Home hypertension : exploring the inverse white coat response

Home hypertension : exploring the inverse white coat response NORBERT DONNER-BANZHOFF YOLANDA CHAN JOHN PAUL SZALAI JOHN HILDITCH SUMMARY Background. The classical white coat response to blood pressure measurement has been studied thoroughly. However, little is known about patients showing a reverse pattern, i.e. who have lower blood pressure readings at the clinic than outside healthcare facilities. Aim. To estimate the proportion of patients whose blood pressure levels as determined by self-measurements at home are higher than those taken at the clinic and to explore possible associations with demographic, clinical, and psychological variables. Method. Patients consecutively attending (n = 214) an academic family medicine department in Toronto, Canada, were eligible. Subjects aged below 16 years and those on psychotropic or blood pressure-lowering agents were excluded. The clinic home blood pressure difference (CHBPD) was calculated for each participating subject by subtracting home blood pressure from clinic blood pressure. Those who had negative values were compared with the rest of the sample. Results. A considerable proportion of patients had lower blood pressure at the clinic than at home (systolic, 34.6%; diastolic, 23.8%). These subjects did not differ from the rest of the sample with regard to age, sex, levels of education attained, immigration status, body mass index, experience of current symptoms, blood pressure levels, or psychological distress. However, in patients with a negative CHBPD, i.e. lower blood pressure at the clinic than at home, readings taken by an automatic, self-inflating device when still at the clinic were higher than in the rest of the sample. Conclusion. The results point to measurement bias being at least partly responsible for higher blood pressure readings outside the clinic. Automatic measurement devices used for self/home blood pressure measurement seem to cause an alerting reaction analogous to the well-described white coat response. N Donner-Banzhoff, MD, MHSc, family practitioner and senior lecturer, Department of Family and Community Medicine, Sunnybrook Health Science Centre, University of Toronto, North York, Canada, and Departments of Family Medicine and Biostatistics and Epidemiology, Centre for Health Science and Methodology, University of Marburg, Germany. Y Chan, RN, research assistant, Department of Family and Community Medicine, Sunnybrook Health Science Centre, University of Toronto, North York. J P Szalai, PhD, director and associate professor, Research Design and Biostatistics and Clinical Epidemiology and Health Care Research Program, Sunnybrook Health Science Centre, University of Toronto, North York, Canada, and Department of Preventive Medicine and Biostatistics, University of Toronto, Canada. J Hilditch, MD, MPH, CCFP, family practitioner and associate professor, Department of Family and Community Medicine, Sunnybrook Health Science Centre, University of Toronto, North York, Canada. Submitted: 17 June 1997; accepted: 10 March 1998. British Journal of General Practice, 1998, 48, 1491-1495. Keywords: blood pressure measurement; blood pressure selfmeasurement; white coat response; primary care; bias. Introduction HE distinction between the so-called white-coat syndrome Tand true or permanent hypertension is now well established in research and in practice. 1,2 This has been made possible by recent advances in blood pressure measurement technology. Easy-to-operate automatic devices are widely available, and most patients can perform their own measurements without healthcare professionals being present. Continuous ambulatory blood pressure monitoring (ABPM) is an alternative that is, however, not always available outside specialist care. In the majority of patients, self-measured or ABPM readings are lower than those taken in clinic. However, a substantial proportion of patients also show a reverse pattern, i.e. they have lower blood pressure levels at the clinic than outside. 3 Little is known about this group, who show an inverse white coat response, as previous research has concentrated on patients with the classical white coat pattern and their characteristics compared with established hypertensive subjects. We studied possible associations between psychological distress and blood pressure in a sample of practice attenders not on blood pressure-lowering drugs. Here, we report on an analysis that addresses the following questions. What proportion of subjects have higher readings at home than in clinic? Do these subjects differ with regard to demographic, clinical, or psychological characteristics? To obtain a continuous variable describing patients reactions to different measurement modes, we calculated the clinic home blood pressure difference (CHBPD) by subtracting home from clinic blood pressure levels. Patients with a negative CHBPD, i.e. higher readings at home, were compared with the rest of the sample. Method Details of the methods used have been reported elsewhere. 4,5 In brief, patients attending the Department of Family and Community Medicine at Sunnybrook Health Science Centre (North York, Ontario, Canada) for consultation were recruited on a consecutive basis. Subjects were excluded if they were less than 16 years old or on blood pressure-lowering or psychotropic medication. Before consulting their doctor, eligible patients filled in a questionnaire on demographic and clinical data as well as psychological distress. Clinic blood pressure measurements were taken by the research assistant, in a quiet private room, usually after consultation with a general practitioner (GP). After five minutes rest in a sitting position, patients blood pressure was measured on the left arm by a desk-model standard mercury device with an appropriately sized cuff (Baumanometer; WA Baum, Copiague, NY, USA). Two more measurements were made after five and 10 minutes. The mean of the last two readings was used for analysis (clinic blood pressure). The research assistant then instructed patients how to carry out self-measurements using a self-inflating oscillometric device (Sunbeam 7650 6 ). After they were familiar with the technique, patients per- British Journal of General Practice, August 1998 1491

formed one measurement by themselves while still in the clinic. Participating patients were required to take two measurements daily, in the morning and at bedtime, for seven consecutive days and record them in a diary. Before data analysis, we decided that patients had to carry out at least 11 measurements to be included in the analysis. Subjects thus had sufficient time to become familiar with the self-measurement procedure. One reading from day 4 could be substituted for one missing value. Patients were excluded if more than one reading was missing from days 5 7, or if they indicated that they did not follow the protocol, e.g. because of travel or shift work. The CHBPD was calculated by subtracting the mean of the last six home readings from the mean of the second and third clinic reading, for each patient, separately for systolic and diastolic blood pressure. After calculating the CHBPD for each subject, we dichotomized the sample according to the sign of this variable. Those whose CHBPD was negative, i.e. who had higher readings at home than in clinic, were compared with those who had a positive response. These comparisons were carried out with regard to a number of basic demographic variables and those that have been suggested to be related to blood pressure or blood pressure reactions by previous research, such as age, sex, education (levels according to Canadian schooling system collapsed into three categories), immigration status (born in Canada versus not born in Canada), experience of current symptoms (this question was adapted from the quantitative dimension of the McGill Pain Questionnaire but extended to any symptom the patient was suffering from, not just pain, on a numerical scale from 0 to 6, 7 dichotomized for analysis with 0 = none or slight, 1 = moderate to severe ), blood pressure levels (grand mean of clinic and home systolic/diastolic blood pressure), body mass index (BMI; kg/m 2 ), and psychological distress (as measured by the General Health Questionnaire, GHQ 8 ). We also compared the two groups with regard to the subdomains of the GHQ, i.e. anxiety, feelings of incompetence, depression, difficulty in coping, and social dysfunction, which have been identified by previous research. 9 Chisquared, t- and Wilcoxon rank sum tests were used depending on the scale and distribution of the variables examined. Adjustments were made for multiple testing. 10 To assess concordance between the dichotomous systolic and diastolic CHBPD, the kappa statistic was calculated. 11 Calculations were made using the SAS statistical package. 12 As mentioned above, the first three readings obtained in the clinic were taken by the standard mercury instrument. For each individual we calculated an expected blood pressure taken at the fourth reading. This was derived from a regression equation calculated for each individual based on the first three readings. Blood pressure was regressed on time. The expected blood pressure was subtracted from each subject s actual fourth reading taken by the automatic device. Intuitively, this provided a measure of an individual s blood pressure reactivity to the automatic device. Like the other variables mentioned above, the difference in the observed minus expected fourth reading was analysed by CHBPD positive/negative. To preserve the continuous nature of the CHBPD variable and to adjust for multiple confounders, we also performed multiple linear regression analyses with the CHBPD as the dependent variable. We entered several independent variables (see Table 4) and the difference in the observed minus expected fourth reading. Whether the latter made a significant contribution to the model was evaluated by an incremental F-test. 13 These comparisons were performed separately for systolic and diastolic blood pressure. Results Patient characteristics Of 325 eligible patients approached, 251 (75%) agreed to take part. Participants and non-participants did not differ significantly with regard to age (mean 48.8 years versus 49.3 years, t = 0.17, P = 0.86) or to sex (%female: 52 versus 61, χ 2 = 2.1, P = 0.15). Thirty-seven participants were subsequently excluded because of psychotropic medication use, missing data for home measurement, or irregular lifestyle during the blood pressure self-measuring period. Table 1 gives the demographic and somatic characteristics of the 214 subjects included in the final analyses. Distribution of the clinic home blood pressure difference As can be seen from Figure 1, a considerable proportion of the sample had higher readings at home than in the clinic, i.e. one third for systolic and nearly one quarter for diastolic pressure. We then examined whether subjects were concordant with their systolic and diastolic CHBPD. To achieve this, we categorized subjects according to whether they had a positive or a negative CHBPD. As Table 2 shows, agreement was moderate with kappa = 0.498 (P < 0.001). Only two patients had systolic home blood pressures above 140 with their clinic levels below that threshold. Another patient produced diastolic home readings of more than 90, while clinic diastolic levels were less than this. Positive versus negative CHBPD Table 1. Patient characteristics. Sex (% female) 51.2 Age: mean (SD) 48.7 (16) Born in Canada (%) 71.2 Level of education attained (%) Not graduated from high school 11.2 High school 9.8 College/trade school 29.8 Undergraduate degree 32 Postgraduate degree 17.2 Current smokers (%) 16 Body mass index: mean (SD) 25.5 (4.6) Current symptom bothering (%) None or mild 72 Discomfort 19 Distressing/horrible/excruciating 9 25 20 15 10 5 0-10 clinic BP < home BP Frequency distribution 0 10 20 clinic BP > home BP mean = 5 std deviation = 11.3 n = 214 Figure 1. (a) The systolic clinic home blood pressure difference. 1492 British Journal of General Practice, August 1998

25 20 Frequency distribution mean = 5.6 std deviation = 8.4 n = 214 subjects with negative clinic home BP difference (means +/- standard errors) all other subjects (means +/- standard errors) 15 10 5 0-10 0 10 20 135 expected 4th reading mercury device automatic device Clinic (systolic - means) Home clinic BP < home BP clinic BP > home BP 125 Figure 1. (b) The diastolic clinic home blood pressure difference. Table 2. Concordance of systolic and diastolic CHBPD. 115 Diastolic CHBPD Positive Negative n (total %) n (total %) M1 M2 M3 A1 A2 A3 A4 A5 A6 A7 A8 Systolic Positive 129 (60) 11 (5) 140 CHBPD Negative 34 (16) 40 (19) 74 163 51 214 κ = 0.498, SE = 0.066, P < 0.001. We compared the positive and the negative CHBPD groups with regard to demographic, clinical, and psychological variables. After adjustment for multiple testing, no significant differences could be shown with regard to age, sex, education, immigration status, experience of current symptoms, BMI, blood pressure levels, or GHQ score (total and subdomains) (details of bivariate analyses can be obtained from first author). Figure 2 shows systolic blood pressure levels over time by positive / negative CHBPD group. As can be seen from the graph, subjects with a negative CHBPD have higher readings than expected on the fourth reading in the clinic. Correspondingly, positive and negative responders differ significantly in their observed minus expected blood pressure levels from the fourth reading (Table 3). Note that, in both groups, home blood pressure levels remain essentially the same over the course of the measurement period, i.e. no accommodation effect is apparent, such as a gradual drop in blood pressure. To avoid a potentially arbitrary dichotomy of negative versus positive CHBPD, we also performed linear regression analyses. We added to the set of potentially confounding variables (Table 4) the difference in the observed minus expected fourth reading. This variable, measuring the reactivity to the automatic device, turned out to be a strong predictor of the CHBPD. It explained an additional 5% of the variance in the systolic CHBPD and an additional 16% of the variance in the diastolic CHBPD (see Table 4). Discussion Among family practice attenders, one third exhibited a systolic inverse CHBPD, and one quarter a diastolic inverse CHBPD. An exploratory analysis of these subjects did not show them to differ from the rest of the sample with regard to important demographic, clinical, or psychological variables. However, subjects exhibiting an inverse CHBPD differed considerably with Figure 2. Blood pressure in the clinic and at home. regard to their blood pressure reactions to measurement by an automatic machine while still in the clinic. Our sample consisted of patients attending a large family medicine group practice. We can, therefore, assume that its composition has been influenced by morbidity and attending patterns. However, to generalize our results to other populations may be permitted, as patients were not selected with regard to specific diagnoses or findings. Still, we have to admit that studies replicating the results in other settings are clearly desirable. The reader should also keep in mind that subjects on blood pressure-lowering medication were excluded. Among non-participants, the proportion of women was higher than among participants, although this did not reach the conventional level of statistical significance. Although, to our knowledge, the negative CHBPD or clinic ambulatory blood pressure difference has not been investigated explicitly, the proportion of patients exhibiting this phenomenon can be calculated from published reports on the distribution of the white coat response and related measures. 3,14 24 Despite considerable variations between studies, the size of this group seems to be far from negligible (a detailed survey of the literature can be obtained from the first author). How can the findings from our exploratory study be interpreted? A possible explanation would be that, using an automatic self-inflating device, some patients exhibit a specific alerting reaction, resulting in a blood pressure rise. Several authors have described phenomena suggesting an alerting response of this kind. Prasad et al 25 performed ambulatory blood pressure monitoring for 48 instead of the usual 24 hours. They found raised systolic and diastolic readings during the first two hours after monitors had been applied. These blood pressure elevations did not occur on the second day that their subjects were wearing their monitors. The results from a study of the disruptive effect of portable monitors on sleep point in a similar direction. 26 Very few studies evaluating the performance of automatic blood pressure measuring devices used sequential designs. The far more common simultaneous comparison method 27 cannot British Journal of General Practice, August 1998 1493

Table 3. Difference observed minus expected fourth reading in clinic by CHBPD category. Variable Positive Negative CHBPD CHBPD Median Median Z P-value Systolic Observed minus expected fourth reading 2 2.5 3.56 0.0004 Diastolic Observed minus expected fourth reading 3.33 1.66 3.98 0.0001 Test statistics refer to Wilcoxon rank sum tests, as the variables analysed were not normally distributed. Table 4. Difference observed minus expected fourth reading in clinic multiple linear regression analysis. Systolic BP Diastolic BP Parameter Standard Parameter Standard Variable estimate error P a estimate error P a Intercept 0.84 7.13 0.9 10.3 4.87 0.04 Age 0.11 0.06 0.07 0.05 0.04 0.15 Sex (f = 0, m = 1) 3.17 1.65 0.06 1.67 1.1 0.13 Education b Dummy no. 1 0.54 1.74 0.8 0.58 1.17 0.62 Dummy no. 2 5.34 2.86 0.06 Born in Canada (yes = 0, no = 1) 1.18 1.77 0.51 0.25 1.22 0.84 BP levels c 0.03 0.07 0.7 0.23 0.07 0.001 BMI 0.16 0.19 0.41 0.2 0.13 0.14 Current symptom (no = 0, yes = 1) 1.15 1.79 0.52 1.02 1.23 0.41 GHQ score 0.05 