OUP user menu

Safety of clopidogrel being continued until the time of coronary artery bypass grafting in patients with acute coronary syndrome: a meta-analysis of 34 studies

Sukhjinder S. Nijjer, Geoff Watson, Thanos Athanasiou, Iqbal S. Malik
DOI: http://dx.doi.org/10.1093/eurheartj/ehr151 2970-2988 First published online: 24 May 2011


Aims Guidelines suggest that patients should discontinue clopidogrel for 5 days prior to coronary artery bypass grafting (CABG) where possible. Those with acute coronary syndrome (ACS) are at elevated risk of further myocardial infarction (MI) and death without clopidogrel. This meta-analysis aims to determine the risk of CABG in ACS patients while continuing clopidogrel.

Method and results Thirty-four studies with 22 584 patients undergoing CABG were assessed. Patients with recent clopidogrel exposure (CL) were compared with those without recent clopidogrel (NC). Although mortality is increased in CL vs. NC [odds ratio (OR) 1.6, 95% CI 1.30–1.96, P < 0.00001], it is influenced by the ACS status and case urgency in these mainly non-randomized studies. In ACS patients, there is no significant difference in mortality (OR 1.44, 95% CI 0.97–2.1, P= 0.07) or in postoperative MI (OR 0.57, 95% CI 0.31–1.07, P = 0.08) and stroke rates (OR 1.23, 95% CI 0.66–2.29, P = 0.52). Combined major adverse cardiovascular event (stroke, MI, and death) was not different in the two groups (OR 1.10, 95% CI 0.87–1.41, P= 0.43). Reoperation rates are elevated on clopidogrel but have reduced over time, and were specifically not different in ACS patients (OR 1.5, 95% CI 0.88–2.54, P= 0.13).

Conclusion Previous studies focused on surrogate endpoints and compared higher risk ACS patients with elective cases. However, many patients have safely undergone CABG on clopidogrel and surgical expertise is growing. Multinational trials are required to fully determine the balance of ischaemia and bleeding. While results are awaited we suggest ACS patients requiring urgent CABG proceed with surgery without delay for a clopidogrel-free period.

  • Coronary artery bypass surgery (CABG)
  • Acute coronary syndrome (ACS)
  • Clopidogrel
  • Bleeding
  • Myocardial infarction
  • Mortality
  • Transfusion


Clopidogrel, a thienopyridine that antagonizes ADP-mediated platelet activation, is well established in the treatment of Acute coronary syndromes (ACSs)1,2 and percutaneous coronary intervention (PCI).3 The majority of ACS patients can be managed medically or with PCI; both groups require aspirin lifelong and clopidogrel for 1 year.2,3 However, 10–15% require surgical revascularization and concerns remain regarding the continuing use of clopidogrel until coronary artery bypass grafting (CABG).4

Observational studies suggest an increased bleeding risk and resternotomy rates if clopidogrel is not discontinued prior to surgery.5 Thus, current guidelines suggest that clopidogrel should be discontinued for 5 days prior to CABG when clinically feasible.6,7 Other reviews and meta-analysis have supported this,810 but focus predominantly patients undergoing elective CABG and not on ACS patients requiring urgent surgery. These patients are likely to be operated upon with recent clopidogrel exposure. Furthermore, registry data on current clinical practice suggest many non-urgent ACS patients undergo CABG without waiting for a clopidogrel-free period.11 Therefore, there remains the specific question of timing clopidogrel discontinuation in the ACS patient who remains at elevated risk of repeat myocardial infarction (MI) and thromboembolic events while awaiting CABG. This ischaemia must be balanced against the bleeding risk post-operatively.

This systematic review aims to determine the risk of mortality, reoperation, perioperative MI, and stroke in the ACS population undergoing surgery while on clopidogrel using all the currently available data. We provide a synthesis of the data and introduce a management algorithm.


Search strategy

The Mesh terms ‘clopidogrel’, ‘coronary artery bypass’, ‘reoperation’ ‘haemorrhage’, ‘MI', and ‘cerebrovascular accident’ were searched in Medline and Embase databases (Figure 1). This was expanded using reference lists from retrieved articles. This meta-analysis was planned, conducted, and reported in accordance with official guidelines for reporting meta-analyses of randomized and observational studies.1214

Figure 1

Embase and medline search strategy.

Data extraction

Two reviewers reviewed all abstracts and extracted the data from the full texts of each eligible study (shown in the results tables). Disagreements regarding data collection were resolved by a third author.

Eligibility criteria

The exposure of interest was clopidogrel use prior to CABG (CL) with the control group being clopidogrel naive or those with a drug-free period prior to surgery (NC). The outcomes were reoperation rates, mortality, postoperative MI, and stroke. Studies were included if they reported at least one outcome measure in such a way that a comparison could be drawn.

Definition of endpoints

Mortality was accepted as reported. Reoperation specifically for bleeding was sought but not always specifically reported. Myocardial infarction was defined by studies as ‘perioperative’ or ‘postoperative’: few gave precise definitions or how MI was diagnosed. Cerebrovascular accident was defined as reported by the studies. Both haemorrhagic and ischaemic stroke were accepted as few differentiated.

Quality assessment

A star rating system was developed to assess the importance of ACS patients in each study (Box 1).15 Newcastle Ottawa quality16 scoring assessment produced identical results as another published meta-analysis8 and is not presented to avoid redundancy.

Risk of bias analysis

A risk of bias analysis was performed, specifically assessing allocation concealment, blinding, assessment of incomplete data, and freedom from selective reporting. Only four studies are randomized controlled trials (RCTs) and met criteria for freedom from bias.27,30,40,45 The remainder observational studies cannot exclude bias and score poorly.

Box 1 Star rating checklist to assess study quality

Study rating checklistYesNo
 1. Prospective data collection1 starNo star
 2. States inclusion of urgent and emergency cases1 starNo star
 3. States inclusion of ACS patients2 star if gives proportion; 1 star if includedNo star
 4. Both groups have more than 50 patients1 starNo star
 5. Two groups are comparable for baseline variables2 star; 1 star if variables not reported but others are controlledNo star
 6. Two groups are comparable for ACS variables2 star; 1 star if variables not reported but others are controlledNo star
 7. Both groups have comparable Asprin use2 starNo star
 8. Both groups have comparable aprotonin use2 starNo star
Outcome Assessment
 9. Hard outcomes assessed (mortality, reoperation)2 star if both; 1 star if oneNo star
 10. Outcomes suitable for MACE assessment (MI, CVA)2 star if both; 1 star if oneNo star
 11. Clear statement of duration of follow-up1 starNo star

Baseline variables: Age, Female, BMI, Urgent/Emergent Surgery, Number of bypass vessels, time on CPB.

ACS Variables: recent ACS, previous MI, previous PCI, history of CHF.

Statistical analysis

Our meta-analysis combines the odds ratio (OR) estimates for mortality, reoperation, MI, and stroke for each study using the random effects model, which accounts for both within- and between-study variation.17 Aggregation of the overall rates of the outcomes of interest was performed with the Mantel–Haenszel Chi-square test.18 The quantitative and graphical methods used to assess, explain, and account for determinants of meta-analytical validity, heterogeneity, and bias have been described previously.15,19 Uni- and multivariate meta-regression was performed for the main outcomes of mortality, reoperation, post-operative MI, and stroke using the logarithm of the OR as the dependent variable. Interaction analysis was performed creating interaction terms between the variable ACS status and the following variables: year of study, the use of on-pump surgery, patient urgency status and use of concomitant antiplatelet agents, including aspirin and GPIIbIIIa antagonists. All the statistical tests were two sided with a threshold of significance being a P< 0.05. Review Manager (RevMan) Version 5.0 (Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2008) and Stata/SE 10.0 for Windows (Stata Corp LP; College Station, Tex) were used to perform the statistical analysis.


