Rapid Responses to:
|
|
Rapid Responses: Rapid Responses published:
-
![[Read Rapid Response]](/icons/misc/sectionDOWN.gif)
Effect of antipsychotics on stroke risk remains unproven
- David Curtis (31 August 2008)
-
Where Are the Patient Counts?
- David Behar (1 September 2008)
-
When we choose atypical antipsycotics?
- giovanni portuesi (3 September 2008)
-
Awaiting a better explanation
- Raghavendar Baburaj, Ranjini Rao (3 September 2008)
-
What is medicine
- Philipp Conradi (5 September 2008)
-
This article fails to control for smoking and is invalid.
- Daniel J Pallin (7 September 2008)
-
Too many flaws to be of any clinical benifit
- Daya Fernandopulle (10 September 2008)
-
Study does not discriminate between treatment and indication for treamtent as risk factor for stroke
- Glen D Harper (14 September 2008)
-
The study does not establish any new information
- Santhana K Gunasekaran, Humber Mental Health NHS Trust, Willerby HU10 6ED (15 September 2008)
-
Is Socrates a woman?
- Liz Miller (15 September 2008)
-
The pragmatic approach to this issue
- Vincent WK Choong (16 September 2008)
-
Re: The pragmatic approach to this issue
- David Curtis (22 September 2008)
-
Rapid responses are much more flawed than the study
- Patrick C Waller (8 October 2008)
-
Reply from study authors
- Ian J Douglas, Liam Smeeth (22 October 2008)
-
Re: Rapid responses are much more flawed than the study
- Julia Hippisley-Cox (28 October 2008)
|
|
|||
|
David Curtis, Honorary Professor of Psychiatry Royal London Hospital, East London (NHS) Foundation Trust, London E1 1BB
Send response to journal:
|
The recent report by Douglas and Smeeth [1] does not, as is widely claimed, demonstrate that both typical and atypical antipsychotics increase the risk of stroke, especially in patients with dementia. What they demonstrate, using the “self-controlled” method, is that of patients who have had a stroke and who were prescribed antipsychotics by their GPs the stroke is more likely to occur during the time antipsychotics are prescribed rather than before or after. The method used is wide open to biases of different sorts. With regard to treatment iniation, the authors mention, but do not give much prominence to, the notion that “time-varying confounders can be be difficult to take into account”. They do not consider at all the possibility that there may be quite direct relations between the clinical situation which leads to the prescription of antipsychotics and the onset of a stroke. For example, it is entirely plausible to conceive of a patient with some degree of on-going deterioration of functioning, which may or may not be recognised as occurring within the context of dementia, who presents with increased behavioural disturbance such as aggression, agitation or wandering who is then prescribed an antipsychotic and who subsequently goes on to develop what is recognised as a full-blown stroke. Nobody would take such a vignette as indicating that the antipsychotic caused the stroke but the study design used would interpret the evidence in this way. Patients are (we hope) prescribed antipsychotics for a reason. If the clinical indicators for the prescription might in any way be associated with an increased risk of subsequent stroke then this completely undermines the inference that the prescription itself increases risk. The problems are perhaps even more evident when we consider how we should interpret the finding that patients are more likely to have a stroke during a period in which the GP prescribes antipsychotics rather than after such a period has finished. Surely it is obvious that there are patients who will be on antipsychotics, will suffer a stroke and the resulting changed clinical picture will mean that the prescription is no longer continued? The most obvious explanation for this to occur is that the patient is in hospital. In fact, I’m going to write that twice. The patient, after a stroke, may be in hospital for months. If so, the GP is not going to be prescribing any medication for them, let alone antipsychotics. However in the study reported this observation will be interpreted as strengthening the association between antipsychotics and stroke risk. Even if the patient does manage to get out of hospital back to the care of the GP, it is not at all difficult to imagine all sorts of reasons why the prescription for antipsychotics they were receiving previously may not be continued. They will have been admitted to hospital seriously ill and will have had a thorough review of medication. They may well suffer from impaired mobility, speech problems and depression. The indications for the original prescription may well no longer be present. The study shows that when patients receive antipsychotics and suffer a stroke they are more likely to have the prescription started before a stroke and/or stopped afterwards. It is