IntroductionMeasurement of platelet function in vitro is the primary clinical method used for the diagnosis of an inherited or acquired platelet function defect in a patient with excessive bleeding, although specific receptor, granule, or DNA studies are often used for confirmation. Light transmission aggregometry (LTA) in platelet-rich plasma (PRP), which has been the gold standard for platelet function testing for fifty years, measures the change in optical density produced when platelets form aggregates in fluid phase. Abbreviations used are defined in Table I. More recently, the simultaneous measurement of adenosine triphosphate (ATP) release as an indicator of platelet activation and granule secretion has been added. Impedance aggregometry allows measurement of aggregation and ATP release in whole blood (WB); it requires that platelets first adhere to a pair of electrodes and then form stationary aggregates which increase the resistance across the electrode gap. There have been few direct comparisons of the various measurement systems available for diagnosis of platelet function defects (Ingerman-Wojenski et al, 1983; Ingerman-Wojenski and Silver, 1986; Riess et al, 1986; Podczasy et al, 1997; Seyfert et al, 2007).
In clinical practice, it is common to test a group of “normal” subjects to set reference ranges for response to specific agonists and to test an individual normal subject with each patient tested to verify appropriate function of the test system (Christie et al, 2008; Hayward et al, 2010; Cattaneo et al, 2013). Numerous drugs have been documented to affect aggregation and ATP release (Rao, 2007). In addition, certain foods, alcohol, caffeine, and smoking have been reported to have effects. Foods rich in flavonoids, such as chocolate, red wine, and tea have been particularly implicated (Pearson et al, 2005; Holt et al, 2005). Certain flavonoids have been shown to inhibit platelet signaling pathways through binding to the thromboxane A2 receptor (Guerrero et al, 2004). In clinical platelet function testing, use of medications is usually considered (Christie et al, 2008; Hayward et al, 2010; Cattaneo et al, 2013); the impact of other exposures has not been evaluated.
In a recent study of women with menorrhagia, using a standardized instrument and protocol for aggregation and ATP release testing at 6 sites, we found 20% of 169 control subjects collected for reference range determination to be abnormal with arachidonic acid alone (Miller et al, 2011). This led us to attempt to define the degree and sources of variation in an unselected group of normal subjects studied over time using four different test systems with prospective collection of exposure data.
MethodsSubjectsSubjects were recruited from among students and employees at Emory University in Atlanta. Each completed a questionnaire on history of bleeding symptoms and chronic disease. Those with no history of excessive bleeding or diagnosed chronic disease were tested for coagulation parameters as previously described (Miller et al, 2011). All subjects had normal coagulation function except one subject with decreased factor XII, who was not excluded. The 64 enrolled subjects ranged in age from 18–35 and included 30 males and 34 females. They self-identified their ethnicities as 34 (53%) White non-Hispanic, 22 (34%) Black non-Hispanic, 3 (5%) Asian, 2 (3%) White Hispanic, and 3 (5%) other. The study was conducted with the approval of the Institutional Review Boards of Emory University and the CDC. Written informed consent was obtained from all participants.
Food and Medication Data CollectionData were collected prospectively on daily drug, food, and alcohol intake and illness through self-recorded standardized diaries which were collected prior to each blood draw and reviewed after testing was concluded. Each specimen was scored as positive or negative for intake of drugs other than oral contraceptives and multi-vitamins within the 2 weeks prior to specimen collection. No subject began or discontinued oral contraceptives during the study. Alcohol and flavonoid intake was scored as positive or negative for two time periods: morning exposure within 6 hours of blood draw and evening exposure within 12–18 hours of blood draw. The list of flavonoid-rich foods was compiled from the literature (Pearson et al, 2005; Holt et al, 2005) and included cocoa, chocolate, tea, grapes and grape products, fish, onions, garlic, broccoli, apples, citrus fruits, nuts, peanuts, soy, and red, blue, and purple berries.
Platelet Function TestsBlood was collected into evacuated siliconized glass tubes (Becton Dickinson, Franklin Lakes, NJ) containing 3.2% sodium citrate in a ratio of 1:9 with blood and maintained at room temperature. For PRP, blood was centrifuged at 22°–25°C for 8 minutes at 200 × g. After transfer of two-thirds of PRP with a plastic pipette to a polypropylene tube, the remaining PRP was centrifuged at 22°–25°C for 20 minutes at 1,600 × g to produce platelet-poor plasma (PPP), which was transferred with a plastic pipette to another polypropylene tube. PRP was standardized to a platelet count of 250 × 109 platelets L−1 by addition of PPP. For WB testing, whole blood was diluted with an equal volume of 0.9%NaCl.
