Results

Overall, 32 346 patients from HFrEF trials and 13 723 patients from HFpEF trials were analysed in this study (see Supplementary data online, Figure S1). The distributions of serum potassium concentrations are shown in Supplementary data online, Figure S2. The 95% interval of serum potassium concentrations was 3.6–5.5 mmol/L in HFrEF and 3.5–5.3 mmol/L in HFpEF. In HFrEF, 502 patients (1.6%) had a potassium concentration <3.5 mmol/L, 3301 participants (10.2%) had a potassium ≥3.5–<4.0 mmol/L, 11 511 (35.6%) potassium ≥4.0–<4.5 mmol/L, 11 963 (37.0%) potassium ≥4.5–<5.0 mmol/L, 4119 (12.7%) potassium ≥5.0–<5.5 mmol/L, and 950 patients (2.9%) had a potassium concentration ≥5.5 mmol/L. In HFpEF, the respective distribution of patients across these potassium categories was: 290 (2.1%) <3 .5 mmol/L, 1968 (14.3%) ≥3.5–<4.0 mmol/L, 5356 (39.0%) ≥ 4.0–<4.5 mmol/L, 4602 (33.5%) ≥ 4.5–<5.0 mmol/L, 1326 (9.7%) ≥ 5.0–<5.5 mmol/L, and 181 (1.3%) ≥ 5.5 mmol/L. Both mean and median potassium values were 4.5 mmol/L in HFrEF and 4.4 mmol/L in HFpEF, respectively (see Supplementary data online, Figure S2). The median follow-up was 24.2 and 36.8 months in HFrEF and HFpEF trials, respectively.

Baseline characteristics according to potassium concentration

HFrEF

The baseline characteristics of each LVEF phenotype are presented in Table 1. From the lowest to the highest potassium category, the proportion of men increased (from 72.5% to 78.5%), as did the proportion of participants with a history of diabetes (from 39.4% to 43.9%), history of myocardial infarction (36.7% to 47.3%) and those in NYHA class III/IV (from 36.7% to 43.6%). When compared to patients with lower potassium levels, those with higher potassium levels had a slightly lower systolic blood pressure and much lower eGFR, and higher mean LVEF.

