Rapid Responses to:
|
|
Rapid Responses: Rapid Responses published:
-
![[Read Rapid Response]](/icons/misc/sectionDOWN.gif)
Results of randomised trials – relative or absolute effect measures?
- Douglas G Altman, Kenneth F Schulz, David Moher (4 September 2002)
-
Responding to issues raised
- Salim Yusuf, Jackie Bosch, Peter Sleight (30 October 2002)
-
demystifying with HOPE
- Malvinder S. Parmar (6 December 2002)
-
Correction to "demystifying with HOPE"
- Malvinder Parmar (23 December 2002)
|
|
|||
|
Douglas G Altman, professor of statistics in medicine Centre for Statistics in Medicine, Oxford, Kenneth F Schulz, David Moher
Send response to journal:
|
Several correspondents have criticised the HOPE investigators for focusing on relative effects when the absolute benefit was very small. Birte Twisselmann, in a summary of many responses on the web site, comments: “It was noted that only the relative risk reduction was given in the study. This should have been accompanied by data on absolute risk reduction and number needed to treat and even number needed to harm (as per CONSORT guidelines).” This sentence may accurately reflect comments on bmj.com but it misrepresents the CONSORT statement. Although we agree that in many cases it is helpful to present results as both reduction in relative risk and reduction in absolute risk, CONSORT makes no such recommendation. Dr Badrinath observes: “If general agreement is reached then the next CONSORT guidelines should include a statement that wherever applicable the results of clinical trials should include the numbers needed to treat.” It seems clear that no such consensus exists, which is why CONSORT does not recommend that the NNT is always given. Rather, the CONSORT explanatory document says simply: “For both binary and survival time data, expressing the results also as the number needed to treat for benefit (NNTB) or harm (NNTH) can be helpful.” [1] The choice between relative and absolute measure of treatment effect continues to cause debate, yet no such choice is needed. In some situations, results can usefully be presented as both relative and absolute effects. However, that could be cumbersome in a journal article for every primary and secondary outcome presented. Authors and journals demand and deserve some flexibility in presenting their results. CONSORT recommends merely the basics and avoids subjective assessments on the presentation details. Simply, we advocate that authors provide “a summary of results for each group, and the estimated effects size and its precision” which entails providing adequate raw data so that readers can calculate either a relative or absolute measure if it is not provided, and they deem that necessary. Unfortunately, there have been other instances of authors incorrectly attributing certain reporting criteria to CONSORT.[2] While we are encouraged by the growing dissemination of the CONSORT statement as a way to help improve the quality of reporting randomised trials, we also encourage a closer examination of its content. 1. Altman DG, Schulz KF, Moher D, Egger M, Davidoff, Elbourne D, Gøtzsche PC, Lang T for the CONSORT Group. The revised CONSORT statement for reporting randomized trials: explanation and elaboration. Ann Intern Med 2001;134:663–694. 2. Schulz KF, Moher D, Altman DG. Interpreting the number needed to treat. JAMA 2002;288:831. Douglas G Altman
Kenneth F Schulz
David Moher
|
|||
|
|
|||
|
Salim Yusuf, Professor of Medicine, Director, Population Health Research Institute, McMaster University Population Health Research Institute, 237 Barton Street East, Hamilton, Ontario, Canada L8L 2X2, Jackie Bosch, Peter Sleight
Send response to journal:
|
We were surprised at the correspondence (BMJ, August 24, 2002) regarding our manuscript that described the effects of ramipril in preventing stroke in the HOPE study, as we had no opportunity to review or respond at the time that the letters were published. Our publication of the effects of ramipril on stroke should be taken in the context of the overall main results publication(1) that described the effects on a number of major clinical outcomes. Moreover, this publication provides all the necessary details such as relative risk reduction, absolute differences, and this information can be used to easily calculate number needed to treat. When one considers all major vascular events prevented, including strokes, myocardial infarction, cardiovascular death, heart failure, prevention of revascularization, as well as the prevention of diabetes, the number needed to treat to prevent one event is extremely small. The table below provides the relevant information:
HOPE: Events Prevented Per 1,000 People Treated
No. of Events Prevented per
1,000
Individuals Treated Correction for
20% Noncompliance
No. Deaths 18 23
Myocardial Infarction 16 20
Stroke 9 11
Revascularization 26 33
Congestive Heart Failure 26 33
Cardiac Arrest 5 6
Diabetes Complications 12 15
New Diabetes 16 20
Total Events 128 NNT*=8 161 NNT*=6
No. People with above 59 NNT**=17 74 NNT**=14
*NNT=Number needed to prevent one event
