Prognostic Impact of Baseline Hemoglobin Levels on Long-Term Thrombotic and Bleeding Events After Percutaneous Coronary Interventions.

Background Association of baseline hemoglobin levels with long-term adverse events after percutaneous coronary interventions has not been yet thoroughly defined. We aimed to assess the clinical impact of baseline hemoglobin on long-term ischemic and bleeding risk after percutaneous coronary intervention. Methods and Results Using the pooled individual patient-level data from the 3 percutaneous coronary intervention studies, we categorized 19 288 patients into 4 groups: high-normal hemoglobin (≥14.0 g/dL; n=7555), low-normal hemoglobin (13.0-13.9 g/dL in men and 12.0-13.9 g/dL in women; n=5303), mild anemia (11.0-12.9 g/dL in men and 11.0-11.9 g/dL in women; n=4117), and moderate/severe anemia (<11.0 g/dL; n=2313). Median follow-up duration was 3 years. Low-normal hemoglobin, mild anemia, and moderate/severe anemia correlated with significant excess risk relative to high-normal hemoglobin for GUSTO (Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Arteries Trial) moderate/severe bleeding, with adjusted hazard ratios of 1.22 (95% CI, 1.04-1.44), 1.73 (95% CI, 1.47-2.04), and 2.31 (95% CI, 1.92-2.78), respectively. Moderate/severe anemia also correlated with significant excess risk relative to high-normal hemoglobin for the ischemic composite end point of myocardial infarction/ischemic stroke (adjusted hazard ratio, 1.33; 95% CI, 1.11-1.60), whereas low-normal hemoglobin and mild anemia did not. However, the excess risk of low-normal hemoglobin, mild anemia, and moderate/severe anemia relative to high-normal hemoglobin remained significant for ischemic stroke and for mortality. Conclusions Decreasing baseline hemoglobin correlated with incrementally higher long-term risk for major bleeding, ischemic stroke, and mortality after percutaneous coronary intervention. Even within normal range, lower baseline hemoglobin level correlated with higher ischemic and bleeding risk.

Clinical Perspective

What Is New?

One third of patients undergoing percutaneous coronary intervention had anemia, as defined per the World Health Organization criteria.

Moderate/severe anemia (hemoglobin <10.9 g/dL) was associated with a markedly higher risk for long‐term ischemic and bleeding events, as well as mortality.

Even mild anemia (hemoglobin 11.0–12.9 g/dL for men and 11.0–11.9 g/dL for women) and low‐normal hemoglobin (hemoglobin 13.0–13.9 g/dL for men and 12.0–13.9 g/dL for women) correlated with significantly higher risk for ischemic stroke, serious bleeding events, such as GUSTO (Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Arteries Trial) moderate bleeding, and mortality compared with high‐normal hemoglobin (hemoglobin ≥14.0 g/dL).

What Are the Clinical Implications?

Given the excess bleeding and ischemic risk of even a small decline in hemoglobin at the threshold of 14 g/dL, we might as well pay attention to the preprocedural hemoglobin value as an indicator of long‐term clinical outcomes in patients planned for percutaneous coronary intervention.

Future studies would be warranted to address the optimal antithrombotic therapy in patients with anemia who have both high ischemic and bleeding risk.

Introduction

Anemia is highly prevalent in patients with cardiovascular diseases.1, 2, 3 Among patients undergoing percutaneous coronary interventions (PCIs), preexisting anemia is known to correlate with a higher risk of short‐ and long‐term mortality,4, 5 major adverse cardiovascular events,6 and major in‐hospital bleeding complications.7, 8 To date, most studies have used the conventional World Health Organization (WHO) thresholds of anemia (<13.0 g/dL for men and <12.0 g/dL for women)7, 9; however, as the WHO definition of anemia encompasses a wide range of hemoglobin values, the severity of anemia (ie, mild, moderate, or severe) should be taken into consideration for the precise risk estimation. In addition, the threshold hemoglobin value correlating with the increased ischemic and/or bleeding risk after PCI could be higher than the WHO criteria of anemia; the different levels of baseline hemoglobin value, even among patients without anemia, might influence on the long‐term ischemic and bleeding outcome. Hence, this study aims to comprehensively assess the clinical impact of the levels of baseline hemoglobin on long‐term ischemic and bleeding risk after PCI.

Methods

The data, analytical methods, and study materials will not be made available to other researchers for purposes of reproducing the results or replicating the procedure.