0.14 0.7 0.08 0.09 0.89 Observed minus expected fourth reading 0.32 0.09 0.0007 0.46 0.07 0.0001 Variance explained/ R 2 with = 0.15 F = 12.0 0.0007 R 2 with = 0.26 F = 38.4 < 0.0001 incremental F-test d R 2 without = 0.1 R 2 without = 0.1 a Probability for H0: Parameter = 0. b Highest level, i.e. at least bachelor s degree, as reference. c Grand mean of clinic and home blood pressure. d To evaluate whether variable observed minus expected fourth reading makes a significant additional contribution to explain the variance of the dependent variable. demonstrate the postulated alerting response, as the standard would only reflect the rise in blood pressure caused by the automatic device. In a population-based study with children aged 10 to 13 years, sequential readings by automatic and mercury devices showed higher systolic blood pressure readings for the automatic machine. However, the pattern was reversed for diastolic blood pressure. 28 Raised systolic blood pressure levels were found in a study testing a coin-operated measuring device. 29 The authors discussed the apprehension experienced by the patient when using a device of this kind. In a sequential comparative study of an automatic and a mercury device, systolic and diastolic readings taken by the automatic device were higher than those taken using the standard device. 30 For diastolic pressure, this response was reduced by previous exposure to automatic blood pressure devices. In the study presented here, no association between the response to automatic blood pressure measurement and demographic, clinical, or psychological variables could be demonstrated. In this respect, there seems to be a parallel with the classical white coat response that seems to be idiosyncratic to the clinic setting rather than being a manifestation of generalized anxiety or increased reactivity. 31 The noise caused by the electric pump, together with the fact that a machine and not a human being is inflating the cuff, may result in discomfort and anxiety that cause a blood pressure reaction of the kind shown above. Devices that require patients to inflate the cuff themselves may not be a good alternative, as the physical effort can also bias the result in a subset of patients. We are confident that the blood pressure-raising effect observed in our study is independent of the type or brand of device used. For the purpose of this study, we chose a device that had been shown to have above average performance characteristics. 6 Previous research has demonstrated bias connected with clinic blood pressure measurement. Our study, however, points to a bias caused by devices used to obtain blood pressure readings outside the clinic. We conclude that no presently available measurement method can claim to produce unbiased estimates of human beings blood pressure, nor do we know what the longterm risk implications of our findings are. The study of blood pressure measurement in experimental and healthcare settings remains a challenge. References 1. Mancia G, Bertinieri G, Grassi G, et al. Effects of blood-pressure measurement by the doctor on patient s blood pressure and heart rate. Lancet 1983; ii: 695-698. 2. Mancia G, Parati G, Pomidossi G, et al. Alerting reaction and rise in blood pressure during measurement by physician and nurse. Hypertension 1987; 9: 209-215. 3. de Gaudemaris R, Chau NP, Maillion JM for the Groupe de la Mesure, French Society of Hypertension. Home blood pressure: variability, comparison with office readings and proposal for reference values. J Hypertension 1994; 12: 831-838. 4. Donner-Banzhoff N, Chan Y, Szalai JP, Hilditch J. Is the clinic home blood pressure difference associated with psychological distress? A primary care based study. J Hypertension 1997; 15: 585-590. 5. Donner-Banzhoff N, Chan Y, Szalai JP, Hilditch J. Low blood pressure associated with low mood a red herring? J Clin Epidemiol 1997; 50: 1175-1181. 6. Anonymous. Blood pressure monitors. Yonkers, NY: Consumers Union, May 1992, pp. 295-299. 7. Melzack R. The short form McGill Pain Questionnaire. Pain 1987; 30: 191-197. 8. Goldberg D. Manual of the General Health Questionnaire. Windsor, Berks: NFER-Nelson; 1978 (used with permission), coded according to Goodchild ME, Duncan-Jones P. Chronicity and the General Health Questionnaire. Br J Psychiat 1985; 146: 55-61. 1494 British Journal of General Practice, August 1998

