Effect of hypertension on anticoagulated patients with atrial fibrillation

Transcription

1 European Heart Journal (2007) 28, doi: /eurheartj/ehl504 Clinical research Arrhythmia/electrophysiology Effect of hypertension on anticoagulated patients with atrial fibrillation Gregory Y.H. Lip 1 *, Lars Frison 2, and Margaretha Grind 2 on behalf of the SPORTIF Investigators 1 Haemostasis Thrombosis and Vascular Biology Unit, University Department of Medicine, City Hospital, Birmingham B18 7QH, UK; and 2 AstraZeneca R&D Mölndal, SE Mölndal, Sweden Received 25 May 2006; revised 11 December 2006; accepted 12 January 2007; online publish-ahead-of-print 8 February 2007 KEYWORDS Hypertension; Atrial fibrillation; Stroke; Ximelagatran; Warfarin Aims To test the hypothesis that stroke and systemic embolic events (SEE) in the Stroke Prevention using an ORal Thrombin Inhibitor in atrial Fibrillation (SPORTIF) III and V trials were related to blood pressure, and that differences in event rates (stroke and SEE, bleeding) could also be related to the degree of hypertension. Methods and results A cross-sectional, longitudinal analysis was conducted, using data from the SPORTIF III and V trials. Results showed an increasing rate of stroke and SEE with increasing quartiles of systolic blood pressure (SBP) in AF patients. For the top quartile of SBP compared with the lowest quartile, the hazard ratio (HR) for stroke and SEE was 1.83 (95% confidence intervals [CI]: ), whereas mortality was lower in the top quartile (HR 0.64; 95% CI: ). In the combined SPORTIF III and V cohort, the event rate for stroke/see increased markedly at mean SBP of.140 mmhg. There was no relationship between bleeding and quartiles of BP. The proportion of subjects with mean systolic BP 140 mmhg was 35.8% (1220/3407) in SPORTIF III and 20.6% (807/3922) in SPORTIF V (P, ). Conclusion Hypertension contributes to increased stroke and SEE in AF. Event rates markedly increase at SBP levels of 140 mmhg. The higher stroke rates observed in SPORTIF III compared with SPORTIF V may be related to the greater proportion of subjects with SBP 140 mmhg during the trial. Introduction Atrial fibrillation (AF) is the most common cardiac arrhythmia and is strongly associated with a substantial risk of stroke and thromboembolism. AF often coexists with advanced age and such cardiovascular conditions as hypertension, heart failure, coronary artery disease (CAD), and diabetes mellitus. Among subjects with AF-associated stroke in a community survey, hypertension was the most common comorbid condition. 1 Hypertension is also the most frequently associated comorbidity among hospitalized patients with AF. 2 Hypertension is an independent risk factor for stroke, with a direct correlation between increasing blood pressure (BP) and stroke risk that extends even into the normotensive range. 3 Hypertension raises the risk of stroke in patients with AF, and the stroke risk in patients with both AF and hypertension is substantially higher than in those with either condition alone. 4 In the aspirin arm of the Stroke Prevention in Atrial Fibrillation (SPAF)-III trial, even subjects with a history of hypertension (but who did not meet the stated high risk criterion of SBP. 150 mmhg) had an increased risk of stroke in AF. 5 Among patients with AF, hypertension is associated with adjusted annual * Corresponding author. Tel: þ ; fax: þ address: g.y.h.lip@bham.ac.uk thromboembolism rates of 4.0 and 2.4% per year in women and men, respectively. 6 Furthermore, uncontrolled hypertension is an important risk factor for bleeding in anticoagulated subjects. 7 Ximelagatran is the first oral direct thrombin inhibitor that was recently tested for stroke thromboprophylaxis in two large Phase III clinical trials, the Stroke Prevention using an ORal Thrombin Inhibitor in AF (SPORTIF) III and V trials. 