Rapid Responses to:
|
|
Rapid Responses: Rapid Responses published:
-
![[Read Rapid Response]](/icons/misc/sectionDOWN.gif)
"Safety of benzodiazepines" not supported by data on hip fractures.
- Andrew Byrne (26 March 2001)
-
Re: "Safety of benzodiazepines" not supported by data on hip fractures: author's response
- Nicholas Moore (29 March 2001)
-
"Duration of exposure to benzodiazepine" should have been sought to improve validity of the study
- S Hemanth Guptha (30 March 2001)
-
Bnezodiazepines do increase the risk of hip fracture
- R M C Herings (8 April 2001)
-
Re: Bnezodiazepines do increase... Author's response
- Nicholas Moore (11 April 2001)
-
Benzodiazepines and hip fractures: cracking evidence
- Giovanni Gambassi (20 April 2001)
|
|
|||
|
Andrew Byrne, General Practitioner, Drug and Alcohol 75 Redfern St, Redfern, 2016, NSW, Australia
Send response to journal:
|
This paper provides further worrying evidence on the dangers of benzodiazepines in the elderly, contradicting the authors' own conclusions on the safety of these drugs. Contrary to much evidence on the subject, this paper's discussion begins with the unreferenced statement: "Exposure to benzodiazepines does not increase the risk of hip fracture in people aged over 65. This applies to benzodiazepines regardless of half life". But they also state: "Only lorazepam ... was significantly associated with hip fractures [and] clobazam, loprazolam, nitrazepam, and temazepam had similarly increased odds ratios ..." and further: "Hip fracture was .. associated with the use of two or more benzodiazepines .." and "The method used to ascertain exposure could influence the results of case- control studies." On my own re-calculations of the data presented, 315 of 817 (38.6%) in the control hospital admission group were taking a benzodiazepine while 116 of 245 (47.3%) of the hip fracture cases were taking such drugs. Thus sedative- takers were over-represented in the hip fracture group. For those taking lorazepam, nitrazepam and temazepam, this over-representation was double in the fracture group [51/817 (6.2%) versus 31/245 (13%)] and appears to be statistically significant when taken as a group. The rates of benzo takers in the fracture group was triple or more for clobazam, loprazolam, prazepam and temazepam. It was double in the cases of lorazepam and nitrazepam. The rate was slightly more for the fracture group regarding bromazepam, oxazepam and zolpidem. These figures seem at odds with the authors' conclusion. Even though statistical significance may not have been reached, the data in this paper seem to be consistent with other findings that there is a strong association between benzo use and falls along with fractures. Many now believe that this association is causative, but such a conclusion cannot be deduced from a hospital case control study, however well conducted it may be. These authors draw a very long bow in suggesting that some benzodiazepines may yield a "non significant protective effect" against fractures by their 'muscle relaxant' properties. The study showed no such protection from my reading, and why they would invoke such a far-fetched scenario is beyond my comprehension, knowing the evidence pointing to harm from these drugs. I am not aware of any studies documenting muscle relaxation from normal doses of benzodiazepines and was not surprised the authors could not find comparative references either. While it is sometimes said that 'drunk people never hurt themselves', experience would seem to indicate the opposite, as any trauma centre can demonstrate. Even further muting their findings, this study excluded a number of fracture cases for whom 'controls' in the general hospital admission lists could not be found. Nearly all of these excluded cases were females over 81 years (c/f mean age 81 and 76% female). I note that fully four references are from Australia, but it is
telling and disappointing that there is no mention of
Campbell's classic pilot randomised study from New Zealand (1999)
indicating protection from falls by withdrawing prescribed sedatives.