Study Population

Using the pooled individual patient‐level data, we constructed a pooled database of 3 Japanese PCI studies, which were conducted after the introduction of a drug‐eluting stent (DES): CREDO‐Kyoto (Coronary Revascularization Demonstrating Outcome Study in Kyoto) PCI/CABG (Coronary Artery Bypass Grafting) Registry Cohort‐2,10 RESET (Randomized Evaluation of Sirolimus‐Eluting Versus Everolimus‐Eluting Stent Trial),11 and NEXT (NOBORI Biolimus‐Eluting Versus XIENCE/PROMUS Everolimus‐Eluting Stent Trial)12 (Figure 1). The design and major results of all studies have been described previously.10, 11, 12 CREDO‐Kyoto enrolled consecutive patients undergoing first PCI or CABG during the study period. In RESET and NEXT, patients scheduled for PCI with DES were to be enrolled without any exclusion criteria. Therefore, there were no exclusion criteria relevant to the current topic. The relevant review boards at all participating centers for each study approved each research protocol for the 3 studies. Because of retrospective enrollment, the requirement for written informed consent from patients was waived in the CREDO‐Kyoto PCI/CABG Registry Cohort‐2; however, we excluded those patients who refused participation in the study when contacted for follow‐up. Written informed consent was obtained from all study patients in RESET and NEXT. Among 19 489 patients undergoing PCI enrolled in all 3 studies combined, the present study population included 19 288 patients from 122 Japanese PCI centers, after excluding 201 patients whose baseline hemoglobin value was unavailable (Figure 1). The follow‐up durations were 5 years in the CREDO‐Kyoto PCI/CABG Registry Cohort‐2 and 3 years in the RESET and NEXT. In this study, the follow‐up was censored at 3 years to standardize the follow‐up duration across 3 studies.

Figure 1

Study flow. CABG indicates coronary artery bypass grafting; CREDO‐Kyoto PCI/CABG Registry, Coronary Revascularization Demonstrating Outcome Study in Kyoto Percutaneous Coronary Intervention/CABG Registry; NEXT, NOBORI Biolimus‐Eluting Versus XIENCE/PROMUS Everolimus‐Eluting Stent Trial; RESET, Randomized Evaluation of Sirolimus‐Eluting Versus Everolimus‐Eluting Stent Trial.

We attained procedural anticoagulation with unfractionated heparin following the local site protocols. Of note, neither glycoprotein IIb/IIIa inhibitor nor bivalirudin was used in any patient. The recommended antiplatelet regimen comprised aspirin (≥81 mg/d) indefinitely and thienopyridines (75 mg of clopidogrel or 200 mg of ticlopidine daily) for ≥3 months for DESs and ≥1 month for bare‐metal stents. However, the actual duration of dual‐antiplatelet therapy (DAPT) was left to the discretion of each attending physician. Likewise, duration of triple antithrombotic therapy of DAPT and warfarin was left to the discretion of each attending physician. The status of antiplatelet therapy was assessed throughout the follow‐up period using the same method across all 3 studies. We defined the discontinuation of DAPT as persistent, if either aspirin or thienopyridine was discontinued for ≥2 months.

Anemia was defined by the standard WHO classification: no anemia (hemoglobin ≥13.0 g/dL for men and ≥12.0 g/dL for women); mild anemia (hemoglobin 11.0–12.9 g/dL for men and 11.0–11.9 g/dL for women); moderate anemia (hemoglobin 8.0–10.9 g/dL); and severe anemia (hemoglobin <8.0 g/dL).9 Patients without anemia were further subdivided into 2 groups: high‐normal hemoglobin (hemoglobin ≥14.0 g/dL) and low‐normal hemoglobin (hemoglobin 13.0–13.9 g/dL for men and 12.0–13.9 g/dL for women). The cutoff value of hemoglobin 14.0 g/dL was selected considering the distribution of hemoglobin value and ease for clinical application. We compared the baseline characteristics and clinical outcomes across the 4 groups, including the high‐normal hemoglobin, low‐normal hemoglobin, mild anemia, and moderate/severe anemia groups.

Definition of Clinical Outcome Measures

In this study, the primary ischemic outcome measure was a composite of myocardial infarction (MI) and ischemic stroke (fatal or nonfatal), whereas the primary bleeding outcome measure was GUSTO (Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Coronary Arteries Trial) moderate/severe bleeding (severe, life‐threatening intracerebral bleeding or bleeding that caused substantial hemodynamic compromise needing treatment; and moderate, bleeding that needed transfusion).13 The secondary outcome measures included all‐cause death, cardiovascular death, noncardiovascular death, MI, definite stent thrombosis, stroke, ischemic stroke, hemorrhagic stroke, GUSTO severe bleeding, GUSTO moderate bleeding, gastrointestinal bleeding, intracranial bleeding, any coronary revascularization, and heart failure hospitalization. Intracranial bleeding included both hemorrhagic stroke and traumatic brain injury. Detailed definitions of the clinical events can be found in Data S1. An independent clinical event committee in each study adjudicated all the end point events.

