META-ANALYSIS
The Risk of Perioperative Bleeding in Patients With Chronic Kidney Disease A Systematic Review and Meta-Analysis Rey R. Acedillo, MD,∗ Mitesh Shah, MBBS,∗ † P. J. Devereaux, MD, PhD,‡§ Lihua Li, MSc,∗ † Arthur V. Iansavichus, MLIS,∗ Michael Walsh, MD, MSc,‡§ and Amit X. Garg, MD, PhD∗ †§ Background: Worldwide, millions of patients with chronic kidney disease undergo surgery each year. Although chronic kidney disease increases the risk of bleeding in nonoperative settings, the risk of perioperative bleeding is less clear. We conducted a systematic review and meta-analysis to summarize existing information and quantify the risk of perioperative bleeding from chronic kidney disease. Methods: We screened 9376 citations from multiple databases for cohort studies published between 1990 and 2011. Studies that met our inclusion criteria included patients undergoing any major surgery, with a sample size of at least 100 patients with chronic kidney disease (as defined by the primary study authors with an elevated preoperative serum creatinine value or a low estimated glomerular filtration rate). Their outcomes had to be compared with a reference group of at least 100 patients without chronic kidney disease. Our primary outcomes were (1) receipt of perioperative blood transfusions and (2) need for reoperation for reasons of bleeding. Results: Twenty-three studies met our criteria for review (20 cardiac surgery, 3 non–cardiac surgery). Chronic kidney disease was associated with a greater risk of requiring blood transfusion (7 studies in cardiac surgery, totaling 22,718 patients) and weighted incidence in patients with normal kidney function was 53% and in chronic kidney disease was 73%; pooled odds ratio, 2.7 (95% confidence interval, 2.1–3.4). After adjustment for relevant factors, the association remained statistically significant in 4 studies. Chronic kidney disease was associated with more reoperation for reasons of bleeding (14 studies in cardiac surgery, totaling 569,715 patients) and weighted incidence in patients with normal kidney function was 2.4% and in chronic kidney disease was 2.7%; pooled odds ratio, 1.6 (95% confidence interval, 1.3–1.8). However, after adjustment for relevant factors (as done in 5 studies), the association was no longer statistically significant. Conclusions: Chronic kidney disease is associated with perioperative bleeding but not bleeding that required reoperation. Further studies should stage chronic kidney disease with the modern system, better define bleeding outcomes, and guide intervention to improve the safety of surgery in this at-risk population.
Keywords: bleeding, chronic kidney disease, perioperative bleeding, reexploration, reoperation, surgery, transfusion (Ann Surg 2013;258:901–913)
C
hronic kidney disease is a major risk factor for perioperative complications.1–3 Under the current staging system, chronic kidney disease affects about 7% of the population and 26% of adults older than 60 years.4–6 As there are about 200 million surgical procedures performed annually worldwide, many millions of patients with chronic kidney disease undergo surgery each year.7 In the nonoperative setting, chronic kidney disease is associated with an increased risk of bleeding.8–13 The pathophysiology of bleeding in chronic kidney disease is multifactorial and is attributed to platelet-vessel wall dysfunction and anemia. The risk of perioperative bleeding in patients with chronic kidney disease has not been well established. Recent narrative reviews have discussed the risk of perioperative bleeding in patients with chronic kidney disease.11,13,14 However, only a few studies referenced in these reviews describe bleeding risk.15–18 We are not aware of any systematic review of the literature that comprehensively evaluates this potential association. Perioperative bleeding can contribute to significant morbidity and mortality, depending on the type of bleeding event and whether or not blood transfusions are received.19–21 In some cases, patients return to the operating room to control bleeding.22 Knowledge of perioperative bleeding risks in patients with chronic kidney disease may help guide physician-patient decision making and informed consent, surgical and blood conservation techniques that are used, and the appropriate level of surveillance to detect bleeding. For these reasons, we conducted this systematic review and meta-analysis to quantify the risk of perioperative bleeding in patients with chronic kidney disease.
