Rapid Responses to:

PRIMARY CARE: Reto Nuesch, Kerstin Schroeder, Thomas Dieterle, Benedict Martina, and Edouard Battegay Relation between insufficient response to antihypertensive treatment and poor compliance with treatment: a prospective case-control study BMJ 2001; 323: 142-146 [Abstract] [Full text]

Rapid Responses: Submit a response to this article

Rapid Responses published:

This is not a case control study

Martin Bland   (24 July 2001)

Need to acknowledge pharmacological properties of antihypertensive drugs in compliance research

Knut Schroeder   (27 July 2001)

Blood pressure and adherence monitoring in a hypertensive population: measurement or intervention?

J J Parienti   (28 July 2001)

Compliance in resistant hypertension

Burnier Michel   (30 July 2001)

Lack of statistical power

Adam Jacobs   (3 August 2001)

Sample size

Joyce Cramer   (7 August 2001)

This is not a case control study 24 July 2001
Martin Bland, Prof. of Medical Statistics St. George's Hospital Medical School Send response to journal: Re: This is not a case control study	I would like to make three points about this report. First, this is not a case-control study in the usual meaning of the term. A case-control study starts with a group of subjects with a disease then finds another group of subjects, without the disease, to which they are compared. In this study, patients with hypertension were recruited and then measured to see whether they were non-responsive (cases) or responsive (controls). We should not label all comparisons of two groups as case-control studies. The term `case-control study' carries with it many implications concerning the analysis and interpretation of the data which are not relevant here. Second, the comparison of proportions of responders by clinic and ambulatory pressure should not be done by Fisher's exact test. The authors appear to have used this to do a comparison of 43/105 clinic responders with 55/105 ambulatory pressure responders. Fishers exact test for this comparison gives p=0.1276 using the doubling of the one-tailed P- value method, which is presumably how the authors got p=0.127. But this counts every subject twice and ignores the fact that they are the same subjects. McNemar's test should be used. There were 12 changes from non- response to response and none from response to non-response, which gives p=0.0015 using the continuity correction. Third, the continuous blood pressure measurements have been dichotomised into response and non-response, and the numerical count of number of times the drug container was opened has been dichotomised into compliant and non-compliant, before any test of significance was done. Had the full data been used, this would have given greater power to detect any relationship. These are minor points and I do not think that they would affect the main conclusions of the study, but I think it is important to get things right. It sets a good example to others.
Need to acknowledge pharmacological properties of antihypertensive drugs in compliance research 27 July 2001
Knut Schroeder, MRC Training Fellow in Health Services Research Division of Primary Health Care, University of Bristol, Canynge Hall, Whiteladies Road, Bristol Send response to journal: Re: Need to acknowledge pharmacological properties of antihypertensive drugs in compliance research	The paper by Nuesch et al suggests that non-compliance with treatment may not be associated with resistance to antihypertensive treatment.[1] Although there is little evidence on this topic,[2] in a recent paper Burnier et al came to the opposite conclusion.[3] Their study involved 41 hypertensive patients resistant to a three-drug regimen. Electronic monitoring of medication compliance alone for two months led to a significant reduction in mean blood pressure from 156/106 mmHg (SD 23/11 mmHg) to 145/97 mmHg (SD 20/15 mmHg, p<0.01). However, this study was small and did not have a control group. Nuesch et al could have strengthened their conclusions by providing some additional information. Side effects of treatment may influence compliance but were not considered in the comparisons of compliance/non- compliance or of response/non-response. It is also important to know whether both study groups were similar with regard to the type and number of antihypertensive drugs being used. Non-compliance with long-acting medications may have a much smaller impact on treatment effect compared to shorter acting agents where the timing of doses may be more important. In our opinion this study was too short and too small to come to any firm conclusions about the relationship between compliance and blood pressure control. Both the methods of measuring compliance and blood pressure are not typical of clinical practice and may have induced powerful effects in some patients, thereby removing any relationship between compliance and blood pressure control. Future studies should acknowledge the different pharmacological properties of various antihypertensive drugs. They should attempt to define what level of medication compliance is required to achieve a desired treatment effect. Electronic monitors can provide valuable additional data that include the exact timing of doses and inter-dose intervals.