extremely alarming that there are widely publicised claims that this supports the notion that antipsychotics increase the risk of stroke, a prospect that I find biologically somewhat implausible. Publishers of influential journals have a responsibility to see that research papers are adequately reviewed and that findings are presented fairly and with appropriate degrees of caution. I believe this has not happened and that publicity surrounding this paper will have adverse consequences for the patients who could well benefit from antipsychotics and for their carers. I hope that in time moderating voices will be heard. David Curtis 1. Ian J Douglas and Liam Smeeth Exposure to antipsychotics and risk of stroke: self controlled case series study BMJ 2008; 337: a1227 Competing interests: None declared |
|||
|
|
|||
|
David Behar, Psychiatrist 206 East Broad Street, Bethlehem, PA 18018, USA
Send response to journal:
|
The authors give only ratios, instead of counts of patients suffering stroke. I want the real numbers, and I want their statistical significances, and I want their statistical significances corrected for the effect of the huge sample size, and for the effect of repeated measures. This article is ridiculous.
If 2 patients suffered a stroke in one group of 1000 patients, and 4 did in another of 1000 patients, yes the risk is double. However, it is so small as to be clinically meaningless.
That nitpicking trivial difference was harped upon in many instances by biased left wing critics of drug companies. Low absolute numbers would rebut the conclusion of the authors, to stop the use of these medications.
I suspect the authors are hiding ridiculously small effects from the reader. What I do not appreciate is the quality of the editing. The peer reviewers and the editors should have made this demand prior to publication. Is the BMJ becoming another left wing, biased, corporation bashing, propaganda rag, as so many famous journals are on both sides of the Atlantic?
Competing interests: I prescribe these medications, and have served as a promotional speakers for some. |
|||
|
|
|||
|
giovanni portuesi, Head of an acute impatients service Ospedale Civile di Pinerolo, Turin, Italy. 10064
Send response to journal:
|
I agree with what is told above by Curtis on the weaknesses of the study. I would like to suggest another possible bias, explaning why atypical antipsychotics (AP) are found relatively more "dangerous" for cerebrovascular accidents ( at least) than the typical AP. In clinical practice, it make sense to use an atypical AP when we are more worried about his general tolerability, i.e. in patient who have a poorer health ( and sure, in a vascular o mixed dementia with parkinsonism). Vascular diseases are the more widespread health problem in old age, so do their risk factors ( diabetes, overweight, etc.) So risk factors for cerebrovascualar diseases are linked to atypical AP by a possible prescribing beahaviour. Competing interests: None declared |
|||
|
|
|||
|
Raghavendar Baburaj, SPR Swansea, Ranjini Rao
Send response to journal:
|
It is commendable that the authors have embarked on this mammoth project, which undoubtedly would benefit our patients. I refer to table 3 which actually displays the risk of stroke [ in exposure periods] for those prescribed antipsychotics. In patients with a recorded diagnosis of dementia,The risk of stroke tends to increase markedly in the first month of prescription of typical and atypical antipsychotics but the risk decreases towards the sixth month. This is something which causes concern . Should we be telling our patients that the risk is greatest in the first month but the risk abates by the following month. I have tried to think of the reasons for this change. It could be that the body is subject to sudden hemodynamic and metabolic changes possibly triggering an Ischaemic event but it would be hard to believe that the risk becomes lesser as time progresses. I am hoping to find the answer for the benefit of my patients . Competing interests: None declared |
|||
|
|
|||
|
Philipp Conradi, GP Otto-Dix-Ring 98 o 01219 Dresden, Germany
Send response to journal:
|
The article highlights a fundamentally flawed perspective of modern medicine. Medicine once served for making patients better, alleviating symptoms and healing disease. Now medicine seems to have degenerate into a risk reducing, patients stratifiying, life years adding bioscience disregarding the individual patients needs. It is simply unethical to deny a patient a good treatment for disturbing and harassing complaints because of worries about possible side effects. Nobody would question prescribing morphins for terminal analgesia.Patients at the end of their life with dementia related behavioural problems should be able to expect proper treatment. To withhold this treatment for spurious and debatable reasons is just ,,madicine'' Competing interests: None declared |
