OUP user menu

Decline in platelet count in patients treated by percutaneous coronary intervention: definition, incidence, prognostic importance, and predictive factors

Axel De Labriolle, Laurent Bonello, Gilles Lemesle, Probal Roy, Daniel H. Steinberg, Zhenyi Xue, William O. Suddath, Lowell F. Satler, Kenneth M. Kent, Augusto D. Pichard, Joseph Lindsay, Ron Waksman
DOI: http://dx.doi.org/10.1093/eurheartj/ehp594 1079-1087 First published online: 19 January 2010


Aims We investigated the incidence, predictors, and prognostic impact of a decline in platelet count (DPC) in patients treated by percutaneous coronary intervention (PCI).

Methods and results A total of 10 146 consecutive patients treated by PCI from 2003 to 2006 were included. According to the magnitude of the DPC, the population was divided into four groups: no DPC (<10%), minor DPC (10–24%), moderate DPC (25–49%), and severe DPC (≥50%). The primary haemorrhagic endpoint was a composite of post-procedure surgical repair major bleeding. The primary ischaemic endpoint was 30-day all-cause mortality–non-fatal myocardial infarction. Among the total population, 36% had a DPC <10%, 47.7% had a DPC of 10–24%, 14% had a DPC of 25–49%, and 2.3% had a DPC ≥50%. On multivariate analysis, moderate and severe DPC were independent predictive factors of the ischaemic outcome. Two procedural practices were identified that, if modified, might reduce the incidence of acquired thrombocytopaenia. Both the intraprocedural use of heparin (as opposed to bivalirudin) and of low molecular weight contrast material were independently associated with severe acquired thrombocytopaenia.

Conclusion Moderate and severe DPC are independent predictors of adverse bleeding and ischaemic outcomes in PCI. Adoption of intraprocedural anticoagulant other than heparin and avoidance of a low molecular weight contrast agent could potentially decrease the occurrence of severe acquired thrombocytopaenia.

  • Platelets
  • Thrombocytopaenia
  • Percutaneous coronary intervention


In recent years advances in pharmacotherapy and catheter-based devices have improved the outcomes of percutaneous coronary intervention (PCI) by reducing the frequency of ischaemic events. While their use contributes to these improved outcomes, heparin,1,2 platelet glycoprotein IIb/IIIa receptor inhibitors,3,4 or use of an intra-aortic balloon pump5 expose patients to the risk of bleeding complications and to the possibility of a decline in platelet count (DPC). When the magnitude of the decline is large enough, one or another definition of ‘acquired thrombocytopaenia’ may be met. In some prior studies that tag has been applied based on the nadir of the absolute post-procedure platelet count (e.g. <100 000 or <150 000/mm3). In others, a decline in absolute count of ≥50% from preprocedure levels has been used. Previous reports from clinical trials indicate that such severe declines are usually accompanied by an increased risk of haemorrhagic complications, but also, not as easily foreseen, an increased risk of ischaemic events.613

Use of the nadir of the absolute platelet count as a marker of PCI effect on platelet function has several potential problems. Minor DPC are frequently encountered. In such instances, none of the cut-points for acquired thrombocytopaenia may be met. Thus, the relationship of DPC <50% to haemorrhagic and ischaemic outcomes has not been systematically evaluated. Similarly, a lack of information exists regarding patients with a marginally low preprocedural platelet count. In such patients, the study definition of acquired thrombocytopaenia, if based on absolute platelet count post-procedure, might be met despite a rather minor effect of the PCI on the platelet count. To avoid this possibility, the HORIZONS-AMI trial11 excluded patients with a platelet count at baseline between 100 000 and 150 000 mm3 from the analysis of the frequency and consequences of acquired thrombocytopaenia.

Because of these considerations, we postulated that the magnitude of DPC during the procedure, rather than the attainment of a binary definition for acquired thrombocytopaenia, may more directly reflect the effect of the procedure on platelet function and thus provide greater insight into the connection between this phenomenon and outcomes. We therefore undertook this analysis to assess the incidence and prognostic impact on ischaemic and haemorrhagic complications in relation to the percent change in platelet count rather than an absolute number. We further sought to identify predictors of clinically significant DPC.


