Peter Hellstern, MD, Institute of Hemostaseology and Transfusion Medicine, Academic City Hospital, D-67063 Ludwigshafen, Germany. E-mail: email@example.com
We investigated the relationship between platelet function analyzer (PFA-100TM) closure times (CT) and bleeding time (BT), platelet aggregation (PA) induced by ADP, arachidonic acid, and collagen, blood cell counts, and von Willebrand factor (VWF) in 120 well-characterised healthy individuals. Pre-analytical and analytical conditions were standardised comprehensively. In a substantial number of cases the differences between duplicate measurements exceeded 15%. The reference range (5th and 95th percentiles) for CT with the collagen/epinephrine (CEPI) and the collagen/ADP (CADP) cartridge was 93–223 s and 64–117 s respectively. Re-examination of 11 individuals with CEPI-CT above the 95th percentile revealed considerable batch-to-batch variation of CEPI-CT. Males had significantly longer CADP than females (P = 0·002). CEPI and CADP-CT measured pm were significantly longer than corresponding values determined am (P = 0·003 and P < 0·0001 respectively). Blood group O was associated with greater CEPI and CADP-CT and lower VWF levels compared with non-O blood groups (P = 0·008, P = 0·0003 and P < 0·0001 respectively). Linear regression analysis revealed association between CEPI-CT, CADP-CT and VWF (P < 0·0001), but no relationship was found between CT and BT or between CT and PA. We conclude that VWF plasma levels modulate PFA-100TM CT to a greater extent than platelet function. Establishment of reliable reference ranges and careful standardisation of pre-analytical and analytical conditions is a prerequisite for obtaining reliable PFA-100TM results. Duplicate measurements are necessary.
The platelet function analyzer, PFA-100TM, manufactured by Dade-Behring, Liederbach, Germany, was developed to assess primary haemostasis in vitro using citrated whole blood (Kundu et al, 1995; Mammen et al, 1995; Harrison, 2004). The device simulates in vivo haemostatic plug formation under high shear flow by measuring the time required to occlude (closure time, CT) a collagen/epinephrine (CEPI) or collagen/ADP (CADP)-coated aperture inserted in a plastic membrane. Most clinical studies regard the PFA-100TM as a simple and easy-to-use tool to detect platelet dysfunction and von Willebrand disease (VWD), and for monitoring treatment with desmopressin, von Willebrand factor (VWF) concentrates and antiplatelet agents (Mammen et al, 1998; Jilma, 2001; Favarolo, 2002; McKee et al, 2002). Conflicting findings from validation studies and in patients with primary haemostasis disorders have, however, raised concerns about the diagnostic performance of the device and its suitability for non-specialised laboratories. The CT depends strongly on VWF plasma levels (Favarolo, 2002), anticoagulant pH value and citrate concentration (Heilmann et al, 1997; Jilma, 2001; Favarolo, 2002), and on the time of day when blood is collected for analysis (Dalby et al, 2000; Wuillemin et al, 2002). Low platelet and leucocyte counts and low haematocrits may lead to prolonged CT (Kundu et al, 1996; Jilma, 2001; Favarolo, 2002). Some studies performing duplicate analysis showed coefficients of variation above 10% (Heilmann et al, 1997; Mammen et al, 1998) or differences between duplicates exceeding 10 or even 15% in a substantial number of cases (Wuillemin et al, 2002). It would seem therefore that earlier statements that duplicate analysis is unnecessary in clinical routine have been premature (Mammen et al, 1998; Böck et al, 1999; Jilma, 2001). The vast majority of clinical studies dealing with the clinical significance of the PFA-100TM system derived their reference ranges from groups of fewer than 40 healthy individuals, and these were often only poorly characterised. This resulted in substantially differing references ranges between study groups (Jilma, 2001). As the upper limits of respective reference ranges are the cutoff levels required to calculate sensitivity, specificity and predictive values in clinical studies, it is not surprising that many (Fressinaud et al, 1998; Cattaneo et al, 1999; Dean et al, 2000; Schlammadinger et al, 2000; Cariappa et al, 2003; Posan et al, 2003) but not all (Wuillemin et al, 2002; Quiroga et al, 2004) trials found the PFA-100TM system to be sufficiently sensitive and superior to the bleeding time (BT) in detecting VWD.
We re-evaluated PFA-100TM testing and investigated those factors influencing CEPI and CADP-CT in a larger number of well-characterised healthy individuals under comprehensively standardised conditions. As there is a lack of data on the relationship between PFA-100TM CT and more specific platelet function tests, we also measured platelet aggregation (PA) induced by ADP, arachidonic acid (AA) and collagen.