8,9 These trials found ximelagatran as effective as well-controlled warfarin in reducing the risk of stroke in moderate-to-high risk patients with AF, and rates of major bleeding on ximelagatran were slightly less than on warfarin. The most common risk factor in both trials was a history of hypertension (requiring anti-hypertensive treatment, but,180/100 mmhg at study randomization), which was present in 70% of participants. Rates of the primary endpoint (stroke and systemic embolic events [SEE]) were 1.6%/year on ximelagatran in both trials, but the event rates on warfarin differed: 2.3%/year in SPORTIF III and 1.2%/year in SPORTIF V. 8,9 In both trials, elevations of serum alanine aminotransferase (ALAT) beyond three times the upper limit of normal (ULN) were seen in 6% of patients treated with ximelagatran. On 14 February 2006, ximelagatran was withdrawn due to concerns about liver safety but recent interest into the value of hypertension control in AF merits an analysis of & The European Society of Cardiology All rights reserved. For Permissions, please journals.permissions@oxfordjournals.org

2 Hypertension and stroke in AF 753 the large, combined SPORTIF trial dataset to ascertain the impact of hypertension and BP control in a large, anticoagulated AF population at moderate-to-high risk of stroke and vascular events. We therefore hypothesized that the rate of stroke and SEE during anticoagulation can be related to the degree of hypertension, and that differences in various event rates between the SPORTIF III and V trials (stroke and SEE, bleeding) could also be related to hypertension. Patients and methods The design and final results of the SPORTIF III and V clinical trials have been published. 8,9 In brief, both trials were designed as noninferiority Phase III trials comparing ximelagatran with warfarin in subjects with non-valvular AF at moderate-to-high risk of thromboembolism, with a pooled analysis of the results of both trials prespecified. SPORTIF V was a double-blind trial conducted in North America, while SPORTIF III was performed open label in 23 countries outside of North America. Inclusion criteria for both trials were similar and included age 18 years, persistent or paroxysmal AF, and at least one of the following risk factors for stroke: hypertension (defined as high BP requiring anti-hypertensive treatment, but which was below 180/100 mmhg on randomization); age 75 years; previous stroke; transient ischaemic attack (TIA) or SEE; left ventricular (LV) dysfunction (ejection fraction,40% or symptomatic heart failure); age 65 years and CAD; and age 65 years and diabetes mellitus. Endpoints were evaluated in a blinded manner by a Central Events Adjudication Committee. To assess the relationship between hypertension and clinical outcomes of interest, patients were categorized into quartiles, according to mean systolic BP (SBP), during the entire follow-up period. For SPORTIF III, the median follow-up was 17.9 months per patient (range 0 26 months); for SPORTIF V, the median follow-up was 20.1 months (range 0 31 months); and for the pooled analysis, the median follow-up was 18.6 months per patient (range 0 31 months). BP was measured at baseline, 1, 3, and 6 months and every 6 months thereafter, and the mean BP for each patient was based, on average, on 6.5 measurements (SD 1.5). Illustration of efficacy results as a function of BP data is complicated by the continuous nature of the BP variable. Rather than categorizing BP data according to typically used intervals, we opted to illustrate unprejudiced relationships through reporting efficacy results vs. a continuously moving average of BP data. The duration of exposure differed between patients according to the time of enrolment, since the study closure period was the same for all patients. Analyses of clinical outcomes were made for all stroke (ischaemic and haemorrhagic) and SEE (the primary endpoint), major bleeding, all-cause mortality, and ALAT elevations (.3 ULN). Statistical analysis Categorical data were evaluated with Fisher s exact test or the x 2 test for more than two categories and continuous data using Student s t-test. All tests were performed as two-sided, adopting a 5% significance level. No adjustments for multiple testing has been done, all results reported are explorative in nature. Annualized event rates formed the basis for analyses, assuming constant event rates