From our on-going registry of patients undergoing PCI at our institution from 2003 to 2006 we selected all in whom at least 1-month follow-up was available. Those with cardiogenic shock on admission, with a platelet count at baseline <100 000, with a periprocedural rise in platelet count, or with missing platelet count data were excluded. Also excluded were patients requiring post-PCI coronary artery surgery14 and those who were treated with thrombolysis. After these exclusions, 10 146 patients were available for analysis. All gave written consent for the PCI procedure and the study was conducted after Institutional Review Board approval.

All data collection, management, and analysis were performed by the dedicated data coordinating center (Data Center, Cardiovascular Research Institute, Washington, DC, USA) that maintains an on-going registry of interventional cases in our laboratory. As is their routine, from retrospective chart review, centre staff abstracted demographic, clinical, and procedural data, together with in-hospital outcome. Data centre staff obtained post-discharge follow-up information through telephone contact with the patient or the referring physician. Source documentation pertaining to the event was obtained and adjudication as to its nature was conducted by physician staff.

Platelet count was measured at admission, 24 h after PCI, and daily thereafter if a decline was observed. Platelet count at baseline and the nadir of platelet count after PCI was used to determine the percent change using the formula: (baseline platelet count − nadir platelet count)/(baseline platelet count) × 100.

The presenting clinical syndrome was classified as either: (i) asymptomatic or stable angina; (ii) unstable angina/non-ST-elevation myocardial infarction (non-STEMI); or (iii) ST-elevation myocardial infarction (STEMI). Angiographic success was defined as the presence of post-procedure of a residual stenosis <30% and a Thrombolysis in Myocardial Infarction flow rate ≥3. Target vessel revascularization (TVR) was recorded when a clinically driven repeat surgical or catheter-based revascularization of the initially treated vessel was undertaken. Death was defined as all-cause mortality. Major adverse cardiac events (MACE) were defined as either death, Q-wave myocardial infarction (MI), or TVR. History of renal insufficiency was said to be present when it had been previously diagnosed or when serum creatinine on admission was >2 mg/dl. Procedural acute renal failure was defined as a rise in creatinine of ≥50% above baseline value. Major bleeding was a composite variable consisting of any of the following: gastrointestinal bleeding, haematocrit drop ≥15 U, a major haematoma (4 × 4 cm or greater), or one requiring transfusion. Surgical repair was defined as surgical intervention for arterial laceration, pseudoaneurysm, or arteriovenous fistula. A neurological event was defined as stroke or transient ischaemic attack.

The primary ischaemic endpoint of the study was 30-day all-cause mortality or non-fatal MI. The primary haemorrhagic endpoint was the composite of post-procedure surgical repair or major bleeding.

Coronary angioplasty was performed using conventional techniques. The interventional strategy was left to the discretion of the operator. In more than 99% a femoral approach was employed. At admission, all patients were treated with aspirin 325 mg prior to PCI and loaded with clopidogrel 300–600 mg, except those who were already on a maintenance dose. During PCI, patients were anticoagulated with either bivalirudin (a bolus of 0.75 mg/kg, followed by an intravenous infusion of 1.75 mg/kg/h) or unfractionated heparin (a bolus of 40 U/kg and additional heparin to achieve an activated clotting time of 250–300 s). The anticoagulants were normally stopped at the end of the PCI. Use of adjunctive devices and platelet IIb/IIIa glycoprotein inhibitors was at the discretion of the operator. Dual-antiplatelet therapy was recommended for ≥4 weeks for those who received bare metal stents and ≥6 months for those treated with drug-eluting stents.

Statistical analysis

The DPC was stratified with regard to its magnitude. Strata were chosen taking into consideration recent studies involving intensive care unit patients,15 in which a DPC of 28% was identified as the best threshold for discriminating between survivors and non-survivors during hospitalization, as well as prior studies that defined heparin-induced thrombocytopaenia16 as a DPC ≥50%. Thus, patients were divided into four strata with regard to these published data—group 1 = DPC <10%; group 2 = DPC 10–24%; group 3 = DPC 25–49%; and group 4 = DPC ≥50%. For the purposes of comparisons with the outcome variables, DPC strata were considered to be a single ‘dummy’ variable.