P> Dr Andrew Byrne,
Benzodiazepines and hip fractures in elderly people: case- control study. Pierfitte C, Macouillard G, Thicoïpe M, Chaslerie A, Pehourcq F, Aïssou M, Martinez B, Lagnaoui R, Fourrier A, Bégaud B, Dangoumau J, Moore N. BMJ 2001;322:704-708 |
|||
|
|
|||
|
Nicholas Moore, Prof Clinical pharmacology University Bordeaux
Send response to journal:
|
Dear sir, I am afraid you must have misread our paper. To take your points one by one: The discussion starts with a summary of the results of our study. References come later. We did not find any association between use of benzodiazepines taken as a whole and hip fractures. In the rest of the discussion we try to understand why we found this and some possible explanations. The initial hypothesis we thought we would confirm was that there was an association, and we were as surprised as you not to find it. Our politic was then to give the whole data and not try to massage it to confirm preconceptions. We looked at indications, we looked at half-life, we looked at plasma concentrations, and found no systematic association. If the overall result is no association, and there were individual associations with some drugs, this means either - that we were guilty of misuse of statistics, and that the significant association we found with lorazepam, and the increased odds-ratios with others were spurious and the results of multiple testing; - or that if some increase the risk, and there is no overall increase, others decrease the risk. Why did you only cite those that have an increased odds ratio, and not those like clorazepate where the odds ratio is less than unity? When we state that fractures were associated with the use of two or more BZD, that was using the questionnaire and medical records data. When plasma data (the objective gold standard in this case) is used the association is decreased to the point of non-significance. This is strongly indicative of data retrieval bias, a common plague of case-control studies: it may that it was because of the hip fracture that we found 2 or more benzos in questionnaires or records, and not the other way around. It is because of this that we ended saying that how you measure things may influence the results, surely something that is not contradictory, especially in pharmacoepidemiological studies. In such studies, hard data for drug exposure such as the plasma assays we used are probably better that looking at medical records, or asking elderly patients what drugs they take, especially if this is not done carefully. In what is that contradictory with what we said earlier? I do not know where you found 116 users of benzodiazepines in cases: there were (table 2) 97 cases with benzos in the plasma, and 98 with benzos in questionaires or records, and respectively 282 and 306 controls with benzo use. The percentages concerned are in the text. How you found 47% is beyond be. If you include the unmatched cases, then it should be 116 out of 345 and not out of 245, resulting in exactly the same percentage. From there on, selective reporting of findings supporting your contention is beyond comment. If you were complete you should have stated also that clorazepate use in controls was almost twice that in cases, and that was the second most used drug after bromazepam. The drugs you cite were used in such small numbers of cases and controls as to be essentially meaningless (nitrazepam was used in 3 controls and 2 cases, loprazolam in four cases and 3 controls ...) Careful and objective reading of our paper shows that overall and whatever way you want to look at it, benzodiazepines users were OVERALL not more frequent in cases that in controls. We should have stopped there, but again because this was surprising, we looked at different subsets, though we should not have done this. Multiple testing in non-significant data will always come up with spurious results. This could be the reason for these apparently increased risks with some drugs and situations, such as 2 or more BZD, or lorazepam, as we were careful to say. "many now believe that this association is causative" - references? - "it is well known that..." Where is the data? You are making the same mistake as others, confusing falls and fractures. All studies of the association between benzodiazepines and FALLS have shown an association that is dose-dependent, etc... and probably causative. Half the studies of the association of benzos and FRACTURES are negative, and the New York State real-life interventional study was negative: no change in the rates of hip fractures, even though the use of benzos in the elderly population was halved, so that the etiological fraction is certainly very small. Why these people taking benzodiazepines fall more but don't break more is one of the subjects of our discussion. Speaking of which, confusing clinical experience with clinical trials or scientific studies is a common error. It may be true that the excessive use of alcohol increases the risk of accidents overall, as any trauma centre will say, but interestingly, no association was found between moderate drinking and hip fracture (Hoidrup et al, am j epid, 1999, 149:993-1001). Another as yet unpublished study in Korea actually shows a protective effect of