Rapid Responses to:
|
|
Rapid Responses: Rapid Responses published:
-
![[Read Rapid Response]](/icons/misc/sectionDOWN.gif)
The need for a community trial
- David J Torgerson, Jill Porthouse (10 January 2003)
-
Another way to think about hip protectors
- Paul J Young (10 January 2003)
-
Use of hip protectors in nursing homes: clarifications requested.
- Susan E Kurrle, Ian D Cameron (11 January 2003)
-
Re: The need for a community trial
- Gabriele Meyer, Andrea Warnke, Ingrid Mühlhauser, and Ralf Bender (11 January 2003)
-
Re: Another way to think about hip protectors
- Gabriele Meyer, Andrea Warnke, Ingrid Mühlhauser, and Ralf Bender (11 January 2003)
-
Re: Use of hip protectors in nursing homes: clarifications requested.
- Gabriele Meyer, Andrea Warnke, Ingrid Mühlhauser, and Ralf Bender (11 January 2003)
-
Re: Re: Another way to think about hip protectors
- Paul J Young (12 January 2003)
-
A clarification of previous submission
- Paul J Young (14 January 2003)
-
Is the hip protector effective?
- Percy V van Eerten (14 January 2003)
-
Re: Is the hip protector effective?
- Gabriele Meyer, Andrea Warnke, Ingrid Mühlhauser, and Ralf Bender (15 January 2003)
-
Effectiveness of the studied hip protector remained uncertain
- Pekka A Kannus (16 January 2003)
-
Re: Effectiveness of the studied hip protector remained uncertain
- Ingrid Mühlhauser, Gabriele Meyer, Andrea Warnke, Ralf Bender (17 January 2003)
-
Re: Re: Effectiveness of the studied hip protector remained uncertain
- Pekka A Kannus (18 January 2003)
|
|
|||
|
David J Torgerson, Director, York Trials Unit University of York, Seebohm Rowntree Building, Heslington, York. YO10 5DD, Jill Porthouse
Send response to journal:
|
The hip protector trial reported by Meyer and colleagues is, methodologically, a significant improvement on previous hip protector trials. However, there are some methodological shortcomings to the study and issues about the study's generalisability to high risk individuals not resident in nursing homes. An important problem, not noted by the authors, is that the study groups have differential loss to follow-up. In the intervention group 64% of the participants completed the 18 month follow-up compared with only 57% of the controls (p = 0.04). This difference can introduce selection bias and could give a false estimate of effectiveness. In addition, there seems to be some evidence that the control group may have been frailer than the intervention group as the death rate was somewhat greater and this may have explained the higher incidence of falls in the control group. These differences could explain some of the apparent effectiveness of the hip protectors. It is interesting that the compliance rate for the hip protectors was very low at only 35%, not 68% as implied by the abstract. This low compliance rate is similar to that we have observed in a study we are undertaking where we are simply posting out hip protectors to high risk participants (reference, Yvonne's Bath abstract). Finally, whilst these data may support the use of hip protectors among nursing home residents we still require evidence for their effectiveness among high risk older people living in the community. Fortunately, we are undertaking a large (4200 participant) individually randomised trial among women at elevated risk of hip fracture living in the community, the results of which will be reported this summer. Competing interests: None declared |
|||
|
|
|||
|
Paul J Young, House Officer (orthopaedics) Middlemore Hospital, Auckland, New Zealand
Send response to journal:
|
Hip protectors clearly reduce the risk of hip fracture. However, whether their use is practical or appropriate is less clear to me. In this study, the rate of hip fractures in the control group is approximately 5.4% per year and approximately 3% per year in the group using hip protectors. The absolute difference between the two groups is 1.8%. Statistically, this means that on average a patient can expect to wear a hip protector for 55 years in order to prevent one hip fracture. Given the fact hip protectors are cumbersome and patients dislike wearing them, is this worth it? Competing interests: None declared |
|||
|
|
|||
|
Susan E Kurrle, Director and Senior Staff Specialist Geriatrician Hornsby Ku-ring-gai Hospital, Hornsby NSW 2077 Australia, Ian D Cameron
Send response to journal:
|
We congratulate Gabriele Meyer and colleagues on their study examining factors influencing the use of hip protectors in nursing homes (1). The study has been carefully conducted and the analysis has adjusted for the effects of clustering. It provides stronger evidence for effectiveness of hip protectors than previously published studies (2). We hope that the authors can provide some further information to assist others in applying the findings of the study. It is unclear what proportion of nursing home residents received (and understood) the educational intervention, and what proportion were simply encouraged to wear hip protectors by the staff who had received the education. In Australia, and other countries, the vast majority of residents of nursing homes have severe cognitive impairment that greatly limits their participation in decision making regarding hip protector use. We would be interested to know what percentage of participants in this study had severe cognitive impairment. The cost of hip protectors is clearly a disincentive to their use in countries, like Germany, where they are not subsidised. Could the authors speculate on how much of the intervention effect was due to the supply of free hip protectors and how much to the educational session? Were three pairs of hip protectors enough for the 15 months of the study? It is our experience that more than three pairs would be required if they are regularly used. The authors touch very briefly on reasons