Study details

The search strategy is presented in Figure 1. Thirty-four studies were included with a total of 22 584 patients (Table 1).5,11,2051 The majority are not blinded with patients treated clinically with full knowledge of prior clopidogrel exposure. Twenty-nine studies were observational reflecting on local experience: 17 were retrospective5,11,22,23,28,32,34,35,37,39,4144,4850 and 12 were prospective.20,21,2426,29,31,33,36,38,47,51 Three studies were subgroup analyses from larger ACS RCTs.27,40,45 Two studies were small RCTs assessing clopidogrel exposure before surgery.30,46 Inclusion and exclusion criteria were not uniformly reported11,22,27,29,39,40,41,44,45,50 making it unclear if higher risk patients were included. Details regarding the operative procedure and pro-coagulant use were greatest in the smaller observational studies.

View this table:
Table 1

Study characteristics

First authorStudy typeGeographical locationPeriod of studyFollow-upStates consecutivePatient/surgery statusACS includedInclusion criteriaExclusion criteriaPrimary endpointSecondary endpointn assessedMean age (SD)% Female
Yende 2001ProsMemphis, USA06/1999–01/20001YAll comersnsK, M, N, O, PTBG, H2455119462.6 (11.7)62.9 (9.7)41.234.5
Hongo 2002ProsSan Francisco, USA03/1999–06/2000nsYnsnsKP, T, N, V, WIB–F2245916566.9 (12.4)67 (10.6)28.826.1
Gansera 2003RetroMunich, Germany01/2000–09/2002nsYUrgent/emergentnsK, O, PnsIB, G128646465 (8)63 (4)21.915.6
Ray 2003RetroHamilton, Canada01/2000–04/2002nsNnsnsKP, T, NGD, E, F1514610565.8 (12.2)66.7 (8.8)30.432.7
Chen 2004ProHouston, USA09/1999–2000nsNElectiveNKX, Y, ZGD, E, F90454558 (8.7)58 (10.1)31.122.2
Chu 2004ProOntario, Canada07/1999–04/20011YUrgent/emergentYO, PT, XB, G, IJ2734123268.3 (10.7)64.9 (11.0)3726
Englberger 2004ProVienna, AustriansnsYnsYKP, T, N, WBG, I50513636966 (10.6)68 (8.4)23.522.5
Fox 2004RCT482 sites, 28 countries12/1998–09/20001NACS patientsYnsnsBns1822912910nsnsnsns
Hekmat 2004RetroCologne, GermanynsnsYAll comersnsKP, T, WGns29014514564 (9)62 (8)119
Karabulut 2004ProIstanbul, Turkey1999–2002nsnsAll comersnsKnsGI162848158066.4 (8.2)59.7 (9.6)3121
Akowuah 2005RCTSheffield, UK06/2002–07/2003nsNACS patientsYKV, X, YGI49252464.2 (9.8)68 (11)2017
Ascione 2005ProBristol, UK01/2001–01/2002nsYUrgent casesYK, LPJB, D, E, F3379124665 (56–70)65 (58–72)1722
Kapetanakis 2005RetroWashington, USA01/2000–06/20021YAll comersnsK, OP, T, V, WBD, E, F2359415194463.9 (11.1)65.0 (10.6)30.826.3
Leong 2005ProAdelaide, Australia07/2000–06/20031nsAll comersnsKTBD, E, F, H, I8107773360.8 (12.7)64.1 (10.4)27.223.2
Nurozler 2005RetroIzmir, Turkey10/2002–10/2004nsYAll comersYK, O, PL, N, V, W, YGD, E, F117496864 (2.4)62 (2.2)2925
von Heymann 2005RetroBerlin, GermanynsnsYnsnsKL, M, V, WIB, D, E, F2613622567 (58–72)66 (59–72)23.130.6
Mehta 2006Retro264 sites, USA01/2003–09/2004nsnsnsYKnsG2565739182665 (56–74)65 (56–74)3028.6
Ouattara 2007ProParis, France11/2003–05/2004nsYAll comersYKP, W, XIB, G2176015766 (9)65 (11)23.323.6
Picker 2007RetroCologne, Germany01/2004–09/20041YElectivensKP, THD, E, F80404067 (11)67 (10)1518
Shim 2007ProSeoul, South Korea11/2005–08/2006nsnsElectivensLP, T, N, V, WID, E, F55203563.7 (8.4)63.8 (8.7)1527.3
Kang 2007RetroBethlehem, USA07/2003–06/2004nsnsnsnsKnsID2802525563.8 (10.6)66.9 (10.1)1630
McLean 2007RCT319 sites, 23 countries02/2003–10/20041nsACS patientsYKnsA, JB1366670nsnsnsns
Berger 2008Retro14 sites, USA01/2004–12/20061nsACS patientsYKT, Y, XBns59629829864.5 (11.24)64.0 (10.88)30.532.9
Filsoufi 2008RetroNew York, USA01/1998–12/20051NnsnsKnsJA–I144727263 (11)63 (11)4040
Maltais 2008RetroMontreal, Canada07/2001–12/2004nsYnsnsLKGns45310135264 (11)64 (10)nsns
Song 2008RetroSeoul, South Korea03/2004–09/2006nsYACS patientsYLKCD, E, F1727010263.5 (9.6)63.4 (9.6)24.342.2
Blasco-Colmenares 2009RetroBaltimore, USA01/2000–06/2003nsnsAll comersnsK, MnsOtherD, E, F1677194148363.4 (11.5)65.6 (10.9)25.827.4
Ebrahimi 2009RCT450 sites, 17 countries08/2003–12/20051NACS patientsYKnsJMany77352424965 (35–87)65 (33–86)2519.7
Firanescu 2009RCTEindhoven, Netherlands06/2006–10/20071YElectivensKV, X, Y, ZGD, E, F, H, I78403862.7 (9.7)65.4 (9.5)1326
Vaccarino 2009ProBuenos Aires, Argentina01/2003–12/2006nsYAll comersYLTBG110412398164.5 (12.9)63 (20)12.816.3
Nesher 2010RetroToronto, Canada04/2005–01/20081YUrgent CABGnsKnsJG45118926266 (11)66 (10)2821
Badreldin 2010RetroCologne, Germany01/2001–12/20041nsElectivensK, NP, T, WB, G, ID, E, F65032532567 (10)68 (10)2621
Herman 2010RetroNovia Scotia, Canada06/2004–08/2008nsYAll comersYK, M, T, O, PnsJG35227422780nsns25.425.2
Dasarathan 2011ProsChennai, India01/2004–12/2008nsYElectivensKL, M, N, XH, IG, B34215119156 (8)54 (8)1213
  • Retro, retrospective study; pro, prospective study; RCT, randomized controlled trial.  A, pre-operative MACE; B, reoperation for bleeding; C, reoperation for other; D, units of RBC transfused; E, units of Platelets transfused; F, units of FFP; G, units of blood products; H, chest drain output over 12 h;  I, chest drain output over 24 h;  J, mortality;  K, conventional CABG; L, off-pump CABG;  M, CABG and valve surgery;  N, redo CABG;  P, emergent surgery;  T, valve surgery alone; V, preoperative exposure to warfarin;  W, preoperative exposure to GPIIbIIIa;  X, bleeding disorder;  Y, end-stage renal disease;  Z, end-stage heart failure.  ns, not stated;  na, not applicable.