 | HFrEF (LVEF ≤ 40%) | P-value for trend
K < 3.5N = 502 | K ≥ 3.5-<4.0N = 3301 | K ≥ 4.0-<4.5N = 11 511 | K ≥ 4.5-<5.0N = 11 963 | K ≥ 5.0-<5.5N = 4119 | K ≥ 5.5N = 950
Age (years) | 63.1 ± 12.5 | 63.9 ± 11.8 | 64.1 ± 11.7 | 65.1 ± 10.9 | 65.6 ± 10.7 | 65.4 ± 10.8 | <.001
Male sex, n (%) | 364 (72.5) | 2531 (76.7) | 8890 (77.2) | 9389 (78.6) | 3236 (78.6) | 746 (78.5) | <.001
Region, n (%) | .015
North America | 115 (22.9) | 566 (17.2) | 1811 (15.7) | 1549 (13.0) | 454 (11.0) | 102 (10.7) | 
Latin and Central America | 80 (15.9) | 494 (15.0) | 1712 (14.9) | 1840 (15.4) | 705 (17.1) | 186 (19.6) | 
Western Europe | 73 (14.5) | 637 (19.3) | 2403 (20.9) | 2584 (21.6) | 776 (18.8) | 139 (14.6) | 
Central and Eastern Europe (including Russia) | 89 (17.7) | 731 (22.1) | 3206 (27.9) | 3977 (33.2) | 1561 (37.9) | 395 (41.6) | 
Asia-Pacific and other | 145 (28.9) | 873 (26.5) | 2379 (20.7) | 2013 (16.8) | 623 (15.1) | 128 (13.5) | 
Race, n (%) | <.001
White | 270 (53.8) | 2037 (61.7) | 7985 (69.4) | 8952 (74.8) | 3188 (77.4) | 750 (79.0) | 
Black | 62 (12.4) | 239 (7.2) | 588 (5.1) | 487 (4.1) | 139 (3.4) | 38 (4.0) | 
Asian | 124 (24.7) | 779 (23.6) | 2144 (18.6) | 1741 (14.6) | 506 (12.3) | 112 (11.8) | 
Other | 46 (9.2) | 246 (7.5) | 794 (6.9) | 783 (6.6) | 286 (6.9) | 50 (5.3) | 
Heart failure characteristics
NYHA functional class III or IV | 184 (36.7) | 1127 (34.2) | 3780 (32.9) | 4065 (34.0) | 1551 (37.7) | 413 (43.6) | <.001
KCCQ-CSS | 69.1 (47.4–83.6) | 73.4 (53.1–89.1) | 75.0 (57.3–89.6) | 75.9 (57.6–89.6) | 74.0 (55.2–88.5) | 69.4 (53.1–86.1) | .76
Physiological measurements
Body mass index (kg/m2) | 27.7 ± 6.3 | 27.9 ± 6.1 | 28.1 ± 5.8 | 28.1 ± 5.5 | 28.0 ± 5.6 | 27.5 ± 5.5 | .078
Heart rate (bpm) | 75.9 ± 13.0 | 73.2 ± 12.2 | 72.0 ± 12.0 | 71.7 ± 12.3 | 71.7 ± 12.1 | 73.1 ± 12.6 | <.001
Systolic blood pressure (mmHg) | 121.5 ± 17.8 | 122.0 ± 17.9 | 121.8 ± 17.1 | 120.6 ± 16.3 | 119.9 ± 15.6 | 119.5 ± 15.6 | <.001
Diastolic blood pressure (mmHg) | 73.8 ± 11.0 | 74.2 ± 11.1 | 73.7 ± 10.6 | 73.1 ± 10.4 | 72.8 ± 10.0 | 72.6 ± 10.3 | <.001
Left ventricular ejection fraction (%) | 27.3 ± 6.9 | 28.0 ± 6.5 | 28.3 ± 6.4 | 28.4 ± 6.3 | 28.7 ± 6.4 | 29.0 ± 6.3 | <.001
Laboratory data
Potassium (mmol/L) | 3.2 ± 0.2 | 3.8 ± 0.1 | 4.2 ± 0.1 | 4.7 ± 0.1 | 5.2 ± 0.1 | 5.8 ± 0.3 | 
Estimated glomerular filtration rate (mL/min/1.73 m2)a | 70.1 ± 29.1 | 70.2 ± 22.4 | 70.5 ± 22.8 | 66.3 ± 21.1 | 60.9 ± 20.5 | 56.6 ± 21.0 | <.001
Estimated glomerular filtration rate <60 mL/min/1.73 m2a | 192 (38.3) | 1086 (32.9) | 3719 (32.3) | 4783 (40.0) | 2122 (51.5) | 564 (59.4) | <.001
Baseline NT-proBNP (ng/L) | 2428 (1121–5299) | 1730 (914–3654) | 1510 (810–2982) | 1486 (805–2929) | 1676 (881–3313) | 1885 (1043–3826) | .78
Haemoglobin (g/dL) | 13.4 ± 1.9 | 13.6 ± 1.7 | 13.8 ± 1.6 | 13.8 ± 1.7 | 13.8 ± 1.7 | 13.6 ± 1.8 | .025
Sodium (mmol/L) | 140.1 ± 5.9 | 140.4 ± 3.3 | 140.4 ± 3.1 | 140.2 ± 3.3 | 140.1 ± 3.6 | 139.4 ± 4.1 | <.001
Serum creatinine (umol/L) | 90.7 ± 50.7 | 87.3 ± 46.2 | 89.4 ± 40.0 | 95.5 ± 40.8 | 108.8 ± 38.9 | 122.0 ± 48.3 | <.001
Comorbidities
Diabetes mellitus | 198 (39.4) | 1091 (33.1) | 3790 (32.9) | 4365 (36.5) | 1639 (39.8) | 417 (43.9) | <.001
Hypertension | 360 (71.7) | 2319 (70.3) | 7739 (67.2) | 8083 (67.6) | 2912 (70.7) | 688 (72.4) | .13
Myocardial infarction | 184 (36.7) | 1337 (40.5) | 4843 (42.1) | 5422 (45.3) | 1911 (46.4) | 449 (47.3) | <.001
Atrial fibrillation/flutter | 189 (37.7) | 1209 (36.6) | 4176 (36.3) | 4452 (37.2) | 1535 (37.3) | 359 (37.8) | .22
Stroke | 49 (9.8) | 280 (8.5) | 1000 (8.7) | 1046 (8.8) | 376 (9.1) | 80 (8.4) | .72
Drugs and devices
Angiotensin-converting enzyme inhibitor/angiotensin receptor blocker/angiotensin receptor–neprilysin inhibitor | 421 (83.9) | 2995 (90.8) | 10 679 (92.8) | 11 254 (94.1) | 3917 (95.1) | 885 (93.2) | <.001
Beta-blocker | 442 (88.1) | 2948 (89.4) | 10 461 (90.9) | 10 949 (91.6) | 3808 (92.5) | 866 (91.2) | <.001
Sodium–glucose cotransporter 2 inhibitor | 4 (0.8) | 14 (0.4) | 72 (0.6) | 81 (0.7) | 42 (1.0) | 4 (0.4) | .043
Mineralocorticoid receptor antagonist | 210 (41.8) | 1409 (42.7) | 5637 (49.0) | 6741 (56.4) | 2577 (62.6) | 647 (68.1) | <.001
Calcium channel blockers | 61 (12.2) | 359 (10.9) | 1080 (9.4) | 959 (8.0) | 333 (8.1) | 77 (8.1) | <.001
Anticoagulant agent | 211 (48.7) | 1320 (46.1) | 4771 (45.6) | 5301 (48.0) | 2013 (50.1) | 503 (53.5) | <.001
Antiplatelet agent | 269 (53.6) | 1774 (53.8) | 6339 (55.1) | 6638 (55.5) | 2352 (57.1) | 535 (56.3) | .004
Diuretics | 457 (91.0) | 2929 (88.8) | 9785 (85.1) | 10 109 (84.5) | 3466 (84.2) | 834 (87.8) | <.001
Loop diuretics | 370 (85.5) | 2386 (83.3) | 8268 (79.0) | 8770 (79.5) | 3167 (78.9) | 779 (82.8) | .013
Dose of oral loop diuretics (furosemide equivalent dose) (mg/day)b | 80 (40–160) | 40 (40–80) | 40 (40–80) | 40 (40–80) | 40 (40–80) | 40 (40–80) | <.001
Thiazide diuretics | 40 (9.2) | 239 (8.3) | 725 (6.9) | 596 (5.4) | 208 (5.2) | 49 (5.2) | <.001
Implantable cardioverter defibrillator | 100 (19.9) | 615 (18.6) | 2244 (19.5) | 2521 (21.1) | 863 (21.0) | 179 (18.8) | .008