Continuous variables are expressed as mean ± SD and were compared using analysis of variance. Categorical variables are expressed as proportions and compared by means of contingency tables. The occurrences of the primary ischaemic endpoint for each strata of the DPC are depicted as Kaplan–Meier curves. Differences between curves were tested for significance by means of the log-rank test. The Cox proportional hazard model was used for risk adjustment for the outcomes of interest after appropriate testing for proportionality. Logistic regression was used to identify predictors of DPC. All variables of interest were evaluated in univariate models (Annex). Variables associated with the outcome of interest with P < 0.2 in the univariate models were entered the final risk adjustment model. A P-value <0.05 indicated statistical significance. All statistical analyses were done with the SAS system version 9.1 (Cary, NC, USA).


Baseline characteristics

As detailed in Table 1, there were major differences in baseline clinical characteristics among the four groups. Patients who presented with the greatest DPC were more likely to be female and to be older. Moreover, they were significantly more likely to have diabetes, renal insufficiency, and findings of severe coronary disease. They more often reported prior myocardial infarction (MI) or heart failure, and a greater proportion presented with STEMI. At the time of diagnostic catheterization, a lower ejection fraction was recorded.

View this table:
Table 1

Baseline clinical characteristics

Decline in platelet countP-value
<10% (n = 3665)10–24% (n = 4837)25–49% (n = 1418)≥50% (n = 226)
CV risk factors
 Age63.9 ± 11.965.2 ± 11.867.3 ± 11.567.6 ± 11.4<0.001
 Men2149 (58.8)2656 (55)682 (48.2)89 (39.4)<0.001
 Smoker2008 (54.8)2652 (54.8)736 (51.9)117 (51.8)0.2
 Hypertension2631 (71.9)3524 (72.9)1058 (74.8)174 (77.3)0.07
 Diabetes1195 (32.8)1542 (32.1)498 (35.4)92 (41.1)0.007
 Hypercholesterolaemia2832 (77.7)3756 (78)1036 (73.6)159 (71)0.001
 Renal insufficiency315 (8.6)479 (9.9)250 (17.8)43 (19.4)<0.001
Cardiac history
 Prior MI1415 (40.4)1966 (42.6)647 (47.9)110 (51.2)<0.001
 Prior CABG967 (26.5)1485 (30.9)489 (34.7)67 (29.9)<0.001
 Prior PCI1121(31.3)1584 (33.5)432 (31.4)66 (30.7)0.1
 Prior CHF550 (15.5)736 (15.8)311 (22.5)64 (29.9)<0.001
Severity of heart disease
 Stable angina1003 (27.4)1363 (28.2)333 (23.5)48 (21.2)<0.001
 Unstable angina/non-STEMI2469 (67.3)3199 (66.1)960 (67.7)147 (65.1)0.7
 STEMI193 (5.3)275 (5.7)125 (8.8)31 (13.7)<0.001
 LVEF48 ± 1348 ± 1345 ± 1444 ± 14<0.001
 Number of diseased vessels1.9 ± 0.81.9 ± 0.82.1 ± 0.82.1 ± 0.8<0.001
Laboratory values
 CK-MB baseline4.7 ± 24.26.89 ± 39.012.3 ± 43.014.4 ± 49<0.001
 Haematocrit baseline39.8 ± 5.240.2 ± 5.039.5 ± 5.437.6 ± 6.0<0.001
 Creatinine baseline1.26 ± 4.51.33 ± 4.91.47±5.82.70 ± 7.70.002
 Platelet baseline229 ± 64239 ± 66252 ± 76259 ± 84<0.001
  • Results are expressed as means ± SD for quantitative variables and as n (%) for qualitative variables. MI, myocardial infarction; CABG, coronary artery bypass graft; PCI, percutaneous coronary intervention; CHF, congestive heart failure; STEMI, ST-elevation myocardial infarction; LVEF, left ventricular ejection fraction.

From a procedural viewpoint (Table 2), patients with a ≥50% DPC were more likely to have complex angiographic lesions, including more left main stenosis, more saphenous vein graft intervention, as well as more American College of Cardiology/American Heart Association type-C target lesions. They were treated somewhat less frequently with a drug-eluting stent. Importantly, angiographic success was achieved in at least 93% of patients in each group.