METHODS We conducted and reported this review in accordance with published guidelines.23,24
Searching
From the ∗ Division of Nephrology, Department of Medicine, Western University, London, Canada; †Department of Epidemiology and Biostatistics, Western University, London, Canada; ‡Department of Medicine, McMaster University, Hamilton, Canada; and §Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, Canada. Disclosure: For this study, ethical approval was not required. The authors declare no conflicts of interest. Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s Web site (www.annalsofsurgery.com). Reprints: Rey R. Acedillo MD, London Kidney Clinical Research Unit, Room ELL-101, Westminster; and London Health Sciences Centre, 800 Commissioners Rd East, London, Ontario, Canada N6A 4G5. E-mail: [email protected]. C 2013 by Lippincott Williams & Wilkins Copyright ISSN: 0003-4932/13/25806-0901 DOI: 10.1097/SLA.0000000000000244
We searched MEDLINE and EMBASE for publications from January 1990 to May 2011. An experienced information specialist developed search strategies using sensitive terms for identifying studies of perioperative risk. We pilot-tested and modified search strategies to identify known eligible articles. The final search strategies included the terms for surgery; intra-, peri-, and postoperative, chronic kidney disease; glomerular filtration rate (GFR); end-stage renal disease; dialysis; bleeding; and blood transfusion. We modified identified terms to correspond to the tree structure and descriptors of different databases (strategies are provided in the Supplemental Digital Content Appendix, available at http://links.lww.com/SLA/A474). We also compiled citations from information provided from citation tracking in SCOPUS and related articles in PubMed and Google Scholar,
Annals of Surgery r Volume 258, Number 6, December 2013
www.annalsofsurgery.com | 901
Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 258, Number 6, December 2013
Acedillo et al
FIGURE 1. Study selection. ∗ Discrete cohorts within 1 article are defined as separate studies. †A study may have been excluded for more than 1 reason. ‡Discrete cohorts were calculated (where possible) from data within case-control articles to analyze the effect of exposure on outcome. §Bleeding defined as a perioperative red blood cell transfusion or surgically related bleeding event. ¶Articles were accepted if the exposure cohort was different from other studies despite using the same database for the referent cohort. CKD indicates chronic kidney disease. as well as reference lists from selected previous reviews9,11 and eligible studies.16,18,25–45 All citations were downloaded into Reference Manager Sente (version 6.5). Two reviewers independently screened each citation and the full text was retrieved for any article considered potentially relevant by either reviewer. Each reviewer independently evaluated the eligibility of each retrieved full text and disagreements were resolved by consensus. When data from the same group of patients or database were described in multiple publications,18,31,42,46,47 we reviewed all publications and included the most representative one.18,31,42
Selection We included all adult cohort studies that compared the risk of perioperative bleeding in patients with and without chronic kidney disease undergoing any major surgical procedure. Although the design of some studies was described as case-control, there were sufficient data to analyze a group of patients having surgery as a cohort. We accepted primary author–defined measures of chronic kidney disease unless the renal impairment was acute or transient. The sample size needed to consist of at least 100 patients with chronic kidney disease. Their outcomes had to be compared to a reference group of at least 100 patients without chronic kidney disease. The study had to report at least 1 of the 2 primary study outcomes: (1) receipt of perioperative blood transfusion and (2) need for reoperation for 902 | www.annalsofsurgery.com
reasons of bleeding. Only studies published in English language were eligible for review. To best reflect modern surgical practices, we focused only on articles published after 1990. We excluded studies that reported only estimated blood loss given potential inaccuracies of this measurement. We also excluded studies with surgical procedures that involved the kidney, such as renal transplant, renal artery bypass, or nephrectomy.
Data Extraction and Validity Assessment Two reviewers independently extracted data from included studies on methods, type of surgery, patient characteristics, definition of chronic kidney disease, and outcomes. Any discrepancies were resolved by consensus. Using recommended validity criteria, 1 reviewer assessed the methodological quality of each study.48 Chronic kidney disease was defined by the primary authors using a variety of measures: estimated glomerular filtration rate (eGFR), creatinine clearance, serum creatinine, or dialysis. When outcomes were reported for more than 2 different categories of renal function, we extracted data for each stratum. We combined study results to compare patients with normal kidney function with those with chronic kidney disease in a dichotomous fashion, using the first cutoff value of chronic kidney disease defined by each study. When possible, we collected data on patients on dialysis separately. For all relevant outcomes of interest, we extracted the raw number of events, the unadjusted association C 2013 Lippincott Williams & Wilkins
Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
C 2013 Lippincott Williams & Wilkins
United States United States
United States
United States
Anderson et al16 (1999) Anderson et al26 (2000)
Charytan et al27 (2010)
Chirumamilla et al28 (2008) Cooper et al29 (2006) Devbhandari et al30 (2006) Filsoufi et al31 (2008)† Foot et al32 (2009) Holzmann et al33 (2005) Ib´an˜ ez et al34 (2007)
China United States United Kingdom United States United States
United States
United States
Lin et al37 (2009) Liu et al18 (2000) Murphy et al38 (2007)§ O’Brien et al39 (2002) O’Hare et al40 (2003)
O’Hare et al41 (2004)
Rahmanian et al42 (2008)† Robinson et al43 (2007)§ Whitson et al44 (2007)§ Witczak et al45 (2005) 10560 2497 106
6449
11051
1288 194 106
245
5943
3817 279 623 6259 4925
313
1345
135 6331 4667 207
379034 386
823
724
683 197
322
No. Patients With Reduced Kidney Function
65 (12) 64 (12) 64 (10)
...
...
... ... ... 62 ( . . . ) ...
...
62 (12)
... 64 ( . . . ) 61 (9) 68 ( . . . )
... ...
...
...
64 ( . . . ) 66 ( . . . )
64 (10)
Mean Age (SD), yr
... ... 83
62
99
84 72 ... 96 99
79
...
62 71 81 55
72 79
74
72
99 97
75
Male (%)
To convert serum creatinine value in units of μmol/L to mg/dL, divide value by 88.4. ∗ Calculated using the Modification of Diet in Renal Disease formula. †Articles were accepted if the exposure cohort was different from other studies despite using the same database for the control cohort. ‡Calculated using the Cockroft-Gault formula. §Discrete cohorts were calculated (where possible) from data within case-control articles to analyze the effect of exposure on outcome. ¶Training data set was used. Validation set/data could not be converted into discrete cohorts of interest to compare bleeding outcomes. . . . indicates data not reported in primary article; CABG, coronary artery bypass graft.
