Department of Pharmacotherapeutics, University of Oslo, Oslo, Norway

Department of Mathematics, University of Oslo, Oslo, Norway

Abstract

Background

Inclusion of unpublished data on the effects of antidepressants on children has suggested unfavourable risk-benefit profiles for some of the drugs. Recent meta-analyses of studies on adults have indicated similar effects. We obtained unpublished data for paroxetine that have so far not been included in these analyses.

Methods

The documentation for drug registration contained 16 studies in which paroxetine had been randomised against placebo. We registered the number of suicides, suicide attempts and ideation. We corrected for duration of medication and placebo treatment and used a standard Bayesian statistical approach with varying priors.

Results

There were 7 suicide attempts in patients on the drug and 1 in a patient on placebo. We found that the probability of increased intensity of suicide attempts per year in adults taking paroxetine was 0.90 with a "pessimistic" prior, and somewhat less with two more neutral priors.

Conclusion

Our findings support the results of recent meta-analyses. Patients and doctors should be warned that the increased suicidal activity observed in children and adolescents taking certain antidepressant drugs may also be present in adults.

Background

The debate about whether the use of antidepressant drugs increases suicidal activity has recently been sharpened after more than 10 years of turmoil

In a February 19^{th }BMJ editorial

Methods

We included only double blind, parallel design studies with patients (all adults) randomised to either paroxetine or placebo. Altogether 16 studies met these criteria (references 79 to 93 and 95 in the Expert Report), containing respectively 916 and 550 paroxetine and placebo treated patients. The study period was in most instances 6 weeks. One important exception was a study (reference 91) with a preponderance of paroxetine use over placebo and lasting for 17 weeks. Patients were excluded from the studies after a suicide-related event. Taking this censoring into account, paroxetine treatment made up 190.7 patient years altogether and placebo 73.3 patient years. Suicide-related events could be found in tables in the Expert Report, in the adverse reactions section in the individual study reports, and in the individual patient descriptions.

We let θ_{p }be the intensity per year of a suicide attempt in the placebo group and θ_{d }the intensity per year in the drug group, for a random patient in the 16 studies; correspondingly, X_{p }and X_{d }represent the total numbers of suicide attempts. We can have at most one suicide attempt for each patient. Taking this censoring into account, we denoted the corresponding patient years in the 16 studies combined by m_{p }and m_{d}. In addition, patients in both the placebo and drug groups are supposed to behave in a similar manner. It then follows that the likelihood of the experiment corresponds to X_{p }and X_{d }having Poisson distributions respectively with parameters (m_{p}θ_{p}) and (m_{d }θ_{d}). In addition, we assume that the two variables were conditionally independent given the parameters. The corresponding observed data are (x_{p, }m_{p}) and (x_{d, }m_{d}), and the prior information is denoted by (x^{o}_{p}, m^{o}_{p}) and (x^{o}_{d}, m^{o}_{d}).

The Bayesian approach is based on the construction of probability distributions for θ_{p }and θ_{d }. This does not mean that these parameters are to be interpreted as random variables, but our knowledge of the parameters is uncertain and we describe this uncertainty with the help of probability distributions. Probability distributions describing our initial uncertainty are called prior distributions (that is, before real data are collected). When the real data are taken into account, the prior distributions are updated by Bayes' formula to posterior distributions. An excellent introduction to Bayesian methods in medicine is given by Spiegelhalter et al.

We assume that the prior distribution for θ_{p }is gamma, with parameters x^{o}_{p }and m^{o}_{p}, while correspondingly θ_{d }has the parameters x^{o}_{d }and m^{o}_{d }and is assumed to be independent of the prior distribution for θ_{p}. Hence, standard Bayesian theory gives the posterior distribution of θ_{p }as gamma, with parameters x^{o}_{p }+ x_{p }and m^{o}_{p }+ m_{p, }while θ_{d }will have the parameters x^{o}_{d }+ x_{d }and m^{o}_{d }+ m_{d}. We performed simulations by making 80000 random draws of θ_{d }and θ_{p }from their independent gamma posterior distributions, computed the logarithms of the ratios θ_{d}/θ_{p, }and constructed diagrams by applying a standard density estimation technique to these logarithms. (The logarithm was introduced to avoid an unwelcome feature of the density estimation method.) Note that the logarithm of the ratio θ_{d}/θ_{p }is greater than zero whenever θ_{d }is greater than θ_{p}. Hence, we calculated the probabilities that medication with paroxetine is associated with an increased intensity of a suicide attempt per year as the proportions of logarithmic ratios greater than zero in the samples. This corresponds to areas below the densities to the right of zero in the diagrams.