Statistical Analysis

Categorical variables are expressed as number and percentage and compared across baseline hemoglobin groups using the χ2 test. Continuous variables are expressed as mean with SD or median with interquartile range and compared across baseline hemoglobin groups using the ANOVA or the Kruskal‐Wallis test, depending on their distributions. We estimated the cumulative 3‐year incidence with the Kaplan‐Meier method and assessed the differences across baseline hemoglobin groups by the log‐rank test. We also performed a landmark analysis at 30 days after index PCI. Surviving patients within 30 days were included for the analysis beyond 30 days. Using the multivariable Cox proportional hazard models, the risks of low‐normal hemoglobin, mild anemia, and moderate/severe anemia relative to high‐normal hemoglobin (reference) on the outcome measures were expressed as hazard ratios and their 95% CIs. We used a dummy code variable for low‐normal hemoglobin, mild anemia, and moderate/severe anemia to assess the hazard ratios relative to high‐normal hemoglobin. Corroborating our previous study,14 we included 34 clinically relevant factors (Table 1) as the risk‐adjusting variables and incorporated the centers and studies as the stratification variables in the multivariable Cox proportional hazard models. We also treated the 4 groups of anemic status as an ordinal variable and estimated the linear trend in the same multivariable Cox proportional hazard models. To determine the risks for hemorrhagic stroke and intracranial bleeding, we constructed parsimonious models with the clinically relevant 13 risk‐adjusting variables (Table 1) because of a small number of patients with event. Proportional hazard assumptions for the risk‐adjusting variables were assessed on the plots of log (time) versus log (−log [survival]), stratified by the variable and verified to be acceptable. We also conducted the subgroup analyses for the primary bleeding outcome measure. The same 34 risk‐adjusting variables used in the entire cohort were included in the multivariable Cox proportional hazard models. For the subgroup analyses stratified by warfarin use, atrial fibrillation, shock, malignancy, and platelet count, we constructed parsimonious models with the same 13 risk‐adjusting variables used in the parsimonious models in the entire cohort because of a small number of patients with outcome. Furthermore, we conducted a sensitivity analysis in which we combined high‐normal hemoglobin group and low‐normal hemoglobin group into one group (no anemia group), so that we could evaluate the risk of mild and moderate/severe anemia relative to no anemia, as defined per the conventional WHO criteria. We also performed another sensitivity analysis in which we divided the high‐normal hemoglobin group further into very high hemoglobin group (hemoglobin ≥16 g/dL) and high‐normal hemoglobin group (hemoglobin 14.0–15.9 g/dL) and evaluated the risk of very high hemoglobin, mild anemia, and moderate/severe anemia using high‐normal hemoglobin group as the reference. All analyses of the Cox proportional hazard model with stratification were performed with SPSS, version 19 (IBM Corporation, Armonk, NY). All other analyses were performed with JMP, version 10.0, software (SAS Institute Inc, Cary, NC) and GraphPad Prism 6.05 (GraphPad Software, Inc, La Jolla, CA). In this study, all reported P values were 2 tailed, and we considered P<0.05 as statistically significant.