for non-adherence with the use of hip protectors. Could they elaborate on these reasons, as measures to address these could well improve the limited adherence rates reported in this study? Some other methodological issues could be clarified. How were hip fractures ascertained? As it is unlikely that this process was blinded, was a careful search for hip fractures conducted with verification from an independent source? Clarification of these issues would assist clinicians working with older people at high risk of hip fractures. There are now a number of review articles that recommend hip protectors for older people who have had multiple falls (3) and further evidence based information is necessary to ensure their appropriate use. References: 1. Meyer G, Warnke A, Bender R, Muhlhauser I. Effect on hip fractures of increased use of hip protectors in nursing homes: cluster randomised controlled trial. BMJ 2003;326:76. 2. Parker MJ, Gillespie LD, Gillespie WJ. Hip protectors for preventing hip fractures in the elderly (Cochrane Review). In: The Cochrane Library, Issue 4, 2002. Oxford: Update Software. 3. Tinetti ME. Preventing Falls in Elderly Persons. NEJM 2003;348:42-49. Competing interests: None declared |
|||
|
|
|||
|
Gabriele Meyer, Research fellow Unit of Health Sciences and Education, University of Hamburg, D-20146 Hamburg, Germany, Andrea Warnke, Ingrid Mühlhauser, and Ralf Bender
Send response to journal:
|
Torgerson is wrong in assuming that the two groups were not comparable. Comparability of groups is reflected by similar baseline characteristics. In fact, the two groups were comparable with respect to the most relevant risk indicators as assessed before randomisation (Table 1 and Table 2). Possible differences during follow up in a randomised controlled trial could be a result of the intervention. Thus, the trend to fewer falls (statistically not significant) and a trend to lower mortality in the intervention group and thus a trend to a longer observation period (statistically not significant) in the intervention group could be a result of the intervention rather than higher frailty in the control group. In contrast to Torgerson's claims we have addressed this aspect in the discussion section and warned from overinterpreting these observations. Adherence was assessed by documentation of hip protector use during a fall. We have addressed extensively the problem of documentation of adherence in the discussion section. In fact, the true proportions of residents who used the hip protector are not known. We have included the estimated figures on request by the BMJ reviewers. We still feel they should not have been included. The proportion of residents who used the hip protector are estimates based upon the observed proportion of fallers who used the hip protector during at least one fall event as documented by fall documentation sheets. We could not document how many of the residents who did not fall during the study have used the protector. Thus, the proportion of residents using a hip protector is certainly an underestimate based on the assumption that no resident without falls during the observation period had used the hip protector. Competing interests: We refer to our article. |
|||
|
|
|||
|
Gabriele Meyer, Research fellow Unit of Health Sciences and Education, University of Hamburg, D-20146 Hamburg, Germany, Andrea Warnke, Ingrid Mühlhauser, and Ralf Bender
Send response to journal:
|
The calculations by Paul J. Young are wrong. We refer the readers to the data provided in Table 3, the results section and the abstract. Competing interests: We refer to our article. |
|||
|
|
|||
|
Gabriele Meyer, Research fellow Unit of Health Sciences and Education, University of Hamburg, D-20146 Hamburg, Germany, Andrea Warnke, Ingrid Mühlhauser, and Ralf Bender
Send response to journal:
|
We thank Susan Kurrle and Ian Cameron for the acknowledgement of the high quality of the study design. The grading system used in the study for assessment of the degrees of disablement does not allow the separate description of functional and cognitive status. Further investigation on cognitive impairment was not performed because this aspect is not important to answer our study question. In the introduction section we have emphasised that adherence will be largely determined by the motivation and competence of nursing home staff. Therefore we educated nurses who were charged to inform residents. We did not intend to quantify the percentage of residents participating in small group educational sessions since educated nurses were responsible to implement the intervention. We have emphasised that the individual effects of the combined intervention of the two components cannot be separated (section intervention). We did not provide more than three hip protectors per residents. Hip protectors were distributed by nursing staff. Therefore nurses could decide to redistribute hip protectors offered to residents who deceased or moved. A cost-effectiveness analysis will be reported elsewhere. We did not advise nursing staff to complete a form on detailed information on non-adherence because these details were expected to be extremely uncertain and are not important to answer our study question. The fall documentation sheet intended to take as little time as possible to ensure commitment of nursing staff. As pointed out in the results section fall documentation in the intervention group indicated a total of 394 falls without the use of hip protectors. Decline of residents was most often mentioned (among 249 falls) followed by other reasons which resulted in non-adherence for example the hip protector could not be provided because of laundering. No radiologist of the various hospitals where residents had been admitted with a hip fracture had been informed about the corresponding