Australia United States Norway
292
Japan 1742 15271 7975 42822 13292
5306
Canada
6314 1109 1888 474
104880 19172
229
1714
3271 637
3276
No. Patients With Normal Kidney Function
Karkouti et al36 (2006)§¶,e Kinoshita et al35 (2010)
United States Australia Sweden Spain
United States United Kingdom
Kuwait
Country
Al-Sarraf et al25 (2011)
Authors, Year of Publication
TABLE 1. Baseline Characteristics of Included Studies
CABG CABG CABG Cardiac Cardiac CABG Cardiac CABG
GFR 133 μmol/L
Cardiac Cardiac Cardiovascular
CABG CABG Cardiac General surgery Lower extremity revascularization Lower extremity amputation Cardiac
CABG
Cardiac
Type of Surgery
Cr >125 μmol/L
Definition of Reduced Kidney Function
Reoperation Blood transfusion Blood transfusion, reoperation
Reoperation
Blood transfusion
Blood transfusion, reoperation Reoperation Reoperation Blood transfusion Blood transfusion Blood transfusion
Reoperation Reoperation Reoperation Blood transfusion, reoperation Blood transfusion
Reoperation Reoperation
Bleeding (tamponade, mediastinal, pericardial), blood transfusion Reoperation
Blood transfusion, reoperation Blood transfusion Blood transfusion
Perioperative Bleeding Outcome Studied
Annals of Surgery r Volume 258, Number 6, December 2013 CKD and Perioperative Bleeding
www.annalsofsurgery.com | 903
Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 258, Number 6, December 2013
Acedillo et al
TABLE 2. Methodological Quality of Included Studies Authors, Year of Publication
Cohort Study Design
Representative Sample
Selection Criteria Defined
Loss to Loss to Follow Follow up up Reported Reported 110 μmol/L for men, Cr > 100 in women. n.r. indicates study did not report whether patients on dialysis were included. kidney disease, it was 1.8%. The unadjusted OR was 2.7 (95% CI, 1.6–4.5). There was significant statistical heterogeneity for the ORs reported across each of the non–cardiac studies (χ 2 = 21.65, P < 0.0001, I2 = 91%). When both cardiac and non–cardiac studies were combined, the weighted incidence in patients with normal kidney function was 1.1% and in patients with chronic kidney disease, it was 2.5%. The unadjusted OR (cardiac and noncardiac surgery combined) ranged from 1.7 to 3.5, with a pooled OR of 2.5 (95% CI, 1.7–3.6). There was significant statistical heterogeneity for the ORs reported across each of the studies (χ 2 = 25.31, P < 0.0001, I2 = 88%). Four studies (2 cardiac surgery, 2 non–cardiac surgery) examined the risk of perioperative red blood cell transfusions in patients with chronic kidney disease, adjusted for baseline, preoperative, and operative characteristics (Tables 3 and 4).25,34,39,40 The factors included in the multivariate model varied greatly across studies. In the non–cardiac study by O’Hare et al,40 the adjusted OR in patients with an eGFR between 30 and 59 mL/min/1.73 m2 , compared with patients with an eGFR of 60 mL/min/1.73 m2 or greater, was 1.34 (95% CI, 1.02–1.76). The non–cardiac study by O’Brien et al39 showed that the adjusted OR for patients with a Cr greater than 265 μmol/L, compared with those with a Cr less than 133 μmol/L, was 1.71 (95% CI, 1.2–2.45). The cardiac study by Al-Sarraf et al25 showed that the adjusted OR in patients on dialysis compared with those with a Cr less than 125 μmol/L was 8.4 (95% CI, 3.8–18.8). The cardiac study by Ib´an˜ ez et al34 showed that the adjusted OR in patients with an eGFR less than 60 mL/min/1.73 m2 , compared with patients with an eGFR of 60 mL/min/1.73 m2 or greater, was 1.09 (95% CI, 1.01–1.7).
Risk of Reoperation for Reasons of Bleeding The risk of reoperation for reasons of bleeding was described only in cardiac surgery and is shown in Figure 4. Fourteen studies totaling 569,715 patients reported this risk.18,25,28–35,37,42,43,45 The weighted incidence of reoperation in patients with normal renal func906 | www.annalsofsurgery.com
tion was 2.4% and in patients with chronic kidney disease, it was 2.7%. The unadjusted ORs ranged from 0.7 to 3.0, with a pooled OR of 1.6 (95% CI, 1.3–1.8). There was significant statistical heterogeneity for the ORs reported in each of the studies (χ 2 = 27.02, P = 0.02, I2 = 52%). Five cardiac studies examined the risk of reoperation for reasons of bleeding, adjusted for baseline, preoperative, and operative characteristics (Tables 5 and 6).18,30,31,33,42 Factors included in the multivariate model also varied greatly across studies, with 1 study using a propensity score model.30 The adjusted ORs ranged from 0.63 to 1.5. None of these ORs reached statistical significance.
Perioperative Outcomes in Patients on Dialysis Perioperative bleeding risk in patients on dialysis is outlined in Figure 5. Two studies totaling 1493 patients on dialysis reported the risk of more than 4 units of red blood cell transfusions (Fig. 5A).40,41 Both studies were conducted in the non–cardiac surgery setting. The weighted incidence in patients with normal renal function was 1% and in patients on dialysis, it was 1.7%. The pooled OR was 2.7 (95% CI, 1.6–4.4). There was mild statistical heterogeneity (χ 2 = 1.40, P = 0.24, I2 = 28%). Five studies totaling 7788 patients on dialysis examined the risk of reoperation for bleeding in patients on dialysis (Fig. 5B).18,25,29,32,42 All the studies were done in cardiac surgery. The weighted incidence in patients with normal renal function was 2.3% and in patients on dialysis, it was 3.4%. The pooled OR was 1.6 (95% CI, 1.3– 1.9). There was mild statistical heterogeneity (χ 2 = 4.24, P = 0.37, I2 = 6%).