View this table:
Table 2

Angiographic and procedural characteristics

Decline in platelet countP-value
Lesion-based(n = 6543)(n = 8752)(n = 2688)(n = 398)
Lesion location
 Left main126 (1.9)261 (3)79 (2.9)18 (4.5)<0.001
 Left anterior descending2220 (33.9)2859 (32.7)898 (33.4)142 (35.7)
 Left circumflex artery1538 (23.5)2020 (23.1)619 (23)86 (21.6)
 Right coronary artery2008 (30.7)2624 (30)757 (28.2)107 (26.9)
 Saphenous vein graft593 (9.1)879 (10)295 (11)41 (10.3)
ACC/AHA type C lesion1375 (22.9)2225 (27.4)702 (29)148 (40.5)<0.001
Number of lesions dilated1.78 ± 1.031.83 ± 1.571.90 ± 1.201.80 ± 1.160.05
Drug-eluting stent2172 (84.4)2665 (82.8)595 (80.2)81 (79.4)0.03
Stent diameter3.1 ± 2.33.1 ± 2.53.0 ± 1.23.0 ± 0.30.8
Stent length20.36 ± 6.320.6 ± 6.620.8 ± 7.019.8 ± 6.80.4
Number of implanted stents1.51 ± 0.871.51 ± 0.891.62 ± 0.991.50 ± 1.130.2
Angiographic success6373 (97.1)8485 (96.6)2564 (96.2)379 (93.3)<0.001
Patient-based(n = 3665)(n = 4837)(n = 1418)(n = 226)
 Low-osmolar contrast agent2468 (69.5)3330 (70.9)1030 (76.3)174 (78.3)<0.001
 High-osmolar contrast agent1002 (28.2)1272 (27.1)297 (22)44 (19.7)<0.001
 Bivalirudin1148 (31.3)1400 (28.9)263 (18.5)40 (17.7)<0.001
 Heparin2318 (63.2)3132 (64.8)1048 (73.9)168 (74.3)<0.001
 Glycoprotein IIb/IIIa445 (12.2)588 (12.2)206 (14.6)40 (17.9)0.008
 Intra-aortic balloon pump79 (2.2)240 (5)202 (14.3)52 (23.1)<0.001
  • Results are expressed as means ± SD for quantitative variables and as n (%) for qualitative variables.

As might be anticipated, use of glycoprotein IIb/IIIa receptor blockers, known to be associated with thrombocytopaenia, was more frequent in patients with a DPC ≥50%. Patients with DPC of ≥25% PCI more frequently received intravenous anticoagulation with heparin during the procedure than those in the strata with lesser degrees of DPC. A low-osmolar contrast agent was employed more frequently in the group with the greatest DPC. Further, an intra-aortic balloon pump was employed more frequently in patients with a DPC of ≥25%.

In-hospital events

Table 3 delineates the post-procedure hospital events stratified by degree of platelet count decline. The frequency of both ischaemic and haemorrhagic complications was importantly related to the magnitude of platelet count decline. Less obviously, patients with the greatest DPC were also far more likely to have post-PCI renal insufficiency, a post-procedural neurological event, and a greater increase in creatinine kinase (CK)-MB. Although the differences were more apparent in patients with the most severe DPC, a trend towards more frequent adverse events was observed in patients with lesser degrees of DPC.

View this table:
Table 3

In-hospital events

Decline in platelet countP-value
<10% (n = 3665)10–24% (n = 4837)25–49% (n = 1418)≥50% (n = 226)
Death9 (0.2)22 (0.5)23 (1.6)21 (9.3)<0.001
Q-wave MI6 (0.2)16 (0.3)10 (0.7)8 (3.6)<0.001
TVR52 (1.4)96 (2)59 (4.2)13 (5.8)<0.001
MACE12 (0.3)31 (0.6)21 (1.5)14 (6.2)<0.001
CK-MB max8.09 ± 22.912.0 ± 42.623.2 ± 63.734.61 ± 78.9<0.001
Procedural acute renal failure46 (1.3)98 (2)86 (6.1)32 (14.5)<0.001
Platelet nadir218.4 ± 61.1199.7 ± 55.3170 ± 52.491.2 ± 48.4<0.001
Major bleeding-surgical repair48 (1.3)102 (2.1)166 (11.7)70 (31.1)<0.001
Transfusion58 (1.7)163 (3.6)246 (18.2)100 (45.5)<0.001
Neurological event23 (0.6)50 (1)39 (2.8)24 (10.6)<0.001
  • Results are expressed as mean ± SD for quantitative variables and as n (%) for qualitative variables. TVR is only PCI–TVR. Patients with CABG in-hospital were excluded from this study. TVR, target vessel revascularization; MACE, major adverse cardiac events.

Major bleeding-surgical repair, the primary haemorrhagic endpoint, occurred progressively more often the greater the degree of DPC. An association was also found with numerous baseline and periprocedural factors. Importantly, after adjusting for these factors, DPC 25–49% and DPC ≥50% both remained as independent predictors with hazard ratios of 3.47 (95% CI = 2.20, 5.50; P < 0.0001) and 4.43 (95% CI = 2.28, 8.61; P < 0.0001), respectively (Table 4). It is noteworthy that transfusion did not remain in the model.

View this table:
Table 4

Predictors of post-procedure bleeding-surgical repair

VariableOdds ratio95% CIP-value
Intra-aortic balloon pump1.571.02–2.440.04
Haematocrit baseline1.121.09–1.16<0.001
Intraprocedure heparin1.991.34–2.970.001
Unstable angina/non-STEMI1.681.11–2.530.01
  • DPC, decline in platelet count. For age, OR and CI are presented per 10-year units.

Appropriate threshold for acquired thrombocytopaenia

An association with bleeding and adverse events appeared with DPC 25–29% suggesting ≥25% as a definition for acquired thrombocytopaenia. To further test this assumption, we constructed receiver-operating curves for the magnitude of DPC against the occurrence of 30-day death and death or non-fatal MI. The most accurate cut-point by this method for 30-day mortality proved to be 20% (c = 0.69303) and 17% (c = 0.64941) for 30-day death or non-fatal MI. Thus, the convenient 25% DPC cut-point used to separate our two most severe strata seems supportable.

Predictors of acquired thrombocytopaenia

Numerous baseline and procedural characteristics were associated with a DPC of ≥25%. On multivariate analysis (Table 5), age, female gender, renal insufficiency, post-procedural renal failure, PCI for acute STEMI, baseline haematocrit, and use of an intra-aortic balloon pump were independent factors of acquired thrombocytopaenia. Two other independent predictors of this level of decline are worthy of notice since they suggest modifications in treatment strategy: (i) patients who received heparin rather than bivalirudin had an approximately 67% increase in risk of developing a DPC ≥ 25%; and (ii) patients who received low-osmolar contrast agent had about a 34% greater risk of reaching that threshold. Glycoprotein IIb/IIIa inhibitor use, although strongly associated univariately, did not continue to be after risk adjustment.

View this table:
Table 5

Independent predictors of decline in platelet count ≥25%

VariableOdds ratio95% CIP-value
History of chronic renal insufficiency1.611.33–1.950.001
Haematocrit baseline1.031.02–1.050.001
Intra-aortic balloon pump3.222.61–3.970.001
Procedural acute renal failure2.491.83–3.380.001
CK-MB max1.01.0–1.00.001
Low-osmolar contrast agent1.341.15–1.560.001
  • For age, OR and CI are presented as per 10-year units.

Thirty-day and 1-year outcomes

The incidence of MACE and its components at 30 days and 1 year are displayed in Table 6. Thirty-day all-cause mortality was nearly 10 times greater in the group with a DPC ≥50% when compared with those of little or no DPC. The same was true of all-cause mortality and non-fatal MI. A lesser but important increase in adverse event rate was also seen in patients with a DPC of 25–49%. Kaplan–Meier curves of the outcome at 30 days are presented in Figure 1. At 1 year, the frequency of adverse clinical outcomes in patients with moderate or severe DPC continued to be significantly worse.

Figure 1

Time-to-event curves through 30 days for (A) death, (B) death–myocardial infarction, (C) target vessel revascularization (TVR)–major adverse cardiac events (MACE), and (D) TVR.