[4] Although methodologically challenging, analysis of such data might provide more detail on the extent and nature of non-compliance and its relationship with blood pressure control. We are currently recruiting 350 patients with uncontrolled essential hypertension for a primary care based randomised controlled trial on the effectiveness of nurse-led and patient-centred compliance counselling, funded by the Medical Research Council. We will use electronic monitors for eight to 12 months after an initial 2-month run-in period prior to the intervention. We hope to be able to contribute new data on the relationship between medication compliance and blood pressure control. Conflict of interest: none Yours sincerely Knut Schroeder, MRC Training Fellow in Health Services Research, Division of Primary Health Care, University of Bristol, Canynge Hall, Whiteladies Road, Bristol BS8 2PR, 7340, email k.schroeder@bristol.ac.uk Shah Ebrahim, Professor in the Epidemiology of Ageing, Department of Social Medicine, University of Bristol, Canynge Hall, Whiteladies Road, Bristol BS8 2PR Alan Montgomery, MRC Training Fellow in Health Services Research, Division of Primary Health Care, University of Bristol, Canynge Hall, Whiteladies Road, Bristol BS8 2PR References 1. Nuesch R, Schroeder K, Dieterle T, Benedict M, Battegay E. Relation between insufficient response to antihypertensive treatment and poor compliance with treatment: a prospective case-control study. BMJ 2001;323:142-146. 2. Ebrahim S. Detection, adherence and control of hypertension for the prevention of stroke. A systematic review. Health Technology Assessment 1998;2: No. 11. 3. Burnier M, Schneider MP, Chiolero A, Stubi CLF, Brunner H. Electronic compliance monitoring in resistant hypertension: the basis for rational therapeutic decisions. J Hypertens 2001;19:335-341. 4. Metry J-M. Measuring compliance in clinical trials and ambulatory care. In: Metry J-M, Meyer UA, eds. Drug regimen compliance: issues in clinical trials and patient management. Chichester: John Wiley & Sons Ltd, 1999;1-21.
Blood pressure and adherence monitoring in a hypertensive population: measurement or intervention? 28 July 2001
J J Parienti INSERM U444 Faculté de médecine Saint Antoine, Paris, France Send response to journal: Re: Blood pressure and adherence monitoring in a hypertensive population: measurement or intervention?	I read with interest the recent article by Nuesch R et al.1 In their study, they demonstrated that electronic monitoring of patient’s adherence had no significant effect on ambulatory blood pressure. Because non- adherence was not more prevalent among patients with resistance to antihypertensive treatment, they concluded that other factors independent of a patient’s willingness to adhere are more relevant in explaining treatment failure, in most patients. However, I believe that these results should be interpreted with caution due to possible selection and differential measurement bias that may have occurred in their study. Patients who accepted inclusion in the study and therefore who were monitored for blood pressure and adherence may be more adherent to antihypertensive treatments than general population. Unfortunately, we have no information about how many eligible patients refused to participate. Furthermore, after 28 days of study and based on the results of ambulatory blood pressure, more patients became responsive to treatment than non- responsive (X2=2.9, p<0.09): 14/49 (28.6%) versus 8/54 (14.8%) respectively. It is important to note that the patients were all following a stable treatment regimen. The only interventions were to monitor blood pressure and adherence, which are both recommended to encourage the patients to take the prescribed medication regularly. Therefore in order to explain this clinically relevant difference, I believe that the impact of electronic measurement on adherence to treatment was stronger among patients with uncontrolled blood pressure at the study entry than among other patients (control group). Considering that adherence was enhanced in the non-responder group (differential bias) during the study, it is not surprising that the prevalence of non-adherence was similar among responder and non-responder patients before the study. Reclassification of patients according to the second measure of blood pressure would not have led to different conclusion. However, as correctly discussed by authors1, interventions to help patients to follow their regimen improving adherence but not treatment outcome are common, particularly among hypertensive patients.2 Finally I agree that other factors that patient’s adherence, such as physician’s non-compliance with current treatment guidelines,3 can play a role in treatment failure of some patients. Nevertheless, in case of insufficient control of hypertension instead of adapted treatment I still believe that non-adherence should be considered first, in most patients. This strategy is likely to prevent the “folie à deux” (madness of two) where the physician’s perception of treatment failure results in both : the physician increasing dose or number of drugs and the patient’s adherence temporarily improving, resulting in drug toxicity. Jean-Jacques Parienti 1. Nuesch R, Shroeder K, Dieterle T, Martina B and Battegay E. Relation between insufficient response to antihypertensive treatment and poor compliance with treatment: a prospective case-control study. BMJ 2001;323:142-46. 2. Haynes RB, McKibbon KA, Kanani R. Systematic review of randomised trials of interventions to assist patients to follow prescriptions for medications. Lancet 1996;383-6. 3. Berlowitz DR, Ash AS, Hickey EC et al. Inadequate management of blood pressure in a hypertensive population. N Engl J Med 1998;339:1957-63.
Compliance in resistant hypertension 30 July 2001