|||
|
|
|||
|
Daniel J Pallin, Director of Research, Department of Emergency Medicine, Brigham and Women's Hospital, Boston 75 Francis St., Boston, MA 02115, USA
Send response to journal:
|
Dear Editor, This article is deeply flawed. I agree with the other responses you have already published, including David Curtis's letter of August 31, 2008, and would add an additional concern. The authors failed to control for tobacco use. They merely state, "While we have successfully controlled for age, data within the database do not generally allow precise measurements of subtle changes in behaviour patterns such as smoking and drinking." There is evidence that antipsychotics increase the urge to smoke. See: Barr AM, Procyshyn RM, Hui P, Johnson JL, Honer WG. Self-reported motivation to smoke in schizophrenia is related to antipsychotic drug treatment. Schizophr Res. 2008 Mar;100(1-3):252-60. Epub 2008 Jan 4. PMID: 18178062 [PubMed - indexed for MEDLINE] Here is a quote from the above-referenced article: "Therefore, as antipsychotic drug dose increased, patients tended to rate specifically these two factors as increasingly important reasons for smoking. There was also an overall positive relationship between chlorpromazine daily dose and the total score on the smoking motive questionnaire [r=0.34; p<0.01]." Competing interests: None declared |
|||
|
|
|||
|
Daya Fernandopulle, Trainee Psychiatrist Mater Hospital, Belfast BT14 6AB
Send response to journal:
|
I want to emphasise the huge selection bias introduced to the study by only including patients having experienced a stroke.So there is a very high likelihood that the study sample consists mainly of people having high risk factors for strokes. The incredible mistake the authors make is generalising the study findings to the whole population of people with dementia,significant proportion of whom may never experience a stroke in their lifetime.I note that Professor Curtis's comment that the study findings would only apply to patients who would experience a stroke and were prescribed antipsychotics. The authors also have analysed the data for two subgroups namely people with dementia and without dementia.However there is no clear comments made with reagards to the methodology used in determining these diagnoses.This opens up a whole debate on the reliability of these 'recorded diagnoses' and the potential for the false positives and negatives.The failure of accountability for the reliability of the diagnoses too,in my opinion discredits any clinical usefulness in the findings of the study. Competing interests: None declared |
|||
|
|
|||
|
Glen D Harper, Consultant Physician North Devon, EX37 9HY
Send response to journal:
|
Whilst Douglas and Smeeth have clearly conducted a very elegant and sophisticated study they cannot imply or conclude that antipsychotics are causal in the increased risk of stroke they have observed. It is equally possible and, indeed, more likely that the cause for the acute mental deterioration requiring the introduction of antipsychotics is associated with an early increased risk of stroke. If this interpretation is correct all patients presenting with symptoms requiring the prescription of an antipsychotic should have a detailed risk assessment for stroke conducted and acted upon at the same consultation. Rather than blame the drugs, lets reduce the risk. Competing interests: None |
|||
|
|
|||
|
Santhana K Gunasekaran, Specialist Registrar (ST5) in Forensic Psychiatry The Humber Centre, Humber Mental Health NHS Trust, Willerby HU10 6ED
Send response to journal:
|
Information about antipsychotic use and its association with stroke is limited and studies in this area should be welcomed. Antipsychotic drugs have also been implicated in venous thromboembolism and sudden cardiac death. For example, in Reilly’s study [1], thioridazine was identified as a risk factor for sudden unexplained death among psychiatric inpatients. Zornberg & Jick [2] highlighted the association of antipsychotics with venous thromboembolism. Several hypotheses have been proposed, including enhanced aggregation of platelets, raised anticardiolipin antibodies, exacerbated venous stasis by sedation, increased adrenaline secretion and hyperhomocysteinaemia. However, in this study [3] the authors have not added any valuable information. Although it has been stated that the analysis is controlled for a few factors there are a number of potential confounders that have not been considered such as smoking and exposure to other drugs. During the study period factors such as other medications taken concomitantly could have accounted for the increased risk. The study design itself fails to establish a causal relationship between stroke and antipsychotics. We now know that often in people diagnosed with dementia