**NNT=Number to prevent an individual developing an event
The side effects due to ramipril are fully described in our main paper, and there are no major side effects that will offset the clinical benefits observed in this study. Dr. Parmar raises the cost effectiveness of ramipril. His calculations are incorrect, as it does not take into account the prevention of numerous other vascular events (described in the table above), nor does it take into account the cost savings that occur from preventing these events and related hospitalization. When a formal cost effectiveness analysis is done(2,3), the use of ramipril for 5 years is cost neutral. Indeed, in some high-risk subgroups (example those over the age of 65 years), the use of ramipril is cost saving. Prof. Yudkin speculates as to whether or not benefits of HOPE can be achieved with other blood pressure lowering agents. This is simply not known. It should be noted that HOPE was designed not as a blood pressure lowering trial, but was based on the observation that ACE inhibitors reduced myocardial infarction, heart failure hospitalizations and cardiovascular death in individuals with heart failure or left ventricular dysfunction(4,5), and normal BP (mean systolic blood pressure of below 120 in these trials). The hypothesis of HOPE was to evaluate whether this effect could be observed in high-risk individuals with vascular disease, but without left ventricular dysfunction or heart failure. While some of the benefits of ramipril seen in HOPE are likely to be due to the drug’s blood pressure lowering effects, a large part remains unexplained simply on the basis of BP lowering. This is fully described in a publication in the Lancet(6). We agree with Prof. Yudkin that the best way to assess whether an agent has a specific cardioprotective effect independent of blood pressure is to compare two agents with different properties but with the same degree of blood pressure lowering. At least two recent trials have confirmed the superiority of blocking the renin angiotensin system as compared to other means of blood pressure lowering in high-risk individuals. This includes the African American Study of Kidney Disease and Hypertension (AASK) trial(7) in which ramipril was superior to amlodipine, a calcium channel blocker, in reducing mortality despite lowering blood pressure to the same extent. Second, the recent Losartan Intervention For Endpoint reduction (LIFE) study(8) indicated that in hypertensive patients with left ventricular hypertrophy, losartan, an angiotensin-2 receptor blocker significantly lowered major vascular events by 13% compared to atenolol, a beta-blocker, despite achieving similar levels of blood pressure lowering. In addition, a meta-analysis of BP lowering trials indicates that ACE-inhibitors are superior to calcium channel blockers in preventing myocardial infarction and heart failure(9) and recent trials of angiotensin receptor blockers demonstrate superiority over calcium blockers in patients with renal dysfunction(10) Thus, there are substantial external data supporting our hypothesis that the benefits of ramipril observed in HOPE are largely independent of blood pressure lowering. At present there are no data that any other blood pressure lowering agent in the population studied in HOPE will produce the same clinical benefits observed with ramipril. Yours sincerely, Salim Yusuf, DPhil, FRCPC
Jackie Bosch, MSc
Peter Sleight, MD, DM, FACC
References: 1. The Heart Outcomes Prevention Evaluation Study Investigators. Effects of an angiotensin-converting-enzyme inhibitor, ramipril, on cardiovascular events in high-risk patients. New Engl J Med 2000; 342(3): 145-53. 2. Lamy A, Gafni A, Pogue J, Yusuf S. Cost-effectiveness of ramipril in high risk patients: analysis of the HOPE Study. 53rd Annual Meeting of the Canadian Cardiovascular Society, October 29-November 1, 2000, Vancouver, British Columbia. Can J Cardiology 2000; 16(Suppl F): 233F 3. Bjorholt I, Andersson FL, Kahan T, Ostergren J. The cost- effectiveness of ramipril in the treatment of patients at high risk of cardiovascular events: a Swedish sub-study of the HOPE study. J Internal Med 2002; 251: 508-17. 4. Yusuf S, Pepine CJ, Garces C et al. Effect of enalapril on myocardial infarction and unstable angina in patients with low ejection fractions. Lancet 1992; 340(8829): 1173-8. 5. Flather MD, Yusuf S, Køber L et al for the ACE-inhibitor Myocardial Infarction Collaborative Group. Long-term ACE-inhibitor therapy in patients with heart failure or left-ventricular dysfunction: a systematic overview of data from individual patients. Lancet 2000; 355(9215): 1575-81. 6. Sleight P, Yusuf S, Pogue J, Tsuyuki R, Diaz R, Probstfield J, for the Heart Outcomes Prevention Evaluation (HOPE) Study Investigators. Blood -pressure reduction and cardiovascular risk in HOPE study. Lancet 2001; 358: 2130-31 7. Agodoa LY, Appel L, Bakris GL et al for the African American Study of Kidney Disease and Hypertension. Effect of ramipril versus amlodipine on renal outcomes in hypertensive nephrosclerosis: a randomized controlled trial. JAMA 2001; 285: 2719-2728. 8. Lindholm LH, Ibsen H, Dahlöf B et al for the LIFE Study Group. Cardiovascular morbidity and mortality in patients with diabetes in the Losartan Intervention For Endpoint reduction in hypertension study (LIFE): a randomised trial against atenolol. Lancet 2002; 359(9311): 1004-10. 9. Blood Pressure Lowering Treatment Trialists’ Collaboration. Effects of ACE inhibitors, calcium antagonists, and other blood-pressure- lowering drugs: results of prospectively designed overviews of randomised trials. Lancet 2000; 355: 1955-64. 10. Lewis EJ, Hunsicker LG, Clarke WR et al for the Collaborative Study Group. Renoprotective effect of the angiotensin-receptor antagonist irbesartan in patients with nephropathy due to type 2 diabetes. New Engl J Med 2001; 345(12): 851-60. Competing interests: All are disclosed in the original paper |
|||
|
|
|||
|
Malvinder S. Parmar, Medical Director, Medical Program Timmins & District Hospital, Timmins, ON. P4N 8R1
Send response to journal:
|