Table 2 shows clopidogrel use and recent ACS status, aspirin and GPIIbIIIa exposure. Fourteen studies assessed patients in whom clopidogrel was stopped ≤5–7 days prior to surgery (‘clopidogrel’ CL), comparing them with clopidogrel naive patients and those with a drug-free ‘washout’ period (‘no clopidogrel’ NC).

View this table:
Table 2

Clopidogrel and acute coronary syndrome characteristics of the assessed studies

First authorClopidogrel exposure related to CABGRecent ACSPrevious MIPrevious PCIAspirin useGPIIb/IIIa useRenal failureBody mass index, mean (SD)
CLNCCL (%)NC (%)CL (%)NC (%)CL (%)NC (%)CL (%)NC (%)CL (%)NC (%)CL (%)NC (%)CLNC
Yende 2001≤5 days>5 daysnsnsnsnsnsns94.167nsns3.99.3nsns
Hongo 2002≤7 days>7 daysnsns52.546.1nsns86.447.3nsnsnsnsnsns
Gansera 2003≤5 days>5 daysnsnsnsnsnsnsnsnsnsnsnsns26.7 (3.7)27.3 (2.7)
Ray 2003≤7 days>7 daysnsnsnsnsnsns1000nsnsnsnsnsns
Chen 2004≤6 daysNevernsnsnsnsnsns6449nsnsnsnsnsns
Chu 2004≤4 days>8 days6144nsnsnsns786500.4nsnsnsns
Englberger 2004≤3 daysNever731667395713nsnsnsns81027.6 (2.9)26.9 (4.1)
Fox 2004≤5 days>5 days10010032.432nsns100100nsnsnsnsnsns
Hekmat 2004≤5 days>5 daysnsnsnsnsnsns1000ExcludedExcludednsns26.727
Karabulut 2004≤2 days>2 daysnsnsnsnsnsns5064.9nsns02.327.3 (4.3)27.6 (7.3)
Akowuah 2005≤ 24 h5 days100100nsnsnsns1000nsnsnsns27.2 (3.6)28.3 (4.3)
Ascione 2005≤5 days>5 daysnsns515111147695nsnsnsnsnsns
Kapetanakis 2005≤7 days>7 daysnsns44.835nsns100100nsns4.33nsns
Leong 2005≤7 days>7 daysnsns44.246.916.98.77771.
Nurozler 2005≤ 3 days≥7 daysnsnsnsnsnsns100100nsnsnsns26.9 (3.9)27.2 (3.7)
Von Heymann 2005≤ 2 days>7 daysnsnsnsnsnsns1000ExcludedExcluded1211.1nsns
Mehta 2006≤5 days>5 days10010021.920.119.113.496.694.94539.257.228.3 (25–32)28.5 (25–32)
Ouattara 2007≤2–5 days>5 days33182214nsns100100nsnsnsns26.5 (3.7)26.3 (4.1)
Picker 2007≤1–7 days>8 daysnsnsnsnsnsns1000nsnsnsns28.128.1
Shim 2007≤ 2 days≥6 daysnsnsnsnsnsns1000nsnsnsns2424.2
Kang 2007≤ 3 days>7 daysnsns5229.7nsnsns100nsnsnsnsnsns
McLean 2007≤5 days>5 days100100nsnsnsnsnsns00nsnsnsns
Berger 2008≤5 days>5 days10010028.519.132.915.49489.916.111.119.717.129.1 (6.1)30.2 (5.9)
Filsoufi 2008≤3 days>3 daysnsns503963319088nsns8428 (6)27 (6)
Maltais 2008≤3 days>3 daysnsnsnsns6.910.810010012.99.1nsnsnsns
Song 2008≤ 3 days3–7 days10010031.436.322.919.6100100nsns8.66.9nsns
Blasco-Colmenares 2009≤5 days>5 daysnsns70.655.441.218.3100100nsns5.75.128.7 (6.6)28.4 (6.7)
Ebrahimi 2009<5 days>5 days10010025.226.125.422.198.297.34141.218.916.3nsns
Firanescu 2009Day 0Day 571655861nsns100100nsnsnsns28.3 (4.7)28.5 (5.5)
Vaccarino 2009≤7 days>7 days7363.441.440.2nsns100100nsns1.63.6nsns
Nesher 2010≤5 days>5 daysnsns71462213nsnsnsns56nsns
Badreldin 2010≤7 days>7 daysnsnsnsnsnsns850excludedexcluded1.20.32727.7
Herman 2010≤5 daysNevernsnsnsnsnsnsnsnsnsns7.75.5nsns
Dasarathan 2011≤5 days>5 daysnsnsnsnsnsnsnsnsnsnsnsnsnsns
  • Key; ns, not stated; na, not applicable.  Figures in parentheses are SD or interquartile ranges.

Recent loading doses of aspirin and clopidogrel are rarely reported, but it is presumed that most patients will load with 300–600 mg. Control populations had variability in their use of aspirin; some control groups had no recent antiplatelet use.23,28,30,35,37,38,49 Reporting of aspirin use overall was variable: some studies did not report aspirin use.22,40,48,50 The proportions of control patients with recent clopidogrel exposure but a washout period of ∼7 days were not reported. Overall, insufficient published data were available to determine whether there was an impact according to the precise day that clopidogrel was stopped. It is also unclear whether clopidogrel naive patients had recent ACS. The use of GPIIbIIIa inhibitors, heparins, and thrombolysis was under reported.

Patient age and gender distribution was matched between CL and NC groups. Table 3 shows the quality assessment of the studies. The median was 10.

View this table:
Table 3

Quality of the studies assessed

First authorSelectionComparabilityOutcome assessmentTotal
Yende 2001**********10*
Hongo 2002********8*
Gansera 2003******6*
Ray 20030
Chen 2004*******7*
Chu 2004****************16*
Englberger 2004***********11*
Fox 2004*************13*
Hekmat 2004*********9*
Karabulut 2004********8*
Akowuah 2005***********11*
Ascione 2005**********10*
Kapetanakis 2005*************13*
Leong 2005***********11*
Nurozler 2005*********9*
Von Heymann 2005***3*
Mehta 2006************12*
Ouattara 2007*************13*
Picker 2007******6*
Shim 2007*****5*
Kang 2007*****5*
McLean 2007************12*
Berger 2008****************16*
Filsoufi 2008***********11*
Maltais 2008*********9*
Song 2008*************13*
Blasco-Colmenares 2009********8*
Firanescu 2009***************15*
Ebrahimi 2009***************15*
Vaccarino 2009*************13*
Nesher 2010*************13*
Badreldin 2010**********10*
Herman 2010******6*
Dasarathan 2011****4*
  • Median rating was 10.