Materials and methods
Selection of healthy subjects and study design
We included 120 apparently healthy individuals (60 males and 60 females aged between 19 and 63 years). Sixty-three individuals were voluntary blood donors from the local blood donation centre. The remaining 57 volunteers were students or members of hospital staff. Individuals were excluded when they had taken aspirin, anti-inflammatory drugs, antibiotics or other drugs with potential to influence platelet function within 10 days prior to study, or had donated blood or blood components within 10 days prior to blood collection. Allowed drugs were oral contraceptives (OC) and thyroid gland hormones. Individuals reporting a history of bleeding on a standardised questionnaire were also excluded. The following data were registered: age, gender, body weight, height, ABO blood group, blood donors or non-donors, smoking habits, all drugs including OC, intake of garlic within 24 h prior to blood collection and time of blood collection.
Blood collection and processing
After a 30-min rest in a sitting position, blood samples were drawn from the antecubital vein using 21-gauge butterfly needles between 8:30 and 11:00 am. A second blood collection was carried out in half of the individuals (30 males and 30 females) on the same day between 2:00 and 5:00 pm. A small lunch and soft drinks were allowed between the am and pm blood collection. Platelet activation was minimised by using a light tourniquet, short and moderate phlebostasis, smooth blood flow, plastic tubes (Sarstedt, Nuembrecht, Germany) and immediate gentle mixing with anticoagulant. The first 3 ml were collected into plastic tubes containing potassium EDTA and used for blood cell counts. Blood for PA testing and for determining fibrinogen and VWF collagen-binding assay (VWF:CBA) was drawn into S-MonovettesTM from Sarstedt, containing 0·106 mol/l unbuffered trisodium citrate (9 parts of blood to 1 part of 0·106 mol/l citrate), followed by a 3·8-ml S-MonovetteTM containing 0·38 ml of 0·129 mol/l buffered sodium citrate for PFA-100TM analysis.
Platelet rich plasma (PRP) was obtained by 10-min centrifugation at 164 × g. Platelet poor plasma (PPP) was prepared by centrifugation at 3500 × g for 10 min. The platelet count of PRP was adjusted to 250 × 109/l by adding PPP to PRP. Aliquots of 0·5 ml of PPP were snap-frozen and stored at −70°C. When required for further analysis (VWF:CBA), samples were thawed at 37°C for 10 min.
Except for the determination of VWF:CBA, blood processing, BT and all assays were performed by one experienced medical technician.
BT was carried out by the Ivy technique using a SurgicuttTM device from International Technidyne (Edison, NJ, USA). CEPI and CADP-CT were determined in duplicate (channels 1 and 2) between 30 min and 1 h after blood collection, using one single lot each of CEPI and CADP cartridges and a PFA-100TM instrument from Dade-Behring (Liederbach, Germany). Another lot of CEPI and CADP cartridges was used to re-examine those individuals with CEPI-CT above the 95th percentile. Device and cartridges were handled in accordance with manufacturer’ s instructions. Measurements were repeated on the same channel of the PFA-100TM instrument if CT exceeded 300 s. The shorter CT was used for further analysis. For CT values greater than 300 s, a value of 300 s was arbitrarily assigned for further statistical analysis.
The PA tests were run on an APACT-4 aggregometer from Labitec (Arensburg, Germany), between 1 and 2 h after blood collection. PRP samples were stimulated by adding 25 μl of ADP or AA to 225 μl of PRP to obtain final concentrations of 5 μmol/l of ADP and 500 μg/ml of AA. Five microlitre of collagen fibrils from equine tendon were added to 245 μl of PRP to achieve a final concentration of 1 μg/ml. ADP and AA were purchased from Progen (Heidelberg, Germany) and collagen from Nycomed (Linz, Austria). Aggregations were recorded for 7 min. The difference in light transmission on a linear scale between PRP and PPP was set to 100% to calculate the maximum changes in light transmission due to PA. Any observed disaggregation was also recorded.
Platelet and leucocyte counts, mean platelet volume (MPV) and Hb were determined on an ADVIA analyzer from Bayer Vital (Fernwald, Germany). We measured VWF:CBA in duplicate, using an enzyme immunoassay from Progen.