Thanks to Yusuf et al for finally responding (seven months after the original publication) to the issues raised concerning their paper1. BMJ editors would be better to answer the concern raised by the authors that enough time was not given to them for responding to the queries before publishing the letters. I believe that it is the policy of BMJ to send the ‘rapid responses’2 to the corresponding author by email regularly. Editors waited for 5-months (from 23 March 2002 to 24 August 2002) before publishing a select rapid responses in the letters section3. I don’t think that this in anyway should surprise anyone, including the authors or the readers. In fact, the editors also wrote a short editorial on how to avoid criticism4 in the same issue. I feel that one way for the authors to avoid criticism is to respond to the queries raised by the reader’s in a timely fashion. We all make mistakes. Admitting mistakes is human and a sign of maturity. I duly respect Prof. Yusuf for his contribution to the field of cardiovascular medicine and for his important role in conducting mega- trials around the world. He mentioned that the results of the Bosch paper should be taken in context with the HOPE5 trial and major clinical outcomes, differences (absolute or relative), complications are mentioned in the original HOPE paper. If we need to take into consideration all the results of HOPE with the Bosch paper under discussion then why publish another paper that cannot stand on its own. This in my opinion is a superfluous and useless publication then. I understand that sometimes papers are presented in parts and part of the information or results are given in each and each part has its own value or impact. The effect of ramipril on stroke was already presented in HOPE trial results and no new information was presented in the paper under discussion. So, bottom line is why publish another paper and essentially reiterate the original results. There is already glut of medical publications. The table shown in this letter6 is from the HOPE trial and not from the paper under discussion. By presenting this table from another paper (of course their own), the authors once again trying to skew the reader’s perception. Various readers2 (see rapid responses to your original paper) and I are not objecting to the data from the HOPE trial but expressed our concerns about the data presented in the current paper. So, please don’t try to convince others by presenting information from HOPE. We are not discussing HOPE. Also, in the table presented, another manipulation of data is presented. If look at the HOPE results in relation to primary endpoints (MI, stroke, death) – the RRR was 0.78, ARR was 3.8, NNT was 26 and 38 events were prevented per 1000 patients treated. However, in the table, authors combined all the primary, secondary end-points and incidental findings (new diabetes prevented) and then came to the conclusion that 128 events were prevented per 1000 patients treated. So, by manipulating data authors increased the total primary endpoints prevented from 38 to 128 (an increase of about 30%). This is another example of how the data can be manipulated or skewed to alter the reader’s and public’s perception. Now, a simple question, how the information presented in the current paper under discussion differs from the HOPE paper and if it doesn’t then what was the real reason for another so called ‘superfluous’ publication? I am not against ramipril. I use it often. I am not saying that ramipril is not effective in preventing cardiovascular events or stroke in high-risk populations but based on the information given in the paper under discussion, I stay corrected that the conclusion of this paper are not supported by the evidence presented in the Bosch paper. I am not a statistician and understand that my ‘simple’ calculations might be wrong (I myself pointed that in my original eletter) as pointed out by Dr. Yusuf but then how would you explain the points raised by other readers7 and in fact comments published in the Journal Watch Neurology8 by Dr. Gorelick. It is interesting that Dr. Gorelick’s conclusions were similar to mine. Authors may support their point by using information from other papers, including HOPE, but still ramipril’s use based on the evidence provided cannot be justified in all but only in a select group of patients at high- risk for stroke. No matter how one manipulates the data, the trial’s persuasiveness is in the eye of the beholder. References: 1. Bosch J. et al: Use of ramipril in preventing stroke: double blind randomized trial. BMJ 2002; 324:699-702 2. Electronic responses. Use of ramipril in preventing stroke: double blind randomized trial. bmj.com 2002. bmj.com/cgi/eletters/324/7399/699 3. Parmar MS: Preventing stroke with ramipril: presentation of data is misleading. BMJ 2002; 325:440-441 4. How to avoid criticism. BMJ 2002; 325:0. 5. The Heart Outcomes Prevention Evaluation Study Investigators. Effect of an angiotensin-converting enzyme inhibitor, ramipril, on cardiovascular events in high-risk patients. New Eng J Med 2000; 342(3):145-53. 6. Yusuf et al: Responding to issues raised. bmj.com 2002. bmj.com/eletters/325/7361/439. 7. Mclaren H. Editorials must be more evidence based. BMJ 2002; 325:280. 8. Gorelick PB: Does ramipril prevents stroke and stroke-related disability? Journal Watch Neurology. May 23, 2002; 2002(523):1. Competing interests: None declared |
|||
|
|
|||
|
Malvinder Parmar, Medical director, Medical Program Timmins & District Hospital, Timmins, ON.
Send response to journal:
|
In the above letter there is a typographical error and by combining all the events, the increase is 330% instead of 30%. Malvinder S. Parmar Competing interests: None declared |
|||