM-III-R Axis I disorder. Case control psychological autopsy study in Northern Ireland. Br J Psychiatry 1999; 175:175–80. Foster T. Dying for a drink. Br Med J 2001;323:817–8. Isacsson G, Bergman U, Rich CL. Epidemiological data suggests antidepressants reduce suicide risk among depressives. J Affect Disord 1996;41:1–8. Isacsson G. Suicide prevention—a medical breakthrough? Acta Psychiatr Scand 2000;102:113–7. Katon W, van Korff M, Lin E, et al. Stepped collaborative care for primary care patients with persistent symptoms of depression. Arch Gen Psychiatry 1999;56:1109–15. Kupinski J, Tiller JWG, Burrows GD, Hallenstein H. Youth suicide in Victoria: a retrospective study. Med J Aust 1994;160:113–6. Linsley KR, Schapira K, Kelly TP. Open verdict v suicide— importance to research. Br J Psychiatry 2001;178:465–8. Lyons HA. Depressive illness and aggression in Belfast. Br Med J 1972;1:342–4. Oydvin K, Kristinnson A. Guidelines for DDD. World Health Organisation Collaborating Centre for Drug Statistics Methodology; 1991. Platt S, Hawton K. Suicidal behaviour and the labour market. In: Hawton K, et al., editors. International handbook of suicide and attempted suicide. Chichester: Wiley; 2000. p. 309–84. Platt S, Micciolo R, Tansella M. Suicide and unemployment in Italy: description, analysis and interpretation of recent trends. Soc Sci Med 1992;34:1191–201. Rich CL, Isacsson G. Suicide and antidepressants in South Alabama: evidence for improved treatment of depression. J Affect Disord 1997; 45:135–43. Rihmer Z. Can better recognition and treatment of depression reduce suicide rates? A brief review. Eur Psychiatry 2001;16:406–9. Rutz W, Van Knorring L, Walinder J. Long-term effect of an educational programme for general practitioners given by the Swedish Committee for the Prevention and Treatment of Depression. Acta Psychiatr Scand 1992;85:83–8. Tondo L, Baldessarini RJ, Hernen J, Flavis G, Silvetti F, Tohen M. Lithium treatment and risk of suicidal behaviour in bipolar disorder patients. J Clin Psychiatry 1998;59:405–14.