The first stages of antidepressant treatment are associated with high drop-out rates (Bogner et al., 2006). In a recent Danish study as many as 47% of adults aged 50 years and over were found to have discontinued treatment after only one prescription of antidepressants (Erlangsen et al., 2008).

In older people reasons for interrupting treatment are fear of addiction, lack of improvement, not considering depression to be a medical illness, and viewing depression as a normal reaction to adverse events (Givens et al., 2006). Side effects, such as nausea, sexual dysfunction, nervousness, insomnia, loss of appetite (Westenberg et al., 2006), or discontinuation syndrome (Stone et al., 2007; Westenberg et al., 2006) might also lead to early discontinuation. Depending on supplemental payments, costs of the medication might influence decisions about whether to continue treatment as well.

Adults over age 50 in Denmark who were in treatment with antidepressants were found to have 4–6 times higher suicide rates than the general population in the same age range, with rates ranging between 153–273 for men and 69–113 per 100,000 for women (Erlangsen et al., 2008). A regional Canadian study, however, found much lower suicide rates among older recipients of antidepressant (Rahme et al., 2008). A 5-fold higher risk of suicide was found in older adults during their first month of treatment with SSRIs compared to TCAs (Juurlink et al., 2006).

The Register of Medicinal Product Statistics records all prescriptions redeemed at pharmacies. In Denmark, antidepressant medications are only available by prescription. Subjects were considered as continuing treatment with antidepressants as long as they kept redeeming prescriptions. Based on the number of pills in the current prescription, we calculated when the next prescription was expected to be filed. A person in ongoing treatment was assumed to consume at least 0.75 pills per day and was allowed to transfer up to 21 ‘unused’ pills from the previous prescription period (to allow for occasional misses and vacation periods). This and a comparable procedure have been recommended and applied in other studies using the same data source (Erlangsen et al., 2008; Hansen et al., 2004; Rosholm et al., 2001).

Early discontinuation was assumed to have occurred when treatment was not continued beyond the first redeemed prescription, i.e. when an individual did not redeem a second prescription of an antidepressant within the expected time frame. Persons who switched to other antidepressants within the treatment phase and continued filing prescriptions with those were considered as continuing treatment.

All persons who commenced a new treatment with antidepressants were included on the date when their first prescription was expected to be replaced by a second prescription (t0). Persons were censored on the date of the following events: 1) migration, 2) admission to psychiatric hospital (except ambulant care), 3) death, and 4) end of study period. Each individual only entered the study once.

The following time-varying covariates were used for stratification: age, marital status, pension, income, contact with general practitioners, somatic comorbidity measured by the Charlson Index (Charlson et al., 1987), drug type, whether the first psychiatric admission was within the last year or previously, whether admitted for depression, and whether the person was in current ambulatory treatment.

In all, 538 persons died by suicide (309 men and 229 women). Men who discontinued treatment early had a suicide rate of 167 per 100,000 compared with a rate of 175 per 100,000 in those who kept on redeeming prescriptions. Their hazard ratio was 0.98 [CI-95%: 0.78–1.23] as shown in Table 1. For women, the rates were 52 and 74 suicides per 100,000 respectively in those who discontinued and continued medication. Women who discontinued treatment early had a hazard ratio of 0.72 [CI-95%: 0.55–0.94]. Other factors associated with elevated suicide risk were a history of previous psychiatric hospitalization, mood disorders, and discontinuation of antidepressants in those in current ambulatory psychiatric treatment.

The rate ratio for men who discontinued treatment was 1.04 [CI-95%: 0.83 – 1.30] in the adjusted regression analysis; for women it was 0.80 [CI-95%: 0.61–1.05]. The marginal structural model showed that early discontinuation of treatment was associated with a rate ratio of 1.12 [CI-95%: 0.89 – 1.39] for men and 0.88 [CI-95%: 0.68 – 1.14] for women. The corresponding values for the propensity score matched models were 1.04 [CI-95%: 0.83 – 1.30] and 0.80 [CI-95%: 0.61–1.05] for men and women, respectively. None of the included covariates predicted early discontinuation of treatment among those who died by suicide (table not shown).

Both people who continued treatment and those who discontinued treatment early had 3–5 fold higher suicide rates than the general population of the same age range (own calculations; BEF1A, 2007; FOD5, 2007).

Men aged 50 years and over who followed treatment had a 2.3-fold higher suicide rate than women who followed treatment, while men who discontinued treatment had a 3.2 fold higher suicide rate than women over aged 50 in the same situation. In other words, men who discontinued treatment had disproportionally higher suicide rates than women. This could be related to women’s more frequent use of primary health care (Vedsted, 2007) and men’s higher suicide risk in general (Erlangsen et al., 2003).

People are prescribed antidepressants by their general practitioners for a wide range of disorders (Creed, 2006; Giuliano, 2007; Rahme et al., 2008; Sindrup et al., 1992). Previous psychiatric in-patients and current ambulatory care patients might be more likely to suffer from severe mood disorders. During a psychiatric hospitalization patients are likely to receive a clinical evaluation for depressive symptoms. People with moderate and severe major depression respond better to antidepressant treatment than those with less severe mood disorders (Blazer, 2003). Previous in-patients and current ambulatory care patients who discontinued treatment were found to have slightly higher risks of suicide compared to those who continued treatment. Although this finding was not statistically significant, it seems probable that this group might need the medication and be at higher risk of suicide when interrupting treatment. The survival curves in Figure 1 support this notion.

Individuals with previous psychiatric treatment had a higher risk of suicide than the rest of the study sample. Psychiatric hospitalization is likely to be an indicator of the severity of psychiatric disorder. The lack of specificity associated with risk factors and the low base rate generally hinders the prediction of suicides (Hawton, 1987; Knox et al., 2004; O’Connell et al., 2004). However, general practitioners usually have information on previous hospitalizations and this could be used to intensify follow-up efforts during the initial period of treatment with antidepressants.

As many as 42% of adults over age 50 seemingly interrupted treatment after just one prescription. Discontinuation can be due to lack of medical effect of the drug (Beyer, 2007) or side effects (Westenberg et al., 2006). Nevertheless, evidence-based research suggest that treatment with a single antidepressant or combination of antidepressants, cognitive behavioral therapy, other types of psychotherapy, and electroconvulsive therapy can reduce depressive symptoms in older adults (Charney et al., 2003; Steinman et al., 2007). Improved adherence to antidepressant treatment has been shown by multimodal interventions focusing on collaborative care (Vergouwen et al., 2003). Furthermore, involvement of case managers who follow and support the treatment course has been linked to lower suicidal ideation (Bruce et al., 2004).

One limitation of our study is the indicator of treatment adherence. We do not know whether pills were actually ingested nor precisely when the treatment was discontinued. Although we controlled for a wide selection of covariates, it would have been preferable to have information on risk modifiers, such as severity of depression, previous suicide attempts, and current psychiatric disorders. Also, having more details regarding the reasons for ending treatment and length of intended treatment would have enabled us to evaluate our findings in more depth. In this regard it is important to acknowledge that, as mentioned above, antidepressants are prescribed for a variety of disorders in adults over age 50.

Underreporting of suicide might be a further limitation, but the registration of suicide in Denmark is generally thought to be reliable (Kolmos et al., 1987). Our findings are not applicable to people who are currently receiving in-hospital treatment.

Advantages of the study include the prospective data collection and longitudinal setting using the entire population for selecting the study sample.