Table 1

Baseline Clinical and Procedural Characteristics and Medications

Variables	High‐Normal Hemoglobin	Low‐Normal Hemoglobin	Mild Anemia	Moderate/Severe Anemia	P Value
	(Hemoglobin ≥14 g/dL)	(13 g/dL ≤ Hemoglobin < 14 g/dL for Men and 12 g/dL ≤ Hemoglobin <14 g/dL for Women)	(11 g/dL ≤ Hemoglobin <13 g/dL for Men and 11 g/dL ≤ Hemoglobin <12 g/dL for Women)	(Hemoglobin <11 g/dL)
	(N=7555; 39.2%)	(N=5303; 27.5%)	(N=4117; 21.3%)	(N=2313; 12.0%)
Hemoglobin, median (IQR), g/dL	14.9 (14.4–15.7)	13.3 (13–13.6)	11.9 (11.5–12.5)	10 (9.2–10.6)	<0.001
Clinical characteristics
Age, mean (SD), y	63.7 (10.4)	69.8 (9.5)	72.5 (9.2)	74.1 (9.8)	<0.001
≥75 ya , b	1115 (15)	1783 (34)	1888 (46)	1270 (55)	<0.001
Mena , b	6898 (91)	2953 (56)	3097 (75)	1245 (54)	<0.001
Body mass index, mean (SD), kg/m2	24.8 (3.3)	23.9 (3.5)	23 (3.4)	22.3 (3.6)	<0.001
<25a , b	4338 (57)	3540 (67)	3096 (75)	1858 (80)	<0.001
Clinical presentation
Acute myocardial infarctiona , b	2395 (32)	1199 (23)	843 (20)	511 (22)	<0.001
ST‐segment–elevation myocardial infarction	2069 (27)	1010 (19)	696 (17)	411 (18)	<0.001
Hypertensiona	6026 (80)	4329 (82)	3398 (83)	1990 (86)	<0.001
Diabetes mellitus	2824 (37)	2018 (38)	1777 (43)	1137 (49)	<0.001
Insulin therapya	384 (5)	407 (8)	439 (11)	422 (18)	<0.001
Lipid‐lowering therapy	3011 (40)	2513 (47)	1733 (42)	822 (36)	<0.001
Current smokera	2992 (40)	1135 (21)	821 (20)	393 (17)	<0.001
History of heart failurea	946 (13)	770 (15)	834 (20)	831 (36)	<0.001
Multivessel coronary diseasea	3757 (50)	2733 (52)	2305 (56)	1392 (60)	<0.001
Mitral regurgitation grade 3/4a	125 (2)	165 (3)	158 (4)	172 (7)	<0.001
Left ventricular ejection fraction, mean (SD), %	58.9 (12.4)	60.4 (12.3)	58.5 (13.3)	56.4 (13.8)	<0.001
Left ventricular ejection fraction ≤40%	540 (9)	352 (8)	383 (11)	270 (14)	<0.001
Prior myocardial infarctiona , b	1112 (15)	839 (16)	810 (20)	472 (20)	<0.001
Prior percutaneous coronary intervention	993 (13)	867 (16)	898 (22)	432 (19)	<0.001
Prior coronary artery bypass grafting	80 (1)	80 (2)	106 (3)	57 (2)	<0.001
Prior strokea , b	593 (8)	527 (10)	545 (13)	384 (17)	<0.001
Hemorrhagic stroke	81 (1)	84 (2)	69 (2)	47 (2)	0.002
Ischemic stroke	526 (7)	457 (9)	488 (12)	343 (15)	<0.001
Peripheral vascular diseasea	370 (5)	352 (7)	431 (11)	378 (25)	<0.001
Estimated glomerular filtration rate <30 mL/min per 1.73 m2, not on dialysisa	58 (0.8)	91 (2)	184 (4)	345 (15)	<0.001
Dialysisa , b	23 (0.3)	67 (1)	258 (6)	471 (20)	<0.001
Atrial fibrillationa , b	552 (7)	389 (7)	358 (9)	249 (11)	<0.001
Platelets, median (IQR), ×109/L	207 (174–245)	206 (173–245)	198 (173–245)	203 (159–254)	<0.001
Chronic obstructive pulmonary diseasea	218 (3)	174 (3)	149 (4)	72 (3)	<0.001
Liver cirrhosisa , b	120 (2)	92 (2)	97 (2)	70 (3)	<0.001
Malignancya , b	398 (5)	394 (7)	483 (12)	347 (15)	<0.001
DAPT score, mean (SD)	1.2 (1.4)	0.6 (1.4)	0.5 (1.5)	0.6 (1.6)	<0.001
≥2	3115 (41)	1354 (26)	987 (24)	609 (26)	<0.001
Procedure characteristics
Stent use	7198 (95)	5070 (96)	3945 (96)	2186 (95)	0.09
Drug‐eluting stenta , b	4832 (64)	3675 (69)	3001 (73)	1572 (68)	<0.001
Sirolimus‐eluting stent	3082 (41)	2164 (41)	1646 (40)	919 (40)	0.68
Paclitaxel‐eluting stent	229 (3)	172 (3)	136 (3)	80 (3)	0.71
Everolimus‐eluting stent	1089 (14)	910 (17)	826 (20)	389 (17)	<0.001
Biolimus‐eluting stent	520 (7)	501 (9)	450 (11)	220 (10)	<0.001
Bare‐metal stent	2857 (38)	1689 (32)	1170 (28)	735 (32)	<0.001
No. of target vessels, mean (SD)	1.3 (0.5)	1.3 (0.5)	1.3 (0.5)	1.3 (0.5)	0.26
No. of target lesions, mean (SD)	1.4 (0.7)	1.4 (0.7)	1.4 (0.7)	1.4 (0.7)	0.79
Target of LAD	4453 (59)	2975 (56)	2213 (54)	1227 (53)	<0.001
Target of proximal LADa	4220 (56)	2822 (53)	2109 (51)	1182 (51)	<0.001
Target of right coronary artery	2771 (37)	2011 (38)	1695 (41)	943 (41)	<0.001
Target of left circumflex coronary artery	2126 (28)	1486 (28)	1079 (26)	585 (25)	0.01
Target of unprotected left main coronary arterya	217 (3)	178 (3)	131 (3)	113 (5)	<0.001
Target of chronic total occlusiona	805 (11)	531 (10)	411 (10)	231 (10)	0.54
Target of restenotic lesion	258 (3)	238 (4)	234 (6)	132 (6)	<0.001
Target of bifurcationa	2282 (30)	1542 (29)	1175 (29)	657 (28)	0.16
Side‐branch stentinga	265 (4)	206 (4)	142 (3)	82 (4)	0.63
Target of aortic ostium	117 (2)	127 (2)	106 (3)	94 (4)	<0.001
Use of intravascular ultrasound	4038 (53)	3058 (58)	2500 (61)	1300 (56)	<0.001
Total stent length, mean (SD), mm	35.0 (27.1)	35.2 (27.2)	36.0 (27.8)	35.7 (26.5)	0.24
>28 mma	3243 (43)	2283 (43)	1788 (43)	1075 (46)	0.02
Minimum stent size, mean (SD), mm	3.0 (0.4)	2.9 (0.4)	2.9 (0.4)	2.9 (0.4)	<0.001
<3.0 mma	2922 (39)	2452 (46)	1844 (45)	1052 (45)	<0.001
Medication at discharge					<0.001
Aspirin	7487 (99)	5242 (99)	4074 (99)	2276 (98)	0.04
Thienopyridines	7410 (98)	5205 (98)	4035 (98)	2241 (97)	0.002
Ticlopidine	5089 (67)	3245 (61)	2396 (58)	1415 (61)	<0.001
Clopidogrel	2290 (30)	1942 (37)	1622 (39)	826 (36)	<0.001
Cilostazola , b	1274 (17)	687 (13)	556 (14)	307 (13)	<0.001
Warfarina , b	622 (8)	390 (7)	359 (9)	189 (8)	0.10
Statinsa	4894 (65)	3328 (63)	2275 (55)	1033 (45)	<0.001
β Blockersa	2536 (36)	1629 (31)	1361 (33)	791 (34)	0.002
Angiotensin‐converting enzyme inhibitors/angiotensin receptor blockers IIa	4519 (60)	3058 (58)	2500 (61)	1358 (59)	0.01
Calcium channel blockersa	2870 (38)	2286 (43)	1851 (45)	1033 (45)	<0.001
Nitratesa	2428 (32)	1721 (32)	1362 (33)	783 (34)	0.41
H2B or PPIc	3386 (53)	2391 (54)	1915 (58)	1185 (62)	<0.001

Data are expressed as number (percentage) of patients unless otherwise indicated. DAPT indicates dual‐antiplatelet therapy; H2B, histamine‐2 receptor blocker; IQR; interquartile range; LAD, left anterior descending coronary artery; PPI, proton‐pump inhibitor.