Table 4 shows the type of surgery performed, including urgency and cross clamp times. Data were not universally available. Some focus specifically on either CABG24,28,29,34,35,46,49,51 or off-pump surgery (OPCAB).38,42,43,47 Those reporting both often combine the results. Patients are referred to as ‘urgent’ or ‘emergent’ as per the individual publications. It is accepted that urgent refers to inpatient surgery when patient stability allows, while emergent refers to immediate life-saving surgery without preparation. Mean cross clamp times were identical in CL and NC groups (50.7 vs. 50.8 min; with two outliers41,48). Mean cardiopulmonary bypass times were 77.9 vs. 77.0min, respectively.

View this table:
Table 4

Surgical characteristics of the assessed studies

First authorConventional CABGOPCABRedo CABGUrgent surgeryEmergent surgeryNumber of bypass vesselsCardiopulmonary bypass time (min)Cross clamp time (min)Intra-operative anti-fibrinolytic therapy
CL (%)NC (%)CL (%)NC (%)CL (%)NC (%)CL (%)NC (%)CL (%)NC (%)CLNCCLNCCLNCCLNC
Yende 200175.376.5nsns11.813.9nsns9.87.7nsns62.9 (46)69 (50.5)42.9 (35.6)46 (35.5)80.479.9
Hongo 2002nsnsnsns00nsnsexclexclnsnsnsnsnsnsnsns
Gansera 2003100100nsnsnsnsnsnsnsns4.1 (0.6)4.0 (0.9)81 (23)83 (21)59 (16)61 (13)nsns
Ray 2003100100nsnsnsnsnsnsnsnsnsnsnsnsnsnsnsns
Chen 2004100100exclexclnsnsnsnsnsns3.0 (0.2)3.3 (0.1)73 (41.6)63 (24.1)nsns4231
Chu 2004nsnsnsns01100100nsnsnsns92.6 (23.5)89.1 (30.8)55.9 (16.2)53.9 (17.4)8181
Englberger 200483.886.716.213.300nsnsexclexcl3.6 (0.3)3.8 (0.5)67 (15)62 (11)43 (13)46 (17)100100
Fox 2004nsnsnsnsnsnsnsnsnsnsnsnsnsnsnsnsnsns
Hekmat 2004100100exclexclnsns22.10exclexcl3.2 (0.9)3.2 (0.9)93 (32)93 (31)49 (17)49 (14)100100
Karabulut 2004100100exclexclnsns2510.1nsns2.6 (1.1)2.8 (1.1)55 (26)63 (25)31 (11)36 (16)00
Akowuah 2005100100nsnsnsns100100nsns3 (0.7)3 (0.5)67.9 (17)74.9 (15.2)45.2 (14.8)44.2 (7.6)1000
Ascione 2005nsnsnsnsnsns100100exclexclnsnsnsnsnsnsnsns
Kapetanakis 200532.335.667.764.465.629.217.3exclexcl3.33.366.5 (39.5)68.0 (43.8)nsns1.71.2
Leong 200590881012nsnsnsns2.46.7nsns55.7 (16.6)56.4 (39.4)34.1 (12.0)32.9 (10.7)nsns
Nurozler 2005100100exclexcl00100100nana2.7 (1.0)2.8 (1.1)42 (14)38 (13)25 (10)28 (11)00
Von Heymann 200591.795.6exclexcl8.34.4nsnsnsns2.5 (1.9–3.1)2.6 (2.0–3.3)80.5 (59.0–93.0)79.1 (65.4–91.9)45.5 (35.2–55.8)47.6 (38.7–56.3)nsns
Mehta 2006nsnsnsnsnsnsnsnsnsnsnsnsnsnsnsnsnsns
Ouattara 2007nsnsnsnsnsns53exclexclnsns73 (26)73 (20)58 (22)57 (17)100100
Picker 2007100100nsnsexclexclexclexclexclexcl2.7 (0.7)3 (0.9)83 (24)82 (30)47 (15)44 (6)100100
Shim 2007exclexcl100100exclexclexclexclexclexcl3.3 (0.9)3.3 (0.7)nananana00
Kang 2007nsnsnsnsnsnsnsnsnsnsnsnsnsns46 (21)37.4 (36.4)nsns
Berger 200872.172.527.927.55.74.468.865.112.173.4 (1.1)3.7 (1.2)95.6 (41.20)94.2 (36.12)nsns66.155.7
McLean 2007100100nsnsnsns100100nsnsnsnsnsnsnsnsnsns
Filsoufi 200874742626nsns54451083.3 (1.2)3.3 (1.3)125 (49)112 (40)95 (32)95 (28)80
Maltais 2008exclexcl100100nsnsnsns11.99.13.18 (0.88)3.19 (0.88)nsnsnsnsnsns
Song 2008exclexcl100100nsns100840163.1 (0.7)3.3 (0.8)nsnsnsnsnsns
Blasco-Colmenares 2009nsnsnsns11.34.8nsns4.65.2nsnsnsnsnsnsnsns
Ebrahimi 200984.3ns12.7nsnsnsnsnsnsnsnsnsnsnsnsnsnsns
Firanescu 2009100100exclexclexclexclexclexclexclexcl4 (1)4 (0.9)68.2 (26.3)71.1 (30.2)46.2 (18.9)50.4 (21.5)100100
Vaccarino 2009exclexcl100100nsns76.533.9nsns3.05 (0.9)3.02 (0.8)nananananana
Nesher 2010nsnsnsns0.20.4100100nsnsnsns115 (34)109 (31)97 (30)90 (28)100100
Badreldin 201097.698.5exclexcl2.41.5exclexclexclexcl2.9 (0.9)3.0 (0.9)84 (30)83 (29)45(16)46 (16)100100
Herman 201091.565.1nsns4.98.421.77.382nsnsnsnsnsnsnsns
Dasarathan 2011100100exclexclexclexclnsnsnsns3.6 (0.7)3.3 (0.7)95.3 (25)88.8 (23.5)49.8 (14.5)48.3 (17.1)00
  • ns, not stated; na, not applicable. Excl, excluded by original study. Numbers in parentheses are SD or interquartile ranges.

Other important factors that affect bleeding and operative risk include diabetes and chronic renal failure (CRF). Diabetes was well reported and matched. The presence of CRF was poorly defined and the majority report it simply as a percentage of patients with ‘CRF’ (Table 4). Body mass index and body surface area were well matched in all cohorts.

Tables 5 and 6 show the outcomes.