9. Huppert FA, Walters DE, Day NE, Elliott BJ. The factor structure of the General Health Questionnaire (GHQ-30). A reliability study on 6317 community residents. Br J Psychiat 1989; 155: 178-185. 10. Bland JM, Altman DG. Multiple significance tests: the Bonferroni method. BMJ 1995; 310: 170. 11. Rosner B. Fundamentals of biostatistics. Belmont, CA: Duxbury Press, 1995. 12. SAS Institute Inc. SAS/STAT user s guide, release 6.03. Cary, NC: SAS Institute Inc., 1988. 13. Kleinbaum DG, Kupper LL. Applied regression analysis and other multivariable methods. Boston: Duxbury Press, 1978, p. 178. 14. Pearce KA, Grimm RH, Rao S, et al. Population-derived comparisons of blood pressure: variability, comparison with office readings and proposal for reference values. J Hypertension 1994; 12: 831-838. 15. Sokolow M, Werdegar D, Kain HK, Hinman AT. Relationship Between level of blood pressure measured casually and by portable recorders and severity of complications in essential hypertension. Circulation 1966; 34: 279-298. 16. Laughlin KD, Sherrard DJ, Fischer L. Comparison of clinic and home blood pressure levels in essential hypertension and variables associated with clinic home differences. J Chronic Dis 1980; 33: 197-206. 17. Floras JS, Jones JV, Hassan MO, et al. Cuff and ambulatory blood pressure in subjects with essential hypertension. Lancet 1981; ii: 107-109. 18. Pickering TG, Harshfield GA, Kleintert HD, Laragh JH. Ambulatory monitoring in the evaluation of blood pressure in patients with borderline hypertension and the role of the defense reflex. Clin Exp Hypertension 1982; A4: 675-693. 19. Pickering TG, Harshfield GA, Kleinert HD, et al. Blood pressure during normal daily activities, sleep and exercise. Comparisons of values in normal and hypertensive subjects. JAMA 1982; 247: 992-996. 20. Perloff D, Sokolow M, Cowan R. The prognostic value of ambulatory blood pressure. JAMA 1983; 249: 2792-2798. 21. Mejia AD, Egan BM, Schork NJ, Zweifler AJ. Artefacts in measurement of blood pressure and lack of target organ involvement in the assessment of patients with treatment-resistant hypertension. Ann Int Med 1990; 112: 270-277. 22. Gosse P, Promax H, Durandet P, Clementy J. White coat hypertension. No harm for the heart. Hypertension 1993; 22: 766-770. 23. Millar JA, Isles CG, Lever AF. Blood pressure, white-coat pressure responses and cardiovascular risk in placebo-group patients of the MRC Mild Hypertension trial. J Hypertension 1995; 13: 175-183. 24. Chatellier G, Battaglia C, Pagny JY, et al. Decision to treat mild hypertension after assessment by ambulatory monitoring and WHO recommendations. BMJ 1992; 305: 1062-1066. 25. Prasad N, MacFadyen RJ, Ogston SA, MacDonald TM. Elevated blood pressure during the first two hours of ambulatory blood pressure monitoring: a study comparing consecutive twenty-four hour monitoring periods. J Hypertension 1995; 13: 291-295. 26. Davies RJO, Jenkins NE, Stradling JR. Effect of measuring ambulatory blood pressure on sleep and on blood pressure during sleep. BMJ 1994; 308: 820-823. 27. O Brien E, Petrie J, Littler W, et al. The British Hypertension Society protocol for the evaluation of automated and semi-automated blood pressure measuring devices with special reference to ambulatory systems. J Hypertension 1990; 8: 607-619. 28. Weaver MG, Park MK, Lee DH. Differences in blood pressure levels obtained by auscultatory and oscillometric methods. Am J Dis Chest 1990; 144: 911-914. 29. Polk BF, Rosner B, Feudo R, Vandenburgh M. An evaluation of the Vita-Stat automatic blood pressure measuring device. Hypertension 1980; 2: 221-227. 30. Donner-Banzhoff N, Spangenberg E, Baum E. Automatic blood pressure measurement. Third European Congress on Family Medicine/General Practice (WONCA), Stockholm, Sweden, 30 June 4 July 1996. 31. Pickering TG. The ninth Sir George Pickering memorial lecture. Ambulatory monitoring and the definition of hypertension. J Hypertension 1992; 10: 401-409. Acknowledgements We should like to thank the members of the Department of Family and Community Medicine and Dr Sheldon Tobe, Sunnybrook Health Science Centre, North York, Ontario, Canada, for helpful feedback and suggestions. We should also like to thank Ms Barb van Maris, Ms Eleanor Boyle, and Ms Marian Mitchell for their support. Dr Denise Lucas, Bad Nauheim, Germany, corrected the manuscript. The study was funded by SAMCOR. Address for correspondence Dr Norbert Donner-Banzhoff, Allgemeinmedizin, Medizinisches Zentrum für Methodenwissenschaften und Gesundheitsforschung, Universität Marburg, Blitzweg 16, D-35033 Marburg, Germany. British Journal of General Practice, August 1998 1495