The D'Agostino-Pearson omnibus test for normality was used to analyse the values. As not all values were normally distributed, data are given as medians, 5th and 95th percentiles or as medians and ranges (minimum–maximum). The 97·5th percentiles were also calculated from the 120 CEPI-CT and CADP-CT values determined am. Percentages of differences between duplicate measurements of all CEPI and CADP-CT (n = 180 each) were calculated as follows: [(CT1−CT2)/average] × 100. The mean ± SD and the numbers of differences >10% and 15% were used to characterise imprecision. To check whether variability was related to the magnitude of measurement, percentages of differences between duplicate measurements were plotted against the average of both measurements (Blant & Altman, 1986).
Paired and unpaired data were compared using the Wilcoxon matched pairs test and the Mann–Whitney U-test respectively. Differences in the proportions of smokers, individuals with blood group O, blood donors and garlic consumers were checked by Fisher's exact test. Any two-tailed probability below 0·05 was considered significant.
Spearman rank correlation coefficients were calculated using the data from all individuals when examined at am (n = 120) and checked by subsequent multiple linear regression analysis. Log-transformed data were used if log-transformation resulted in a better adjustment to normal distribution according to the D'Agostino-Pearson omnibus test. A correlation matrix was constructed to discover any multicolinearity between parameters. Two-tailed P < 0·01 were regarded as statistically significant.
Analyses were conducted using InStat for Windows software, version 3.0, and GraphPad software, version 4.02 (GraphPad Software Inc, San Diego, CA, USA). Blant–Altman plots without regression line were created using GraphPad software, version 4.02 (GraphPad Software Inc.).
Baseline data for all individuals, men and women are shown in Table I. Females and males did not differ in age or in the proportion of smokers, whole blood donors, blood group O and garlic consumption. As expected, females had significantly lower body mass indices than males (P = 0·0005). Age was significantly associated with body mass index (r = 0·39, P < 0·0001; data not shown). The statement of all included individuals that they had taken neither aspirin nor any other drug potentially influencing platelet function for 10 days prior to blood collection was confirmed by AAPA, which was within the normal range (am values) and showed no disaggregation in all individuals (data not shown).
Table I. Baseline data of 120 healthy individuals.
P-value (F versus M)
Data are expressed as median value (minimum–maximum).
F, females; M, males.
*Within 24 h prior to blood collection.
Body mass index (kg/m2)
Blood group O/non-O
Garlic consumers/ non-consumers*
Duplicate analysis of PFA-100TM CT showed a mean difference of 0·05 ± 10·2% for the CEPI cartridge and of 1·8 ± 9·5% for the CADP cartridge. However, 76 of 180 CEPI-CT differences exceeded 10% and 21 of 180 exceeded 15%. The respective data for CADP-CT were 48 of 180 greater than 10% and 19 of 180 greater than 15%. Blant–Altman plots revealed similar imprecision of duplicate measurements over the full measuring range (Fig 1). CEPI-CT determined on channels 1 and 2 of the PFA-100TM equipment were not significantly different (P = 0·818), whereas CADP-CT measured on channel 1 were slightly greater than respective values obtained on channel 2 (medians, 86 s vs. 85 s; P = 0·002).
CEPI-CT, CADP-CT, BT, PA induced by ADP, AA and collagen, platelet and leucocyte count, MPV, Hb, fibrinogen and VWF:CBA determined in the 120 individuals are shown in Table II. The reference range (5th and 95th percentiles) of CEPI-CT and CADP-CT determined am was 93–223 s and 65–117 s respectively. The 97·5th percentiles for CEPI-CT and CADP-CT were 254 s and 129 s respectively. Ten (8·3%) of the 120 subjects had CEPI-CT greater than 200 s. The results obtained in 11 individuals with CEPI-CT exceeding the 95th percentile, using two different lots of CEPI and CADP cartridges, respectively, are shown in Table III. CEPI-CT measured with the second lot (579568) were significantly shorter (P = 0·018), but CADP-CT, BT, PA and VWF:CBA were not significantly different between the two measurements.
Table II. Collagen/epinephrine (CEPI) and collagen/ADP (CADP) platelet function analyzer (PFA-1700TM) closure times (CT), bleeding time, platelet aggregation (PA) induced by ADP (ADPPA), arachidonic acid (AAPA) and collagen (collagen PA), blood cell counts, mean platelet volume (MPV), Hb, fibrinogen and von Willebrand factor collagen binding assay (VWF:CBA) in 120 healthy subjects.