View this table:
Table 6

Adverse cardiac events

Decline in platelet countP-value
<10% (n = 3665)10–24% (n = 4837)25–49% (n = 1418)≥50% (n = 226)
 Death28 (0.8)38 (0.8)36 (2.5)4 (10.6)<0.001
 Death–MI70 (1.2)88 (1.2)75 (5.2)38 (16.8)<0.001
 TVR50 (1.4)59 (1.2)28 (2.0)10 (4.7)<0.001
 TVR–MACE75 (2.0)99 (2.0)69 (4.9)35 (15.5)<0.001
 Death146 (4.6)214 (5.0)135 (10.7)46 (23.3)<0.001
 Death or MI154 (4.8)230 (5.4)146 (11.6)50 (25.4%)<0.001
 TVR439 (13.7)597 (14.0)179 (14.2)26 (14.2)0.5
 TVR–MACE566 (17.6)795 (18.5)304 (24.1)72 (36.6)<0.001
  • Results are expressed as n (%).

Association of DPC with adverse outcomes

Table 7 lists the independently associated predictors of the primary ischaemic endpoint (all-cause mortality or MI at 30 days). A DPC 25–49% and a DPC ≥50% were both predictors of combined outcome, death, and MI at 30 days.

View this table:
Table 7

Independent predictors of death and death or MI at 30 days

VariableDeathDeath or MI
Hazard ratio95% CIP-valueHazard ratio95% CIP-value
History of MI1.71.2–2.40.002
History of PCI0.60.4 0.90.02
History of heart failure2.41.4–4.00.001
Unstable angina/non-STEMI1.51.0–2.40.06
Haematocrit baseline1.00.9–1.00.05
Left main3.11.5–6.30.002
Saphenous vein graft lesion1.71.0–2.80.03
Intra-aortic balloon pump1.61.1–2.50.02
Procedural acute renal failure2.21.1––3.10.03
CK-MB max1.01.0–1.0<0.0011.01.0–1.0<0.001
Neurological event4.22.0–8.7<0.0012.21.2–4.00.01
  • For age, OR and CI are presented as per 10-year units.

It is important to note that despite close univariate association of major bleeding-surgical repair with death and non-fatal infarction at 30 days, neither that complication (univariate HR 4.9, 95% CI 3.4–6.9, P < 0.001) nor the need for transfusion (univariate HR 5.3, 95% CI 3.9–7.1, P < 0.001) remained associated with either after adjustment for the DPC.


Results of this study provide two insights into the association between PCI-related thrombocytopaenia and adverse outcomes. First, they confirm in a large unselected (‘real world’) group of patients that post-procedure thrombocytopaenia is independently associated with an increased incidence of both adverse ischaemic events and haemorrhagic complications. Second, this investigation is the first to report that this association, while strongest with a severe reduction in platelet count, is not limited to the usual criteria for acquired thrombocytopaenia, but is also present with a more modest degree of DPC.

Effect of findings on the interpretation of trials

Previous reports on acquired thrombocytopaenia in PCI have used a variety of definitions.5,8,12,13 Most often the criterion was either a platelet count nadir (e.g. <100 000/mm3) or a DPC ≥50%. In this investigation we focused on procedure-related DPC rather than on absolute count values. We reasoned that the magnitude of DPC might better reflect PCI-related changes in platelet function than the absolute platelet nadir reached post-procedure. This approach allowed us to identify individuals with substantial falls in platelet count but in whom the count never reached ‘thrombocytopaenic’ levels, and excluded individuals with low normal counts in whom a minimal decline accompanying the PCI resulted in absolute levels <100 000/mm3.

As indicated in the Methods section in this analysis, the magnitude of DPC for each patient determined the strata to which the individual was assigned. These strata were defined based on previously published criteria for acquired thrombocytopaenia. In these papers, acquired thrombocytopaenia defined in these ways was used to assess a potential association with procedural bleeding complications. DPC is, however, a continuous variable and cut-points other than those adopted for this paper could also be considered. Our use of ≥25% DPC as a threshold was supported by receiver-operating curve analysis. Thus, the convenience of that number makes it attractive as a useful definition for both clinical practice and for trials. It clearly offers advantages over more restrictive definitions.

By stratifying patients with regard to the magnitude of the DPC we were able to uncover a significant relationship between adverse events and moderate DPC (25–49%). Thus, patients with a DPC in the moderate range whose mean nadir platelet count was 170 ± 52.4 (Table 3) experienced more frequent complications of the PCI. This was true with regard to both ischaemic and haemorrhagic events; even though in previous reports these patients were not considered to have acquired thrombocytopaenia and their outcomes may not have been associated with platelet effects.