Severity of Chronic Kidney Disease and Perioperative Bleeding The relationship between the severity of chronic kidney disease and perioperative bleeding (defined in 3 ways) was C 2013 Lippincott Williams & Wilkins
Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 258, Number 6, December 2013
CKD and Perioperative Bleeding
TABLE 3. Adjusted Risk of Perioperative Blood Transfusions Authors, Year of Publication Al-Sarraf et
al25 (2011)∗
Ib´an˜ ez et al34 (2007)∗ O’Hare et
al40
(2003)‡
O’Brien et al39 (2002)‡
Definition of Bleeding
Definition of Renal Function
No. Patients
Blood transfusion
Cr 125 μmol/L Dialysis GFR >60 mL/min/1.73 m2 † GFR 60 mL/min/1.73 m2 † GFR 30–59.9 mL/min/1.73 m2 GFR 4 units RBC
>4 units RBC
Adjusted Odds Ratio (95% CI)
424 42822 4945§ 1314
P
1.3 (0.99–1.7) 8.4 (3.8–18.8)
0.064 200 μmol/L Cr ≤ 221 μmol/L∗ Cr > 221 μmol/L GFR ≥ 90 mL/min/1.73 m2∗ GFR 60–90 mL/min/1.73 m2 GFR 30–60 mL/min/1.73 m2 GFR < 30 mL/min/1.73 m2 Nondialysis∗ Dialysis Nondialysis∗ Dialysis
(2006)
Filsoufi et al31 (2008) Holzmann et al33 (2005)
Liu et al18 (2000) Rahmanian et al42 (2008)
Adjusted Odds Ratio (95% CI)
P
1.5 (0.9–2.7)
0.14
1.4 (0.55–3.55)
0.476
19172 386 6314 135 1888 3212 1391 64 15271 279 6449 245
0.9 (0.6–1.3) 1.0 (0.6–1.7) 1.4 (0.5–3.9)
n.r. n.r. n.r.
1.3 (0.69–2.5)
0.401
0.62 (0.23–1.73)
0.364
∗
Threshold for each study’s definition of normal and abnormal renal function. n.r. indicates not reported.
TABLE 6. Factors Considered in the Multivariate Models for the Risk of Reoperation for Reasons of Bleeding After Cardiac Surgery in Patients With Chronic Kidney Disease
Source∗ Definition of CKD
Devbhandari et al30 (2006) Filsoufi et al31 (2008) Cr > 200 μmol/L Cr > 221 μmol/L
Risk of reoperation for bleeding Demographics Age Sex Prior cardiac surgery Comorbidities Hypertension Diabetes Coronary artery disease/ACS Congestive heart failure Cerebrovascular disease Dyslipidemia Peripheral vascular disease Liver disease Cancer Respiratory disease Obesity/weight Body surface area Preoperative characteristics Anemia Ejection fraction/LV function Hemodynamic stability Intra-aortic balloon pump Antiplatelet/anticoagulant Operative characteristics Procedure type Surgical priority Comorbidities Hypertension Diabetes Coronary artery disease/ACS Congestive heart failure Cerebrovascular disease Dyslipidemia Peripheral vascular disease Liver disease Cancer Respiratory disease Obesity/weight Body surface area
√ √ √ √ √ √ √
√ √
√ √ √ √ √ √ √
Holzmann et al33 (2005) GFR < 90 mL/min/1.73 m2 √ √
√ √
√ √ √
√
√ √ √
√ √ √
√ √
Rahmanian et al42 (2008) Dialysis
√ √ √
√ √ √
√ √
√ √ √ √
√ √ √ √ √ √ √
√ √ √ √
Liu et al18 (2000) Dialysis
√ √ √ √ √
√ √
√ √ √
√
√ √
√ √ √ √
√ √ √
√ √
√ √ √ √ √ √ √ √
∗
Not all baseline characteristics, preoperative characteristics, and operative characteristics used in the multivariate analysis within each study are listed (see references). ACS indicates acute coronary syndrome; CKD, chronic kidney disease; LV, left ventricle.
C 2013 Lippincott Williams & Wilkins
www.annalsofsurgery.com | 909
Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 258, Number 6, December 2013
Acedillo et al
FIGURE 5. Unadjusted risk of perioperative bleeding in patients on dialysis. Top figure (A) represents patients requiring red blood cell transfusions in the non–cardiac surgery setting. Bottom figure (B) represents the risk of reoperation for reasons of bleeding in the cardiac surgery setting. The size of each square is proportional to the variability of the study estimate. n represents the number of patients who had a perioperative bleeding event and N represents the number of patients in the cohort. ∗ Articles were accepted if the exposure cohort was different from other studies despite using the same database for the control cohort.
FIGURE 6. GFR and perioperative bleeding. The area of each circle is proportional to the number of patients in GFR strata for each study. Meta-regression was used to create best-fit lines with 95% CIs. Four studies in cardiac surgery describe the risk of receiving red blood cell transfusion. Three studies in non–cardiac surgery describe the risk of receiving at least 4 units of red blood cell transfusions. Six studies in cardiac surgery describe the risk of reoperation for reasons of bleeding.
910 | www.annalsofsurgery.com
C 2013 Lippincott Williams & Wilkins
Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 258, Number 6, December 2013
CKD and Perioperative Bleeding
TABLE 7. Perioperative Transfusion Requirements in Cardiac Surgery Study Size
Definition of Renal Function
Units of RBC Transfusions
(1999)
3954
μmol/L∗
Anderson et al26 (2000)
834
al34
681
1.9 3.5 4.1 3.3 6.0 2.3 (2.6)‡ 3.8 (3.6)‡ 0 (0–2)|| 2 (0–4) 3 (2–5) 2.3 (4.1)‡ 5.4 (6.5)‡
Authors, Year of Publication Anderson et
Ib´an˜ ez et
al16
(2007)
Charytan et al27 (2010)¶
2438
Witczak et al45 (2005) 212
Cr 265 μmol/L Cr 60 mL/min/1.73 m2∗ GFR 60 mL/min/1.73 m2∗ GFR >30–59.9 mL/min/1.73 m2 GFR
The Risk of Perioperative Bleeding in Patients With Chronic Kidney Disease A Systematic Review and Meta-Analysis Rey R. Acedillo, MD,∗ Mitesh Shah, MBBS,∗ † P. J. Devereaux, MD, PhD,‡§ Lihua Li, MSc,∗ † Arthur V. Iansavichus, MLIS,∗ Michael Walsh, MD, MSc,‡§ and Amit X. Garg, MD, PhD∗ †§ Background: Worldwide, millions of patients with chronic kidney disease undergo surgery each year. Although chronic kidney disease increases the risk of bleeding in nonoperative settings, the risk of perioperative bleeding is less clear. We conducted a systematic review and meta-analysis to summarize existing information and quantify the risk of perioperative bleeding from chronic kidney disease. Methods: We screened 9376 citations from multiple databases for cohort studies published between 1990 and 2011. Studies that met our inclusion criteria included patients undergoing any major surgery, with a sample size of at least 100 patients with chronic kidney disease (as defined by the primary study authors with an elevated preoperative serum creatinine value or a low estimated glomerular filtration rate). Their outcomes had to be compared with a reference group of at least 100 patients without chronic kidney disease. Our primary outcomes were (1) receipt of perioperative blood transfusions and (2) need for reoperation for reasons of bleeding. Results: Twenty-three studies met our criteria for review (20 cardiac surgery, 3 non–cardiac surgery). Chronic kidney disease was associated with a greater risk of requiring blood transfusion (7 studies in cardiac surgery, totaling 22,718 patients) and weighted incidence in patients with normal kidney function was 53% and in chronic kidney disease was 73%; pooled odds ratio, 2.7 (95% confidence interval, 2.1–3.4). After adjustment for relevant factors, the association remained statistically significant in 4 studies. Chronic kidney disease was associated with more reoperation for reasons of bleeding (14 studies in cardiac surgery, totaling 569,715 patients) and weighted incidence in patients with normal kidney function was 2.4% and in chronic kidney disease was 2.7%; pooled odds ratio, 1.6 (95% confidence interval, 1.3–1.8). However, after adjustment for relevant factors (as done in 5 studies), the association was no longer statistically significant. Conclusions: Chronic kidney disease is associated with perioperative bleeding but not bleeding that required reoperation. Further studies should stage chronic kidney disease with the modern system, better define bleeding outcomes, and guide intervention to improve the safety of surgery in this at-risk population.
Keywords: bleeding, chronic kidney disease, perioperative bleeding, reexploration, reoperation, surgery, transfusion (Ann Surg 2013;258:901–913)
C
hronic kidney disease is a major risk factor for perioperative complications.1–3 Under the current staging system, chronic kidney disease affects about 7% of the population and 26% of adults older than 60 years.4–6 As there are about 200 million surgical procedures performed annually worldwide, many millions of patients with chronic kidney disease undergo surgery each year.7 In the nonoperative setting, chronic kidney disease is associated with an increased risk of bleeding.8–13 The pathophysiology of bleeding in chronic kidney disease is multifactorial and is attributed to platelet-vessel wall dysfunction and anemia. The risk of perioperative bleeding in patients with chronic kidney disease has not been well established. Recent narrative reviews have discussed the risk of perioperative bleeding in patients with chronic kidney disease.11,13,14 However, only a few studies referenced in these reviews describe bleeding risk.15–18 We are not aware of any systematic review of the literature that comprehensively evaluates this potential association. Perioperative bleeding can contribute to significant morbidity and mortality, depending on the type of bleeding event and whether or not blood transfusions are received.19–21 In some cases, patients return to the operating room to control bleeding.22 Knowledge of perioperative bleeding risks in patients with chronic kidney disease may help guide physician-patient decision making and informed consent, surgical and blood conservation techniques that are used, and the appropriate level of surveillance to detect bleeding. For these reasons, we conducted this systematic review and meta-analysis to quantify the risk of perioperative bleeding in patients with chronic kidney disease.
METHODS We conducted and reported this review in accordance with published guidelines.23,24
Searching
From the ∗ Division of Nephrology, Department of Medicine, Western University, London, Canada; †Department of Epidemiology and Biostatistics, Western University, London, Canada; ‡Department of Medicine, McMaster University, Hamilton, Canada; and §Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, Canada. Disclosure: For this study, ethical approval was not required. The authors declare no conflicts of interest. Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s Web site (www.annalsofsurgery.com). Reprints: Rey R. Acedillo MD, London Kidney Clinical Research Unit, Room ELL-101, Westminster; and London Health Sciences Centre, 800 Commissioners Rd East, London, Ontario, Canada N6A 4G5. E-mail: [email protected]. C 2013 by Lippincott Williams & Wilkins Copyright ISSN: 0003-4932/13/25806-0901 DOI: 10.1097/SLA.0000000000000244
We searched MEDLINE and EMBASE for publications from January 1990 to May 2011. An experienced information specialist developed search strategies using sensitive terms for identifying studies of perioperative risk. We pilot-tested and modified search strategies to identify known eligible articles. The final search strategies included the terms for surgery; intra-, peri-, and postoperative, chronic kidney disease; glomerular filtration rate (GFR); end-stage renal disease; dialysis; bleeding; and blood transfusion. We modified identified terms to correspond to the tree structure and descriptors of different databases (strategies are provided in the Supplemental Digital Content Appendix, available at http://links.lww.com/SLA/A474). We also compiled citations from information provided from citation tracking in SCOPUS and related articles in PubMed and Google Scholar,
Annals of Surgery r Volume 258, Number 6, December 2013
www.annalsofsurgery.com | 901
Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 258, Number 6, December 2013
Acedillo et al
FIGURE 1. Study selection. ∗ Discrete cohorts within 1 article are defined as separate studies. †A study may have been excluded for more than 1 reason. ‡Discrete cohorts were calculated (where possible) from data within case-control articles to analyze the effect of exposure on outcome. §Bleeding defined as a perioperative red blood cell transfusion or surgically related bleeding event. ¶Articles were accepted if the exposure cohort was different from other studies despite using the same database for the referent cohort. CKD indicates chronic kidney disease. as well as reference lists from selected previous reviews9,11 and eligible studies.16,18,25–45 All citations were downloaded into Reference Manager Sente (version 6.5). Two reviewers independently screened each citation and the full text was retrieved for any article considered potentially relevant by either reviewer. Each reviewer independently evaluated the eligibility of each retrieved full text and disagreements were resolved by consensus. When data from the same group of patients or database were described in multiple publications,18,31,42,46,47 we reviewed all publications and included the most representative one.18,31,42
Selection We included all adult cohort studies that compared the risk of perioperative bleeding in patients with and without chronic kidney disease undergoing any major surgical procedure. Although the design of some studies was described as case-control, there were sufficient data to analyze a group of patients having surgery as a cohort. We accepted primary author–defined measures of chronic kidney disease unless the renal impairment was acute or transient. The sample size needed to consist of at least 100 patients with chronic kidney disease. Their outcomes had to be compared to a reference group of at least 100 patients without chronic kidney disease. The study had to report at least 1 of the 2 primary study outcomes: (1) receipt of perioperative blood transfusion and (2) need for reoperation for 902 | www.annalsofsurgery.com
reasons of bleeding. Only studies published in English language were eligible for review. To best reflect modern surgical practices, we focused only on articles published after 1990. We excluded studies that reported only estimated blood loss given potential inaccuracies of this measurement. We also excluded studies with surgical procedures that involved the kidney, such as renal transplant, renal artery bypass, or nephrectomy.
Data Extraction and Validity Assessment Two reviewers independently extracted data from included studies on methods, type of surgery, patient characteristics, definition of chronic kidney disease, and outcomes. Any discrepancies were resolved by consensus. Using recommended validity criteria, 1 reviewer assessed the methodological quality of each study.48 Chronic kidney disease was defined by the primary authors using a variety of measures: estimated glomerular filtration rate (eGFR), creatinine clearance, serum creatinine, or dialysis. When outcomes were reported for more than 2 different categories of renal function, we extracted data for each stratum. We combined study results to compare patients with normal kidney function with those with chronic kidney disease in a dichotomous fashion, using the first cutoff value of chronic kidney disease defined by each study. When possible, we collected data on patients on dialysis separately. For all relevant outcomes of interest, we extracted the raw number of events, the unadjusted association C 2013 Lippincott Williams & Wilkins
Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
C 2013 Lippincott Williams & Wilkins
United States United States
United States
United States
Anderson et al16 (1999) Anderson et al26 (2000)
Charytan et al27 (2010)
Chirumamilla et al28 (2008) Cooper et al29 (2006) Devbhandari et al30 (2006) Filsoufi et al31 (2008)† Foot et al32 (2009) Holzmann et al33 (2005) Ib´an˜ ez et al34 (2007)
China United States United Kingdom United States United States
United States
United States
Lin et al37 (2009) Liu et al18 (2000) Murphy et al38 (2007)§ O’Brien et al39 (2002) O’Hare et al40 (2003)
O’Hare et al41 (2004)
Rahmanian et al42 (2008)† Robinson et al43 (2007)§ Whitson et al44 (2007)§ Witczak et al45 (2005) 10560 2497 106
6449
11051
1288 194 106
245
5943
3817 279 623 6259 4925
313
1345
135 6331 4667 207
379034 386
823
724
683 197
322
No. Patients With Reduced Kidney Function
65 (12) 64 (12) 64 (10)
...
...
... ... ... 62 ( . . . ) ...
...
62 (12)
... 64 ( . . . ) 61 (9) 68 ( . . . )
... ...
...
...
64 ( . . . ) 66 ( . . . )
64 (10)
Mean Age (SD), yr
... ... 83
62
99
84 72 ... 96 99
79
...
62 71 81 55
72 79
74
72
99 97
75
Male (%)
To convert serum creatinine value in units of μmol/L to mg/dL, divide value by 88.4. ∗ Calculated using the Modification of Diet in Renal Disease formula. †Articles were accepted if the exposure cohort was different from other studies despite using the same database for the control cohort. ‡Calculated using the Cockroft-Gault formula. §Discrete cohorts were calculated (where possible) from data within case-control articles to analyze the effect of exposure on outcome. ¶Training data set was used. Validation set/data could not be converted into discrete cohorts of interest to compare bleeding outcomes. . . . indicates data not reported in primary article; CABG, coronary artery bypass graft.