moderate drinking. Since the risk of falling is increased even in moderate users, the situation is the same as that we found with benzos. We excluded non-matched cases because that was the design of our study. This did not change the results: unmatched analysis found exactly the same results, with less power. This was indicated in our results: the use of benzos in these excluded cases was not greater than in the matched cases or in controls. Finally, though I looked again on Medline this morning, I could not find the "classic" study you mention. What are the exact references? Of course, we do not say that benzos are without danger in the elderly and can be used ad lib, as a group or individually. Some seem more at risk than others, and finding why may be a way to develop safer drugs. Nicholas Moore |
|||
|
|
|||
|
S Hemanth Guptha, Specialist Registrar, General Medicine Edith Cavell Hospital, Peterborough
Send response to journal:
|
Sir Pierfitte et al found no association between benzodiazepines in plasma (except lorazepam) and hip fractures1. However their results have to be interpreted with caution. Several studies have shown an increased risk of falls in the elderly with the use of benzodiazepines 2,3. Although these studies did not ascertain exposure by checking plasma levels of the benzodiazepines, some of them were conducted in the controlled environment of a nursing home or a hospital inpatient ward under direct exposure. The risk of falling was found to be particularly high in the first few weeks of exposure to benzodiazepines and decreased with a longer duration of exposure 3. Pierfitte et al did not measure the duration of exposure to benzodiazepine in their patients before their admission with fracture or fall, which could have differentially affected the rate of falling in the two study groups, and may have possibly lead to bias. Pierfitte et al also conclude that they did not find any association with plasma level of benzodiazepine and rate of hip fracture with different benzodiazepines. But several of the benzodiazepines their patients used had active metabolites. They used high-pressure liquid chromatography with diode array to identify benzodiazepines which identifies only some of the active metabolites4. Although the plasma levels for most of the benzodiazepines did not significantly differ between the two groups, the rate of metabolism can vary widely between patients depending on genetic factors, co-existing diseases and co- prescriptions. Unless the active metabolites are accounted for, it cannot be ascertained which amongst the various benzodiazepines are more harmful than the others. References: 1) Pierfitte C, Macouillard G, Thicoipe M et al. Benzodiazepines and hip fractures in elderly people: case - control study. BMJ 2000; 322: 704- 8. 2) Passaro A, Volpato S, Ramagnoni F, Manzoli N, Zuliani G, Fellin R. Benzodiazepines with different half-life and Falling in a hospitalised population. The GIFA study. J Clin Epidemiol 2000; 53(12):1222-9. 3) Neutel CI, Hirdes JP, Maxwell CJ, Patten SB. New evidence on benzodiazepine use and falls: the time factor. Age Ageing 1996; 25(4): 273 -8. 4) He W, Parissis N, Kiratzidis T. Determination of benzodiazepines in forensic samples by HPLC with photo-diode array detection. J of forensic science; 43(5): 1061-7. |
|||
|
|
|||
|
R M C Herings, PHARMO Institute Utrecht
Send response to journal:
|
Dear Sir, With interest, I read the paper of Pierfitte et al. They conclude that use of benzodiazepines was not associated with hip fractures. This conclusion might be misleading, despite the impressive way of exposure assessment by plasma level analysis. They at least suggest that blood- plasma analysis is more accurate exposure measure than questionnaires. One might doubt if this is true for several reasons. First, they assessed plasma levels at the time of admission, which is only a proxy of the blood-plasma level the time patients fell. It may takes several hours before people are admitted to the hospital and the metabolism is active for several hours. Bromazepam, the most used benzodiazepines in their study has a T-max of only 2 hours and may degrade very fast in active metabolites that are not perse different from metabolites from other benzodiazepines. If they didn't detect bromazepam, did they conclude that the patient was not exposed to bromazepam at fall- time? In (hospital)controls it is even more unclear how we should interprete benzodiazepine plasma levels, especially because no-fall related plasma levels can be obtained at all. Timing of intake, fall and admission is not reported in this study. Moreover, results show that virtually none of the cases is admitted at night. This is a remarkable observation, because at least in the Netherlands 10-15% of all admissions for injury related to falls are admitted at night. May be in this in region in France it takes a lot of time to admit patients or there is a habit to wait until the next morning. If so, the abovementioned problem with timing of fall and admission is even more important. Secondly, if they detect more than one benzodiazepine in the plasma, did they detect and attribute active metabolites to the original used benzodiazepine or did they conclude that the patient is