over time. Patients were categorized into quartiles according to their overall mean SBP-levels. Comparisons between quartiles have been conducted assuming that neither study drug nor length of follow-up influenced SBP levels, thus making it appropriate to use each patient s mean SBP level during the study as an explanatory variable. Unless otherwise stated, all analyses were performed following pooling of the two studies. With one exception, the analyses were conducted without allowing for a study factor this was deliberate, since there were systematic differences in SBP-control in the two studies, resulting from more aggressive antihypertensive treatment in the study centres based in North America (SPORTIF V). Thus, the study centre factor is intrinsically confounded with SBP-control. All randomized patients were included in the intention-to-treat (ITT) population. An on-treatment (OT) analysis accounted for time until the therapy was interrupted for up to 30 consecutive days (or up to 60 days for cardioversion) or 60 days cumulatively. Analyses confined to stroke, SEE, and death were analysed according to ITT. Assessment of major bleeding used the OT approach. No censoring was made of any ALAT elevations. Hazard ratios (HR) between treatments were obtained through Cox regression modelling. Moving averages of event rates over time are given as a function of BP, together with smoothing splines to illustrate unprejudiced relationships. All analyses were performed in SAS (Version 8.2; SAS Institute, Inc., Cary, NC, USA). More detailed descriptions of the statistical analyses used in the SPORTIF III and V trials have been published previously. 8,9 Results The mean BP and patient characteristics at study entry are summarized in Tables 1 and 2. The patients in the higher BP quartile were more likely to have more stroke risk factors (P ¼ for 2 vs..2 risk factors) as well as higher mean body mass index (BMI) (P ¼ ). Pre-specified stroke risk factors and mean SBP for patients in SPORTIF III, V, and the pooled analysis, stratified according to high vs. low BP values, are summarized in Table 3. The prevalence of prior stroke or TIA was higher in groups with higher SBP (P ¼ 0.02). As expected, those with high BP were older, but the prevalence of LV dysfunction (LV ejection fraction,40% or clinical heart failure) was lower (both Table 1 Mean (SD) BP and heart rate in SPORTIF III and V, based on the mean of each patient s measurements during the study SPORTIF III SPORTIF V Pooled N ¼ 3407 N ¼ 3922 N ¼ 7329 SBP (mmhg) Total (13.8) (13.0) (13.6) Male (13.9) (13.1) (13.7) Female (13.3) (12.6) (13.2) DBP (mmhg) Total 79.8 (7.6) 75.5 (7.5) 77.5 (7.9) Male 79.7 (7.7) 75.6 (7.7) 77.5 (7.9) Female 80.0 (7.4) 75.3 (7.2) 77.5 (7.7) Heart rate (b.p.m.) Total 76.6 (11.4) 73.5 (9.7) 75.0 (10.7) Male 76.0 (11.5) 72.9 (9.5) 74.3 (10.6) Female 78.1 (11.1) 74.8 (10.0) 76.4 (10.7)

3 754 G.Y.H. Lip et al. Table 2 Patient characteristics at entry: SPORTIF III, V, and pooled number and percentage of patients by quartiles of mean SBP during study (ITT population) SBP quartile P-value a I II III IV Sex Male 1407 (77%) 1283 (70%) 1213 (66%) 1169 (64%) Female 411 (23%) 552 (30%) 625 (34%) 669 (36%), Race Caucasian 1645 (90%) 1699 (93%) 1696 (92%) 1717 (93%) Black 46 (3%) 32 (2%) 27 (1%) 23 (1%) (Caucasian vs. other groups) Oriental 116 (6%) 101 (6%) 110 (6%) 95 (5%) Other 11 (1%) 3 (0%) 5 (0%) 3 (0%) Age (years), (27%) 425 (23%) 355 (19%) 296 (16%) (39%) 742 (40%) 747 (41%) 766 (42%) (34%) 668 (36%) 736 (40%) 776 (42%) Mean(SD) 69.6 (9.8) 70.5 (9) 71.5 (8.5) 72.1 (8.1), Range Number of risk factors b 0 5 (0%) 11 (1%) 1 (0%) (31%) 556 (30%) 500 (27%) 426 (23%) (31%) 574 (31%) 630 (34%) 621 (34%) (22%) 430 (23%) 398 (22%) 460 (25%) (2 vs..2) (11%) 193 (11%) 215 (12%) 225 (12%) 5 70 (4%) 60 (3%) 86 (5%) 83 (5%) 6 18 (1%) 11 (1%) 8 (0%) 21 (1%) 7 2 (0%) (0%) BMI (kg/m 2 ), (28%) 449 (24%) 415 (23%) 395 (21%) (39%) 717 (39%) 731 (40%) 766 (42%) (32%) 661 (36%) 683 (37%) 674 (37%) Mean(SD) 28.4 (5.8) 29.1 (6) 29.2 (5.8) 29.2 (5.6) Range Calculated CrCl,30 9 (0%) 5 (0%) 11 (1%) 7 (0%) (ml/min) 30 to (14%) 212 (12%) 223 (12%) 224 (12%) 50 to (38%) 706 (38%) 717 (39%) 756 (41%) (47%) 903 (49%) 875 (48%) 840 (46%) Mean(SD) 84 (35.8) 87 (38.9) 85 (36.8) 83.8 (34.7) 0.14 Range Modified Rankin Score (76%) 1419 (77%) 1398 (76%) 1392 (76%) (15%) 274 (15%) 295 (16%) 320 (17%) (9%) 137 (7%) 136 (7%) 120 (7%) (2 vs..2) 3 7 (0%) 1 (0%) 8 (0%) 5 (0%) (0%) 0 1 (0%) CrCl, creatinine clearance using Cockcroft Gault formula. 