Thirty‐four variables incorporated into the multivariable analysis as the full‐adjusting model.

Thirteen variables incorporated into the multivariable analysis as the parsimonious model for hemorrhagic stroke and intracranial bleeding.

Data were available in 16 093 patients (83%).

Results

Baseline Characteristics Based on the Levels of Hemoglobin

In the entire study population, the median baseline hemoglobin value measured on median 1 (interquartile range, 0–5) day before index PCI was 13.4 (interquartile range, 12.1–14.6) g/dL. Overall, 6430 patients (33.3%) had anemia: mild (N=4117; 21.3%), moderate (N=2152; 11.2%), or severe (N=161; 0.8%). Among the remaining 12 858 patients without anemia, the median baseline value of hemoglobin was 14.2 g/dL; the baseline hemoglobin was ≥14 g/dL in 7555 patients. Accordingly, the study population consisted of the 4 groups: high‐normal hemoglobin (≥14.0 g/dL) group: N=7555 (39.2%); low‐normal hemoglobin (13.0–13.9 g/dL in men and 12.0–13.9 g/dL in women) group: N=5303 (27.5%); mild anemia (11.0–12.9 g/dL in men and 11.0–11.9 g/dL in women) group: N=4117 (21.3%); and moderate/severe anemia (<11.0 g/dL) group: N=2313 (12.0%) (Figure 2).

Figure 2

Distribution of hemoglobin at the baseline. All patients were categorized into 4 groups: high‐normal hemoglobin (≥14.0 g/dL), low‐normal hemoglobin (13.0–13.9 g/dL for men 12.0–13.9 g/dL for women), mild anemia (hemoglobin 11.0–12.9 g/dL for men and 11.0–11.9 g/dL for women), and moderate/severe anemia (hemoglobin <10.9 g/dL).

The baseline patient characteristics were markedly different across the 4 groups. Overall, with decreasing hemoglobin value, incrementally higher proportion of patients had advanced age, lower body mass index, and comorbidities, including hypertension, diabetes mellitus, end‐stage renal disease not on dialysis with an estimated glomerular filtration rate <30 mL/min per 1.73 m2, and malignancy. In addition, the proportions of patients with a history of heart failure, multivessel disease, reduced left ventricular ejection fraction, history of MI, stroke, peripheral vascular disease, and atrial fibrillation were incrementally higher with the increasing severity of anemia (Table 1). Conversely, the proportion of patients who presented as having acute MI was higher in the high‐normal group compared with that in the other 3 groups. For the lesion and procedural characteristics, the target lesion was more likely to be right coronary artery lesion, restenotic lesion, aortic ostial lesion, and bifurcation lesion with increasing severity of anemia. In addition, the prevalence of cilostazol use was higher in the high‐normal group than in the other 3 groups, whereas the prevalence of statin use was higher in the high‐ and low‐normal groups compared with the mild and moderate/severe anemia groups. In all 4 groups, <10% of patients received warfarin, with no significant difference across the groups. Furthermore, the prevalence of calcium channel blockers, nitrates, and histamine‐2 receptor blocker or proton‐pump inhibitor use was incrementally higher with the increasing severity of anemia (Table 1).

Long‐Term Ischemic Outcomes Based on the Levels of Hemoglobin

Median follow‐up duration was 3 years; clinical follow‐up at 3 years was completed in 94.6% of patients overall. The rate of persistent discontinuation of DAPT through 3‐year follow‐up was only slightly, but significantly different across the 4 groups (Figure S1). Median (interquartile range) duration of DAPT was 571 (67–1095) days, 674 (91–1095) days, 630 (106–1095) days, and 439 (45–1095) days in the high‐normal hemoglobin, low‐normal hemoglobin, mild anemia, and moderate/severe anemia groups, respectively (P<0.001).

The cumulative 3‐year incidence of the primary ischemic outcome measure (a composite of MI and ischemic stroke) was incrementally higher with decreasing baseline hemoglobin (Figure 3A). After adjusting for confounders, the excess risk of the moderate/severe anemia group relative to the high‐normal hemoglobin group remained significant for the primary ischemic outcome measure, whereas it was no longer significant in the low‐normal hemoglobin and mild anemia groups (Table 2 and Figure 4). However, the adjusted risk for ischemic stroke was significantly higher in the low‐normal hemoglobin, mild anemia, and moderate/severe anemia groups than in the high‐normal hemoglobin group, whereas decreasing baseline hemoglobin did not affect the risk for MI (Table S1).

Figure 3

Kaplan‐Meier curves for the primary ischemic and bleeding outcome measure. A, Myocardial infarction or ischemic stroke. B, GUSTO (Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Arteries Trial) moderate/severe bleeding. PCI indicates percutaneous coronary intervention.