Pacemaker | 40 (8.0) | 340 (10.3) | 1216 (10.6) | 1229 (10.3) | 424 (10.3) | 84 (8.8) | .67
 | HFpEF (LVEF ≥50%) | P-value for trend
K < 3.5N = 290 | K ≥ 3.5–<4.0N = 1968 | K ≥ 4.0–<4.5N = 5356 | K ≥ 4.5–<5.0N = 4602 | K ≥ 5.0–<5.5N = 1326 | K ≥ 5.5N = 181
Age (years) | 72.6 ± 10.0 | 72.3 ± 9.1 | 72.1 ± 8.8 | 72.6 ± 8.5 | 72.3 ± 8.1 | 72.3 ± 8.7 | .60
Male sex, n (%) | 120 (41.4) | 847 (43.0) | 2366 (44.2) | 2032 (44.2) | 642 (48.4) | 85 (47.0) | .006
Region, n (%) | <.001
North America | 108 (37.2) | 683 (34.7) | 1386 (25.9) | 881 (19.1) | 180 (13.6) | 29 (16.0) | 
Latin and Central America | 31 (10.7) | 208 (10.6) | 562 (10.5) | 589 (12.8) | 153 (11.5) | 18 (9.9) | 
Western Europe | 64 (22.1) | 436 (22.2) | 1273 (23.8) | 1124 (24.4) | 351 (26.5) | 47 (26.0) | 
Central and Eastern Europe (including Russia) | 47 (16.2) | 385 (19.6) | 1542 (28.8) | 1574 (34.2) | 516 (38.9) | 71 (39.2) | 
Asia-Pacific and other | 40 (13.8) | 256 (13.0) | 593 (11.1) | 434 (9.4) | 126 (9.5) | 16 (8.8) | 
Race, n (%) | <.001
White | 217 (74.8) | 1523 (77.4) | 4483 (83.7) | 3986 (86.6) | 1180 (89.0) | 157 (86.7) | 
Black | 27 (9.3) | 151 (7.7) | 232 (4.3) | 142 (3.1) | 19 (1.4) | 4 (2.2) | 
Asian | 36 (12.4) | 217 (11.0) | 461 (8.6) | 310 (6.7) | 82 (6.2) | 15 (8.3) | 
Other | 10 (3.5) | 77 (3.9) | 180 (3.4) | 164 (3.6) | 45 (3.4) | 5 (2.8) | 
Heart failure characteristics
NYHA functional class III or IV | 109 (37.6) | 863 (44.0) | 2249 (42.0) | 1859 (40.4) | 544 (41.0) | 81 (44.8) | .17
KCCQ-CSS | 63.8 (44.0–80.2) | 64.7 (47.4–82.3) | 68.8 (51.0–84.9) | 69.8 (52.6–84.4) | 71.2 (55.7–84.9) | 72.4 (54.2–84.4) | <.001
Physiological measurements
Body mass index (kg/m2) | 31.2 ± 7.1 | 31.2 ± 6.9 | 30.7 ± 6.1 | 30.5 ± 5.9 | 30.2 ± 5.3 | 30.1 ± 5.7 | <.001
Heart rate (bpm) | 72.1 ± 12.9 | 71.4 ± 11.8 | 70.6 ± 11.4 | 70.3 ± 11.3 | 70.1 ± 11.7 | 71.7 ± 12.3 | .001
Systolic blood pressure (mmHg) | 131.2 ± 18.5 | 131.8 ± 16.4 | 132.0 ± 15.5 | 132.0 ± 15.7 | 132.2 ± 15.3 | 130.8 ± 16.4 | .97
Diastolic blood pressure (mmHg) | 74.1 ± 11.3 | 75.0 ± 10.8 | 75.2 ± 10.6 | 75.5 ± 10.4 | 75.4 ± 10.5 | 75.8 ± 11.0 | .039
Left ventricular ejection fraction (%) | 58.9 ± 6.6 | 59.5 ± 7.2 | 59.3 ± 7.2 | 59.4 ± 7.3 | 59.1 ± 7.0 | 58.5 ± 7.6 | .31
Laboratory data
Potassium (mmol/L) | 3.3 ± 0.2 | 3.8 ± 0.1 | 4.2 ± 0.1 | 4.7 ± 0.1 | 5.1 ± 0.1 | 5.7 ± 0.3 | 
Estimated glomerular filtration rate (mL/min/1.73 m2)a | 62.8 ± 20.7 | 66.1 ± 20.6 | 67.4 ± 21.4 | 64.0 ± 20.8 | 59.7 ± 20.5 | 54.7 ± 21.3 | <.001
Estimated glomerular filtration rate <60 mL/min/1.73 m2a | 137 (47.2) | 829 (42.1) | 2121 (39.6) | 2158 (46.9) | 745 (56.2) | 121 (66.85) | <.001
Baseline NT-proBNP (ng/L) | 901 (378–1849) | 752 (272–1692) | 565 (241–1342) | 585 (249–1279) | 543 (233–1255) | 692 (338–1527) | <.001
Haemoglobin (g/dL) | 13.0 ± 1.7 | 13.4 ± 1.7 | 13.4 ± 1.7 | 13.4 ± 1.7 | 13.4 ± 1.8 | 13.4 ± 1.8 | .092
Sodium (mmol/L) | 140.5 ± 3.9 | 140.5 ± 3.1 | 140.5 ± 2.9 | 140.3 ± 3.1 | 140.1 ± 3.3 | 139.3 ± 4.0 | <.001
Serum creatinine (umol/L) | 98.9 ± 30.2 | 94.6 ± 29.6 | 92.7 ± 28.2 | 96.8 ± 29.8 | 104.6 ± 34.6 | 114.6 ± 36.9 | <.001
Comorbidities
Diabetes mellitus | 113 (39.0) | 703 (35.7) | 1985 (37.1) | 1773 (38.5) | 591 (44.6) | 98 (54.1) | <.001
Hypertension | 265 (91.4) | 1822 (92.6) | 4860 (90.7) | 4182 (90.9) | 1219 (91.9) | 164 (90.6) | .43
Myocardial infarction | 44 (15.2) | 355 (18.0) | 1155 (21.6) | 978 (21.3) | 322 (24.3) | 40 (22.1) | <.001
Atrial fibrillation/flutter | 163 (56.2) | 979 (49.8) | 2484 (46.4) | 2044 (44.4) | 574 (43.3) | 80 (44.2) | <.001
Strokec | 41 (14.2) | 269 (13.7) | 556 (10.4) | 466 (10.1) | 137 (10.3) | 26 (14.4) | .003
Drugs and devices
Angiotensin-converting enzyme inhibitor/angiotensin receptor blocker/angiotensin receptor–neprilysin inhibitor | 187 (64.5) | 1228 (62.4) | 3547 (66.2) | 3205 (69.7) | 950 (71.6) | 130 (71.8) | <.001
Beta-blocker | 209 (72.1) | 1423 (72.3) | 3901 (72.9) | 3455 (75.1) | 985 (74.3) | 114 (63.0) | .17
Sodium–glucose cotransporter 2 inhibitor | 19 (6.6) | 84 (4.3) | 201 (3.8) | 153 (3.3) | 39 (2.9) | 9 (5.0) | .007
Mineralocorticoid receptor antagonist | 17 (5.9) | 127 (6.5) | 486 (9.1) | 597 (13.0) | 274 (20.7) | 41 (22.7) | <.001
Calcium channel blockers | 115 (40.0) | 895 (45.5) | 2113 (39.5) | 1686 (36.6) | 477 (36.0) | 50 (27.6) | <.001
Anticoagulant agent | 137 (47.2) | 770 (39.1) | 1843 (34.4) | 1457 (31.7) | 409 (30.8) | 57 (31.5) | <.001
Antiplatelet agent | 79 (27.2) | 672 (34.2) | 1731 (32.3) | 1434 (31.2) | 408 (30.8) | 51 (28.2) | .056
Diuretics | 279 (96.2) | 1867 (94.9) | 4880 (91.1) | 4169 (90.6) | 1219 (91.9) | 168 (92.8) | <.001
Loop diuretics | 158 (76.3) | 1167 (73.4) | 3339 (71.2) | 3003 (71.7) | 947 (72.7) | 136 (76.0) | .81
Dose of oral loop diuretics (furosemide equivalent dose) (mg/day)d | 40 (40–80) | 40 (40–80) | 40 (20–60) | 40 (20–40) | 40 (20–40) | 40 (40–40) | <.001
Thiazide diuretics | 63 (30.4) | 408 (25.6) | 1035 (22.1) | 902 (21.5) | 262 (20.1) | 30 (16.8) | <.001
Implantable cardioverter defibrillator | 4 (1.4) | 16 (0.8) | 30 (0.6) | 22 (0.5) | 5 (0.4) | 1 (0.6) | .026
Pacemaker | 35 (12.1) | 166 (8.4) | 414 (7.7) | 345 (7.5) | 106 (8.0) | 20 (11.1) | .26

Patients with higher potassium levels were more frequently treated with renin-angiotensin system inhibitors, an angiotensin receptor–neprilysin inhibitor (ARNI), and, especially, MRAs. Loop diuretic use was only slightly more common in patients with low versus higher potassium concentrations, although loop diuretic dose and use of thiazide diuretics were greater in patients with lower potassium levels.

HFpEF

Among individuals with HFpEF, those with higher potassium levels were more often male, had a higher prevalence of diabetes and myocardial infarction, and a greater proportion in NYHA class III/IV, but a lower prevalence of atrial fibrillation/flutter, compared to those with lower potassium concentrations. Patients with higher potassium levels had a lower average eGFR, sodium concentration, and NT-proBNP level than those with lower potassium concentrations. People with higher potassium levels were more likely to be treated with a renin-angiotensin system blocker and MRA, but less likely to be treated with a calcium channel blocker and thiazide diuretic (although the median dose of oral loop diuretic was similar across all potassium groups).

Clinical outcomes according to potassium level in HFrEF and HFpEF

The incidence rates for each outcome of interest, according to serum potassium concentration, illustrated as a continuous variable, are shown in Figures 1 and Figure 2 and Supplementary data online, Figures S3–S5. Cumulative incidence rate plots for each outcome of interest, according to potassium category, are depicted in Supplementary data online, Figures S6 and S7, and the results of the time-to-first-event analyses are shown in Table 2. The HRs for each outcome, according to potassium level shown as a continuous variable, are illustrated in Supplementary data online, Figures S8–S11.

Figure 1: Incidence rate for each outcome across the range of potassium levels in HFrEF (analysed as a continuous variable). (A) All-cause death, (B) CV death, (C) Sudden death, (D) Pump failure death, (E) Non-CV death, (F) First HF hospitalization, (G) First HF hospitalization or all-cause death, and (H) First HF hospitalization or CV death. The X axis shows potassium, and the Y axis shows the unadjusted incidence rate for each outcome. The solid line shows a continuous incidence rate, and the interrupted lines on either side of this show the 95% CI. All the incidence rates are shown per 100 patient-years. CV indicates cardiovascular, HF, heart failure; and HFrEF, heart failure with reduced ejection fraction

Figure 2: Incidence rate for each outcome across the range of potassium levels in HFpEF (analysed as a continuous variable). (A) All-cause death, (B) CV death, (C) Sudden death, (D) Pump failure death, (E) Non-CV death, (F) First HF hospitalization, (G) First HF hospitalization or all-cause death, and (H) First HF hospitalization or CV death. The X axis shows potassium, and the Y axis shows the unadjusted incidence rate for each outcome. The solid line shows a continuous incidence rate, and the interrupted lines on either side of this show the 95% CI. All the incidence rates are shown per 100 patient-years. CV indicates cardiovascular, HF, heart failure; and HFpEF, heart failure with preserved ejection fraction

 | Crude incidence rate, 100 patients-year (95% CI) | Unadjusted hazard ratiosa(95% CI) | Adjusted hazard ratiosb(95% CI)
 | HFrEF (LVEF ≤ 40%)
All-cause death
K < 3.5 (n = 502) | 17.7 (15.1–20.6) | 1.86 (1.58–2.18) | 1.49 (1.27–1.76)
K ≥ 3.5–<4.0 (n = 3301) | 11.1 (10.3–11.9) | 1.19 (1.10–1.29) | 1.07 (0.98–1.16)
K ≥ 4.0–<4.5 (n = 11 511) | 9.4 (9.0–9.7) | Reference | Reference
K ≥ 4.5–<5.0 (n = 11 963) | 9.3 (9.0–9.7) | 0.98 (0.93–1.04) | 0.94 (0.89–1.00)
K ≥ 5.0–<5.5 (n = 4119) | 10.5 (9.9–11.2) | 1.07 (0.99–1.16) | 0.96 (0.89–1.04)
K ≥ 5.5 (n = 950) | 13.8 (12.2–15.6) | 1.32 (1.16–1.51) | 1.02 (0.89–1.16)
Cardiovascular death
K < 3.5 | 15.4 (13.0–18.1) | 1.98 (1.67–2.36) | 1.58 (1.32–1.88)
K ≥ 3.5–<4.0 | 9.4 (8.7–10.1) | 1.23 (1.13–1.34) | 1.10 (1.00–1.20)
K ≥ 4.0–<4.5 | 7.6 (7.3–8.0) | Reference | Reference
K ≥ 4.5–<5.0 | 7.4 (7.1–7.8) | 0.96 (0.90–1.02) | 0.92 (0.87–0.98)
K ≥ 5.0–<5.5 | 8.5 (7.9–9.1) | 1.07 (0.98–1.16) | 0.96 (0.88–1.04)
K ≥ 5.5 | 10.4 (9.0–11.9) | 1.25 (1.08–1.45) | 0.95 (0.81–1.11)
Non-cardiovascular death
K < 3.5 | 2.0 (1.2–3.1) | 1.36 (0.84–2.18) | 1.15 (0.70–1.91)
K ≥ 3.5–<4.0 | 1.4 (1.1–1.7) | 0.97 (0.78–1.22) | 0.92 (0.73–1.16)
K ≥ 4.0–<4.5 | 1.4 (1.3–1.6) | Reference | Reference
K ≥ 4.5–<5.0 | 1.6 (1.4–1.8) | 1.09 (0.95–1.26) | 1.03 (0.89–1.19)
K ≥ 5.0–<5.5 | 1.6 (1.3–1.9) | 1.05 (0.86–1.28) | 0.93 (0.76–1.14)
K ≥ 5.5 | 2.6 (2.0–3.5) | 1.61 (1.19–2.18) | 1.26 (0.92–1.72)
Sudden death
K < 3.5 | 5.4 (4.1–7.1) | 1.87 (1.40–2.50) | 1.58 (1.18–2.12)
K ≥ 3.5–<4.0 | 3.4 (3.0–3.8) | 1.18 (1.02–1.37) | 1.07 (0.92–1.24)
K ≥ 4.0–<4.5 | 2.9 (2.7–3.1) | Reference | Reference
K ≥ 4.5–<5.0 | 2.8 (2.6–3.0) | 0.99 (0.90–1.10) | 0.98 (0.88–1.09)
K ≥ 5.0–<5.5 | 2.8 (2.4–3.1) | 0.98 (0.85–1.14) | 0.93 (0.80–1.07)
K ≥ 5.5 | 3.5 (2.7–4.4) | 1.24 (0.96–1.61) | 1.05 (0.81–1.36)
Pump failure death
K < 3.5 | 6.4 (4.9–8.2) | 2.38 (1.82–3.12) | 1.71 (1.29–2.25)
K ≥ 3.5–<4.0 | 3.4 (3.0–3.9) | 1.38 (1.19–1.60) | 1.15 (0.98–1.33)
K ≥ 4.0–<4.5 | 2.5 (2.3–2.7) | Reference | Reference
K ≥ 4.5–<5.0 | 2.3 (2.2–2.5) | 0.90 (0.81–1.01) | 0.86 (0.76–0.96)
K ≥ 5.0–<5.5 | 2.9 (2.6–3.3) | 1.04 (0.89–1.20) | 0.92 (0.79–1.07)
K ≥ 5.5 | 3.5 (2.8–4.5) | 1.12 (0.87–1.45) | 0.84 (0.65–1.09)
First heart failure hospitalization
K < 3.5 | 17.2 (14.5–20.4) | 1.66 (1.39–1.98) | 1.22 (1.02–1.46)
K ≥ 3.5–<4.0 | 12.3 (11.5–13.2) | 1.28 (1.18–1.39) | 1.13 (1.04–1.23)
K ≥ 4.0–<4.5 | 9.5 (9.1–9.9) | Reference | Reference
K ≥ 4.5–<5.0 | 8.9 (8.6–9.3) | 0.91 (0.85–0.96) | 0.88 (0.83–0.94)
K ≥ 5.0–<5.5 | 10.0 (9.4–10.8) | 0.97 (0.90–1.05) | 0.92 (0.84–1.00)
K ≥ 5.5 | 11.3 (9.8–13.0) | 0.98 (0.85–1.14) | 0.81 (0.70–0.95)
First heart failure hospitalization or all-cause death
K < 3.5 | 29.3 (25.7–33.4) | 1.77 (1.55–2.02) | 1.37 (1.20–1.58)
K ≥ 3.5–<4.0 | 19.6 (18.5–20.7) | 1.24 (1.16–1.32) | 1.11 (1.03–1.18)
K ≥ 4.0–<4.5 | 15.8 (15.3–16.3) | Reference | Reference
K ≥ 4.5–<5.0 | 15.4 (14.9–15.9) | 0.96 (0.91–1.00) | 0.92 (0.88–0.97)
K ≥ 5.0–<5.5 | 17.4 (16.5–18.4) | 1.04 (0.98–1.11) | 0.96 (0.90–1.02)
K ≥ 5.5 | 21.6 (19.4–23.9) | 1.19 (1.07–1.33) | 0.95 (0.85–1.06)
First heart failure hospitalization or cardiovascular death
K < 3.5 | 26.8 (23.4–30.7) | 1.77 (1.54–2.04) | 1.36 (1.18–1.58)
K ≥ 3.5–<4.0 | 18.1 (17.1–19.2) | 1.26 (1.18–1.35) | 1.12 (1.04–1.20)
K ≥ 4.0–<4.5 | 14.4 (13.9–14.9) | Reference | Reference
K ≥ 4.5–<5.0 | 13.9 (13.4–14.3) | 0.94 (0.90–0.99) | 0.91 (0.87–0.96)
K ≥ 5.0–<5.5 | 15.6 (14.8–16.5) | 1.03 (0.96–1.10) | 0.95 (0.89–1.02)
K ≥ 5.5 | 18.5 (16.5–20.7) | 1.12 (1.00–1.26) | 0.89 (0.79–1.01)
 | HFpEF (LVEF ≥ 50%)
All-cause death
K < 3.5 (n = 290) | 7.4 (5.7–9.5) | 1.30 (1.00–1.70) | 1.14 (0.87–1.48)
K ≥ 3.5–<4.0 (n = 1968) | 5.8 (5.3–6.5) | 1.11 (0.98–1.26) | 1.07 (0.94–1.21)
K ≥ 4.0–<4.5 (n = 5356) | 5.1 (4.8–5.5) | Reference | Reference
K ≥ 4.5–<5.0 (n = 4602) | 5.3 (4.9–5.7) | 1.06 (0.96–1.17) | 1.03 (0.93–1.13)
K ≥ 5.0–<5.5 (n = 1326) | 5.8 (5.1–6.5) | 1.21 (1.05–1.40) | 1.05 (0.91–1.22)
K ≥ 5.5 (n = 181) | 9.0 (6.8–11.9) | 1.92 (1.43–2.57) | 1.29 (0.96–1.74)
Cardiovascular death
K < 3.5 | 3.9 (2.8–5.6) | 1.31 (0.91–1.88) | 1.15 (0.79–1.65)
K ≥ 3.5–<4.0 | 3.5 (3.1–4.1) | 1.21 (1.03–1.43) | 1.19 (1.01–1.41)
K ≥ 4.0–<4.5 | 2.8 (2.6–3.1) | Reference | Reference
K ≥ 4.5–<5.0 | 3.0 (2.7–3.3) | 1.07 (0.94–1.22) | 1.05 (0.91–1.20)
K ≥ 5.0–<5.5 | 3.7 (3.2–4.3) | 1.36 (1.14–1.64) | 1.17 (0.97–1.42)
K ≥ 5.5 | 5.4 (3.8–7.8) | 2.01 (1.38–2.93) | 1.25 (0.84–1.85)
Non-cardiovascular death
K < 3.5 | 3.1 (2.1–4.6) | 1.48 (0.98–2.23) | 1.30 (0.86–1.96)
K ≥ 3.5–<4.0 | 1.9 (1.6–2.3) | 1.01 (0.82–1.26) | 0.95 (0.76–1.18)
K ≥ 4.0–<4.5 | 1.9 (1.7–2.1) | Reference | Reference
K ≥ 4.5–<5.0 | 2.0 (1.7–2.2) | 1.08 (0.92–1.27) | 1.02 (0.86–1.20)
K ≥ 5.0–<5.5 | 1.8 (1.4–2.2) | 1.05 (0.81–1.35) | 0.91 (0.70–1.18)
K ≥ 5.5 | 2.6 (1.6–4.4) | 1.55 (0.91–1.20) | 1.18 (0.68–2.03)
Sudden death
K < 3.5 | 0.7 (0.3–1.6) | 0.75 (0.33–1.69) | 0.71 (0.31–1.62)
K ≥ 3.5–<4.0 | 0.9 (0.7–1.2) | 0.97 (0.71–1.31) | 0.96 (0.70–1.30)
K ≥ 4.0–<4.5 | 1.0 (0.8–1.1) | Reference | Reference
K ≥ 4.5–<5.0 | 1.2 (1.0–1.3) | 1.20 (0.97–1.50) | 1.17 (0.94–1.46)
K ≥ 5.0–<5.5 | 1.2 (0.9–1.6) | 1.29 (0.94–1.77) | 1.06 (0.76–1.47)
K ≥ 5.5 | 2.6 (1.6–4.4) | 2.76 (1.59–4.77) | 1.87 (1.07–3.28)
Pump failure death
K < 3.5 | 1.6 (0.9–2.8) | 1.79 (1.01–3.17) | 1.47 (0.83–2.63)
K ≥ 3.5–<4.0 | 1.1 (0.9–1.5) | 1.39 (1.04–1.86) | 1.28 (0.95–1.73)
K ≥ 4.0–<4.5 | 0.8 (0.7–1.0) | Reference | Reference
K ≥ 4.5–<5.0 | 0.8 (0.6–0.9) | 0.97 (0.75–1.25) | 0.98 (0.75–1.27)
K ≥ 5.0–<5.5 | 0.9 (0.7–1.2) | 1.22 (0.85–1.76) | 1.10 (0.76–1.60)
K ≥ 5.5 | 2.1 (1.1–3.7) | 2.83 (1.53–5.24) | 1.46 (0.77–2.79)
First heart failure hospitalization
K < 3.5 | 8.3 (6.4–10.7) | 1.25 (0.97–1.63) | 1.09 (0.84–1.42)
K ≥ 3.5–<4.0 | 7.4 (6.7–8.2) | 1.24 (1.10–1.39) | 1.14 (1.01–1.28)
K ≥ 4.0–<4.5 | 5.8 (5.5–6.2) | Reference | Reference
K ≥ 4.5–<5.0 | 5.5 (5.2–6.0) | 0.97 (0.88–1.07) | 0.96 (0.87–1.06)
K ≥ 5.0–<5.5 | 6.2 (5.4–7.0) | 1.14 (0.98–1.32) | 1.00 (0.86–1.17)
K ≥ 5.5 | 7.9 (5.7–10.9) | 1.44 (1.03–2.00) | 0.99 (0.70–1.39)
First heart failure hospitalization or all-cause death
K < 3.5 | 13.6 (11.1–16.5) | 1.29 (1.05–1.58) | 1.13 (0.92–1.39)
K ≥ 3.5–<4.0 | 11.1 (10.2–12.0) | 1.15 (1.04–1.26) | 1.08 (0.98–1.19)
K ≥ 4.0–<4.5 | 9.4 (9.0–9.9) | Reference | Reference
K ≥ 4.5–<5.0 | 9.4 (8.8–9.9) | 1.01 (0.94–1.09) | 0.99 (0.91–1.07)
K ≥ 5.0–<5.5 | 10.3 (9.3–11.4) | 1.17 (1.04–1.31) | 1.02 (0.91–1.15)
K ≥ 5.5 | 14.5 (11.4–18.4) | 1.63 (1.28–2.09) | 1.15 (0.89–1.48)
First heart failure hospitalization or cardiovascular death
K < 3.5 | 10.9 (8.8–13.6) | 1.29 (1.03–1.63) | 1.13 (0.90–1.43)
K ≥ 3.5–<4.0 | 9.4 (8.6–10.3) | 1.20 (1.08–1.34) | 1.13 (1.02–1.26)
K ≥ 4.0–<4.5 | 7.6 (7.2–8.1) | Reference | Reference
K ≥ 4.5–<5.0 | 7.5 (7.1–8.0) | 1.00 (0.92–1.09) | 0.99 (0.91–1.08)
K ≥ 5.0–<5.5 | 8.7 (7.8–9.7) | 1.22 (1.08–1.38) | 1.07 (0.94–1.21)
K ≥ 5.5 | 11.3 (8.6–14.8) | 1.56 (1.18–2.05) | 1.07 (0.80–1.43)

Clinical outcomes according to potassium concentration in HFrEF

Mortality

In HFrEF, the highest cumulative incidence rate of all-cause death was observed among those in the lowest potassium category (i.e. potassium <3.5 mmol/L) and the lowest incidence rate was observed among those in the potassium category ≥4.5–<5.0 mmol/L (Table 2). Compared to those in the potassium category ≥4.0–<4.5 (reference group), the aHR for all-cause death were 1.49 (95% CI 1.27–1.76), 1.07 (95% CI 0.98–1.16), 0.94 (95% CI 0.89–1.00), 0.96 (95% CI 0.89–1.04), and 1.02 (95% CI 0.89–1.16) in potassium categories <3.5 mmol/L, ≥3.5–<4.0 mmol/L, ≥4.5–<5.0 mmol/L, ≥5.0–<5.5 mmol/L, and ≥5.5 mmol/L, respectively (Table 2 and Supplementary data online, Figure S6). No meaningful differences were observed in the estimates between the two-stage meta-analysis and the one-stage meta-analysis stratified by trial; therefore, the results of the one-stage meta-analysis stratified by trial are presented in the following section (see Supplementary data online, Table S4). The significantly higher risk in patients with serum potassium level <3.5 mmol/L was observed even after further adjustment for the dose of oral loop diuretic (see Supplementary data online, Table S5). When potassium level was analysed as a continuous variable, the unadjusted spline curve for the incidence rate of all-cause death showed a reverse J-curve relationship (Figure 1). The corresponding spline curves for HRs are shown in Supplementary data online, Figures S8 and S9.

The association between potassium concentration and cardiovascular death was similar to that observed for all-cause death (Figure 1, Supplementary data online, Figures S8 and S9). Incidence rates and aHR for both sudden death and pump failure death were also highest in the lowest potassium category (potassium <3.5 mmol/L) (Table 2). Examination of these outcomes according to continuous potassium concentration demonstrated a reverse J-curve association, with significantly elevated aHR for potassium <4.5 mmol/L, while the association with hyperkalaemia was attenuated and not statistically significant (see Supplementary data online, Figure S8).

By contrast, the incidence of death from non-cardiovascular causes was low and varied little across potassium categories (Table 2 and Figure 1).

