Article Figures & Data
Figures
- Figure 1.
The course of disease may mediate (top diagram) or confound (bottom diagram) the association between duration of predialysis care and mortality after dialysis start. Causal diagrams representing two possible scenarios of the influence of the course of disease on the association between duration of predialysis care and mortality after dialysis start. (A) The course of disease is an effect mediator on the causal pathway between duration of predialysis care and mortality after dialysis start. Under this assumption, the total effect of duration of predialysis care is estimated ignoring the influence of markers of disease course. (B) The course of disease may be both a mediator and a confounder of the relationship of interest. Conditioning the analysis on markers of disease course (adjustment by regression or stratification) in these situations could introduce collider bias (course of disease is a collider on the causal path: duration of predialysis care → course of disease ← unknown factors → mortality after dialysis start). Marginal methods have been proposed to control for this form of confounding and minimize collider bias (21).
- Figure 2.
The association between duration of predialysis care and mortality after dialysis start differed by setting of dialysis start. Unadjusted mortality by duration of predialysis care and setting of dialysis start. The numbers of deaths for people with 0, 1–119, 120–364, and ≥365 days of predialysis care were 200, 76, 78, and 456, respectively. Error bars represent SEM.
- Figure 3.
The survival benefit associated with longer predialysis care was weaker and no longer significant after accounting for markers of disease course. Hazard ratios (HRs) and 95% confidence intervals (95% CIs) for mortality associated with duration of predialysis care. Models included all participants (A). Models included participants with predialysis care (B). (1) Standard Cox model adjusted for age at initiation of dialysis, sex, dialysis programs, the last available serum albumin before starting dialysis, and ten comorbidities (diabetes, congestive heart failure, cancer, other cardiac disease, cerebrovascular disease, coronary artery disease, peripheral vascular disease, chronic obstructive lung disease, polycystic kidney disease, and gastrointestinal bleeding). (2) Inverse probability of treatment weighted Cox model accounting for the same covariates as in the above Cox model and markers of disease course (inpatient or outpatient of dialysis start and the last available eGFR before starting dialysis). Analyses of inverse probabilities of treatment weighting are reported in Supplemental Figure 1 and Supplemental Tables 2 and 3.
Tables
Characteristic Overall, n=3152 Predialysis Care Any Predialysis Care, d No, n=738 Any, n=2414 1–119, n=264 120–364, n=305 ≥365, n=1845 Demographics Age, yr 64±15 63±16 65±15 63±16 64±16 65±15 Age ≥65 yr 1726 (55) 377 (51) 1349 (56) 140 (53) 153 (50) 1056 (57) Men 1928 (61) 472 (64) 1456 (60) 152 (58) 195 (64) 1109 (60) BMI, kg/m2 28.4±7.2 28.3±7.7 28.4±7.0 27.7±7.4 27.5±6.4 28.6±7.1 Presence of comorbidities Diabetes 1656 (53) 292 (40) 1364 (57) 135 (51) 174 (57) 1055 (57) Coronary artery disease 1067 (34) 251 (34) 816 (34) 77 (29) 95 (31) 644 (35) Congestive heart failure 937 (30) 248 (34) 689 (29) 72 (27) 87 (29) 530 (29) Other cardiac disease 907 (29) 273 (37) 634 (26) 58 (22) 85 (28) 491 (27) Cerebrovascular disease 490 (16) 100 (14) 390 (16) 32 (12) 45 (15) 313 (17) Peripheral vascular disease 532 (17) 121 (16) 411 (17) 42 (16) 49 (16) 320 (17) Cancer 633 (20) 173 (23) 460 (19) 74 (28) 63 (21) 323 (18) Chronic obstructive lung disease 215 (7) 60 (8) 155 (6) 15 (6) 14 (5) 126 (7) Polycystic kidney disease 111 (4) 3 (0.4) 108 (4) 4 (2) 5 (2) 99 (5) Gastrointestinal bleeding 299 (9) 91 (12) 208 (9) 20 (8) 20 (7) 168 (9) Setting of dialysis starts Outpatient 1400 (44) 15 (2)a 1385 (57) 111 (42) 169 (55) 1105 (60) Inpatient 1752 (56) 723 (98) 1029 (43) 153 (58) 136 (45) 740 (40) Modality of dialysis initiation CRRT 228 (7) 182 (25) 46 (2) 5 (2) 5 (2) 36 (2) HD 2302 (73) 543(74) 1759 (73) 234 (89) 237 (78) 1288 (70) PD 622 (20) 13 (2) 609 (25) 25 (9) 63 (21) 521 (28) Laboratory measurements eGFR, ml/min per 1.73 m2 7.5 [5.6–10.2] 8.2 [5.2–14.7] 7.3 [5.7–9.6] 7.2 [5.5–9.7] 7.3 [5.6–9.6] 7.4 [5.7–9.6] eGFR, ml/min per 1.73 m2, categoryb <10 2310 (74) 430 (59) 1880 (78) 201 (76) 237 (78) 1442 (78) 10–15 532 (17) 121 (17) 411 (17) 41 (16) 47 (15) 323 (18) >15 296 (9) 178 (24) 118 (5) 21 (8) 20 (7) 77 (4) Missing 14 9 5 1 1 3 Serum albumin, g/dl 3.2±0.7 2.7±0.7 3.3±0.7 3.1±0.7 3.2±0.7 3.3±0.7 Serum albuminb <3.6 g/dl 2070 (67) 613 (88) 1457 (61) 181 (71) 197 (65) 1079 (59) Missing 78 42 36 8 4 24 Hemoglobin, g/dl 9.7±1.6 9.3±1.7 9.8±1.6 9.3±1.7 9.7±1.6 9.9±1.5 Values for categorical variables are given as count (percentage); values for continuous variables are given as mean±SD or median [interquartile range]. BMI, body mass index; CRRT, continuous RRT; HD, hemodialysis; PD, peritoneal dialysis.
↵a These 15 patients had no outpatient nephrology care before starting dialysis, but they were seen by a nephrologist in an emergency department and stable enough to start dialysis in an outpatient setting without being admitted to hospital.
↵b Values do not sum up to subtotal due to missing data.
Supplemental Data
In this issue