All (am) (n = 120)
Males (am) (n = 60)
Females (am) (n = 60)
BGO (am) (n = 52)
BG non-O (am) (n = 68)
am (n = 60)
pm (n = 60)
M versus F*
*BGO versus non-O
BG, blood group; M, males; F, females.
Medians (5th and 95th percentiles).
†Wilcoxon matched pairs test.
PFA-100 CEPI-CT (s)
PFA-100 CADP-CT (s)
Bleeding time (min)
ADPPA (% light transmission)
AAPA (% light transmission)
Collagen PA (% light transmission)
Leucocyte count (×109/l)
Table III. Collagen/epinephrine (CEPI) and collagen/ADP (CADP) platelet function analyzer (PFA-100TM) closure times (CT), bleeding time, platelet aggregation (PA) induced by ADP (ADPPA), arachidonic acid (AAPA) and collagen (collagen PA) and von Willebrand factor collagen binding assay (VWF:CBA) in 11 of the 12 subjects with CEPI-CT values exceeding the 90th percentile at the first examination.
Values are given as median (range)
*Wilcoxon matched pairs test
PFA-100 CEPI-CT (s)
PFA-100 CADP-CT (s)
Bleeding time (min)
ADPPA (% light transmission)
AAPA (% light transmission)
Collagen PA, % light transmission
Males had significantly longer CADP-CT than females (P = 0·002). CEPI and CADP-CT measured pm were significantly longer than corresponding values determined am (P = 0·003 and <0·0001 respectively). BT, MPV and leucocyte counts also increased significantly from am to pm, and PA induced by ADP, AA or collagen dropped significantly during the day. Blood group O was significantly associated with greater CEPI-CT and CADP-CT and with lower VWF:CBA levels compared with non-O blood groups (P = 0·008, P = 0·0003 and P < 0·0001 respectively). No significant differences between blood group O and non-O individuals or between males and females were found with respect to BT and PA. We noted no significant differences in any assay results between whole blood donors and non-donors (data not shown). OC users had significantly stronger collagen PA values than OC non-users (P = 0·027) and garlic consumption resulted in a significant impairment of ADPPA (P = 0·014; data not shown).
Simple linear regression analysis revealed a significant association between CEPI-CT and CADP-CT (r = 0·61; P < 0·0001), between CEPI-CT and VWF:CBA (r = −0·44; P < 0·0001) and between CADP-CT and VWF:CBA (r = −0·47; P < 0·0001; Table IV). In contrast, CEPI-CT and CADP-CT did not correlate significantly with BT or PA. The BT correlated significantly with ADPPA and collagen PA, and PA induced by ADP, AA and collagen correlated significantly with each other. Platelet and leucocyte counts, MPV and fibrinogen did not influence CT significantly. Results from simple linear regression analysis were predominantly confirmed by subsequent multiple linear regression (Table III). Results from linear regression analysis were also confirmed by comparing individuals having CEPI-CT or CADP-CT above the respective 90th percentiles (n = 12 each) with the remaining 108 healthy subjects. Individuals with CEPI-CT above the 90th percentile had significantly longer CADP-CT (medians, 101 s vs. 85 s, P = 0·001) and lower VWF:CBA levels (medians, 83·5 U/dl vs. 114·5 U/dl; P = 0·012) than the remaining persons. When individuals with CADP-CT above the 90th percentiles were compared with the remaining 108 subjects, significant differences in CEPI-CT (medians, 175 s vs. 135 s; P = 0·008) and VWF:CBA levels (medians, 69 U/dl vs. 117 U/dl; P < 0·0001) were found. In contrast, BT and PA values in subjects with CEPI-CT or CADP-CT above the respective 90th percentiles did not differ significantly from respective values found in individuals with CT below the 90th percentiles. No significant differences could be found between the 14 individuals with BT above the 95th percentile and/or any PA below the respective 5th percentile and the other 106 subjects (CEPI-CT medians, 140 s vs. 138 s, P = 0·673; CADP-CT medians, 88 s vs. 82 s, P = 0·603). The 95th percentiles of CEPI and CADP-CT in the 106 individuals with BT below the 95th percentile and/or PA values above the respective 5th percentile were 218 and 124 s, respectively, which was very similar to the 223 s (CEPI-CT) and 117 s (CADP-CT) observed in the total cohort of 120 individuals (Table II).