The grounds for a pessimistic prior have been given by Healy and Whitaker ^{o}_{d}) events with paroxetine during 50 (m^{o}_{d}) patient years and one (x^{o}_{p}) with placebo during 50 (m^{o}_{p}) patient years, adding up to 3 attempts per 100 patient years, which is similar to our observed average value for paroxetine and placebo taken together. We based the calculations on a total of only 100 (m^{o}_{d }+ m^{o}_{p}) patient years in the prior, compared to 264 (m_{d }+ m_{p}) patient years in the real data, in order to increase the importance of the real data over the prior information. The slightly optimistic and slightly pessimistic priors represent respectively a paper by Lapierre

Results

There were no suicides in the 16 studies with paroxetine randomised against placebo. Suicidal activities are listed in table

Suicide attempts and ideation in randomised clinical trials with paroxetine against placebo. Extracted from "APPLICATION FOR MARKETING AUTHORIZATION: SEROXAT" 1989. * Referring to list in Part I, Volume 3

**Patient identification number**

**Study reference number***

**Suicidal attempt**

**Suicidal ideation**

**Medication**

**Tabulated**

**Individually described**

02 01 009

79

X

Placebo

Yes

Yes

02 04 089

82

X

Paroxetine

Yes

Yes

03 002 034

84

X

Placebo

No

Yes

04 02 056

84

X

Paroxetine

Yes

Yes

1 09 021

90

X

Washout

Yes

Yes

09 01A 005

91

X

Paroxetine

Yes

Yes

09-01A-006

91

X

Paroxetine

No

Yes

09 01E 260

91

X

Paroxetine

Yes

No

09 01J 573

91

X

Paroxetine

Yes

Yes

09-OU-620

91

X

Paroxetine

No

Yes

09-01G-405

91

X

Paroxetine

No

Yes

037

93

X

Paroxetine

No

Yes

07 01A 001

95

X

Paroxetine

Yes

Yes

The three prior distributions of the logarithm of the ratio θ_{d}/θ_{p }are shown in figure

Prior intensity of suicidal attempt

**Prior intensity of suicidal attempt. **Distributions of three different prior (see text) logarithmic intensity ratios ln(θd/θp) (logarithmic intensity of a suicide attempt on drug minus logarithmic intensity of a suicide attempt on placebo).

Posterior intensity of suicidal attempt

**Posterior intensity of suicidal attempt. **Distributions of posterior logarithmic intensity ratios ln(θ_{d}/θ_{p}) with three different priors.

Discussion

We believe that the chosen studies are similar enough to be pooled for analysis. This view is supported by the similarities of the protocols for the various studies, although the populations that were studied differed considerably. We also believe it is best to count patient years rather than patients, although claims have been made for the contrary

Another statistical approach would have been to express the prior distributions in terms of independent priors for θ_{p }and the ratio θ_{d}/θ_{p}, with a common, relatively weak prior for θ_{p }in all three formulations. This was our approach in a previous publication using basically the same method

Conclusion

Although we report a small data set, by taking various priors into account the data strongly suggest that the use of SSRIs is connected with an increased intensity of suicide attempts per year. The two meta-analyses and our contribution taken together make a strong case for the conclusion, at least with a short time perspective, that adults taking antidepressants have an increased risk of suicide attempts. We also conclude that the recommendation of restrictions on the use of paroxetine for children and adolescents recently conveyed by regulatory agencies

Competing interests

The author(s) declare that they have no competing interests.

Authors' contributions

IA collected the data, presented the problem and drafted the manuscript along with BN, who suggested the statistical solution based on earlier work by the present authors

Acknowledgements

Journalist Ane Hoyem, the Norwegian Broadcasting Corporation (NRK), encouraged this investigation while researching for the medical information program "Puls".

Pre-publication history

The pre-publication history for this paper can be accessed here: