How does hypertension cause stroke: usual BP, variability in BP or both? PM Rothwell Professor of Clinical Neurology University of Oxford BHS; September 2011 CVS Reactivity Ultrasound: Carotid + TCD Cold Pressor Test Mental Arithmetic Pulse Wave Analysis Analysis Neurovascular Physiology Aortic BP Aortic stiffness Peripheral stiffness IMT MCA stiffness Cerebral Autoregulation Beat-to-Beat BP Postural BP Baroreceptor Gain Induced Hypotension Measures of variability / lability 250 Statistics • SD • CV • VIM • ASV • RSD • Peak size • Trough size 200 150 100 0 1 2 3 4 5 Variability period Minutes (e.g. within-visit) Hours (e.g. ABPM) Days (e.g. home monitoring) Weeks (e.g. visit-to-visit) Variability in clinic or home BP • Is small compared with variance in “true” mean BP • Is “random” and “noise” • Anyway: – usual BP already explains all variance in risk – benefit of antihypertensive drugs is already fully explained by effects on mean BP mean “Clinic readings are a surrogate marker for a patient’s true BP (the average level over prolonged periods of time), which is thought to be the most important component of BP in determining its adverse effects.” AHA Guideline 2005 “In the absence of markedly raised BP, repeat readings should be obtained over several months to define the patient’s usual BP as accurately as possible”. Joint European Guideline 2007 Patient A 175 Systolic Diastolic 150 125 100 75 01JAN1998 01JAN2000 0101JAN- JUN2002 2002 15102626SEP- OCT- NOV- APR2003 2003 2003 2004 07JAN2005 112316FEB- MAY- DEC2005 2005 2005 1905JUL- OCT2006 2007 200 150 100 50 100 150 200 250 250 200 150 100 50 300 50 120 100 80 60 40 40 60 80 100 120 140 Baseline DBP (mm Hg) Howard SC, Rothwell PM. Second m easurem ent of DBP (m m Hg) 140 100 150 200 250 Dutch TIA 300 250 200 150 100 50 300 50 160 140 120 100 80 60 40 160 40 60 80 100 120 Baseline DBP (mm Hg) J Clin Epidemiol 2003; 56: 1084-91 & 140 100 150 200 250 300 Baseline SBP (m m Hg) Baseline SBP (m m Hg) Baseline SBP (m m Hg) 160 Second m easurem ent (m 250 UK-TIA 300 Second m easurem ent of DBP (m m Hg) Second measurement (m Second measurement of SBP Hg) SBP (mm (mm Hg) ECST 300 50 Second m easurem ent of DBP (mm Hg) DBP (m m Hg) Second measurement of Variation in BP – baseline vs first follow-up 160 140 120 100 160 80 60 40 40 60 80 100 120 Baseline DBP (mm Hg) Stroke 2006; 37: 2776-83 140 160 Variability in clinic or home BP • Is small compared with variance in “true” mean BP • Is “random” and “noise” • Anyway: – usual BP already explains all variance in risk – benefit of antihypertensive drugs is already fully explained by effects on mean BP Took over chair Lecture 2 200 Lecture 1 SBP (mmHg) Lecture 3 150 100 Gym Break Board meeting fast Lunch 3 Lectures Time of day Formal Walk meeting Walk + dinner Sleep Comparison of visit-to-visit variability in SBP with other measures in ASCOT Correlation • Within-visit variability – r=0.21, p<0.0001 • Daytime variability on ABPM – r=0.35, p<0.0001 • Morning surge on ABPM – r=0.15, p=0.008 Lancet 2010; 375: 895-905 % Predictive value 10% 50% 20% Comparison of day-to-day and week-to-week variability Weeks 1 - 30 180 180 160 160 140 120 100 140 120 100 80 80 60 60 0 1 2 3 4 5 6 25 0 7 5 10 180 180 160 160 BP (mm Hg) BP (mm Hg) 20 25 30 Week Day B 15 140 120 100 0 1 2 3 4 Day Unpublished data 5 6 7 15 10 5 0 5 10 15 120 Within-w eek SD SBP 100 60 60 20 0 140 80 80 Week-to-week SD SBP BP (mm Hg) A BP (mm Hg) Week 1 0 5 10 15 Week 20 25 30 20 25 Variability in clinic or home BP • Is small compared with variance in “true” mean BP • Is “random” and “noise” • Anyway: – usual BP already explains all variance in risk – benefit of antihypertensive drugs is already fully explained by effects on mean BP Relevant epidemiology of stroke • • • • • • • • • Age Acute hypertension Diurnal variation Triggers Personality Sex Race Renal failure, diabetes Vascular dementia Lancet 2010; 375: 938-48 Diurnal pattern of stroke incidence in Oxfordshire, UK Regression Dilution Bias Prospective Studies Collaboration Count “Usual” value Measured value Value Risk “Usual” value Measured value Lancet 2002; 360: 1903-13 Value Problems in interpretation of adjustment for RDB 1. The more variable BP - the greater is the degree of adjustment BUT - the more variable BP is the less credible it becomes to argue that the underlying usual BP is likely to be pathologically relevant. 2. Adjustment is driven by patients with variable BP and assumes that the predictive value of usual BP is independent of variability BUT - this assumption has never been tested – and is false 3. Assumes that all of the prognostic value of a single baseline BP reading is attributable to usual BP – variability being “random” BUT - variability is of prognostic value and reproducible (not random); AND - a single high BP reading is more related to variability than to mean BP. Lancet 2010; 375: 938-48 Relative contribution of mean vs SD Relative strength of association of mean vs SD SBP (7 clinic visits) with baseline systolic BP in the UK-TIA Trial 100 90 80 70 60 Mean 50 SD 40 30 20 10 0 1 2 4 5 Quintile of baseline SBP Contributions from the Wald statistics from multinomial logistic regression (outcome = probability of being in a particular quintile, relative to quintile 3). Lancet 2010; 375: 938-48 Variability in BP versus stroke risk in UK-TIA Trial Excluding cases with prior stroke or CT infarction Hazard ratio (95% CI) 12 10 8 6 Adjusted for mean SBP 4 2 0 1 2 3 4 5 6 7 8 9 10 Decile of SD SBP Lancet 2010; 375: 895-905 Cuffe RL, Rothwell PM. Medium-term variability in systolic blood pressure is an independent predictor of stroke. Cerebrovasc Dis 2005; 19 (suppl 2): 51. 1.44 (0.58 – 3.57) 10 readings 13.04 (1.66 – 102.6) 0.43 0.015 Predictive value of residual visit-to-visit variability in SBP in ASCOT-BPLA HR for top quintile of mean SBP HR for VIM SBP 5 2.82 (1.67 – 4.76) 4 4 readings 3 6 <0.0001 HR for SD SBP 2.67 (1.74 – 4.11) 6 2 readings 3.07 (1.62 – 5.83) 0.001 1 8 readings 2.68 (1.29 – 5.56) 0.008 66 7 10 7 88 99 10 2.26 (0.98 – 5.17) Decile of measure 0.055 0 10 readings0 11 22 33 44 55 1.09 (0.73 – 1.62) 0.67 1.50 (0.90 – 2.48) 0.12 1.98 (1.05 – 3.77) 0.036 5.00 (1.75 – 14.30) 0.003 5 4 <0.0001 6 readings 2 HR for top quintile of CV SBP 3 2 1 0 0 1 1 22 33 4 4 5 5 6 6 7 7 8 8 9 910 10 HR for top quintile of mean SBP 13.05 (1.74 – 97.66) Decile of measure 0.012 HR for top quintile of VIM SBP 2 readings 2.86 (1.88 – 4.36) <0.0001 1.25 (0.86 – 1.82) 0.25 4 readings 3.18 (1.90 – 5.33) <0.0001 1.59 (1.00 – 2.54) 0.053 6 readings 3.70 (1.97 – 6.94) <0.0001 2.31 (1.26 – 4.23) 0.007 8 readings 3.70 (1.81 – 7.56) <0.0001 6.04 (2.14 – 17.03) 0.001 10 readings 3.31 (1.46 – 7.47) 0.004 15.35 (2.08 – 113.1) 0.007 Lancet 2010; 375: 895-905 SYST-EUR: Stroke (fatal or non-fatal) (n = 197) 12 Hazard ratio (95% CI) Ha z a rd ratio (9 5% C I) 7 6 5 4 3 2 1 0 10 8 6 4 2 0 1 2 3 4 5 6 7 8 D e c il e of m e a n S B P Unpublished data 9 10 1 2 3 4 5 6 7 8 Decile of CV SBP 9 10 Hazard ratios (95% CI) for the risk of subsequent stroke in relation to maximum SBP and minimum SBP in the UK-TIA trial Maximum SBP exceeds mean SBP by: 7 measurements 10 measurements 0 - 9% 10 - 19% 20 – 29% 30 - 39% ≥40% 1 1.21 (0.69 – 2.14) 1.84 (1.01 – 3.35) 3.24 (1.54 – 6.78) 6.21 (2.29 – 16.9) 1 2.24 (0.67 – 7.46) 4.10 (1.22 – 13.8) 6.16 (1.70 – 22.4) 9.31 (2.07 – 41.8) 1 0.98 (0.43-2.25) 0.90 (0.39-2.06) 1.72 (0.75-3.95) 1.64 (0.66-4.07) 1 1.19 (0.40-3.52) 1.02 (0.35-2.95) 2.17 (0.76-6.17) 2.59 (0.91-7.36) Minimum SBP fell below mean SBP by: 0-4% 5-9% 10-14% 15-19% ≥20% Lancet 2010; 375: 895-905 C e re b ra l b lo o d flo w A rte rio la r c a lib re r C e re b ra l b lo o d flo w P e rfusio n pressure A rte rio la r c a lib re r: SBP (mm Hg) 170 150 130 110 P e rfusio n pressure Time Blood pressure parameters and their predictive values (hazard ratios and 95% CI for risk of events relative to the stable normotension group) in the four groups of patients from the UK-TIA Trial cohort based on the patterns of their blood pressure over the first seven clinic visits. Patients with mean usual SBP ≥180mmHg are excluded Stable1 Episodic moderate normotension hypertension2 Cases 241 Mean (SD) of measures 1-7 of SBP Individual mean 123.5 (7.6) Individual SD 8.5 (3.3) Individual CV 6.9 (2.8) Individual VIM 11.5 (4.9) Individual maximum 134.3 (7.6) Individual minimum 112.0 (9.9) 601 141.1 (8.6) 12.8 (4.1) 9.1 (3.1) 13.9 (5.0) 159.4 (8.5) 124.2 (10.6) Episodic severe hypertension3 263 Stable hypertension4 Heterogeneity 154 157.9 (8.7) 21.4 (5.7) 13.6 (3.6) 19.1 (5.2) 190.2 (12.3) 130.2 (9.5) 167.3 (7.2) 13.4 (5.8) 7.9 (3.2) 10.6 (4.1) 187.8 (16.2) 151.7 (4.9) Events after 7th measure of SBP Stroke 9 (3.7%) HR (95% CI) 1.00 37 (6.2%) 1.68 (0.81-3.47) 36 (13.7%) 4.18 (2.01-8.68) 7 (4.5%) 1.22 (0.45-3.27) p=0.00003 Coronary event HR (95% CI) 22 (9.1%) 1.00 66 (11.0%) 1.40 (0.86-2.28) 39 (14.8%) 1.72 (1.02-2.91) 13 (8.4%) 0.84 (0.42-1.70) p=0.12 All vascular events 31 (12.9%) 103 (17.1%) 75 (28.5%) 20 (13.0%) 1 2 3 4 p=0.000007 All values ≤140mmHg At least one BP ≤140mmHg, at least one >140mmHg, but all <180mmHg At least one BP ≤140mmHg and at least one BP ≥180mmHg. All BPs >140mmHg Lancet 2010; 375: 895-905 Postural instability (lying – standing) vs visit-to-visit SD SBP (sitting) 40 10 Average postural change 35 Stroke risk P<0.001 9 30 8 25 7 20 6 Stroke risk (%) 15 5 4 3 10 5 0 0 10 20 30 visit-to-visit SD SBP Unpublished data 40 2 1 0 3 2 1 1 2 V-to-V SD Tertile of visit-to-visit 3 PCTe Variability in clinic or home BP • Is small compared with variance in “true” mean BP • Is “random” and “noise” • Anyway: – usual BP already explains all variance in risk – benefit of antihypertensive drugs is already fully explained by effects on mean BP The relationships between change in SBP and the effect on vascular risk in RCTs of BP lowering JA Staessen et al. Lancet 2001; 358: 1305-15 Inter-individual SD SBP (mmHg) Group SD SBP in the MRC Trial in elderly hypertensive patients 20 19 18 17 Placebo 16 beta-blocker diuretic 15 0 5 10 15 Months Lancet Neurol 2010; 9: 469-80 20 25 Mean (SD) SBP (mmHg) at baseline and during follow-up stratified by randomised treatment in the ALLHAT trial Follow-up visit Treatment group Significance (p) of difference in SD Amlodipine (A) Chlorthalidone (C) Lisinopril (L) A vs L C vs L 0.5 Baseline 146.2 (15.7) 146.2 (15.7) 146.4 (15.7) 0.5 1 year 138.5 (14.9) 136.9 (15.8) 140.0 (18.5) 9 x 10 2 years 137.1 (15.0) 135.9 (15.9) 138.4 (17.9) 3 x 10 3 years 135.6 (15.2) 134.8 (15.4) 136.7 (17.3) 9 x 10 4 years 134.8 (15.0) 133.9 (15.7) 135.5 (17.2) 1 x 10 5 years 134.7 (14.9) 133.9 (15.2) 135.9 (17.9) 1 x 10 Lancet 2010; 375: 938-48 -79 7 x 10 -55 -48 1 x 10 -25 2 x 10 -24 2 x 10 -24 8 x 10 -28 -25 -14 -25 Distributions of SBP at the one year follow-up visit in ASCOTBPLA stratified by randomised treatment group Am lo d ip in e g ro u p Ate n o lo l g ro u p 3000 N u m b e r o f s u b je c ts 2500 S D = 1 5 .8 S D = 1 9 .2 2000 1500 1000 500 0 270 250 230 S B P (m m H g ) 210 190 170 150 130 110 90 70 50 270 250 230 Lancet Neurol 2010; 9: 469-80 210 190 170 150 130 110 90 70 50 S B P (m m H g ) All large RCTs of CCBs vs beta-blockers or ACE-inhibitors in which the mean (SD) SBP during follow-up has been reported by treatment group Stroke risk Events/Patients CCB Drug B Odds Ratio 159 / 5410 196 / 5471 0.81 Mean SBP 95% CI 95% CI 0.66-1.01 3.70 3.07, 4.33 -2.36, -1.44 Difference Trial NORDIL (vs BB/D)98 ASCOT (vs BB)64 327 / 9639 422 / 9618 0.77 0.66-0.89 -1.90 VALUE (vs ARB)99 281 / 7596 322 / 7649 0.87 0.74-1.03 -1.80 -4.92, 1.32 176 / 11267 201 / 11309 0.88 0.72-1.08 0.00 -0.27, 0.27 377 / 9048 457 / 9054 0.82 0.71-0.94 -1.30 -1.78, -0.82 -1.20, 2.40 -0.41, -0.01 INVEST (vs BB)97 ALLHAT (vs ACE)95 CAMELOT (vs ACE)63 TOTAL 6 / 663 8 / 673 0.76 0.26-2.20 0.60 1326 / 43623 1606 / 43774 0.82 0.76-0.88 -0.21 0.5 Odds Ratio (95%CI) 1.5 SBP at follow-up Mean (sd) CCB Drug B Variance Ratio 155.2 (16.3) 151.5 (17.4) 0.88 Mean SBP 95% CI 95% CI 0.83-0.93 3.70 3.07, 4.33 -2.36, -1.44 Difference Trial NORDIL (vs BB/D)98 138.4 (14.8) 140.3 (17.8) 0.69 0.67-0.72 -1.90 138.2 (13.8) 140.0 (16.2) 0.73 0.68-0.77 -1.80 -4.92, 1.32 INVEST (vs BB)97 131.0 (11.0) 131.0 (13.0) 0.83 0.80-0.86 0.00 -0.27, 0.27 ALLHAT (vs ACE)95 137.1 (15.0) 138.4 (17.9) 0.70 0.67-0.73 -1.30 -1.78, -0.82 0.74 0.64-0.86 0.60 -1.20, 2.40 0.76 0.74-0.77 -0.21 -0.41, -0.01 ASCOT (vs BB)64 VALUE (vs ARB)99 CAMELOT (vs ACE)63 TOTAL Lancet 2010; 375: 938-48 124.2 (15.5) 123.6 (18.0) 0.5 Variance Ratio (95%CI) 1.5 Effect of treatment of group versus individual variability in SBP Blood Pressure Lowering Trialists’ Collaboration G-VR 2.0 2.0 G-VR R2 = 0.87, p<0.0001 I-VR 0.5 2.0 0.5 Unpublished data 2.0 0.5 0.5 4.5 Atenolol Amlodipine Effect of treatment on group SD SBP in ASCOT-BPLA 4.1 3.9 3.7 Within-visit individual SD 3.5 Atenolol 22 Amlodipine 20 Follow-up 18 16 14 Ba se l in e 3 m on th s 1 ye ar 2 ye ar 3 ye ar 4 ye ar 5 ye ar 6 ye ar 12 Follow -up Lancet Neurol 2010; 9: 469-80 Average intra-ABPM daytime SD SBP Inter-individual SD SBP 24 Group SD Average within-visit CV SBP 4.3 ABPM daytime SD 12.5 12 11.5 11 Atenolol 10.5 Amlodipine 10 0.5 1.5yr 1.5 2.5yr 2.5 3.5yr 3.5 4.5yr 4.5 5.5yr Tim e from random isation > 5.5yr No. of patients Effect of treatment allocation on withinindividual variability in SBP in ASCOT 1400 