the pathology is a mixture of vascular and alzheimer's dementia. Without considering these factors, I find the study's conclusions to be premature. I agree with the other responses you have already published, including David Curtis's letter of 31 Aug 2008 and Daniel Pallin's letter dated 7 Sep 2008. Competing interests: None declared References 1. Thioridazine and sudden unexplained death in psychiatric in-patients. Reilly JG, Ayis SA, Ferrier IN, Jones SJ, Thomas SHL. British Journal of Psychiatry 2002; 180: 515-522 2. Zornberg GL, Jick H. Antipsychotic drug use and risk of first-time idiopathic venous thromboembolism: a case control study. Lancet 2000; 356:1219-23 3. Ian J Douglas and Liam Smeeth Exposure to antipsychotics and risk of stroke: self controlled case series study BMJ 2008; 337: a1227 Competing interests: None declared |
|||
|
|
|||
|
Liz Miller, Occupational Health Physician London SW6 4PH
Send response to journal:
|
The conclusions, are in my opinion, flawed. Is this the medical equivalent of suggesting that because all women are mortal and Socrates is mortal, Socrates is a woman? Epidemiology is dangerous without a strong theory to back it up. So far no one has a theory to why antipsychotics might be associated with strokes. It would behove us all to remember a) Association is not causation. This paper suggests more people taking antipsychotics have more strokes. However agitation and psychosis may be caused by the same condition that causes strokes. For example, patients with an infection often take antibiotics. Patients with an infection often have a fever. Unless the authors of an epidemiological study understand that infections cause fevers, they could easily assume that antibiotics cause fevers. b) Larger numbers may improve vital statistics but not a paper's validity. The larger the group, the more likely you are to cross a cultural divide, and find answers to questions to you did not ask. For example, “what happens to patients of Northern doctors north who follow protocol A, compared with patients of Southern doctors who follow protocol B?” Incidentally, Northern doctors use more atypical antipsychotics than Southern Doctors. The major difference between protocol A and B is that Northern patients read the newspapers and made sure they took their daily aspirin. The larger the groups the harder it is to ignore regional effects. c) To quote the authors “Differences between patients are of little relevance as the risks comparisons are made entirely within patients”. The differences between patients matter. Take this to the point of absurdity. Imagine comparing a charging elephant, a starving chimpanzee and an orange, and see what happens when each is given an antipsychotic. The charging elephant calmsdown and starts eating everything in sight. The starving chimp calms down, stops looking for food and dies. The orange would still be an orange but ends up in the bin because it tastes strange. Patient differences matter. And finally d) “Patients with dementia are twice as likely to get a stroke with antipsychotic medication as those without dementia.” Agitated patients with dementia may be simply be twice as likely to have a stroke as calm patients with dementia, regardless of any medication they receive. What causes a patient with dementia to become agitated? We don’t know. Ergo: antipsychotics are not ideal drugs for many reasons, Sometimes it is right to prescribe them. People need individual treatment depending on their condition and their circumstances. Even in the same patient, agitation from hallucinations needs different management from agitation due to urinary retention. Putting everything together into one large study muddies the water. Competing interests: None declared |
|||
|
|
|||
|
Vincent WK Choong, SpR Old Age Psychiatry Friarage Hospital, Northallerton DL6 1JG
Send response to journal:
|
Dear Editor, Whilst this study adds weight to others, drawing the same conclusion that antipsychotics increases incidence of strokes, like others, it has made no suggestion of the mechanisms of action or medical alternatives. I absolutely agree with Philipp Conradi's views of the current 'risk- reducing' approach in patient management. Whilst it is easy for one to say it is adverse to use antipsychotics to manage hehaviour that challenges in dementia, it is often not that simple in real life clinical situations, where all non-pharmacological attempts are exhausted. I find by carefully explaining the rationale of using antipsychotics and disclosing the fact that it may increase the risk of strokes, I have not had any refusal from patients' next of kin. Most relatives are distressed seeing their loved ones in extreme agitated states and are relieved when antipsychotics are suggested knowing to potential dangers. However, this cannt be said the same for the proportion of patients who have no relatives to make decisions on their behalf.) Competing interests: Have prescribed antipsychotics for management of BPSD in Dementia |