Table IV. Simple linear regression (Spearman rank correlation coefficients, level of significance P < 0·01); collagen/epinephrine (CEPI) and collagen/ADP (CADP) platelet function analyzer (PFA-100TM) closure times (CT), bleeding time, platelet aggregation (PA) induced by ADP (ADPPA), arachidonic acid (AAPA) and collagen (collagen PA), blood cell counts, mean platelet volume (MPV), Hb, Fibrinogen and von Willebrand factor collagen binding assay (VWF:CBA) determined am in 120 healthy subjects.
We attached great importance to obtaining a representative sample of normal individuals who had not taken any drugs that might impair platelet function within 10 days prior to examination. In addition to careful exploration of medical history, we excluded intake of aspirin or other drugs influencing platelet thromboxane synthesis by measuring AAPA, which showed no significant inhibition in any individual. To avoid haemodilution or haemoconcentration because of contemporary blood or blood component donation, only non-donors or subjects who had not donated within 10 days preceding blood collection were included. Baseline data showed that females and males did not differ significantly in age, blood group distribution, proportion of whole blood donors and smoking habits.
In agreement with previous work (Wuillemin et al, 2002), our duplicate analyses revealed relative differences between duplicates exceeding 10% or even 15% in a substantial number of cases, with similar imprecision over the full range of measurement. These results strongly indicate that duplicate testing is required if reliable results are to be obtained, contradicting statements that single determinations are sufficient (Mammen et al, 1998; Böck et al, 1999; Jilma, 2001). We found no marked differences between CT values determined in channels 1 and 2 of the PFA-100TM device, which agrees with previous observations (Mammen et al, 1998).
We found no differences in CT between whole blood donors and non-donors. This contradicts two previous studies observing a substantial greater percentage of prolonged CEPI-CT and CADP-CT in platelet donors than in controls (Jilma-Stohlawetz et al, 2001; Harrison et al, 2004). Jilma-Stohlawetz et al (2001) attributed this, at least in part, to decreased platelet thromboxane B2 formation and frequent platelet donation, although controls showed significantly lower thromboxane B2 levels than donors. Our subjects donated whole blood only at intervals varying between 2 and 6 months, and all individuals included in this study had normal BT and PA induced by ADP, AA and collagen. Thus, intake of aspirin or other drugs influencing platelet function could largely be excluded.
As already noted previously (Dalby et al, 2000; Wuillemin et al, 2002), PFA-100TM CT and BT increased significantly from morning to afternoon (Table II). We also demonstrated a concomitant increase in BT, MPV and leucocyte count and a marked decrease in PA induced by ADP, AA and collagen. Previous small studies examining PA in 10–25 healthy individuals had lacked statistical power and yielded inconclusive or contradictory results (Tofler et al, 1987; Brezinski et al, 1988; Haus et al, 1990; Jovicic & Mandic, 1991; Malyszko et al, 1994). Our data suggest that the time of day when blood is collected for determining PFA-100TM CT, PA, MPV and leucocyte count or when BT is measured has to be kept within a narrow range.
This study had several significant limitations. Although we tried to get a representative group of healthy males and females, we were unable to avoid bias because of different proportions of blood group O individuals in our subgroups. We are confident that our normal subjects were more meticulously characterised than in any previous study. Nevertheless, we cannot completely rule out that some individuals were taking drugs that influenced platelet function at the time of blood collection. None of our subjects had impaired PA indicating platelet dysfunction. However, more sensitive assays such as platelet thromboxane formation might have revealed individuals consuming antiplatelet drugs. The lack of significant linear correlations between PFA-100TM CT and some other variables, particularly BT and PA, may be caused by the low differences between minimum and maximum values seen in normal individuals. As the correlation coefficients depend strongly on the ranges covered by the single values, inclusion of values outside normal ranges might have resulted in substantially higher correlation coefficients.
In conclusion, the clinical performance of the PFA-100TM system is critically dependent on carefully standardised pre-analytical and analytical conditions and establishing reference ranges derived from a sufficient number of well-characterised individuals. Duplicate testing is an essential requirement to obtain reliable results. That CT test results may vary from batch to batch must also be taken into consideration. Like BT, PA, leucocyte count and MPV, PFA-100TM CT are markedly influenced by the time of day when blood is collected. VWF plasma levels, ABO blood group and possibly gender are additional variables influencing PFA-100TM CT. The significant association between PFA-100 CT and VWF plasma levels and the lack of correlation between CT and PA in healthy individuals suggests that PFA-100TM results more significantly reflect VWF activity than platelet function. The use of the PFA-100TM system as a screening test for primary haemostasis outside specialised laboratories cannot be recommended.