We acknowledge Dr. Kavirajan for his interest in our work and for addressing interesting methodological issues, some of which require additional information.
- First of all, it is important not to confuse statistical and causal associations. Despite our results confirming the association previously found between long-term benzodiazepine (BZD) use and an increased risk of Alzheimer’s disease (AD), we never stated that this association was direct or causal. On the contrary, we devoted a significant part of the Discussion to reviewing various alternative hypotheses including a residual protopathic bias. We even mentioned that BZDs could simply be early markers of a risk factor for AD such as depression, anxiety or insomnia. Having said that, there was no reason not to consider a possible contribution of BZDs to this excess risk of AD, at least owing to the tremendous public health impact such an option would have.
- Concerning the figure of 30% given by Dr. Karavijan for the proportion of BZD users presenting with anxiety, depression or insomnia, there may have been a misinterpretation of our Table 2 comparing the characteristics of cases and controls (and not of BZD users and non-users). Indeed, for cases the proportion of anxiety recorded as a diagnosis in the database was 21.4%. As the proportion of BZD users was less than 50% in the group of cases, it is obvious that the percentage of anxiety in those using BZDs was much higher than 21.4%. The same could be said for depression and insomnia. Moreover, given the therapeutic effects of BZDs on anxiety and insomnia, a lower incidence of these symptoms was expected in prevalent users. As to the comparison with the data of Pelissolo and colleagues,1 it does not seem to be relevant. Indeed, that study (i) was conducted in another country (France), and (ii) included persons aged 18 and over (versus 66 and over in our study), those having used BZDs for at least 6 months i.e. exceeding the duration recommended (real-life use was considered in our study), those having given their written informed consent, etc.
- No clear-cut hypothesis or conclusion can be put forward concerning a putative confounding by alcohol consumption. Confounding would only ensue from an association of alcohol with both BZD use and the risk of AD. The latter is far from being established. Indeed, according to the literature, for regular and « moderate » drinkers, the association wavers between « no effect » and « protective ». Therefore, not considering this variable for adjustment tends rather to be conservative.
- The most challenging comment concerns the as yet unresolved issue of the type of association, if any, between depression, anxiety or insomnia and AD. This point was discussed in our article. First, it seems important not to confuse the protopathic bias, i.e. BZDs having been, at least in part, prescribed to treat pre-diagnosis symptoms of AD, and the possible role of these symptoms as risk factors for this disease. To minimize this bias, it seems sensible to censor prescriptions initiated a few years before the diagnosis of AD. A censored time-window of 5 or 6 years, as in our study, seems to be a realistic compromise. Studying the possible role of depression, anxiety or insomnia identified 10, 20 years before or even more on the risk of AD is another challenge that is probably not feasible at present. To our knowledge, no database or cohort study offers a 30-year follow-up with a sample size large enough to allow subgroup analyses as well as reliable data on (i) drug exposure, (ii) comorbidities and risk factors, and (iii) cognitive disorders. Failing this, the study of interactions and the adjustment on these variables, although not perfect, seems to be the way to go. Whatever the process involved, a conclusion remains: long-term BZD use is associated with an increased risk of AD and the precautionary principle would require considering the worst-case scenario, especially when a suspected risk factor is mostly alterable.
1. Pelissolo A, Maniere F, Boutges B, Allouche M, Richard-Berthe C, Corruble E. [Anxiety and depressive disorders in 4,425 long term benzodiazepine users in general practice]. L'Encephale 2007;33(1):32-8.
Competing interests:
No competing interests
13 October 2014
Bernard Bégaud
Professor
Sophie Billioti de Gage, PhD student, U657-Pharmacoepidemiology, Université de Bordeaux, F-33000 Bordeaux, France
INSERM, U657-Pharmacoepidemiology, Université de Bordeaux, F-33000 Bordeaux, France
Université de Bordeaux, INSERM U657, 146 rue Léo Saignat, F-33076, Bordeaux cedex, France
Rapid Response:
We acknowledge Dr. Kavirajan for his interest in our work and for addressing interesting methodological issues, some of which require additional information.
- First of all, it is important not to confuse statistical and causal associations. Despite our results confirming the association previously found between long-term benzodiazepine (BZD) use and an increased risk of Alzheimer’s disease (AD), we never stated that this association was direct or causal. On the contrary, we devoted a significant part of the Discussion to reviewing various alternative hypotheses including a residual protopathic bias. We even mentioned that BZDs could simply be early markers of a risk factor for AD such as depression, anxiety or insomnia. Having said that, there was no reason not to consider a possible contribution of BZDs to this excess risk of AD, at least owing to the tremendous public health impact such an option would have.
- Concerning the figure of 30% given by Dr. Karavijan for the proportion of BZD users presenting with anxiety, depression or insomnia, there may have been a misinterpretation of our Table 2 comparing the characteristics of cases and controls (and not of BZD users and non-users). Indeed, for cases the proportion of anxiety recorded as a diagnosis in the database was 21.4%. As the proportion of BZD users was less than 50% in the group of cases, it is obvious that the percentage of anxiety in those using BZDs was much higher than 21.4%. The same could be said for depression and insomnia. Moreover, given the therapeutic effects of BZDs on anxiety and insomnia, a lower incidence of these symptoms was expected in prevalent users. As to the comparison with the data of Pelissolo and colleagues,1 it does not seem to be relevant. Indeed, that study (i) was conducted in another country (France), and (ii) included persons aged 18 and over (versus 66 and over in our study), those having used BZDs for at least 6 months i.e. exceeding the duration recommended (real-life use was considered in our study), those having given their written informed consent, etc.
- No clear-cut hypothesis or conclusion can be put forward concerning a putative confounding by alcohol consumption. Confounding would only ensue from an association of alcohol with both BZD use and the risk of AD. The latter is far from being established. Indeed, according to the literature, for regular and « moderate » drinkers, the association wavers between « no effect » and « protective ». Therefore, not considering this variable for adjustment tends rather to be conservative.
- The most challenging comment concerns the as yet unresolved issue of the type of association, if any, between depression, anxiety or insomnia and AD. This point was discussed in our article. First, it seems important not to confuse the protopathic bias, i.e. BZDs having been, at least in part, prescribed to treat pre-diagnosis symptoms of AD, and the possible role of these symptoms as risk factors for this disease. To minimize this bias, it seems sensible to censor prescriptions initiated a few years before the diagnosis of AD. A censored time-window of 5 or 6 years, as in our study, seems to be a realistic compromise. Studying the possible role of depression, anxiety or insomnia identified 10, 20 years before or even more on the risk of AD is another challenge that is probably not feasible at present. To our knowledge, no database or cohort study offers a 30-year follow-up with a sample size large enough to allow subgroup analyses as well as reliable data on (i) drug exposure, (ii) comorbidities and risk factors, and (iii) cognitive disorders. Failing this, the study of interactions and the adjustment on these variables, although not perfect, seems to be the way to go. Whatever the process involved, a conclusion remains: long-term BZD use is associated with an increased risk of AD and the precautionary principle would require considering the worst-case scenario, especially when a suspected risk factor is mostly alterable.
1. Pelissolo A, Maniere F, Boutges B, Allouche M, Richard-Berthe C, Corruble E. [Anxiety and depressive disorders in 4,425 long term benzodiazepine users in general practice]. L'Encephale 2007;33(1):32-8.
Competing interests: No competing interests