Australia United States Norway
292
Japan 1742 15271 7975 42822 13292
5306
Canada
6314 1109 1888 474
104880 19172
229
1714
3271 637
3276
No. Patients With Normal Kidney Function
Karkouti et al36 (2006)§¶,e Kinoshita et al35 (2010)
United States Australia Sweden Spain
United States United Kingdom
Kuwait
Country
Al-Sarraf et al25 (2011)
Authors, Year of Publication
TABLE 1. Baseline Characteristics of Included Studies
CABG CABG CABG Cardiac Cardiac CABG Cardiac CABG
GFR 133 μmol/L
Cardiac Cardiac Cardiovascular
CABG CABG Cardiac General surgery Lower extremity revascularization Lower extremity amputation Cardiac
CABG
Cardiac
Type of Surgery
Cr >125 μmol/L
Definition of Reduced Kidney Function
Reoperation Blood transfusion Blood transfusion, reoperation
Reoperation
Blood transfusion
Blood transfusion, reoperation Reoperation Reoperation Blood transfusion Blood transfusion Blood transfusion
Reoperation Reoperation Reoperation Blood transfusion, reoperation Blood transfusion
Reoperation Reoperation
Bleeding (tamponade, mediastinal, pericardial), blood transfusion Reoperation
Blood transfusion, reoperation Blood transfusion Blood transfusion
Perioperative Bleeding Outcome Studied
Annals of Surgery r Volume 258, Number 6, December 2013 CKD and Perioperative Bleeding
www.annalsofsurgery.com | 903
Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 258, Number 6, December 2013
Acedillo et al
TABLE 2. Methodological Quality of Included Studies Authors, Year of Publication
Cohort Study Design
Representative Sample
Selection Criteria Defined
Loss to Loss to Follow Follow up up Reported Reported 110 μmol/L for men, Cr > 100 in women. n.r. indicates study did not report whether patients on dialysis were included. kidney disease, it was 1.8%. The unadjusted OR was 2.7 (95% CI, 1.6–4.5). There was significant statistical heterogeneity for the ORs reported across each of the non–cardiac studies (χ 2 = 21.65, P < 0.0001, I2 = 91%). When both cardiac and non–cardiac studies were combined, the weighted incidence in patients with normal kidney function was 1.1% and in patients with chronic kidney disease, it was 2.5%. The unadjusted OR (cardiac and noncardiac surgery combined) ranged from 1.7 to 3.5, with a pooled OR of 2.5 (95% CI, 1.7–3.6). There was significant statistical heterogeneity for the ORs reported across each of the studies (χ 2 = 25.31, P < 0.0001, I2 = 88%). Four studies (2 cardiac surgery, 2 non–cardiac surgery) examined the risk of perioperative red blood cell transfusions in patients with chronic kidney disease, adjusted for baseline, preoperative, and operative characteristics (Tables 3 and 4).25,34,39,40 The factors included in the multivariate model varied greatly across studies. In the non–cardiac study by O’Hare et al,40 the adjusted OR in patients with an eGFR between 30 and 59 mL/min/1.73 m2 , compared with patients with an eGFR of 60 mL/min/1.73 m2 or greater, was 1.34 (95% CI, 1.02–1.76). The non–cardiac study by O’Brien et al39 showed that the adjusted OR for patients with a Cr greater than 265 μmol/L, compared with those with a Cr less than 133 μmol/L, was 1.71 (95% CI, 1.2–2.45). The cardiac study by Al-Sarraf et al25 showed that the adjusted OR in patients on dialysis compared with those with a Cr less than 125 μmol/L was 8.4 (95% CI, 3.8–18.8). The cardiac study by Ib´an˜ ez et al34 showed that the adjusted OR in patients with an eGFR less than 60 mL/min/1.73 m2 , compared with patients with an eGFR of 60 mL/min/1.73 m2 or greater, was 1.09 (95% CI, 1.01–1.7).
Risk of Reoperation for Reasons of Bleeding The risk of reoperation for reasons of bleeding was described only in cardiac surgery and is shown in Figure 4. Fourteen studies totaling 569,715 patients reported this risk.18,25,28–35,37,42,43,45 The weighted incidence of reoperation in patients with normal renal func906 | www.annalsofsurgery.com
tion was 2.4% and in patients with chronic kidney disease, it was 2.7%. The unadjusted ORs ranged from 0.7 to 3.0, with a pooled OR of 1.6 (95% CI, 1.3–1.8). There was significant statistical heterogeneity for the ORs reported in each of the studies (χ 2 = 27.02, P = 0.02, I2 = 52%). Five cardiac studies examined the risk of reoperation for reasons of bleeding, adjusted for baseline, preoperative, and operative characteristics (Tables 5 and 6).18,30,31,33,42 Factors included in the multivariate model also varied greatly across studies, with 1 study using a propensity score model.30 The adjusted ORs ranged from 0.63 to 1.5. None of these ORs reached statistical significance.
Perioperative Outcomes in Patients on Dialysis Perioperative bleeding risk in patients on dialysis is outlined in Figure 5. Two studies totaling 1493 patients on dialysis reported the risk of more than 4 units of red blood cell transfusions (Fig. 5A).40,41 Both studies were conducted in the non–cardiac surgery setting. The weighted incidence in patients with normal renal function was 1% and in patients on dialysis, it was 1.7%. The pooled OR was 2.7 (95% CI, 1.6–4.4). There was mild statistical heterogeneity (χ 2 = 1.40, P = 0.24, I2 = 28%). Five studies totaling 7788 patients on dialysis examined the risk of reoperation for bleeding in patients on dialysis (Fig. 5B).18,25,29,32,42 All the studies were done in cardiac surgery. The weighted incidence in patients with normal renal function was 2.3% and in patients on dialysis, it was 3.4%. The pooled OR was 1.6 (95% CI, 1.3– 1.9). There was mild statistical heterogeneity (χ 2 = 4.24, P = 0.37, I2 = 6%).