15 a Comparison between quartiles I and II vs. III and IV. b Among the seven risk factors specified in the study protocols (see Table 3). P, ) in patients with hypertension. In the pooled cohort, fewer subjects with known CAD had high BP (P, ). Anti-hypertensive medication at entry by quartiles of mean SBP is shown in Table 4. Neither study drug nor length of follow-up significantly influenced SBP levels (Table 5). For patients categorized by mean SBP, diastolic blood pressure (DBP), and pulse pressure, selected outcome variables are summarized in Tables The increasing stroke rate with increasing quartiles of SBP (Figure 1) is less apparent for DBP. The highest mortality rate occurred in patients in the lowest quartiles of SBP and DBP. When compared by quartiles of SBP, the HR between the SPORTIF III and V trials was not significantly different from unity (Table 8). The top quartile of SBP was associated with a greater risk of stroke and SEE than the lowest quartile (HR 1.83; 95% CI: ), whereas mortality was lower in the top quartile (HR 0.64; CI: ). There was a similar relationship between stroke/see and pulse pressure, and a trend for major bleeding (Table 9). The increased mortality in the lower quartile for SBP was not seen for pulse pressure. We assessed whether LV dysfunction was a confounder in explaining the mortality pattern in relation to SBP control. The increased mortality rate in the lowest SBP quartile was largely driven by a larger proportion of patients with heart failure who had a higher mortality rate (see Supplementary material online, Figure S1). Addition of CAD into the analysis did not alter this observation (data not shown). Among the patients with neither of the baseline risk factors (i.e. LV dysfunction/cad/stroke/tia/age 75 years,,20% of the total cohort) mortality rate was about equally low in all SBP quartiles: 1.62, 1.30, 1.53, and 1.55%/year, respectively.

4 Hypertension and stroke in AF 755 Table 3 Medical history for patients in SPORTIF III and V by quartiles of mean SBP (ITT population) SBP quartile P-value a I II III IV n (%) n (%) n (%) n (%) Previous stroke No 1609 (89) 1618 (88) 1616 (88) 1568 (85) Yes 209 (11) 217 (12) 222 (12) 270 (15) SEE No (95) 1755 (96) 1753 (95) 1763 (96) Yes 88 (5) 80 (4) 85 (5) 75 (4) 0.53 Previous TIA No 1638 (90) 1639 (89) 1656 (90) 1620 (88) Yes 180 (10) 196 (11) 182 (10) 218 (12) 0.40 Previous stroke and/or TIA No 1466 (81) 1462 (80) 1471 (80) 1391 (76) Yes 352 (19) 373 (20) 367 (20) 447 (24) 0.02 Age 75 No 1194 (66) 1167 (64) 1102 (60) 1062 (58) Yes 624 (34) 668 (36) 736 (40) 776 (42), Hypertension No 688 (38) 490 (27) 326 (18) 200 (11) Yes 1130 (62) 1345 (73) 1512 (82) 1638 (89), LV dysfunction b No 868 (48) 1181 (64) 1272 (69) 1327 (72) Yes 950 (52) 654 (36) 566 (31) 511 (28), Diabetes mellitus No 1410 (78) 1425 (78) 1403 (76) 1366 (74) Yes 408 (22) 410 (22) 435 (24) 472 (26) Age 65 and diabetes mellitus No 1523 (84) 1528 (83) 1496 (81) 1442 (78) Yes 295 (16) 307 (17) 342 (19) 396 (22) CAD No 924 (51) 988 (54) 1044 (57) 1093 (59) Yes 894 (49) 846 (46) 794 (43) 745 (41), Age 65 and CAD No 1103 (61) 1130 (62) 1154 (63) 1178 (64) Yes 715 (39) 705 (38) 684 (37) 660 (36) Clinically significant bleeding No 1696 (93) 1730 (94) 1724 (94) 1718 (93) Yes 122 (7) 105 (6) 114 (6) 119 (6) 0.77 a Comparison between quartiles I and II vs. III and IV; Fisher s exact test (two-sided). b LV dysfunction, either LVEF 40% or symptomatic CHF, congestive heart failure. Table 4 Anti-hypertensive medication at entry to SPORTIF III and V by quartiles of mean SBP (ITT population) SBP quartile P-value a I II III IV n (%) n (%) n (%) n (%) ACE-Inhibitors Yes 959 (53) 834 (45) 819 (45) 917 (50) 0.11 ARBs Yes 162 (9) 174 (9) 229 (12) 257 (14), Calcium antagonists Yes 504 (28) 627 (34) 687 (37) 720 (39), Beta blockers Yes 977 (54) 863 (47) 784 (43) 770 (42), Diuretics Yes 1043 (57) 913 (50) 893 (49) 945 (51) Agents per patient (7) 161 (9) 146 (8) 142 (8) (24) 514 (28) 531 (29) 453 (25) (2 vs..2) (35) 669 (36) 670 (36) 684 (37) (31) 406 (22) 424 (23) 453 (25) 4 66 (4) 84 (5) 66 (4) 101(6) 5 1 (0) 1 (0) 1 (0) 5 (0) a Comparison between quartiles I and II vs. III and IV; Fisher s exact test (two-sided). There was an increased event rate for haemorrhagic stroke with increasing quartiles of BP (P-value for Q1þ2 vs. Q3þ4 is 0.013). The total number of such events being 1 (0.04%/year), 2 (0.07%/year), 5 (0.17%/year), and 9 (0.32%/year) in the four SBP-quartiles, respectively, with similar trends for both study drugs. There was no relationship between major bleeding, major plus minor bleeding or ALAT elevations to quartiles of BP (Tables 10 and 11). Relationships between SBP-control and bleeding rates were fairly similar between the warfarin and ximelagatran (data not shown). Figure 2 shows the cumulative stroke/see rate in SPORTIF III and V combined, with patients categorized by quartiles of mean SBP over the course of the studies.

5 756 G.Y.H. Lip et al. Table 5 Descriptive statistics for sitting SBP (mmhg) by study drug over time Ximelagatran Warfarin 3 months N Median Mean SD months N Median Mean SD Closing visit N Median Mean SD Smoothed curves around moving average rates of stroke/ SEE, major bleed, and mortality, in SPORTIF III and V combined, pooled over study drugs, according to mean SBP are provided in Supplementary material online, Figure S2. The greatest cumulative event rates occurred in association with the highest quartile of SBP, while the Kaplan Meier curves for Q1 and Q2 are almost identical and the curve for Q3 shows a slight increase in the cumulative event rate after 12 months. The increasing event rate with increasing quartiles of BP was similar in the warfarin and ximelagatran treatment groups in the whole cohort (see Supplementary material online, Table S1) and in the individual studies (Table 11). HRs are given for comparison between the two studies for each quartile (Table 11). Differences in rates of stroke/see between SPORTIF III and V for each quartile of BP did not reach statistical significance. Sensitivity analyses show that the relationships seen between SBP quartiles and the outcomes analysed remained largely unaffected when allowing Table 6 Efficacy and safety results by quartiles of the mean of individual patient SBP values Mean SBP quartile Event rate a (per PY) Stroke/SEE Mortality Major bleed ALAT.3 ULN b /2770 ¼ 1.30% 146/2795 ¼ 5.22% 59/2431 ¼ 2.43% 60/951 ¼ 6.3% /2841 ¼ 1.27% 68/2865 ¼ 2.37% 51/2520 ¼ 2.02% 68/915 ¼ 7.4% /2847 ¼ 1.62% 88/2874 ¼ 3.06% 52/2539 ¼ 2.05% 52/928 ¼ 5.6% /2775 ¼ 2.38% 94/2813 ¼ 3.34% 55/2464 ¼ 2.23% 44/870 ¼ 5.1% a Pooled analysis over treatments (except for ALAT) and studies SPORTIF III and V. b Ximelagatran patients only; rates per patient (not per year). Table 7 Efficacy and safety results by quartiles of the mean of individual patient DBP values Mean DBP quartile Event rate a (per PY) Stroke/SEE Mortality Major bleed ALAT.3 ULN b /2755 ¼ 1.71% 150/2785 ¼ 5.39% 72/2328 ¼ 3.09% 55/960 ¼ 5.7% /2883 ¼ 1.49% 101/2911 ¼ 3.47% 44/2592 ¼ 1.70% 57/883 ¼ 6.5% /2826 ¼ 1.52% 76/2853 ¼ 2.66% 57/2540 ¼ 2.24% 58/929 ¼ 6.2% /2770 ¼ 1.84% 69/2797 ¼ 2.47% 44/2493 ¼ 1.76% 54/892 ¼ 6.0% a Pooled analysis over treatments (except for ALAT) and studies SPORTIF III and V. b Ximelagatran patients only; rates per patient (not per year). Table 8 Efficacy results by quartiles of the mean of individual patient SBP values: comparison of SPORTIF III and V cohorts Mean SBP quartile Stroke/SEE rate a (per PY) SPORTIF III SPORTIF V SPORTIF III/V HR b (95% CI) /908 ¼ 1.32% 24/1862 ¼ 1.29% 0.98 ( ); P ¼ /1075 ¼ 1.67% 18/1766 ¼ 1.02% 1.50 ( ); P ¼ /1283 ¼ 1.71% 24/1564 ¼ 1.53% 1.08 ( ); P ¼ /1621 ¼ 2.71% 22/1153 ¼ 1.91% 1.31 ( ); P ¼ 0.30 Total 96/4 886 ¼ 1.96% 88/6 346 ¼ 1.39% 1.33 ( ); P ¼ a Pooled analysis over treatments. b Study drug has been accounted for in the statistical model.

6 Hypertension and stroke in AF 757 Table 9 Efficacy and safety results by pulse pressure quartiles (mean SBP mean DBP) Mean pulse pressure: quartile Event rate a (per PY) Stroke/SEE Mortality Major bleed ALAT.3 ULN b /2801 ¼ 1.29% 105/2825 ¼ 3.72% 49/2509 ¼ 1.95% 78/943 ¼ 8.3% /2816 ¼ 1.38% 97/2842 ¼ 3.41% 45/2530 ¼ 1.78% 47/939 ¼ 5.0% /2828 ¼ 1.27% 90/2847 ¼ 3.16% 57/2491 ¼ 2.29% 50/872 ¼ 5.7% /2788 ¼ 2.62% 104/2832 ¼ 3.67% 66/2424 ¼ 2.72% 49/910 ¼ 5.4% a Pooled analysis over treatments (except for ALAT) and studies SPORTIF III and V. b Ximelagatran patients only; rates per patient (not per year). Table 10 HRs a (and 95% CIs) for comparisons in Table 6 Comparison Stroke/SEE Mortality Major bleed ALAT.3 ULN b Q2/Q ( ) 0.46 ( ) 0.83 ( ) 7.4/6.3% P ¼ 0.92 P, P ¼ 0.32 P ¼ 0.36 Q3/Q ( ) 0.59 ( ) 0.84 ( ) 5.6/6.3% P ¼ 0.32 P, P ¼ 0.37 P ¼ 0.56 Q4/Q ( ) 0.64 ( ) 0.91 ( ) 5.1/6.3% P ¼ P ¼ P ¼ 0.63 P ¼ 0.27 Q3/Q ( ) 1.29 ( ) 1.01 ( ) 5.6/7.4% P ¼ 0.27 P ¼ 0.11 P ¼ 0.95 P ¼ 0.13 Q4/Q ( ) 1.41 ( ) 1.10 ( ) 5.1/7.4% P ¼ P ¼ P ¼ 0.63 P ¼ Q4/Q ( ) 1.10 ( ) 1.07 ( ) 5.1/5.6% P ¼ P ¼ 0.54 P ¼ 0.74 P ¼ 0.67 SBP continuous covariate c ( ) d ( ) ( ) ( ) P ¼ P, P ¼ 0.95 P ¼ 0.15 a Study drug has been accounted for in the statistical model. b Analysed by Fisher s exact test rather than Cox regression due to time-specific pattern. The vast majority of elevations occurred during the first 6 months of treatment. c Cox regressions (logistic regression for ALAT. 3 ULN b ), the estimates reported for the HR (or odds ratio) are for each increase of 1 mmhg in SBP. In addition to study drug, study has also been accounted for in these models. d Due to the apparent J-shaped relationship between SBP and mortality, SBP-levels,130 mmhg have been transformed according to SBP-transf ¼ (130/SBP) 2 i.e. SBP-values deviating in either direction from 130 are assumed to be associated with increased risk for mortality, with a steeper relationship for lower values. This transformation resulted in a good fit relative to the observed relationship. for relevant risk factors (e.g. age, gender, and previous stroke/tia) (see Supplementary material online, Table S2). A moving average for rates of stroke/see in the pooled cohort for both study drugs, according to mean SBP and DBP, is shown in Figure 3A and B, respectively. The data suggest that the rate of primary events rate increases markedly at mean SBP 140 mmhg. The corresponding relationship was not apparent for DBP. The proportion of subjects with mean SBP 140 mmhg was 35.8% (1220/3407) in SPORTIF III and 20.6% (807/3922) in SPORTIF V (P, ). Major bleeding rates in the four SBP quartiles were 2.43, 2.02, 2.05, and 2.23%/year, respectively. Corresponding rates by pulse pressure quartiles were 1.95, 1.78, 2.29, and 2.72%/year. Discussion This is the most extensive analysis of the cumulative effects of BP on stroke and SEE in patients with AF currently available, involving serial measurements in anticoagulated patients over more than patient-years (PY). The Table 11 Rates of major plus minor bleeding by quartiles of the mean individual patient SBP values Mean SBP quartile Event rate a (per PY) Minor/major bleed /1857 ¼ 36.4% /1944 ¼ 33.0% /1924 ¼ 34.9% /1852 ¼ 36.5% a Pooled analysis over treatments and studies SPORTIF III and V. data clearly show that rates of stroke and SEE increase substantially at mean SBP levels of 140 mmhg, and the event rates were highest in participants with the highest average SBP readings. The higher proportion of subjects with average SBP 140 mmhg in the SPORTIF III trial may account for the higher event rates than in SPORTIF V, where hypertension was better controlled. The data further suggest that event rates markedly increase at SBP