nursing homes being part of a study. In Germany, it is obligatory for nursing staff to document all falls and fractures for legal reasons in the residents records. Any fracture, suspicion of fracture and significant injury requires consultation of a physician for legal reasons. In Germany, there is substantial overuse rather than underuse of x-rays in case of suspected fractures. Hip fractures have been documented not only by nursing staff, but also evaluated by personal visits by the investigators every two months. At the end of the study, one investigator and the nominated study co-ordinator of each cluster re-evaluated all records including hospital discharge reports and radiologists reports to ascertain whether a hip fracture had occurred and to verify the completeness of the fracture (and fall) data (section study outcomes). Competing interests: We refer to our article. |
|||
|
|
|||
|
Paul J Young, House Officer (orthopaedics) Middlemore Hospital, Auckland, New Zealand
Send response to journal:
|
Apologies for the previous submission - the calculations were indeed wrong! In the hip protector group 4.6% have hip fractures over the 15 month period of follow-up corresponding to 3.68% having a hip fracture per year. In the control group 8.1% have a have hip fractures over the 14 month period of follow-up corresponding to 6.94% having a hip fracture per year. The absolute risk difference is 3.26% per year. The number of people needed to treat for one year to prevent one hip fracture is 30.7. Stastically, a patient can expect to wear a hip protector for just over 30 years to prevent one hip fracture. It is still not clear whether this is worthwhile. Competing interests: None declared |
|||
|
|
|||
|
Paul J Young, House Officer (orthopaedics) Middlemore Hospital, Auckland, New Zealand
Send response to journal:
|
Given that hip protectors are used by both groups in this study, it more accurate to say that the intervention would need to be employed on a patient over a period of 30 years on average. The key point I wish to raise is that the presentation of the data emphasised in terms of relative risk reduction (see the highlighted table) is misleading. A relative risk reduction of 40% is in fact a small absolute difference when outcomes occur in only a small percentage of patients. Competing interests: None declared |
|||
|
|
|||
|
Percy V van Eerten, trauma surgeon MMC-E Eindhoven Netherlands
Send response to journal:
|
I would like to complement Gabriele Meyer and co-workers for there well conducted study. Although I have a vew commnets on the interpertation of the results. The discussion opens with the statement of a risk reduction of 40% at borderline significance can be expected from the hip protector. To my perception the confidence interval from -7.3 to 0.3 should be interpreted as not significant at all since it crosses the zero line. Probably due to the relative low number of hip fractures in both groups. A proper power calculation given this incidence can solve this problem. Percy van Eerten Competing interests: None declared |
|||
|
|
|||
|
Gabriele Meyer, Research fellow Unit of Health Sciences and Education, University of Hamburg, D-20146 Hamburg, Germany, Andrea Warnke, Ingrid Mühlhauser, and Ralf Bender
Send response to journal:
|
We refer the readers to the sample size section and the statistical methods section. We have used an one sided hypothesis. However, to follow statistical convention results for two sided tests are reported. Competing interests: We refer to our article. |
|||
|
|
|||
|
Pekka A Kannus, Chief Physician Accident & Trauma Research Center, 33500 Tampere, Finland
Send response to journal:
|
Meyer and colleagues are to be congratulated on their hip protector study. However, the trial left the main issue or effectiveness of the studied hip protector uncertain or unanswered. First, I fully agree with Torgeson that the two groups were not comparable since significantly more elderly persons in the controls withdrew from the trial during the follow-up and since the controls had a clearly (29%) higher incidence of falling than those in the intervention group. Whether these differences already existed at baseline or developed during the trial remained unknown, especially since many important baseline characteristics were not reported in the paper. Second, since difference in the risk of hip fracture in the two groups was statistically insignificant, the fracture protective effect of the hip protector remained uncertain. This is also seen as an open end in the 95% confidence interval of the NNT calculation. Third, the authors did not provide statistical analysis without cluster randomisation so it remained unknown whether adjustment for this procedure had any effect on the relative fracture risk and its p-value. For falls, this adjustment had relatively little effect. Fourth, since the authors could not provide the fracture data per protected and unprotected falls (in neither group), it also remained unclear what was the biomechanical efficacy of the studied protector in actual falls. This information would have been essential in interpreting the relative risk obtained, especially when about half of the falls in the protector group were without the protector. Competing interests: None declared |
|||
|
|
|||
|
Ingrid Mühlhauser, Professor University Hamburg, Unit of Health Science and education, Gabriele Meyer, Andrea Warnke, Ralf Bender
Send response to journal:
|
With regard to basic principles of clinical epidemiology we refer Kannus to our response to Torgerson. We are not aware of important characteristics others than those already included in baseline assessment. In addition, adequate concealed randomisation as performed in our study should prevent significant differences in baseline characteristics that have not been assessed.