|||
|
|
|||
|
David Curtis, Honorary professor of psychiatry Royal London Hospital, East London (NHS) Foundation Trust, London E1 1BB
Send response to journal:
|
It is worrying that an SpR in Old Age Psychiatry seems to think that relatives can make decisions on behalf of patients. This has never been the case in the UK. The situation has changed with the introduction of the Mental Capacity Act in that a patient now can appoint somebody to make such decisions but that person does not have to be a relative. Competing interests: None declared |
|||
|
|
|||
|
Patrick C Waller, Consultant in Pharmacoepidemiology Southampton, SO30 2NY, UK
Send response to journal:
|
The casual reader of the rapid responses published so far would surely take away the message that the study of Douglas and Smeeth is deeply flawed. Those of Curtis and Conradi have been published in the print issue without comment from the authors and I feel it is about time there was some redress. Please note my competing interests before reading further but it is not my purpose here to defend the authors. Rather there seems to be a need to defend the discipline of pharmacoepidemiology against clinicians who appear to find it an inconvenience to their practice. In particular, I profoundly disagree with Dr. Conradi’s contribution. I wouldn’t question the use of morphine for terminal care but this example is hardly comparable in any way. Are anti-psychotics really a “good” treatment for behavioural symptoms of dementia and does “worries about possible side-effects” adequately encapsulate this risk? Treatment with all medicines is a question of balancing expected benefit against possible harms. When evidence for the latter outweighs the former, the opposite of what Dr. Conradi says is likely to be true – i.e. it is probably unethical to use such a treatment. One aspect of this issue that seems not to have been properly appreciated is that prior concerns about stroke with anti-psychotics came from a series of randomised, placebo-controlled trials conducted for the purpose of evaluating benefit-risk in the treatment of behavioural symptoms of dementia. These trials studied risperidone and olanzapine, and although they have not been properly published, information about them is available and was cited as references 2 and 20 in the original paper. It is important to realise that the reason why a series of fairly small trials was able to identify such an adverse effect was because the absolute excess risk of cerebrovascular events was high, the studies being conducted in a population at a high level of baseline risk. These studies were also short-term but nevertheless they were able, collectively, to estimate a number need to harm of about 37 during that short period i.e. treating 37 patients for 2-3 months would cause one additional cerebrovascular event. These trials did not establish any kind of benefit which would justify exposing patients to such a risk. That is why regulatory authorities refused to grant the indication and actively warned against unlicensed use for this purpose. This level of risk is not a “small” effect as Dr. Behar attempts to suggest, and he and Dr. Fernandopulle do not appear to have understood the study design despite the excellent accompanying editorial by Whitaker published in the same issue. Patient counts are included in the original paper – there were 6,790 patients in total, all of whom had a recorded diagnosis of stroke and at least one recorded prescription for an anti- psychotic drug. Dr. Fernandopulle also raises a common (but often invalid) criticism thrown at pharmacoepidemiological studies conducted using such databases i.e. the reliability of the diagnosis of stroke. Despite the auditing process described and a generally high level of data validity, there is little doubt that some of the patients included will not actually have actually experienced a stroke. It is therefore worth considering what effect this might have on the observed association between antipsychotic drugs and stroke. In general, inclusion of patients with other diagnoses will weaken positive associations (i.e. lead to underestimation) because such errors are usually non-systematic and unrelated to drug treatment. Misdiagnosis could only be an explanation for a substantially increased risk if (a) it was frequent and systematic i.e. large numbers of patients didn’t have a stroke but had another specific diagnosis and (b) that specific diagnosis was strongly associated with use of the relevant drugs. So, fairly clearly, this point does not “discredit” the study, as was suggested. Professor Curtis raises some good points about possible biases but appears to base his judgement of whether or not this