Thus, consideration should be given to modifying the current definition of acquired thrombocytopaenia in clinical practice and in research trials. Based on our results we suggest defining DPC ≥25% as a threshold for defining acquired thrombocytopaenia. A DPC 25–49% could be considered as moderate acquired thrombocytopaenia and a DPC ≥50% as severe acquired thrombocytopaenia, no matter the absolute value of the nadir. Such a change would significantly increase the frequency with which that complication is encountered. In prior studies limited to patients treated with PCI for an acute coronary syndrome12,17,18 acquired thrombocytopaenia was reported to develop in 1.0–7.3%. In our unselected series of PCI-treated patients, the incidence of clinically significant decline reached 16.2% (13.9% with a DPC 25–49% and 2.3% with a DPC ≥50%).

Relationship of DPC to intermediate or late outcomes

Several prior reports have indicated an association between post-procedural haemorrhagic complications and intermediate or late outcomes.18,19 The current study suggests a similar relationship with procedure-related DPC. The question must be asked of course as to whether or not thrombocytopaenia results from the adverse events or has a pathogenetic role in their production. Moreover, it can be suggested that blood transfusion, often required in treating haemorrhagic complications of PCI, might in itself be detrimental.11 Importantly, in this large study the link between DPC and the adverse outcomes at 30 days is, on multivariate analysis, independent of other confounding variables including major bleeding-surgical repair and transfusion (Table 7). While the possibility of unmeasured confounders must be acknowledged, these observations could be taken to suggest the possibility of a direct pathogenetic influence of DPC on adverse events.

Potentially preventable risk factors for DPC

Since it is plausible that DPC is independently associated with adverse outcome, it is important to gain an understanding of patient characteristics and procedural strategies associated with such a decline. In particular it would be important to identify any factors that may be anticipated and prevented. Many are not modifiable (e.g. older age, female sex, renal insufficiency, intra-aortic balloon pump use, and presence of acute MI). However, two opportunities to modify therapeutic strategies are suggested (Table 5). First, the data suggest that use of a low-osmolar radiographic contrast agent (as opposed to a high-osmolar agent) may lead to thrombocytopaenia. Use of a low-osmolar agent increased the risk of DPC by about 30%. In this regard, a few reports of transient thrombocytopaenia associated with low-osmolar agent use have appeared.20,21 An antiplatelet and anticoagulant effect through activation of the C3 fraction of complement has been suggested.22 However, no randomized trials dealing with contrast agents have specifically studied this endpoint, and such studies are mandatory to confirm or deny these exploratory data. Second, in this analysis the use of heparin (instead of bivalirudin) increased the risk of thrombocytopaenia by 67%. Use of heparin is known to result in thrombocytopaenia from immunological and non-immunological mechanisms.2325 Our data do not allow any conclusions regarding mechanisms for this association. It is, however, plausible that immune mechanisms similar to the ones that produce overt thrombotic events in classic heparin-induced thrombocytopaenia were set in motion by using heparin for anticoagulation during the PCI.

It is noteworthy that our findings are congruent with randomized trials directly comparing heparin and bivalirudin.26,27 In REPLACE 2,26 a randomized trial involving 6002 patients undergoing elective or urgent PCI, a lower risk of thrombocytopaenia was found in the bivalirudin arm compared with the arm receiving unfractionated heparin plus glycoprotein IIb/IIIa (0.7% vs. 1.7%, P < 0.001). In HORIZONS AMI,11 the risk of thrombocytopaenia was also higher in the heparin group compared with the bivalirudin group (4.2% vs. 1.4%, P < 0.001). However, the relationship of the acquired thrombocytopaenia and the outcome in these studies was not completely defined. In our analysis the use of glycoprotein IIb/IIIa receptor blockers was univariately associated with DPC, but the difference was no longer significant after adjustment for other factors. Use of these agents is relatively infrequent in our series and the analysis lacked the power to detect a small difference.