Burnier Michel, Associate Professor CHUV, Lausanne, Switzerland Send response to journal: Re: Compliance in resistant hypertension	DEAR EDITOR - We have read with interest the recent paper by Nuesch et al (1) which evaluated the relationship between insufficient response to antihypertensive treatment and poor compliance to therapy. In this paper, the authors conclude that non-compliance with treatment is not more prevalent in patients with treatment-resistant hypertension, and hence that non-compliance is not a relevant cause of resistance to anti- hypertensive treatment. We feel that the data presented in this paper do not support these conclusions which actually provide a misleading clinical message. Indeed, non-compliance may be equally frequent in treatment-responsive and in treatment-resistant patients. However, this does by no means indicate that poor adherence to therapy is not an important issue in resistance to therapy. How can we explain that a poorly compliant patient has a normal blood pressure ? Non-compliance in a treatment responsive patient suggests that the patient is either over-treated or never needed anti-hypertensive therapy, i.e. he suffers from white coat hypertension.. Clinically, it is of course more important to detect non-compliance in treatment resistant patients because of the potential impact on cardiovascular complications. In this respect, it is rather unacceptable that the authors do not provide the blood pressure values measured before and after compliance monitoring in all subgroups since blood pressure control is after all the ultimate goal of the whole intervention. They do also not emphasize that one third of the treatment resistant patients could have been reclassified to treatment-responsive following the introduction of the compliance monitoring which per se improves compliance and thereby blood pressure control. One must also question the validity of the arbitrarily chosen cut-off for compliance at 80%. This has often been used in the literature but there exists no evidence that this number has any clinical relevance. The data presented by the authors would actually indicate that this percentage has no clinical meaning since some patients could be controlled with less than 80% compliance and others could not. In addition, the investigators did not take into account the dynamic aspect of compliance. Patients often improve compliance immediately before the consultations (white coat compliance). Thus, it is important to analyze the distribution of the days of non-compliance in relation with the corresponding blood pressure measurements. Finally, to assess the relevance of non-compliance in explaining treatment resistance in hypertensive patients, one should at least try to actively enhance compliance and evaluate the impact of this intervention on blood pressure control, over an observation period of at least 6 months. Thus, this study was too short and lacked an active intervention. In our opinion, the inherent message provided by the authors borders on absurdity : “it does not matter whether the patient takes his treatment regularly or not”. Do we really question the causal relationship that should exist between the ingestion of the antihypertensive medication and blood pressure control ? M. Burnier, Associate Professor and H.R. Brunner, Professor of Medicine Division of Hypertension and Vascular Medicine, CHUV, Lausanne, Switzerland E-mail address: Michel.Burnier@chuv.hospvd.ch Reference: 1. Nuesch R, Schroeter K, Diertele T, Martina B, Battegay E. Relation between insufficient response to antihypertensive treatment and poor compliance with treatment: a prospective case-control study. BMJ 2001:323:142-146.
Lack of statistical power 3 August 2001
Adam Jacobs, Director Dianthus Medical Limited Send response to journal: Re: Lack of statistical power	It seems to me that the conclusions that can be drawn from Nuesch et al’s study are limited by the lack of statistical power. The authors state that they did a power calculation before the study, but give no details. What did they consider to be a clinically relevant difference in compliance between the groups? What were their assumptions about the proportions of responders and non responders, and compliant and non- compliant patients? What sample size did they calculate that they needed, and did they recruit that number of patients? Their results were that 46 of 54 (85%) responsive patients were compliant, compared with 40 of 49 (82%) non-responsive patients. This gives a difference between the proportions of 3.6%, but with a 95% confidence interval of –11% to 18%. In other words, their results are statistically compatible with an 18% higher compliance rate in the responders, which could easily be seen as a clinically important difference. For those more comfortable with odds ratios, the odds ratio for compliance with treatment in responsive versus non-responsive patients was 1.3 with a 95% confidence interval of 0.47 to 3.6. Because of the wide confidence intervals around the results of the study, it is difficult to draw any firm conclusions. Incidentally, I am puzzled about the figure of P = 0.33 quoted for the difference between the proportions using Fisher’s exact test. I make it P = 0.79.
Sample size 7 August 2001
Joyce Cramer, Associate Research Scientist, Dept. Psychiatry Yale University School of Medicine, New Haven, CT USA Send response to journal: Re: Sample size	Nuesch et al report on a study of patients with treatment-resistant hypertension, finding that few were poorly compliant with medication. The report makes a contribution to the field of compliance research by the justification that giving patients an electronic monitor for their medication does not affect compliance. However, they note without explanation that classification of 22 patients was changed after the second ambulatory blood pressure assessment. My main concern with the report is the comment about a sample size calculation based on their finding of 3% difference in compliance rates. What they should have described in the methods section was the basis for the sample size of 103 subjects with expectations of differences. Their search of the literature was inadequate because they did not cite a paper by their neighbors in Lausanne that found significantly improved blood pressure when poor compliance was identified (Burnier et al J Hypertension, 2001; 19: 335-341).