View this table:
Table 5

Transfusion and bleeding outcomes

First authorDrain output (ml)RBC units transfusedPlatelet units transfusedFFP units transfusedCryoprecipitate
MeanSDMeanSDMean unitsSDMean unitsSDMean unitsSDMean unitsSDMean unitsSDMean unitsSDMean unitsSDMean unitsSD
Yende 2001nsnsnsns331.62.14.3010.001.704.401.102.200.601.802.405.901.204.60
Hongo 2002122411198406212.512.411.742.160.861.200.240.600.681.690.240.850.191.310.171.2
Gansera 20039776287883892.
Ray 2003nsnsnsns2.
Chen 2004132910339654904.
Chu 200410497485823244.2ns1.1ns3.50ns0.60ns3.60ns0.70ns0.6ns0.1ns
Englberger 200414853107501054.
Fox 2004nsnsnsnsnsnsnsnsnsnsnsnsnsnsnsnsnsnsnsns
Hekmat 200411235378743512.
Karabulut 20047192656123500.50.90.41nsnsnsns1.101.200.901.10nsnsnsns
Akowuah 20054462877024950.321.21.001.680.
Ascione 20051000675–1450700550–925nsnsnsnsnsnsnsnsnsnsnsnsnsnsnsns
Kapetanakis 2005nsnsnsns26.3241.
Leong 2005720200–378061550–450010–3010–210.000–50.000–70.000–40.000–11nsnsnsns
Nurozler 200511802806601103.
Von Heymann 200518401230–3710280185–7650.54ns0.3ns0.16ns0.06nsnsnsnsnsnsnsnsns
Mehta 2006nsnsnsns2(0–4)1(0–3)nsnsnsnsnsnsnsnsnsnsnsns
Ouattara 2007375150–850350175–8752.871.452.450.957.000.007.503.503.
Picker 20079408614125904.
Shim 20076272577564081.
Kang 200718111223172012585.
McLean 2007nsnsnsnsnsnsnsnsnsnsnsnsnsnsnsnsnsnsnsns
Berger 20086685155573394.
Filsoufi 20089603628515705.
Maltais 2008865312604432nsnsnsnsnsnsnsnsnsnsnsnsnsnsnsns
Song 20086013126274260.
Blasco-Colmenares 20097005175534672.
Ebrahimi 2009600360–860550320–850nsnsnsnsnsnsnsnsnsnsnsnsnsnsnsns
Firanescu 20099294716643120.
Vaccarino 2009892586840537nsnsnsnsnsnsnsnsnsnsnsnsnsnsnsns
Nesher 20107174307354652.232.541.703.001.906.00nsnsnsnsnsnsnsns
Badreldin 20108027205054453.
Herman 2010nsnsnsnsnsnsnsnsnsnsnsnsnsnsnsnsnsnsnsns
Dasarathan 20113892694052830.751.010.741.50.441.020.320.800.741.200.591.3nsnsnsns
  • ns, not stated; na, not applicable. Those with ranges in SD column are median and interquartile ranges.

View this table:
Table 6

Postoperative outcomes

First authorITU stayHospital stayReoperationPostoperative CVAPerioperative MiMortality
Days (SD)DaysDays (SD)Days (SD)Number (%)Number (%)%%%%%%
Yende 20013.3 (ns)3.6 (ns)11.3 (ns)12.1 (ns)5 (9.8%)3 (1.5%)nsnsnsns2 (3.9%)7 (3.6%)
Hongo 2002nsnsnsns4 (6.8%)1 (0.6%)2 (3.4%)8 (4.8%)0 (0%)6 (3.6%)1 (1.7%)6 (3.6%)
Gansera 20031.36 (0.92)1.16 (0.76)nsns5 (7.8%)0 (0%)nsnsnsns0 (0%)0 (0%)
Ray 2003nsnsnsnsnsnsnsnsnsnsnsns
Chen 2004nsns8.3 (1.1)6.6 (0.3)4 (8.9%)1 (2.2%)nsnsnsns1 (2.2%)0 (0%)
Chu 2004nsns976 (14.6%)4 (1.7%)4 (9.8%)7 (3%)2 (4.9%)9 (3.9%)1 (2.4%)10 (4.5%)
Englberger 20041.9 (1.3)1.1 (0.7)9.6 (2.1)7.9 (2.0)8 (5.9%)5 (1.2%)nsns6 (4.3%)11 (3.1%)2 (1.4%)3 (0.8%)
Fox 2004nsnsnsns18 (4.1%)11 (2.3%)nsnsnsnsnsns
Hekmat 20043.5 (5.7)2.8 (5.3)13.3 (9.4)10.2 (4.9)4 (2.8%)5 (3.4%)nsns3 (2.1%)9 (6.2%)4 (2.8%)3 (2.1%)
Karabulut 20040.83 (0.12)0.91 (0.56)5.5 (1.7)5.4 (2.1)0 (0%)16 (1%)nsnsnsnsnsns
Akowuah 20051.35 (1.03)1.93 (2.29)7.3 (3.5)10.9 (12.8)3 (12%)3 (12.5%)nsns002 (8%)0 (0%)
Ascione 2005nsns7 (6–11)6 (5–8)11 (12%)9 (4%)4 (4%)0 (0%)2 (2%)4 (2%)7 (8%)2 (0.8%)
Kapetanakis 20051 (6.8)1 (12.5)5 (20.8)5 (31.8)24 (1.7%)25 (1.3%)nsns9 (2.2%)14 (0.7%)7 (1.7%)27 (1.4%)
Leong 20051 (ns)1 (ns)6 (ns)6 (ns)2 (2.6%)9 (1.2%)nsns1 (1.3%)20 (2.8%)1 (1.3%)6 (0.8%)
Nurozler 20051.6 (0.7)1.1 (0.5)5.7 (1.4)5.4 (1.6)3 (6.1%)1 (1.4%)1 (2%)1 (1.4%)1 (2%)2 (2.8%)nsns
Von Heymann 2005nsnsnsns2 (5.6%)5 (2.2%)nsnsnsnsnsns
Mehta 2006nsns65nsns12 (1.6%)26 (1.4%)11 (1.5%)51 (2.8%)26 (3.5%)53 (2.9%)
Ouattara 2007nsnsnsns02 (1%)nsnsnsnsnsns
Picker 20071.7 (1.3)1.7 (1.4)11.6 (3.9)10.4 (2.3)8 (20%)3 (7.5%)nsns0 (0%)3 (7.5%)1 (2.5%)0 (0%)
Shim 20072.7 (0.7)2.9 (0.7)10.1 (2.2)12.9 (7.0)0 (0%)2 (5.7%)nsnsnsnsnsns
Kang 20072.06 (2.65)2.17 (3.25)nsns2 (8%)11 (4%)nsnsnsns2 (8.0%)8 (3.1%)
McLean 2007nsnsnsns6 (9.1%)10 (14.3%)0 (0%)0 (0%)1 (1.5%)1 (1.4%)3 (4.5%)3 (4.3%)
Berger 20082.7 (3.17)2.4 (2.52)7.2 (5.53)6.3 (3.87)14 (4.7%)4 (1.3%)5 (1.7%)3 (1%)nsns4 (1.3%)1 (0.3%)
Filsoufi 20082.5 (2.7)1.4 (0.9)9.9 (11)6 (2.5)0 (0%)0 (0%)100 (0%)0 (0%)6 (9%)1 (1%)
Maltais 2008nsns8.18 (5.95)7.61 (9.05)7 (6.9%)14 (4.0%)0 (0%)6 (1.7%)2 (2%)13 (3.7%)7 (6.9%)14 (4%)
Song 2008nsnsnsns1 (1.4%)1 (1.0%)nsns1 (1.4%)1 (1.0%)1 (1.4%)0 (0%)
Blasco-Colmenares 2009nsnsnsns5 (2.6%)15 (1%)nsnsnsns5 (2.6%)15 (1%)
Ebrahimi 2009nsns6.905.88 (1.5%)2 (0.8%)nsnsnsns24 (4.6%)6 (2.4%)
Firanescu 20090.7 (0.9)0.4 (0.8)4.7 (1.6)4.9 (1.5)1 (2.5%)0 (0%)nsns0 (0%)1 (3%)0 (0%)0 (0%)
Vaccarino 2009nsns8 (5–10)5 (4–7)5 (4%)14 (1.4%)nsns4 (3.2%)14 (1.4%)6 (4.8%)27 (2.7%)
Nesher 20101.75 (2)1.3 (1.8)9 (6)9 (9)10 (5%)13 (5%)8 (4%)8 (3%)8 (4%)10 (3.8%)6 (3%)2 (2%)
Badreldin 20102.8 (4.6)2.1 (2.2)11.5 (7.2)10.3 (4.9)34 (10.4%)13 (4%)nsns12 (3.7%)15 (4.6%)9 (2.8%)6 (2%)
Herman 2010nsnsnsns36 (5.3%)105 (3.8%)nsnsnsns45 (6.0%)110 (4.0%)
Dasarathan 2011nsnsnsns0 (0%)0 (0%)nsnsnsnsnsns
  • ns, not stated.