Table 2

Clinical Outcome Through 3 Years

Variable	High‐Normal Hemoglobin	Low‐Normal Hemoglobin	Mild Anemia	Moderate/Severe Anemia
Myocardial infarction/ischemic stroke
Cumulative 3‐y incidence (%)	486 (6.6)	405 (7.8)	352 (9.0)	265 (12.7)
Unadjusted HR (95% CI), P value	Reference	1.20 (1.05–1.37), P=0.007	1.38 (1.20–1.58), P<0.001	1.96 (1.68–2.27), P<0.001
Adjusted HR (95% CI), P value	Reference	1.14 (0.99–1.32), P=0.07	1.16 (1.00–1.35), P=0.05	1.33 (1.11–1.60), P=0.003
GUSTO moderate/severe bleeding
Cumulative 3‐y incidence (%)	329 (4.5)	316 (6.1)	374 (9.6)	383 (18.0)
Unadjusted HR (95% CI), P value	Reference	1.39 (1.19–1.62), P<0.001	2.20 (1.89–2.55), P<0.001	4.42 (3.81–5.12), P<0.001
Adjusted HR (95% CI), P value	Reference	1.22 (1.04–1.44), P=0.02	1.73 (1.47–2.04), P<0.001	2.31 (1.92–2.78), P<0.001

The outcomes were adjusted by 34 full‐adjusting covariates listed in Table 1. GUSTO indicates Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Coronary Arteries Trial; HR, hazard ratio.

Figure 4

A hazard ratio plot showing the adjusted excess risk of the low‐normal hemoglobin and mild and moderate/severe anemia groups relative to high‐normal hemoglobin group for the primary ischemic and bleeding outcome measures. We included 34 clinically relevant factors indicated in Table 1 as the risk‐adjusting variables and incorporated the centers and studies as the stratification variables in the multivariable Cox proportional hazard models. GUSTO indicates Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Arteries Trial; HR, hazard ratio.

Long‐Term Bleeding Outcomes Based on the Levels of Hemoglobin

The cumulative 3‐year incidence of the primary bleeding outcome measure was incrementally higher with decreasing baseline hemoglobin (Figure 3B). After adjusting for confounders, the excess risk of the low‐normal hemoglobin, mild anemia, and moderate/severe anemia groups relative to the high‐normal hemoglobin group remained highly significant for the primary bleeding outcome measure and GUSTO moderate bleeding (Table 2, Figure 4, and Table S1). The moderate/severe anemia group was also associated with significant excess risk for intracranial bleeding (Table S1). By the landmark analysis at 30 days after index PCI, the cumulative 3‐year incidence of the primary bleeding outcome measure was incrementally higher, with decreasing baseline hemoglobin both within and beyond 30 days (Figure S2A). Adjusted excess risk of the mild and moderate/severe anemia groups relative to the high‐normal hemoglobin group for the primary bleeding outcome measure remained significant both within and beyond 30 days, whereas the risk of the low‐normal hemoglobin group was significant within 30 days, but it was no longer significant beyond 30 days (Figure S2B).

For the relationship between bleeding events and DAPT status, 68% (952 events) of the primary bleeding events occurred while under DAPT. The proportion of primary bleeding events while under DAPT to all primary bleeding events was incrementally higher with the decreasing baseline hemoglobin (Figure S3A). The proportion of primary bleeding events that needed blood transfusion and distribution of bleeding sources was not significantly different across the 4 groups (Figure S3B and S3C).

Long‐Term Mortality Based on the Levels of Hemoglobin

The cumulative 3‐year incidence of all‐cause death was also incrementally higher with decreasing baseline hemoglobin. After adjusting confounders, the excess risk of the low‐normal hemoglobin, mild anemia, and moderate/severe anemia groups relative to the high‐normal hemoglobin group remained significant for all‐cause death, driven by the excess risk for both cardiovascular and noncardiovascular death (Table S1).

Subgroup Analysis

There was significant interaction between those subgroup factors, such as sex, shock, renal function, and atrial fibrillation, and the effect of hemoglobin levels on the primary bleeding outcome measure. The effect of decreasing baseline hemoglobin on the primary bleeding outcome measure was more prominent in men, patients without shock, patients with estimated glomerular filtration rate ≥60 mL/min per 1.73 m2, and patients without atrial fibrillation. There was no interaction in other subgroups, such as age, body mass index, malignancy, platelet counts, use of histamine‐2 receptor blocker or proton‐pump inhibitor, DAPT score, and use of DES (Figure 5). Of note, despite difference in baseline clinical and procedural characteristic between CREDO‐Kyoto PCI/CABG Registry Cohort‐2 and RESET/NEXT (Table S2), there was no significant interaction between the study and the effect of hemoglobin levels on the primary bleeding outcome measure (Figure 5).