First heart failure hospitalization and composite outcomes

The association between serum potassium concentration and the incidence of HF hospitalization was very similar to the association observed for all-cause death and cardiovascular death, and this was also seen for composites of HF hospitalization and all-cause and cardiovascular mortality (Table 2 and Figure 1).

Serum potassium level and clinical outcomes stratified by kidney function

Patients with an eGFR <60 mL/min/1.73 m2 had higher incidence rates for all outcomes compared to those with an eGFR ≥60 mL/min/1.73 m2 (see Supplementary data online, Figure S3). Across all-cause death, cardiovascular death, non-cardiovascular death, the composites of HF hospitalization and all-cause and cardiovascular mortality, there were significant interactions across strata of eGFR (see Supplementary data online, Table S6). In patients with eGFR <60 mL/min/1.73 m2, the serum potassium level (4.8 mmol/L) corresponding to the nadir of the incidence rate for these outcomes was higher than among those with an eGFR ≥60 mL/min/1.73 m2 (4.3–4.4 mmol/L). However, in both strata, these potassium levels fell within the overall optimal range (4.2–5.0 mmol/L).

Serum potassium level and clinical outcomes stratified by treatment arm or MRA use

In categorical analyses of serum potassium, no significant interactions were observed with either treatment arm or MRA use (see Supplementary data online, Tables S7 and S8). When serum potassium was analysed as a continuous variable, no interaction was observed with treatment arm, but significant interactions by MRA use were observed for all-cause death and for cardiovascular death, with a higher nadir of potassium concentration for the incidence of these events in the MRA group compared with the no MRA group (see Supplementary data online, Figure S5). However, the difference in the serum potassium level corresponding to the nadir of the incidence rate between patients receiving MRAs and those not receiving MRAs was small (4.5 vs 4.4 mmol/L) (see Supplementary data online, Table S6).

Potassium concentration associated with lowest risk

For all-cause death, binary recursive partitioning identified 3.7–5.0 mmol/L as the lowest-risk range, whereas potassium associated with the nadir of the aHR was 4.7 mmol/L and for nadir of the incidence rate it was 4.4 mmol/L (see Supplementary data online, Table S9). When considering all clinical outcomes, the overlapping range associated with the lowest risk was 4.2–5.0 mmol/L.

Clinical outcomes according to potassium concentration in HFpEF

Mortality

Compared to HFrEF, crude incidence rates were substantially lower across all potassium categories. In contrast to HFrEF, the cumulative incidence rate for all-cause death in HFpEF was highest in those in the highest potassium category (i.e. potassium ≥5.5 mmol/L) (see Supplementary data online, Figure S7). The aHR for all-cause death compared with potassium ≥4.0–<4.5 (reference category) were 1.14 (95% CI 0.87–1.48), 1.07 (95% CI 0.94–1.21), 1.03 (95% CI 0.93–1.13), 1.05 (95% CI 0.91–1.22), and 1.29 (95% CI 0.96–1.74) in potassium categories <3.5 mmol/L, ≥ 3.5–<4.0 mmol/L, ≥ 4.5–<5.0 mmol/L, ≥5.0–<5.5 mmol/L, and ≥5.5 mmol/L, respectively (Table 2). Examination of potassium level as a continuous variable showed a flattened U-shape relationship for all-cause mortality, with relatively little variation across the baseline potassium level (Figure 2, Supplementary data online, Figures S10 and S11). The picture was similar for cardiovascular death and pump failure death, with an even flatter relationship for sudden death, the incidence of which was very low (Table 2 and Figure 2).

First heart failure hospitalization and composite outcomes

There was a flattened U-shape association between first HF hospitalization, and the composite outcomes analysed, in patients with HFpEF (Figure 2 and Supplementary data online, Figure S10).

Serum potassium level and clinical outcomes stratified by kidney function

Similar to HFrEF, HFpEF patients with an eGFR <60 mL/min/1.73 m2 had higher incidence rates for all outcomes compared to those with an eGFR ≥60 mL/min/1.73 m2, but the associations between potassium concentration and outcomes were similar in each eGFR stratum, with no significant interaction (see Supplementary data online, Figure S4 and Supplementary data online, Table S5).

Serum potassium level and clinical outcomes stratified by treatment arm or MRA use

In analyses of serum potassium levels and clinical outcomes stratified by treatment arm or MRA use, no significant interactions were observed with either treatment arm or MRA use in the categorical analyses (see Supplementary data online, Tables S10 and S11). Similarly, when serum potassium was analysed as a continuous variable to estimate incidence rates, no significant interactions with treatment arm or MRA use were observed (see Supplementary data online, Table S6).

Potassium concentration associated with lowest risk

In HFpEF, the nadir of the aHR and incidence rate converged around a potassium concentration of 4.3–4.6 mmol/L, while binary recursive partitioning produced broader boundaries, e.g. 3.8–5.3 mmol/L for all-cause death (see Supplementary data online, Table S9). Considering all clinical outcomes, the overlapping range associated with the lowest risk was 4.2–5.1 mmol/L.

Serious adverse events related to hyperkalaemia and hypokalaemia across trials

Serious adverse events (SAEs) related to hyperkalaemia were reported in 203 of 26 515 patients (0.8%) in the control group and 238 of 28 878 patients (0.8%) in the intervention group; among these, fatal SAEs occurred in 4 (0.02%) and 10 (0.03%) patients, respectively.

Similarly, SAEs related to hypokalaemia were reported in 108 of 26 515 patients (0.4%) in the control group and 75 of 28 878 patients (0.3%) in the intervention group; among these, fatal SAEs occurred in 1 patient in each group (0.004% in the control and 0.003% in the intervention group) (see Supplementary data online, Table S12).