Acute coronary syndrome patients

Twelve studies specifically state inclusion of ACS patients.5,11,2527,30,36,40,43,4547 Some also classified unstable angina separately from ACS, while the CURE study considered both confirmed infarction and unstable angina together. Studies also report ‘previous MI’ or ‘recent MI’ without definition.5,11,21,26,27,3133,36,39,41,4348 Percutaneous coronary intervention status is similarly reported. Potentially, some may represent recent ACS, but insufficient information regarding timing is available. The use of clopidogrel-naive patients in the control groups creates significant differences between the CL and NC groups regarding ACS and PCI status at baseline.5,26,32,33,36,39,41,43,44,48 Few specifically reported whether the surgery was performed as inpatient or outpatient after stabilization following ACS, and the times from ACS to surgery are not reported.

Assessment of mortality

Those undergoing CABG with recent clopidogrel exposure have an increased mortality compared with the control populations used (OR 1.6, 95% CI 1.30–1.96, P< 0.00001). Control populations were often elective stable patients. Univariate meta-regression showed that there were marginal trends of the following variables on the log-OR of mortality: study sample size (P= 0.068), female sex (P= 0.055), and a significant effect for ACS status in studies recruiting more than 50% patients with ACS (P= 0.042) (Figure 2). However, when assessed in a multivariate analysis, none of variables was found to have a significant effect including on-pump surgery, renal impairment, weight, aspirin use, glycoprotein use, study quality, aprotonin, and number of vessels grafted. Subgroup analysis of studies recruiting only ACS patients showed a similar increased mortality rate (OR 1.44, 95% CI 0.97–2.1, P= 0.07). The smaller subgroup size has reduced the level of statistical significance. Interaction analysis showed a significant effect of the interaction between urgency of the surgery and recruitment of less than 50% of ACS patients into the study (P= 0.03). Following univariate meta-regression, it was found that the effect size of mortality in studies where ACS patients are not evenly distributed across the CL and NC groups may be affected by the urgency status of the surgery. This is an expected finding. No significant interaction effect was seen when the proportions of ACS patients were balanced in both CL and NC groups and when only ACS was recruited (P> 0.05). Therefore, the ACS status of patients may drive an increase in mortality on those having urgent surgery with recent clopidogrel exposure.

Figure 2

Mortality rates. Studies explicitly reporting mortality rates are shown. (A) Studies that include only acute coronary syndrome patients. (B) Studies that include an average of >50% acute coronary syndrome patients.

Assessment of postoperative myocardial infarction

Two single-centre studies suggest there is an increased risk of MI following CABG performed with recent clopidogrel exposure.31,32 The mechanism is unclear but increased transfusion requirements was postulated. Other studies do not show significantly increased rates of MI in CL groups.2530,34,4044,4749 Assessment of studies recruiting only ACS patients (2873 patients; OR 0.57, 95% CI 0.31–1.07, P= 0.08) or studies with >50% ACS patients (4328 patients; OR 0.91, 95% CI 0.48–1.74, P= 0.79) shows that there is no increased risk of MI (Figure 3). While this may reflect the difficulty in diagnosing post-CABG infarction and the differing diagnostic criteria for MI used among studies, it may reflect the biologically plausible antiplatelet effect of clopidogrel reducing coronary thrombosis. No significant effect was seen for patients performed on bypass (OR 0.89, 95% CI 0.58–1.36; P= 0.58) and off-pump (OR 1.18, 95% CI 0.51–2.73, P= 0.71). Mutivariate meta-regression showed that the effect size of MI (log OR of MI) was not affected by aspirin use, GPIIbIIIa use, quality of study, year of study, sample size, on-pump surgery, urgency status, number of vessels grafted, and aprotonin use in a multivariate model (all P> 0.05). Acute coronary syndrome status did influence the effect size of MI in univariate analysis (P= 0.04), but this was not significant in multivariate analysis. No significant effect from interaction was seen between ACS status with urgency, on-pump surgery, and GPIIbIIIa use (all P> 0.05). No significant interaction effect was seen on the log OR of MI between studies with <50% and >50% ACS patients recruited and the variables: year of study, on-pump surgery, urgency status, and use of concomitant aspirin or GPIIbIIIa antagonists.

Figure 3

Postoperative myocardial infarction rates. Studies explicitly reporting postoperative myocardial infarction rates are shown. (A) Studies that include only acute coronary syndrome patients. (B) Studies that include an average of >50% of acute coronary syndrome patients.

Assessment of postoperative stroke and major adverse cardiovascular event

Fewer studies report postoperative stroke rates. One study specifically described transient ischaemic attacks (TIA).31 It is unclear whether other studies included patients having TIA as 'stroke' or were included at all. Directionally, there is an increase in the stroke rate in the CL group, but it is not significant (OR 1.46, 95% CI 0.91–2.33, P= 0.11). This is also true in the studies recruiting 100% ACS patients (OR 1.23, 95% CI 0.66–2.29, P= 0.52). Acute coronary syndrome status was a predictor of stroke in univariate meta-regression analysis (p = 0.04) and GPIIbIIIa antagonist use demonstrated marginal significance (P= 0.065). Neither was significant in multivariate analysis. No other significant interaction terms were identified.

Assessing a combined major adverse cardiovascular event (MACE) rate (MIs, stroke, and death) after CABG reveals that there is no increased risk in patients with recent clopidogrel exposure (Figure 4; OR 1.10, 95% CI 0.87–1.41, P= 0.43). Caution is required in interpreting this statistic because it is post hoc assessment of published studies but it is reassuring.

Figure 4

Combined major adverse cardiovascular event rate (postoperative death, myocardial infarction, stroke) after coronary artery bypass grafting with recent clopidogrel exposure. Only studies explicitly reporting all three components are shown.

Assessment of chest drain output

The majority of the studies are not adequately powered to find differences in rare outcomes, including reoperation and mortality. Many focus on chest drain output either at 12 or 24 h. Drain output follows a skewed distribution in the majority and is increased in the CL group (mean difference of 219 mL, 95% CI 129–308 mL, P< 0.00001). However, there is marked heterogeneity (I2 97%). Based on surgical experience (using the year that surgery was performed) stratification into early (1999–2002), mid (2003–2006), and recent (>2007) demonstrates that more recent publications have reduced drain output (Figure 5). Recent patient cohorts have an equal amount of drainage in both NC and CL groups, suggesting improving surgical expertise and haemostasis. Heterogeneity prevents firm conclusions and this finding may be due to a smaller subgroup size.