Severity of Chronic Kidney Disease and Perioperative Bleeding The relationship between the severity of chronic kidney disease and perioperative bleeding (defined in 3 ways) was C 2013 Lippincott Williams & Wilkins
Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 258, Number 6, December 2013
CKD and Perioperative Bleeding
TABLE 3. Adjusted Risk of Perioperative Blood Transfusions Authors, Year of Publication Al-Sarraf et
al25 (2011)∗
Ib´an˜ ez et al34 (2007)∗ O’Hare et
al40
(2003)‡
O’Brien et al39 (2002)‡
Definition of Bleeding
Definition of Renal Function
No. Patients
Blood transfusion
Cr 125 μmol/L Dialysis GFR >60 mL/min/1.73 m2 † GFR 60 mL/min/1.73 m2 † GFR 30–59.9 mL/min/1.73 m2 GFR 4 units RBC
>4 units RBC
Adjusted Odds Ratio (95% CI)
424 42822 4945§ 1314
P
1.3 (0.99–1.7) 8.4 (3.8–18.8)
0.064 200 μmol/L Cr ≤ 221 μmol/L∗ Cr > 221 μmol/L GFR ≥ 90 mL/min/1.73 m2∗ GFR 60–90 mL/min/1.73 m2 GFR 30–60 mL/min/1.73 m2 GFR < 30 mL/min/1.73 m2 Nondialysis∗ Dialysis Nondialysis∗ Dialysis
(2006)
Filsoufi et al31 (2008) Holzmann et al33 (2005)
Liu et al18 (2000) Rahmanian et al42 (2008)
Adjusted Odds Ratio (95% CI)
P
1.5 (0.9–2.7)
0.14
1.4 (0.55–3.55)
0.476
19172 386 6314 135 1888 3212 1391 64 15271 279 6449 245
0.9 (0.6–1.3) 1.0 (0.6–1.7) 1.4 (0.5–3.9)
n.r. n.r. n.r.
1.3 (0.69–2.5)
0.401
0.62 (0.23–1.73)
0.364
∗
Threshold for each study’s definition of normal and abnormal renal function. n.r. indicates not reported.
TABLE 6. Factors Considered in the Multivariate Models for the Risk of Reoperation for Reasons of Bleeding After Cardiac Surgery in Patients With Chronic Kidney Disease
Source∗ Definition of CKD
Devbhandari et al30 (2006) Filsoufi et al31 (2008) Cr > 200 μmol/L Cr > 221 μmol/L
Risk of reoperation for bleeding Demographics Age Sex Prior cardiac surgery Comorbidities Hypertension Diabetes Coronary artery disease/ACS Congestive heart failure Cerebrovascular disease Dyslipidemia Peripheral vascular disease Liver disease Cancer Respiratory disease Obesity/weight Body surface area Preoperative characteristics Anemia Ejection fraction/LV function Hemodynamic stability Intra-aortic balloon pump Antiplatelet/anticoagulant Operative characteristics Procedure type Surgical priority Comorbidities Hypertension Diabetes Coronary artery disease/ACS Congestive heart failure Cerebrovascular disease Dyslipidemia Peripheral vascular disease Liver disease Cancer Respiratory disease Obesity/weight Body surface area
√ √ √ √ √ √ √
√ √
√ √ √ √ √ √ √
Holzmann et al33 (2005) GFR < 90 mL/min/1.73 m2 √ √
√ √
√ √ √
√
√ √ √
√ √ √
√ √
Rahmanian et al42 (2008) Dialysis
√ √ √
√ √ √
√ √
√ √ √ √
√ √ √ √ √ √ √
√ √ √ √
Liu et al18 (2000) Dialysis
√ √ √ √ √
√ √
√ √ √
√
√ √
√ √ √ √
√ √ √
√ √
√ √ √ √ √ √ √ √
∗
Not all baseline characteristics, preoperative characteristics, and operative characteristics used in the multivariate analysis within each study are listed (see references). ACS indicates acute coronary syndrome; CKD, chronic kidney disease; LV, left ventricle.
C 2013 Lippincott Williams & Wilkins
www.annalsofsurgery.com | 909
Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 258, Number 6, December 2013
Acedillo et al
FIGURE 5. Unadjusted risk of perioperative bleeding in patients on dialysis. Top figure (A) represents patients requiring red blood cell transfusions in the non–cardiac surgery setting. Bottom figure (B) represents the risk of reoperation for reasons of bleeding in the cardiac surgery setting. The size of each square is proportional to the variability of the study estimate. n represents the number of patients who had a perioperative bleeding event and N represents the number of patients in the cohort. ∗ Articles were accepted if the exposure cohort was different from other studies despite using the same database for the control cohort.
FIGURE 6. GFR and perioperative bleeding. The area of each circle is proportional to the number of patients in GFR strata for each study. Meta-regression was used to create best-fit lines with 95% CIs. Four studies in cardiac surgery describe the risk of receiving red blood cell transfusion. Three studies in non–cardiac surgery describe the risk of receiving at least 4 units of red blood cell transfusions. Six studies in cardiac surgery describe the risk of reoperation for reasons of bleeding.
910 | www.annalsofsurgery.com
C 2013 Lippincott Williams & Wilkins
Copyright © 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 258, Number 6, December 2013
CKD and Perioperative Bleeding
TABLE 7. Perioperative Transfusion Requirements in Cardiac Surgery Study Size
Definition of Renal Function
Units of RBC Transfusions
(1999)
3954
μmol/L∗
Anderson et al26 (2000)
834
al34
681
1.9 3.5 4.1 3.3 6.0 2.3 (2.6)‡ 3.8 (3.6)‡ 0 (0–2)|| 2 (0–4) 3 (2–5) 2.3 (4.1)‡ 5.4 (6.5)‡
Authors, Year of Publication Anderson et
Ib´an˜ ez et
al16
(2007)
Charytan et al27 (2010)¶
2438
Witczak et al45 (2005) 212
Cr 265 μmol/L Cr 60 mL/min/1.73 m2∗ GFR 60 mL/min/1.73 m2∗ GFR >30–59.9 mL/min/1.73 m2 GFR