Figure 5

Subgroup analyses for the primary bleeding outcome in the clinically relevant subgroups. A, Hazard ratio of low‐normal hemoglobin, mild anemia, and moderate/severe anemia relative to high‐normal hemoglobin and P value for interaction between subgroup factors and the effect of hemoglobin levels on the primary bleeding outcome measure. The same 34 risk‐adjusting variables used in the entire cohort were included in the multivariable Cox proportional hazard models. B, Continued, *We constructed parsimonious models with the 13 risk‐adjusting variables used in the parsimonious models in the entire cohort. C, Continued, *We constructed parsimonious models with the 13 risk‐adjusting variables used in the parsimonious models in the entire cohort. CREDO‐Kyoto indicates Coronary Revascularization Demonstrating Outcome Study in Kyoto; DAPT, dual‐antiplatelet therapy; H2B, histamine‐2 receptor blocker; NEXT, NOBORI Biolimus‐Eluting Versus XIENCE/PROMUS Everolimus‐Eluting Stent Trial; PPI, proton‐pump inhibitor; RESET, Randomized Evaluation of Sirolimus‐Eluting Versus Everolimus‐Eluting Stent Trial.

Sensitivity Analysis

In the sensitivity analysis with 3 groups (ie, no anemia [combination of high‐normal and low‐normal hemoglobin], mild anemia, and moderate/severe anemia), adjusted excess risk of moderate/severe anemia relative to no anemia remained significant for both the primary ischemic and bleeding outcome measures, whereas the risk of mild anemia relative to no anemia remained significant for the primary bleeding outcome measure, but not for the primary ischemic outcome measure (Table S3). In the sensitivity analysis with 5 groups (ie, very high hemoglobin, high‐normal hemoglobin, low‐normal hemoglobin, mild anemia, and moderate/severe anemia), adjusted excess risk of moderate/severe anemia relative to high‐normal hemoglobin remained significant for both the primary ischemic and the bleeding outcome measures, whereas the risk of mild anemia and low‐normal hemoglobin relative to high‐normal hemoglobin remained significant for the primary bleeding outcome measure, but not for the primary ischemic outcome measure (Table S4). Adjusted excess risk of very high hemoglobin relative to high‐normal hemoglobin was not significant for both the primary ischemic outcome measure and the primary bleeding outcome measure (Table S4).

Discussion

The primary findings in this study are as follows: (1) One third of patients undergoing PCI had anemia, as defined per the WHO criteria. (2) With the increasing severity of anemia, patients trended to be older, have lower body mass index, and have more severe comorbidities. (3) Moderate/severe anemia was associated with a markedly higher risk for ischemic and bleeding events, as well as mortality. (4) Even mild anemia and low‐normal hemoglobin correlated with significantly higher risk for ischemic stroke, serious bleeding events, such as GUSTO moderate bleeding, and mortality.

Corroborating the prior studies of patients undergoing PCI,15 the incidence of anemia in this study was 33%, which is considerably higher than that reported in patients in a primary care setting,16 possibly reflecting the higher prevalence of comorbidities predisposing to anemia in patients with coronary artery disease. Indeed, we observed that patients with lower hemoglobin correlated with more advanced age, lower body mass index, history of heart failure, and more severe comorbidities, including diabetes mellitus, reduced renal dysfunction, peripheral vascular disease, and history of malignancy.

Consistent with previous studies,7, 8 this study demonstrated a robust and strong correlation between the baseline hemoglobin levels and the subsequent bleeding outcome. Even a mild degree of anemia was associated with marked excess risk for major bleeding compared with high‐normal hemoglobin. Most strikingly, even among patients without anemia, the lower level of baseline hemoglobin value was associated with higher long‐term bleeding risk. The findings corroborated and expanded a previous report in patients with acute coronary syndromes, demonstrating that the 30‐day rates of major bleeding progressively increase from the highest (>16 g/dL) to the lowest (10 g/dL) levels of baseline hemoglobin.17 In patients with lower baseline hemoglobin levels, even a small decrease in hemoglobin might lead to transfusion at an earlier stage, which might result in increased bleeding events. Indeed, the incidence of GUSTO moderate bleeding (bleeding that needs transfusion) was incrementally higher with the increasing severity of anemia in this study. In addition, preprocedural anemia could result from the presence of long‐standing unrecognized hemorrhagic diathesis, which could result in an increased future risk for bleeding. Of note, the presence of a small decline in hemoglobin might often be regarded as “not severe” and dismissed. However, given the excess bleeding risk of even a small decline in hemoglobin at the threshold of 14 g/dL, we might as well pay attention to the low‐normal hemoglobin value as an indicator of high bleeding risk in patients planned for PCI.

Previous studies in the eras of bare‐metal stent have shown that baseline anemia was associated with increased risk of in‐hospital or short‐term (<30 days) incidence of MI or recurrent ischemia.5, 18, 19 More recent study, including 6528 patients treated with DES with 4‐year follow‐up, demonstrated that the patients with baseline severe anemia (5.5% of the entire cohort) were associated with increased risk of MI compared with no/mild anemia.4 In the present study, moderate/severe anemia was associated with excess risk for the long‐term ischemic outcome measure, but this was driven by higher incidence of ischemic stroke rather than MI. The discrepancy of the results might be caused by the different categorization of severity of anemia. We combined moderate anemia group and severe anemia group into one group because proportion of patients with severe anemia in our study was small (<1%). Rather, our study more focused on the patients with mild anemia or without anemia. We observed that even mild anemia and low‐normal hemoglobin were associated with significant excess risk relative to high‐normal hemoglobin for the ischemic stroke. Several underlying mechanisms are plausible. As the tissue oxygen supply is limited because of the reduced hemoglobin value, patients would be more susceptible to ischemic events.20 Alternatively, patients with anemia might exhibit a hypercoagulable state, which could exacerbate the risk of ischemic events.21 Furthermore, advanced age and comorbidities related to anemia might result in the higher incidence of ischemic events. High ischemic and bleeding events, along with anemia‐related comorbidities, could likely underlie the correlations between low hemoglobin levels and mortality. Unlike some previous studies dominantly enrolling the patients with acute coronary syndrome and examining short‐term outcomes,15, 17, 19 those with hemoglobin >16 g/dL in the present study did not appear to be at significantly increased risk for the long‐term bleeding and ischemic outcomes relative to those with hemoglobin 14.0 to 15.9 g/dL.