Figure 5

Chest drain output stratified according to the operating experience (year of procedure). Studies explicitly reporting chest drain output are included and shown. Englberger et al.,26 Hekmat et al.28 and von Heymann et al.35 are not included because they do not state when data collection occured. Year of procedures are given in Table 1.

OPCAB studies show a mean difference of drain output of 154 ml (95% CI −80 to 389, P= 0.20) with no clinically significant difference in drain output between NC and CL groups. However, heterogeneity is not removed by any stratification, reflecting the diverse patient population, recording, and reporting of drain output and surgical technique.

Assessment of transfusion

Blood transfusion has been reported as the percentage of patients receiving transfusion, or the units or millilitres given. Transfusions were converted into units where possible. Studies had differing triggers for transfusion creating heterogeneity. Larger RCTs have used variations of ‘TIMI major bleeding’ and do not report transfusion. The majority report packed red blood cell (RBC) transfusion post-operatively. Clopidogrel increases RBC transfusion (by 1.1 units, 95% 0.58–1.64 units, P< 0.0001; Figure 6), platelet (by 0.18 units, 95% CI 0.14–0.21, P< 0.00001), and FFP (by 0.49 units, 95% CI 0.14–0.84, P= 0.006) transfusion requirements. However, heterogeneity for all three variables (I2 94–99%) is raised, reflecting the weakness of this outcome measure.

Figure 6

Blood transfusion—packed red blood cells (RBCs)—requirements post-CABG. (A)–(C) show stratified according to operating experience. Year of procedures are given in Table 1.

Assessment of reoperation

Reoperation rates are increased in the CL group compared with NC group (OR 2.32, 95% CI 1.76–3.06, P< 0.00001). As similarly shown elsewhere,8 the risk of reoperation has diminished with recent surgical experience. Early cases (1999–2002) demonstrate an increased reoperation risk on clopidogrel (OR 3.67, 95% CI 2.49–5.42), while mid-term cases (2003–2006) demonstrate a more favourable risk (OR 1.76, 95% CI 1.26–2.41), and recent cases (2007–2009) no longer show a significant difference (OR 1.14, 95% CI 0.50–2.59). The lack of blinding in the observational studies allows bias and alters surgical behaviour: awareness of clopidogrel use may encourage reoperation. Assessing studies reporting results of predominantly urgent CABG cases (>50% of case-mix) shows an elevated risk of reoperation in the CL group similar to the overall cohort. Stratification for only ACS patients shows a similar but not-significant raised reoperation risk (OR 1.50, 95% CI 0.88–2.54, P= 0.13) (Figure 7). Meta-regression showed no significant effect on the log odds of reoperation from renal impairment, weight, aspirin use, female sex, quality of study, sample size, on-pump surgery, urgency status, number of grafts, and aprotonin use (all P> 0.05). The year of surgery and ACS status had a significant effect in univariate analysis (P= 0.01 and P= 0.02, respectively) with a marginal significance in multivariate analysis for the year of surgery, with higher rates of reoperation seen in earlier studies (P= 0.095). No significant effect from interaction variables was identified.

Figure 7

Reoperation rates. (A) Studies that include only ACS patients. (B) Studies that include an average of >50% of ACS patients.

Assessment of ITU stay

Clopidogrel does increase ITU stay length by a mean difference of 0.31 days (95% CI 0.06–0.56, P= 0.01), but this is unlikely to be clinically significant. Duration of ITU stay is highly variable according to local protocols and bed availability for step-down. Importantly, in many of the studies suggesting large drain output and transfusion requirement, no significant ITU or hospital stay differences were seen suggesting that the clinical impact is small.

Recent RCT data

Within the PLATO study, 1899 patients underwent CABG with 1261 patients having surgery within 7 days of clopidogrel or ticagrelor.52 No aspirin only group was included in PLATO and thus it is not included in our meta-analysis. In 629 patients exposed to clopidogrel before CABG, post-operative MI occurred in 5.7% (35 patients), stroke in 2.1% (11), and all-cause mortality in 9.7% (58). Reoperation due to bleeding was 3.3% (21). Median chest tube output within 24 h was 540 mL (320–810 IQR). 56.5% (351) had transfusion within 7 days of CABG (51.9% packed red cells or whole blood, 17.5% platelets, and 24.1% FFP). Nearly, all the bleeding events occurred within 24 h post-CABG.52 Overall, these findings were consistent with ACUITY study featured in this analysis.

The design of the TRITON-TIMI-38 trial assessing the efficacy of prasugrel compared with clopidogrel in ACS meant that few patients underwent CABG with a recent dose of either drug.53 In the clopidogrel arm, 189 patients underwent CABG with 3.2% (6) having TIMI major bleeding.


This meta-analysis demonstrates that present guidelines are based on limited evidence which do not definitively determine the safety of continuing clopidogrel until CABG in ACS patients. Comparisons have been made of ACS patients requiring urgent surgery with more stable and elective cases. While it is accepted that ACS patients must be operated on aspirin,54,57 many studies have compared bleeding outcomes and reoperation in ACS patients on dual antiplatelet therapy to stable patients on none.23,28,30,35,37,38,41,49 Furthermore, none has considered recent loading doses, known to influence bleeding,55 and few consider the implications of GPIIbIIIa inhibitors, commonly used alongside clopidogrel in high-risk ACS. The majority of the studies are small and retrospective with limited subgroup analysis from RCTs. The bleeding and transfusion data are heterogeneous and with potential for treatment and ascertainment bias. This meta-analysis demonstrates the need for a RCT assessing different discontinuation times prior to CABG to definitively answer the question. In the meantime, the data do demonstrate that many patients have undergone CABG safely with recent clopidogrel exposure and this practice can continue in expert hands in ACS patients who need to continue clopidogrel.

Mortality and reoperation

Accepting its limitations, the data suggest that continuing dual antiplatelet therapy until the day of CABG reduces the risk of recurrent ischaemic events in ACS patients and reduces MI post-operatively. However, there may be opposing effect in post-operative stroke. While mortality and reoperation rates are increased after recent clopidogrel exposure, event rates are still low. Multivariate meta-regression including interaction terms did not find a strong signal that ACS status and urgency of patients drove the mortality increase but even meta-analysis has limited power to detect such signal. Overall MACE rates did not appear to be significantly increased in the clopidogrel group.

Reoperation rates are likely influenced by the experience of operating on clopidogrel and correspondingly recent cohorts have seen a reduction in reoperation rates.46,50,51 In this meta-analysis, taken as a collective registry, a 2.87% mortality rate is seen in those on clopidogrel (in a total of 6019 patients) vs. 1.87% in a mixed control population of 16 565 patients. Reoperation occurs in 3.9 and 1.85% patients, respectively. These rates are driven by the steep learning curve of earlier surgical experience (most studies reflect early experience, circa 1998–20032027,2935,42,44,49) and some recruitment bias in the larger studies.5 They are within acceptable rates of resternotomy of 2–6% from the pre-clopidogrel era.56 The case volume and size of the surgical faculty must also influence reoperation rates. Rigorous attention to haemostasis is required because half of all reoperations reveal a surgical bleeding point suggesting inadequate technique.25 Generalized ooze represents the consequences of dual antiplatelet therapy and responds to blood products. Few studies categorize the findings on reoperation. Previously, surgeons resisted operating on aspirin but adapted with great success making it the standard of care.57,58 It is expected that adaption to operating on aspirin and clopidogrel will also occur.