7 758 G.Y.H. Lip et al. Figure 1 Rates of stroke/see rate in SPORTIF III and V by quartiles of individual patient mean SBP values ( , , , and mmhg). Figure 2 Cumulative rates of stroke/see in SPORTIF III and V combined, pooled over study drugs, by quartiles of mean SBP during the studies. levels over 140 mmhg, and that control of hypertension to lower SBP is associated with a uniformly lower thromboembolic risk. Although hypertension raises the risk of stroke in patients with AF, this analysis suggests controlled hypertension (mean SBP,140 mmhg) is associated with a low stroke risk compared with patients with poorly controlled hypertension, in which the risk markedly increases with higher BP. The present analysis also supports SBP as being a more powerful determinant of stroke risk as compared with DBP among the AF patients. Hypertension is an important risk factor in most stratification schemes used to identify patients with AF who require anticoagulation 10 and maintains predictive power for the development of ischaemic events, even in patients with AF receiving anticoagulant therapy. 11 Hypertension is also a risk factor for bleeding complications of anticoagulation, including intracranial haemorrhage (OR 2.69; 95% CI: ). 7,12,13 Bleeding can also be related to the duration and intensity of anticoagulant therapy. 13 We were not able to confirm a relationship between quartiles of SBP and rates of major or major plus minor bleeding in the SPORTIF cohorts, but this may be a reflection of the optimization of risk factor management and close monitoring associated with a clinical trial setting. Clinical practice usually advocates that patients with hypertension require especially close monitoring of anticoagulation and optimum control of BP 13 during chronic anticoagulant therapy. Figure 3 Moving average rates of stroke/see in SPORTIF III and V combined, pooled over study drugs, according to mean (A) SBP and (B) DBP. The 7329 patients were sorted according to their individual mean SBP (during the entire study) from lowest to highest. The first data point to the left represents the 800 patients (about 1200 PY) with the very lowest mean SBP, plotted at their overall mean. For the second point, the 40 patients with the lowest SBP are removed and replaced with those having the 801st to 840th lowest mean SBP, and so forth, until the 800 patients with highest mean SBP are represented at the right-hand end of the figure. Of interest is our observation that mortality was highest in the lowest quartile of BP. There has been a long-standing debate on the existence of a J-curve in relation to the relationship between BP and mortality, 14 especially among the elderly. The present post hoc analysis of the SPORTIF trials adds to this debate, but only among an AF population.