is a causal association on the Douglas and Smeeth study alone. If this study was all we had, I would agree with him, but in the face of the randomised trial evidence, a case-control study (also apparently unpublished but discussed within reference 6), and this new study, all of which shows a similar level of relative risk, the overall body of data supporting causation is strong. Although Curtis finds the association “biologically somewhat implausible”, possible mechanisms e.g. through a hypotensive effect, have been suggested. In any case, understanding why a risk is raised is not a prerequisite for accepting that an association is probably causal. If Dr. Pallin is correct in implying that antipsychotic drugs increase the urge to smoke thereby explaining the observed association, this would argue for a casual association, albeit an indirect one. The notion put forward by Dr. Gunaserkaran that the new case-series study does not add anything to existing knowledge is unhelpful because there were important limitations to the original randomised trial evidence (e.g. it only related to two drugs, one high-risk sub-population and short durations of exposure), and further studies such as this one using a different design and in a wider population do add useful information. The studies of Gill et al. and Hermann et al. cited as references 4 and 5 in the original paper should not be interpreted as counter-evidence, as they are comparisons of the risk of typical and atypical drugs (rather than against no treatment). Their negative findings merely imply that the relative risk is likely to be similar with all antipsychotic drugs. Douglas and Smeeth did find a marginally higher risk with atypical drugs but, overall, the evidence suggests little or no difference. To conclude, most of the objections to the study of Douglas and Smeeth raised so far do not stand up to critical analysis of all the relevant evidence. Rather they largely seem to be the response of clinicians who wish to fly in the face of the evidence so that they can keep using treatments which are likely to be seriously harmful in vulnerable patients and which have, at best, limited benefits. The suggestion of Dr. Conradi that withholding such treatment is unethical is preposterous. Competing interests: I have been an honorary senior lecturer at the London School of Hygiene and Tropical Medicine since 1998 but have had no involvement with this study or contact with authors about it. I have held a consultancy contract with the MHRA since 2002 and was previously employed by them. These views are entirely my own. |
|||
|
|
|||
|
Ian J Douglas, Epidemiologist London School of Hygiene & Tropical Medicine, Liam Smeeth
Send response to journal:
|
We are pleased to see our study on the risk of stroke associated with antipsychotics has generated much interest. However, it is clear that the study was not well understood by some people and we would like to answer some of the questions raised. A consideration of both the beneficial and harmful effects of prescribed drugs is of great importance. We find it particularly disturbing, in fact shocking that Dr Conradi believes it is unethical that "worries about possible side effects" should influence decisions around treatment choices. Most people would agree that the art of successful patient care involves a careful balance of the potential benefits and harms of any intervention. The benefits of a drug are generally well established by randomised clinical trials, but trials are limited in their ability to give us detailed information about possible adverse effects. They are generally small, of short duration and restricted to a sub-section of the population deemed eligible to take part in a trial. In the case of antipsychotics, separate clinical trials of olanzapine and risperidone in elderly patients with dementia detected an increased risk of stroke in treated patients. That these relatively small trials were able to detect such a risk suggests that this effect is not rare. The consistency of this result across trials means it is unlikely the finding was due to chance. These results were made available to doctors in 2004 and 2006 and so it is worrying to hear psychiatric practitioners voicing doubt over the plausibility of such a risk. The lack of a clearly understood mechanism behind an accepted adverse drug reaction is not unprecedented (e.g. tendon rupture with macrolide antibiotics). The UK's Committee on Safety of Medicines issued a warning that antipsychotics should not be used for the treatment of behavioural symptoms in patients with dementia. However, important questions remained, including whether the risk was present for atypical and typical antipsychotics and in patients with and without dementia. Further large scale clinical trials in this area are extremely unlikely - they would rightly be considered