Postoperative stroke

Directionally, there appears to be an increase in stroke, although it is not statistically significant. We expect haemorrhagic stroke in those on clopidogrel when exposed to bypass and full anticoagulation. However, stroke has been poorly reported in the studies and an increase in TIAs and embolic strokes is difficult to interpret given dual antiplatelet therapy should reduce thrombus. The wide confidence intervals seen likely reflect the small numbers of rare events.

Postoperative bleeding and transfusion

Postoperative bleeding and chest-drain output remain a serious concern. However, as an outcome measure, it is highly variable with some elective series reporting higher levels of output than that seen in the clopidogrel-exposed patients assessed here.59 Adverse outcomes are more associated with blood product transfusion than anaemia post-CABG,60 and minimizing transfusion is essential. Intra-operatively, good haemostasis and appropriate use of platelets will reduce transfusion requirements. Aprotonin may yet find a new role in patients with clopidogrel exposure.61 Algorithms reduce transfusion rates even in the setting of clopidogrel exposure24 and should be encouraged given that transfusion is more strongly associated with the performing surgeon than clopidogrel exposure.62 Clearly, transfusion is a surrogate marker and future studies should focus on hard outcomes.

Pre-operative infarction risks

Patients with ACS have elevated risks of re-infarction prior to revascularization. Patients delayed for a clopidogrel-free period (‘washout’) can suffer prolonged inpatient waits for CABG, which in the UK is often much greater than 5 days and may be weeks.63 Re-infarction during this time may force emergency angioplasty or intra-aortic balloon pump therapy, with potentially suboptimal long-term results for those with established surgical disease. Discontinuation of clopidogrel does trigger a real ‘rebound phenomena’64 and even in stable CAD without the presence of fresh thrombus, early discontinuation of antiplatelets increases the incidence of MI within 10 days.65

While anecdotally many patients have further MIs while awaiting CABG after clopidogrel withdrawal, few studies report the MACE rate for this period. In CURE, patients awaiting CABG having had clopidogrel had numerically lower MACE rates than the placebo group (2.9 vs. 4.7%, 1.8% risk reduction, RR 0.56, 95% CI 0.29–1.08).27 Those having clopidogrel withdrawn are likely exposed to this increased 1–2% risk of MI. Combining the bleeding risk and MACE reduction, CURE suggested 18 pre-CABG events per 1000 patients treated were prevented by treating with clopidogrel prior to CABG.27 In a small study randomizing ACS patients to stopping clopidogrel prior to CABG, 12.5% of the group stopping clopidogrel had an MI during the 5-day wait for CABG.30 The observational studies do not report the MACE rate prior to CABG; these patients may not even enter the surgical registry. There is a need for accurate data to be collected, ideally in an RCT setting. While we await such data, we suggest high-risk patients undergo CABG at the earliest opportunity without a delay for washout and accepting a small increase in mortality and reoperation when compared with elective cases. Clopidogrel should be reinstituted after CABG as soon as chest drain output allows to protect from further events and improve graft survival.66,67

Anti-platelet and genetic testing

Significant literature has been dedicated to the inter-individual variability in response to clopidogrel. Daily platelet-function testing may guide antiplatelet therapy and help find an operating window based on platelet reactivity.6870 Genetic testing may help identify variants of liver enzymes: CYP2C19*2 carriers are hypo-responsive to clopidogrel predisposing recurrent ischaemia, while CYP2C19*17 carriers are hyper-responsive predisposing bleeding.71 In reality, clinical decisions may be forced in urgent cases, but these tests may gain utility in more elective cases.

The future: new antiplatelet agents

The role of third-generation thienopyridines is likely to increase. Prasugrel use prior to CABG lead to a 15% absolute increase in major bleeding compared with clopidogrel (18.8 vs. 2.7%, P= 0.003).72 This dramatic increase suggests that a prasugrel-free period is mandated before CABG. In contrast, PLATO suggested ticagrelor substantially reduced mortality after CABG compared with clopidogrel while maintaining an equivalent bleeding rate. The mechanisms remain undetermined.73 Shorter-acting reversible antiplatelet agents, such as the oral ticagrelor,52 or the intravenous cangrelor may have a role in bridging ACS patients from the acute phase, admission, and investigation until surgery can be performed. This will be delineated in the BRIDGE trial (NCT00767507).74

Clinical implications: a new algorithm

The paradigm continues to be a balance between the risk of preoperative ischaemia and peri-operative bleeding. For elective patients without ACS, the bleeding risk is more important and clopidogrel should be stopped for 5 days as per current guidance. However, we suggest ACS patients with surgical anatomy have CABG at the earliest clinical opportunity, while on clopidogrel without delay for washout (Box 2). Those discussed at a multidisciplinary ‘Heart Team’ meeting and considered at high bleeding risk or those having redo surgery should have clopidogrel withdrawn cautiously. Platelet function testing may help determine an operating window.

In those with drug-eluting stents less than 1 year old, surgeons should consider operating on clopidogrel. If the surgical risk is considered too high, intravenous tirofiban with or without heparin may provide bridging cover until RCTs demonstrate an alternative approach.75

Box 2 Algorithm for treating acute coronary syndrome patients that require coronary artery bypass grafting

Stable elective CABG with no drug eluting stent (DES)Stop clopidogrel >5 days
Stable elective CABG in presence of DES <1 year oldConsider operating on clopidogrel or switch to IV tirofiban ± heparin as bridge to surgery
Non-elective CABG (ACS/urgent/emergent)Operate on clopidogrel unless:
 Redo surgery/bleeding disorder
 Troponin negative
 Multidisciplinary decision


Only five RCTs were included in this analysis and only two focused on addressing the issue of clopidogrel use prior to CABG.30,46 Both are too small to assess relatively rare outcomes. Three studies are subgroups of larger RCTs and may contain bias as data collection did not focus on the outcomes assessed here.27,40,45 Only two specifically randomized to clopidogrel vs. no clopidogrel27,40 and only one sufficient sample size to provide outcome data.27

The remaining studies are small to medium sized registries, with potential for bias. Without access to the individual patient data, we cannot make a detailed assessment of confounders. We are reliant on the original publication's summary statistics. Heterogeneity prevents firm conclusions and highlights the variability in practice and outcomes throughout the world. Viewed positively, this meta-analysis strongly encourages a definitive RCT to specifically answer this question. Based on the expected major adverse (death, MI, CVA) event rate of 5% a trial would require at least 2000 patients with 80% power to detect a relative risk reduction of 0.5. Given adverse outcome rates are low, detecting more realistic but smaller relative risk reductions will necessitate larger sample sizes. This would require international support.


Clopidogrel is well established in ACS, but current guidelines on stopping clopidogrel in ACS patients prior to CABG are based on limited data. Clopidogrel reduces MI in ACS patients prior to revascularization but does increase bleeding. Modern surgical techniques should aim to reduce reoperation rates and minimize unnecessary transfusion. A registry of patients having major cardiovascular events while awaiting inpatient CABG is required. We strongly urge a large international RCT to assess the question of clopidogrel in ACS patients undergoing CABG. Until then we support selected ACS patients undergoing CABG without stopping clopidogrel.

Conflict of interest: none declared.


View Abstract