8 Hypertension and stroke in AF 759 Further observations on the existence of a J-curve in larger, non-trial AF populations would be needed to address this issue. A major limitation of this study was its reliance upon post hoc analysis of the SPORTIF III and V clinical trial dataset. The selection phenomenon for trial inclusion may partly account for the unusual distribution of risk factors; for example, hypertension is a risk factor for LV dysfunction, heart failure, and CAD, but patients with hypertension had a lower prevalence of LV dysfunction. Patients were also not randomized on the basis of severity of antecedent hypertension or quality of BP control during the trials, and hypertension management was left to the discretion of the patients personal physicians. Apart from the differences in BP control between SPORTIF III and V, there were also geographical differences within the SPORTIF III trial cohort, with patients in some countries achieving better BP control than others. The relatively small numbers of patients in these subsets, however, preclude meaningful analyses in relation to outcomes. The SPORTIF V study was conducted exclusively in North America, where secondary prevention strategies (including control of hypertension) were more systematically applied than in the array of countries involved in the SPORTIF III study, and other therapies (e.g. lipid-lowering medications) were more often given concurrently. Hence, relationships between SBP-control and events may intrinsically be confounded with other differential effects between the two studies. These correlations between study and SBP-control caused us to exclude study (that is, SPORTIF III or V) as a factor in most analyses. In conclusion, hypertension contributes to increased rates of stroke and SEE in AF. Event rates markedly increased at SBP levels of 140 mmhg, and control of hypertension to lower levels was associated with uniformly lower rates of stroke and SEE. The higher stroke rates in SPORTIF III may have been related to the greater proportion of subjects with inadequate BP control. Supplementary material Supplementary material is available at European Heart Journal online. Acknowledgements The SPORTIF III and V investigators are listed in references Olsson 8 and SPORTIF Executive Committee for the SPORTIF V Investigators, 9 respectively. Conflict of interest: G.Y.H.L. has received funding for research, educational symposia, consultancy, and lecturing from different manufacturers of drugs used for the treatment of hypertension, AF, and thrombosis, including AstraZeneca, who manufacture ximelagatran. L.F. and M.G. are employees of AstraZeneca. References 1. Tsang TS, Petty GW, Barnes ME, O Fallon WM, Bailey KR, Wiebers DO, Sicks JD, Christianson TJ, Seward JB, Gersh BJ. The prevalence of atrial fibrillation in incident stroke cases and matched population controls in Rochester, Minnesota: changes over three decades. JAmColl Cardiol 2003;42: Wattigney WA, Mensah GA, Croft JB. Increasing trends in hospitalisation for atrial fibrillation in the United States, 1985 through 1999: implications for primary prevention. Circulation 2003;108: Wang JG, Staessen JA, Franklin SS, Fagard R, Gueyffier F. Systolic and diastolic blood pressure lowering as determinants of cardiovascular outcome. Hypertension 2005;45: Gage BF, Waterman AD, Shannon W, Boechler M, Rich MW, Radford MJ. Validation of clinical classification schemes for predicting stroke: results from the National Registry of Atrial Fibrillation. JAMA 2001; 285: Zabalgoitia M, Halperin JL, Pearce LA, Blackshear JL, Asinger RW, Hart RG. Transesophageal echocardiographic correlates of clinical risk of thromboembolism in nonvalvular atrial fibrillation. Stroke Prevention in Atrial Fibrillation III Investigators. J Am Coll Cardiol 1998;31: Fang MC, Singer DE, Chang Y, Hylek EM, Henault LE, Jensvold NG, Go AS. Gender differences in the risk of ischaemic stroke and peripheral embolism in atrial fibrillation: the AnTicoagulation and Risk factors In Atrial fibrillation (ATRIA) study. Circulation 2005;112: Hart RG, Tonarelli SB, Pearce LA. Avoiding central nervous system bleeding during antithrombotic therapy: recent data and ideas. Stroke 2005; 36: Olsson SB, Executive Steering Committee on behalf of the SPORTIF III Investigators. Stroke prevention with the oral direct thrombin inhibitor ximelagatran compared with warfarin in patients with non-valvular atrial fibrillation (SPORTIF III): randomised controlled trial. Lancet 2003;362: SPORTIF Executive Committee for the SPORTIF V Investigators. Ximelagatran versus warfarin for stroke prevention in patients with nonvalvular atrial fibrillation. JAMA 2005;293: Lip GY, Boos C. Antithrombotic therapy for atrial fibrillation. Heart 2006; 92: Poli D, Antonucci E, Cecchi E, Marcucci R, Liotta AA, Cellai AP, Lenti M, Gensini GF, Abbate R, Prisco D. Culprit factors for the failure of wellconducted warfarin therapy to prevent ischaemic events in patients with atrial fibrillation: the role of homocysteine. Stroke 2005;36: Huber J, Stollberger C, Finsterer J, Schneider B, Langer T. Quality of blood pressure control and risk of cerebral bleeding in patients with oral anticoagulation. Acta Med Austriaca 2003;30: Berwaerts J, Webster J. Analysis of risk factors involved in oral-anticoagulant-related intracranial haemorrhages. QJM 2000;93: Boutitie F, Gueyffier F, Pocock S, Fagard R, Boissel JP; INDANA Project Steering Committee. INdividual Data ANalysis of Antihypertensive intervention. J-shaped relationship between blood pressure and mortality in hypertensive patients: new insights from a meta-analysis of individualpatient data. Ann Intern Med 2002;136: Cockcroft DW, Gault MH. Prediction of creatinine clearance from serum creatinine. Nephron 1976;16:31 41.