unethical given the findings to date. This leaves us with only one plausible option for exploring this area further; observational studies. As many of the responses to our paper have rightly stated, observational studies are prone to a number of biases. This doesn't mean we should avoid them altogether, but rather we should design them as best we can and interpret their findings with due caution. Before our study, others had explored the association between antipsychotics and stroke using traditional epidemiological study designs. Whilst these studies generated evidence consistent with the findings from clinical trials, the possibility that the results were due to confounding remained. Observational studies of drug effects are particularly prone to "confounding by indication" whereby patients given a drug differ from those not given the drug in important ways which are related to their underlying health. We chose to use the case-series design because it avoids much of the confounding by indication described above. This is achieved by all comparisons being within individuals, with no comparisons being made between individuals. It is purely for this reason that we stated the differences between patients are of little relevance; Dr Miller appears not to have understood this. Naturally in a wider context, the differences between patients are of enormous importance, but in terms of the inferences we can make from statistical analyses performed with a case series, they are not; comparisons are made purely within individuals, measuring the risk of an event during exposed versus unexposed periods. The accompanying editorial by Heather Whitaker explains the advantages of this approach in detail. Dr Behar appears to have missed the fact that all 6,790 patients in the study had a stroke and received antipsychotics. For this reason we cannot calculate an absolute risk of stroke using this study design. However, other studies show the risk of stroke in the elderly to be much more common than the 2 in 1,000 as suggested by Dr Behar. Hollander et al (1) showed a rate of greater than 20 per 1,000 patient years in an elderly Danish population. Applying the relative rate of stroke detected in our study we can see the impact of antipsychotic treatment may be considerable. We strongly disagree with the statements by Dr Behar that a tripling of such a risk is “clinically meaningless” or a “nitpicking trivial difference”. We do not mind being accused of being part of a left wing anti-industry conspiracy in league with the BMJ and other journals, but have no knowledge of such a conspiracy. Dr Baburaj asks about the interpretation of the 6 month post- treatment period. This period refers to the "washout" period after patients finished taking antipsychotics. The risk returns gradually to the baseline level over this period. Dr Pallin raises an interesting point about smoking which we are unable to test because we do not have detailed daily smoking data. We believe it is extremely unlikely the entire association between antipsychotic treatment and stroke in our study is due to temporary increases in smoking. However, if such an effect on smoking did occur, this would actually provide us with a possible causal mechanism for some of the observed association between antipsychotics and stroke rather than being a source of bias. We would agree with Dr Fernandopulle that people at high risk of stroke are more likely to have strokes, and are therefore more likely, in absolute terms, to be affected by any increased risk of stroke associated with antipsychotic use. However, it goes against all scientific thinking to suggest the risk is confined entirely to people who have a stroke. This is like arguing that the risk of cancer associated with smoking is only observed among people who smoke and get cancer, therefore it is safe for people to smoke. In addition, concerns about the reliability of dementia diagnoses do not in any way discredit the findings. Firstly, dementia status was determined using diagnoses recorded in the patients' notes; previous studies of patients with dementia using the GPRD have found the validity of this diagnosis to be high (2). Secondly if dementia status was misclassified this would in fact dilute any real difference in the effect of antipsychotics between those with and without dementia. Professor Curtis, Dr Portuesi, Dr Harper and Dr Gunasekaran raise interesting questions about possible biases. The suggestion that we underestimated the exposed period in patients admitted to hospital after a stroke would be a valid concern. However, in this scenario we would expect the risk in the period immediately after exposure to be lower than the baseline, since by definition a large group of patients would be unable to have their first stroke during this period. We do not see this in the results for the washout