exposed to two or more benzodiazepines. In a patients who used diazepam, active metabolites can be detected in the blood-plasma such as temazepam, nordazepam and oxazepam. If temazepam is detected, do we attribute temazepam as original exposed to diazepam or to temazepam ? Again timing between intake, fall and admission and control selection is important for misclassification between risk estimates of individual original used benzodiazepines. Thirdly, the increased propensity of falling among benzodiazepine users is most likely a central effect. Concentrations in the brain do not linear correlate with blood-plasma levels and differ substantial between individual benzodiazepines. Given the problems with timing of intake and problems with control selection, dose-response as assessed in blood-plasma levels is a very weak predictor of dose dependent effects. I doubt whether assessment blood-plasma levels of benzodiazepines are a superior way of assessment of exposure for studying associations with hip fracture risk. If timing of intake, fall and admission is not taken into consideration and corrected for, the applied method is the reason why they did not find any association between benzodiazepine use and hip fractures. |
|||
|
|
|||
|
Nicholas Moore, Univ Bordeaux2 CHU Pellegrin-Carreire
Send response to journal:
|
Dear Dr Herings, All the points you mention were carefully considered and analyzed, and also raised by the referees during the editorial process. Space constraints did not allow us to explain all at length, though we did give indications for all in the paper. 1) if the Tmax of, eg, bromazepam is two hours, its half-life is 20 hours, so that it takes over 4 days for it to clear from the bloodstream, metabolites notwithstanding. There is no problem detecting its presence even after a few hours, especially after chronic use as was the case in most patients. As you rightly point out, time of admission from fall does not apply for controls, which is why we took time of admission as a surrogate for time from ingestion: since most patients take their drugs at about the same time of day, (ie, morning and noon for anxiolysis, after dinner for hypnotics), the fact that the cases and controls were admitted at the same time of day is an indicator that the same delay probably elapsed after drug ingestion for cases and controls. The median time between fracture and admission was two hours, and 65% were admitted within three hours. Of course in some cases exact time of fracture was not known, but was within 24 hours. Most cases were admitted in the afternoon and evening, up to midnight. This corrsponds to hip fracture during the diurnal activity period, which seems reasonable for active elderly. Not all subjects break their hips when getting up in the middle of the night. About 10% of the fractures were admitted between midnight and 8 in the morning, figures similar to those cited for the netherlands, where 85-90% of fall injuries are not admitted at night. What is the mean delay between fracture and admission in Holland? Admitting people late at night in the Netherlands could also indicate longer delays from occurence of an evening fall. 2) we did not consider active metabolites, but attributed these to individual drugs, not being able to identifiy which are metabolites and which are the drugs themselves. This would not change much since his would be the case for cases and for controls. There is the same number of users with single drugs found, and there may be increased risk when two drugs are found. This could be an explanation as to why the odds ratio for use of two or more drugs is considered decreased when plasma, though why this would apply only to cases and not to controls I do not understand. Most of the drugs that were found unexpectedly concentrate on clorazepate (to be published soon in CPT), a drug which is not a metabolite, which is used as a hypnotic, and is found more often in controls-it could have a protective role by keeping patients safely in bed? 3) I agree with you that probably plasma concentrations are a poor predictor of brain concentrations. Plasma presence of a benzodiazepine ensure that at least one was present in the bloodstream, and presumably in the brain, within the time-frame of the fracture occurrence. Considering the pharmacokinetics of these drugs, not finding it in the plasma (except for the very short acting drugs like triazolam, suspected in only one case, but not found) indicates that the patient was not exposed to the drug at the time of fracture, taking into account as we did the timing of fracture and drug ingestion. There is no other way to affirm this, and certainly not from health insurance or pharmacy databases, considering the way benzodiazepines are used. Using questionnaires and medical records, the usual way of ascertaining drug use, in fact led to exactly the same results. Plasma assays were used in the hope of finding covert use of benzodiazepines in patients with fractures and not in controls, explaining lack of association in other studies. This we did not find. In any event, pleading a difference between plasma and brain concentrations to explain absence of association is a weak explanation for not finding what is obviously a weak