In this study, we found that the rate of persistent discontinuation of DAPT through 3‐year follow‐up was only slightly different across the 4 groups, suggesting that baseline severity of anemia might not be taken into consideration in the decision making for DAPT duration. We are confronted with a difficult situation in deciding the intensity of antithrombotic therapy of patients with anemia who have both high ischemic and bleeding risk. The current guidelines generally recommend less intensive antithrombotic therapy in patients with high bleeding risk.22, 23 Future studies would be warranted to address the optimal antithrombotic therapy in patients with anemia.

Limitations

This study has several limitations. First, this study did not address the causes of anemia. In addition, anemia was evaluated only at baseline. Some patients might have subclinical causes of bleeding, such as subclinical malignancies, that could lead to anemia later on. However, long‐term change in hemoglobin and its correlation with the prognosis remain unclear. Second, we did not have the reference for the low‐normal and high‐normal hemoglobin groups; the current study does not provide the definite threshold of hemoglobin to stratify the long‐term ischemic and bleeding events. Nevertheless, the patients with low‐normal hemoglobin still face marked excess risk for bleeding and stroke compared with those with high‐normal hemoglobin. Third, despite comprehensive statistical adjustment for potential confounders and a wide range of interaction tests, unmeasured confounders could still have affected the study results. Finally, several patients in this study underwent PCI using bare‐metal stent, the first‐generation DES, and older antiplatelet agents, which are less frequently used in the current practice. In addition, the duration of DAPT was considerably long in the current study because the importance of DAPT has been stressed in each study era, whereas the results of randomized clinical trials of DAPT duration after PCI have not been published.24, 25 Thus, findings in this study should be cautiously applied to the current practice.

Conclusions

Decreasing baseline hemoglobin was associated with incrementally higher long‐term risk for not only major bleeding, but also ischemic stroke and mortality after PCI. Even among patients without anemia, the lower level of baseline hemoglobin value was associated with higher long‐term ischemic and bleeding risk.

Sources of Funding

CREDO‐Kyoto PCI/CABG (Coronary Revascularization Demonstrating Outcome Study in Kyoto Percutaneous Coronary Intervention/Coronary Artery Bypass Grafting) Registry Cohort 2 is funded by Pharmaceuticals and Medical Devices Agency in Japan, RESET (Randomized Evaluation of Sirolimus‐Eluting Versus Everolimus‐Eluting Stent Trial) by Abbott Vascular Japan, Co, Ltd, and NEXT (NOBORI Biolimus‐Eluting Versus XIENCE/PROMUS Everolimus‐Eluting Stent Trial) by Terumo Japan, Co, Ltd.

Disclosures

Dr Kozuma reports honoraria from Abbott Vascular, Daiichi Sankyo, and Sanofi. Dr Tanabe reports honoraria from Abbott Vascular, Terumo Japan, Daiichi Sankyo, Sanofi, and AstraZeneca Japan. Dr Morino reports honoraria from Abbott Vascular and Terumo Japan. The remaining authors have no disclosures to report.

Appendix S1. List of the participating centers and the investigators.

Data S1. Supplemental methods.

Table S1. Relationship Between Baseline Hb Value and the Secondary Outcome Measures Through 3 Years

Table S2. Baseline Clinical and Procedural Characteristics Across the Studies

Table S3. Sensitivity Analysis: Clinical Outcome Across No Anemia (a Composite of High‐Normal Hb and Low‐Normal Hb), Mild Anemia and Moderate/Severe Anemia

Table S4. Sensitivity Analysis: Clinical Outcome Across Very High Hb, High‐Normal Hb, Low‐Normal Hb, Mild Anemia and Moderate/Severe Anemia

Figure S1. The Kaplan–Meier curves for the persistent discontinuation of dual antiplatelet therapy among patients with high‐normal Hb, low‐normal Hb, mild, and moderate/severe anemia.

Figure S2A. Thirty‐day landmark analysis of the primary bleedinh outcome measure.

Figure S2B. A hazard ratio plot showing the adjusted excess risk of the low‐normal Hb, mild and moderate/severe anemia groups relative to high‐normal Hb group for the primary bleeding outcome measure within and beyond 30 days.

Figure S3. A, The proportion of primary bleeding events while under DAPT to all primary bleeding events. DAPT indicates dual antiplatelet therapy; Hb, hemoglobin. B, The proportion of primary bleeding events requiring blood transfusion to all primary bleeding events. C, The proportion of bleeding sources.

Click here for additional data file.