We want to draw the readers' attention to the fact, that in contrast to Kannus' assumption participants did not "withdraw" from the study, but rather had died during the study. In about 90% death was the reason for early study termination a finding not unexpected among a nursing home population with a mean age of approx. 87 years to start with. With respect to the second remark of Kannus we want to draw the readers' attention to our response to Eerten. We have used a one-sided hypothesis, but nevertheless reported two-sided test results. Because the study was cluster randomised only statistical analysis taking cluster randomisation into account is appropriate. This procedure weakens the statistical significance of differences. Thus, analysis not taking cluster randomisation into account yielded lower p-values. We don't understand the last point raised by Kannus. Competing interests: We refer to our article |
|||
|
|
|||
|
Pekka A Kannus, Chief Physician Accident & Trauma Research Center, UKK Institute, 33500 Tampere, Finland
Send response to journal:
|
In reply to Meyer et al's response to my letter: Many important baseline characteristics (risk factors of hip fracture)were not reported in the Table 2 of the study. For example, subjects' height, weight, BMI, mental status, and the most common medical conditions (heart disease, diabetes, osteoarthritis, eye diseases etc.) were not included. Randomization, although a good procedure as such, does not fully quarantee group comparability, and, if between-group differences are found, the results should be adjusted for these variables. In statistics, "withdrawal" means premature study termination irrespective of its passive or active causes and that I meant in my letter. Whatsoever the reasons for this termination were in Meyer et al's study, a fact still is that significantly more control elderly persons could not be followed to the end thus leaving a possibility that the controls were in general more frail (and thus more attentive to hip fracture) than their counterparts in the the hip protector group. Concerning one-sided hypothesis, very few hip protector researchers would accept one-sided hypothesis, since at least two previous RCTs made with this particular protector have shown no effect (Cameron et al. Age and Ageing 2001, Hildreth and Torgeson Osteoporos Int Suppl 2, 2001). Thus, the only correct way is to use two-sided hypothesis(as the authors did),and, the result was that the risk of hip fracture did not differ statistically significantly between the two study groups. I agree that it is reasonable to take into account the possible cluster effect. My point was that readers would like to know how much the cluster effect was in the fracture data. Concerning falls, it was relatively small: 58% of the falls in the intervention group occurred with the protector, and after adjustment with cluster randomisation, the figure was 54%. In controls, these numbers were 11% and 8%. So, how much the cluster effect, if there, changed the fracture risks in the Table 1? My final comment concerned the true efficacy analysis of the studied protector; that is, the risk of hip fracture (or other fractures)analysed per protected and unprotected falls in the protector group (and in the control group, too, since 11% of their falls were with the protector!). So, how many hip fractures occurred in the protector group as the result of their 552 protected falls vs. their 394 unprotected falls? Accordingly, how many fractures occurred in the controls per their 160 protected falls vs. their 1249 unprotected falls? These calculations can be easily done provided that information concerning the protector use during each fall was adequately collected and reported. This analysis would provide valuable information about the biomechanical efficacy of the studied protector in the actual fall situations, since the claimed 40% intention- to-treat risk reduction in the protector group can be true only if there was a clearly higher risk reduction in actual falls(about half of the falls in the protector group were without the protector and yet the above noted 40% risk reduction compared to controls). I could not find this efficacy analysis information from the paper of Meyer et al. so it seems that the protector use in the injurious and noninjurious falls were not recorded. Competing interests: None declared |
|||