period, suggesting this is not a major source of bias in our study. The possibility that antipsychotics are initiated in patients at a point in time when their risk of a stroke has recently increased for entirely unrelated and unrecorded reasons cannot be ruled out. However, the washout period in our study demonstrates that the risk of stroke returns to baseline after antipsychotics are withdrawn, arguing against suggestions that the increased risk during treatment reflects a close temporal link between treatment initiation and more permanent changes in underlying vascular pathology. Like Professor Curtis, we are concerned that Dr Choong seems to suggest relying on relatives to make therapeutic decisions for patients. We thank Dr Waller for his response, and simply take the opportunity to say that we have had no contact with Dr Waller about this issue. By their nature, no single study can, or should claim, to provide final definitive answers to the questions they set out to investigate. However, our study adds to a growing body of work on the risk of stroke with antipsychotics in a number of ways; it gives specific measures of relative risk in patients with and without dementia and confirms an increased risk for both typical and atypical drugs. The design of the study also means it is unlikely to suffer from the same biases as more traditional designs. If we look at the evidence as a whole, the results of clinical trials and several observational studies give a consistent message that antipsychotics are associated with an increased risk of stroke. The trial evidence in particular suggests this association is causal. Our study suggests the risk is higher in patients with dementia, reinforcing the decision of regulators to advise against the use of antipsychotics in this patient group. We agree with Dr Miller, Dr Choong and others that real life clinical situations are often not straightforward and that people need individual treatment based on a consideration of all factors. With this in mind we suggest looking again at the main summary in the discussion section of our paper: “Since the background risk of stroke in elderly patients is relatively high, we reaffirm that the risks associated with antipsychotic use in patients with dementia generally outweigh the potential benefits, and in this patient group, use of antipsychotic drugs should be avoided wherever possible. When the use of antipsychotic drugs is deemed necessary, our results suggest that typical antipsychotics may be preferable to atypicals, with respect to serious cerebrovascular outcomes. By contrast, the results presented here suggest a much more modest association between antipsychotics and stroke in patients without dementia, regardless of age (risk ratio=1.41; 95% CI (1.29-1.55). This implies that in elderly patients without dementia, the use of antipsychotics may be acceptable, although as with all treatment choices, a wider consideration of all potential risks and benefits would need to be taken into account” We would prefer it if antipsychotic drugs were perfectly safe: unfortunately they appear to be associated with an increased risk of stroke. What matters is that therapeutic decisions are made in the context of the best available evidence. As previously highlighted by the MHRA, the evidence around the risk of stroke associated with different antipsychotic drugs and among people with and without dementia was not totally clear. The aim of our study was to strengthen the evidence base – an aim that was successfully met. 1) Hollander M et al Incidence, risk, and case fatality of first ever stroke in the elderly population. The Rotterdam Study. J Neurol Neurosurg Psychiatry 2003;74:317–321 2) Seshadri S et al Postmenopausal estrogen replacement therapy and the risk of Alzheimer disease. Arch Neurol. 2001 58:435-40 Competing interests: Study authors |
|||
|
|
|||
|
Julia Hippisley-Cox, Professor Clinical Epidemiology University of Nottingham
Send response to journal:
|
Douglas and Smeeth have presented a good study. The question was clear and important. The data were appropriate as was the study design. The discussion was fair and balanced. No study is perfect and all research of this kind is descriptive and there will always be unmeasured confounding. like Waller, I have had no contact wiht the authors about this study but I concerned that the poor quality of the some of the rapid responses could distract from what is otherwise a valuable contribution to the literature. Could the BMJ introduce some quality assurance to the rapid responses to avoid the responses potentially misleading the general reader? Competing interests: I am a clinical epidemiologist undetaking research into drug safety and use similar data sources and methods to answer similar questions. I am also on the BMJ editorial board but did not review this article. |
|||