The present study has several strengths. It is large and includes the ‘real world’ experience of an active, high-volume interventional practice. Its size provides an opportunity for statistical adjustment for the inevitable variation in patient and procedural characteristics found in a retrospective analysis. On the other hand, its observational and retrospective nature carries all the limitations inherent in such an investigation. While this analysis utilized a registry in which all data were prospectively recorded in accordance with prespecified definitions of data fields, and statistical techniques were used to adjust for differences in baseline variables, the potential for unaccounted confounders is always present in observational studies. Moreover, we are unable to compare outcomes between patients whose platelet count returned to normal with those in which recovery did not occur.

Heparin and bivalirudin were used as was customary at the time of the PCI. In general, neither is used beyond the end of the procedure itself. It should be noted that an unknown number of patients who received bivalirudin during the PCI got heparin prior to and/or subsequent to the intervention. If heparin was used before the procedure, any benefit from avoidance of heparin during the PCI may have been blunted. On the other hand, use of heparin after the procedure could increase the risk of bleeding and of DPC.

Like all retrospective analyses, the principal value of these data is in the generation of hypotheses. In particular, decisions to make changes in antithrombotic strategies during PCI must take into account other reports and ultimately must await additional data and randomized trials.


DPC is strongly associated with ischaemic and haemorrhagic complications of PCI. The association of adverse outcomes and DPC was found with smaller declines from baseline than had been previously reported. Avoidance of heparin may reduce the risk of DPC and thereby the risk of haemorrhagic and ischaemic complications. The relationship between DPC and low osmolar contrast agents needs additional confirmatory studies. Our data further suggest that the definition for clinically significant acquired thrombocytopaenia should include a DPC of 25–49% (moderate acquired thrombocytopaenia) and a DPC ≥50% (severe acquired thrombocytopaenia) without necessarily taking into account the nadir platelet count value.

Conflict of interest: none declared.


Variables tested in univariate analysis for the ischaemic endpoint

Clinical characteristics (gender, age); clinical presentation (stable angina, unstable angina-non STEMI, STEMI); medical history (history of MI, history of PCI, history of coronary artery bypass grafting (CABG), history of renal failure, history of cardiac failure); cardiovascular risk factors (hypertension, diabetes, hypercholesterolaemia, any smoking); procedural characteristics (number of vessels diseased, left main, left anterior descending artery, circumflex, right coronary artery, saphenous vein graft, type C lesions, number of vessels dilated, angiographic success, IABP, low-osmolar contrast agents, high-osmolar contrast agents); post-procedural complications (procedural acute renal failure, CK-MB max, neurological event, surgical repair-major bleeding, transfusion); medication used (bivalirudin, heparin, IIb/IIIa glycoprotein inhibitor use); and biological parameters (haematocrit baseline, CK-MB baseline).

Variables tested in univariate analysis for the haemorrhagic endpoint

Clinical characteristics (gender, age); clinical presentation (stable angina, unstable angina-non STEMI, STEMI); medical history (history of MI, history of percutaneous transluminal coronary angioplasty (PTCA), history of CABG, history of renal failure, history of cardiac failure); CV risk factors (hypertension, diabetes, hypercholesterolaemia, any smoking); procedural characteristics (number of vessels diseased, left main, left anterior descending artery, circumflex, right coronary artery, saphenous vein graft, type C lesions, number of vessels dilated, angiographic success, IABP, low-osmolar contrast agents, high-osmolar contrast agents); post-procedural complications (procedural acute renal failure, CK-MB max); medication used (bivalirudin, heparin, IIb/IIIa glycoprotein inhibitor use); and biological parameters (haematocrit baseline, CK-MB baseline).

Variables tested in univariate analysis for predictors of thrombocytopaenia

Clinical presentation (stable angina, unstable angina-non STEMI, STEMI); medical history (history of MI, history of PTCA, history of CABG, history of renal failure, history of cardiac failure); CV risk factors (hypertension, diabetes, hypercholesterolaemia, any smoking); procedural characteristics (number of diseased vessels, left main, left anterior descending artery, circumflex, right coronary artery, saphenous vein graft, type C lesions, number of vessel dilated, angiographic success, IABP, low-osmolar contrast agents, high-osmolar contrast agents); post-procedural complications (procedural acute renal failure, CK-MB max); medication used (bivalirudin, heparin, IIb/IIIa glycoprotein inhibitor use); and biological parameters (haematocrit baseline, CK-MB baseline).


View Abstract