effect. This was not needed to find the association with falling: as stated, most if not all studies find an association between benzodiazepines and falling. Most studies of fractures do not find an association between benzodiazepines and hip fractures, or only in specific subsets of drugs. Contradictions between studies abound, each finding in fact an association with a different subset of drugs, related to indication, half-life or other characteristics, or as we did with individual drugs. This resembles very much the random effect of multiple testing when global tests are negative. For me the clinch is still the new york state study, where decreasing by half the use of benzodiazepines in the elderly population over a number of years led to no visible change in number of admissions for hip fractures. This again contrasts with the results of interventional studies on falls, which show clear positive results. So that YES benzodiazepine use should be discouraged in the elderly because they increase the risk of falling, impair cognition and probably increase the risk of Alzheimer's disease, and for any number of other good reasons, but NOT because they increase the risk of fracture. |
|||
|
|
|||
|
Giovanni Gambassi, associate professor of medicine Catholic University, Rome, Italy
Send response to journal:
|
Pierfitte and colleagues1 completed a case-control study with 245 incident hip fractures concluding that benzodiazepine use does not confer any added risk. The study is unique in that benzodiazepine plasma concentrations were determined, nevertheless, there are a number of issues that merit discussion. Since fracture is considered the result of either an increased tendency to fall and/or to fracture, it would have been appropriate to gather information on history of fall. In fact, it is clear that benzodiazepine use may lead to hip fracture by increasing the risk of falling and by affecting reaction time rather than by an intrinsic effect on bone. Regardless, do the authors obtained data on osteoporosis? and how many patients were on a osteoporosis medication? The article concludes dismissing the role of dosage and elimination half-life. In contrast, we reported that dose and elimination of benzodiazepines have an independent and additive effect to increase the risk of falls2 and hip fracture.3 In addition, in a case-control study considering 9752 incident hip fractures, we demonstrated that, among long half-life benzodiazepines only oxidative agents are associated with an increased risk, and that, at high dosages, this is strongly related to age.3 The authors also affirm that hypnotic and anxiolytic benzodiazepines were equally innocuous. This finding contradicts our previous report on 8851 cases of hip fractures.4 We showed that use of long half-life hypnotic benzodiazepines confers a two-fold increased risk, while there appears to be no added risk for users of any anxiolytic agents. Our study confirms an increased risk for users of lorazepam, although much smaller (OR 1.2; 95% CI 1.1-1.7), but it also cautions about the interpretation of odds ratios for individual agents. Among 571 users of temazepam (105 cases and 466 controls) we observed an OR of 1.0 (95% CI 0.9-1.1), while in the present study with 8 users, equally divided between cases and controls, OR was 2.7. Data about associated diseases is not reported. The issue is not irrelevant since there appears to be a number of diseases, such as diabetes mellitus and hypertension, that are characterized by an increased bone loss and reduced bone mineral density. While both cases and controls used over four non-benzodiazepine drugs, the authors reported on the effect of diuretics, tricyclic antidepressants and antacids. Yet, other medications may be potentially affecting their outcome.5 Recent studies have found that statins may protect against risk of fracture, whereas, several anticonvulsants, corticosteroids and oral anticoagulants may have a moderate effect to reduce bone mineral density. Antonio Sgadari, MD Centro di Medicina dell’Invecchiamento Università Cattolica del Sacro Cuore Rome, Italy Claudio Pedone, MD, PhD, MPH Center for Gerontology & Health Care Research Brown University School of Medicine Providence, RI, USA Graziano Onder, MD Sticht Center on Aging Wake Forest University Medical School Winston Salem, NC, USA Giovanni Gambassi, MD Centro di Medicina dell’Invecchiamento Università Cattolica del Sacro Cuore Rome, Italy 1. Pierfitte C, Macouillard G, Thicoipe M, et al. Benzodiazepines and hip fractures in elderly people: case-control study. BMJ 2001;322:704-8 2. Sgadari A, Bernabei R, Gambassi G, Landi F. Benzodiazepines and the risk for falls: a role for metabolic pathway? The Gerontologist 1998;38(A244):251 3. Sgadari A, Lapane KL, Mor V, Landi F, Bernabei R, Gambassi G. Oxidative and nonoxidative benzodiazepines and the risk of femur fracture. J Clin Psychopharmacol 2000;20:234-9 4. Sgadari A, Lapane KL, Mor V, Bernabei R, Gambassi G. Hypnotic and anxiolytic benzodiazepines and the risk of femur fracture. (in press) 5. Lapane K, Gambassi G, Hume A, Sgadari A, Mor V. Which antidepressants increase the risk of femur fracture in long term care? American Geriatric Society 1998 Annual Meeting. Abstract book, p.142 |
|||