Atenolol 1200 Amlodipine 1000 800 600 400 200 0 1 Lancet 2010; 375: 895-905 2 3 4 5 6 7 Decile of CV SBP 8 9 10 Effect of treatment on risk of stroke in ASCOT-BPLA Adjustment for mean BP and variability in BP Model SBP HR (95% CI) p Treatment 0.78 (0.67 – 0.90) 0.001 Treatment + mean 0.84 (0.72 – 0.98) 0.025 Treatment + SD 0.94 (0.81 – 1.10) 0.47 Treatment + CV 0.94 (0.79 – 1.07) 0.27 Treatment + VIM 0.93 (0.77 – 1.04) 0.16 Treatment + mean + SD 0.98 (0.82 – 1.12) 0.59 Treatment + mean + CV 0.97 (0.82 – 1.11) 0.55 Treatment + mean + VIM 0.98 (0.82 – 1.12) 0.58 Lancet Neurol 2010; 9: 469-80 Systematic review of all RCTs of BP-lowering drugs 1372 Eligible Reports 910 phase 2 trials excluded due to inadequate reporting 94 Phase 3 Trials 377 Phase 2 Trials with follow-up SD (>100 patients per group for >1 year) (9 reports included 2 trials) + 21 phase 3 trials 93 trials of 578456 patients Meta-analyses of OR between drugs Lancet 2010; 375: 906-15 28 comparisons in 188564 patients 685 drug groups of 157505 patients Metaregressions of VR vs OR Change in VR or %CV 361 comparisons in 150663 patients VR or %CV on different agents Pooled estimates of variance ratio between drug classes in crossover trials Lancet 2010; 375: 906-15 220 200 B lo o d p re s su re (m m H g ) OXVASC Bluetooth Home BP monitoring F e lo d ip in e 2 .5 m g F e lo d ip in e sto p p e d F e lo d ip in e 5 m g 180 160 140 120 100 80 60 40 1 2 3 4 5 6 7 8 9 10 11 12 W eeks A m lodipine 5m g 220 A m lodipine 10m g B lo o d p re s s u re (m m H g ) 200 COMMIT Study 180 160 - 310 patients 140 120 - 1284 drug changes 100 80 - 92% uptake 60 40 1 2 3 4 5 W eeks 6 7 8 Lancet 2010; 375: 938-48 Pooled estimates of VR in parallel group trials compared with drug-class effects on stroke risk Lancet 2010; 375: 906-15 Meta-regressions relating effect of treatment on group variability in SBP to effect of treatment on risk of stroke 1.4 1.4 1.2 ALLHAT II ALLHAT Adjusted for difference in mean SBP HOT<85 1.2 UKPDS 39 MOSES 11 OR (EXP/REF) 1 10 ALLHAT II IPPPSH 1 HOT<80 HAPPHY LIFE 0.8 INVEST NORDIL SCOPE CAMELOT NICS PATS ECOST STOP I CAMELOT 23 0.8 ASCOT 11 3 22 6 12 21 13 4 18 20 EWPHE 14 1 15 7 9 FEVER SHEP 0.6 23 1 25 2 5 24 VALUE 16 0.6 19 17 UKPDS 38 8 0.4 0.4 USPHS 0.2 SCAT 0.2 PREVEND IT 0 0 0.6 0.8 1 1.2 1.4 1.6 0.6 Variance Ratio (REF/EXP) Lancet 2010; 375: 906-15 0.8 1 1.2 1.4 1.6 Pooled estimates of VR in parallel group trials in BPLTC: Intra-individual versus inter-individual variability Unpublished data Drug vs control Drug A vs Drug B OR (all stroke) A) B) 1.5 1.5 1.0 1.0 0.5 -4 0.5 0 16 1.5 OR (all stroke) Reduction in SBP (mmHg) 1.5 1.0 1.0 Unpublished data 0.5 Reduction in SBP (mmHg) 1.5 0.5 -4 1.0 0.5 0 16 Change in I- VR Reduction in SBP (mmHg) 1.5 1.0 0.5 Change in I-VR Variability in clinic or home BP • Is small compared with variance in “true” mean BP • Is “random” and “noise” • Anyway: – usual BP already explains all variance in risk – benefit of antihypertensive drugs is already fully explained by effects on mean BP Implications • Diagnosis of “hypertension” • Decision to treat • • • • • • Risk prediction Choice of agent Combinations of agents Monitoring of BP on Rx Development of new agents Aetiology of stroke? 200 Systolic BP (mmHg) – Not currently treating the right patients 250 150 100 250 200 150 100 0 1 2 3 Months 4 5