LTA was measured in PRP using a BioData Platelet Aggregation Profiler, Model PAP-4 (BD) (BioData Corp.) and a Chrono-Log platelet lumi-aggregometer Model 560-CA (CL) (Chrono-Log Corp, Haverton, PA, USA) by change in optical density and expressed as % maximal aggregation. WBA was measured by change in impedance and expressed as ohms in the CL and in aggregation units (AU) in the Multiplate analyzer (MP) (Dynabyte GmbH, Munich, Germany). In CL-PRP and CL-WB, REL was measured by luminescence using luciferin-luciferase reagent (Chrono-Log Corp.) added at a ratio of 50 microliters (μL) to 450 μL PRP or 100 μL to 900 μL diluted WB. REL was calculated by comparison of peak luminescence recorded from the subject sample with that of a 2 μM ATP standard (Chrono-Log Corp) and expressed in μmoles (μM). Reactions were initiated by the addition of agonists to produce the final concentrations recommended by the manufacturers, as shown in Table II.
Statistical MethodsDistributions were checked for normality by the D’Agostino and Pearson omnibus normality check. Multivariate analysis was performed for each method and agonist using gender and race. A P-value was calculated for multiple comparisons by the Bonferroni correction. For 56 comparisons at a cut-off of 0.025, a P-value of <0.0004 was required to achieve significance. Coefficient of variation (CV) was calculated for each subject with four or more specimens for each agonist and method. Reference ranges were calculated with a lower limit of 2.5 percentile and upper limit of 97.5 percentile using results from the first blood draw on each subject with no drugs other than oral contraceptives and multi-vitamins reported during the previous two weeks. Chi-square and Fisher’s exact test were used for comparisons. Statistical analysis was carried out using SAS Version 9.2 (SAS Institute, Cary, NC, USA) and GraphPad Prism 5 (GraphPad Software, Inc., LaJolla, CA, USA).
ResultsPlatelet aggregation was performed on the same blood specimen by each of four methods: LTA using platelet-rich plasma (PRP) in two instruments, one with simultaneous measurement of ATP release (REL), and WBA in two instruments, one with simultaneous measurement of REL. Key findings are summarized in Table III.
Specimen characteristicsSixty-four subjects were studied at two-week intervals over a three month period. A total of 349 specimens were tested, an average of 5.4 specimens per subject. Subjects were not asked to abstain from drug use but were asked to record drugs and foods ingested during the two-week period prior to each blood draw. Exposure to drugs was reported for 217 of 349 specimens (62.2%), 59 (16.9%) with drugs known to affect platelet function. All subjects reported drug use prior to at least one specimen. Specimens with a single drug exposure included 35 with ibuprofen, 10 with acetaminophen, 7 with acetylsalicylic acid, and 5 with loratadine. A total of 162 specimens had multiple drug exposures. Results on specimens with drug exposures will be reported in a separate publication.
A total of 129 specimens from 41 subjects with no drug exposure reported in the previous two weeks other than oral contraceptives or multi-vitamins were classified as drug-free. Characteristics of these subjects were similar to those of the entire study group. Seven specimens from 4 subjects with platelet counts below 150 × 109 L−1 were excluded from analysis, leaving 122 drug-free specimens. Distributions of test results for these specimens are shown in Figure 1.
Evaluation of test profilesReference ranges for each test were calculated using the first drug-free specimen from each individual (Table IV). The remaining 81 drug-free specimens were used as a test group. The profile of tests from a single method, excluding ristocetin, was evaluated for each specimen (Table V). For REL in CL-PRP, ADP and EPI results were excluded, because their reference ranges included zero. The numbers of profiles with one or more abnormal AGG measurements were 21% for CL-PRP, 15% for CL-WB, 13% for BD-PRP, and 6% for MP-WB. When REL was included, 28% of CL-PRP and 30% of CL-WB profiles had at least one abnormality. If an abnormality with 2 or more agonists was required to consider a profile as abnormal, as suggested by Hayward et al (2009), AGG profiles were abnormal in 8.6% of CL-PRP, 3.7% of CL-WB, 1.7% of BD-PRP, and no MP-WB profiles, and profiles combining AGG and REL were abnormal in 9.9% (CL-PRP) and 13.6% (CL-WB).
Effects of race and genderTo investigate sources of the variability observed, a multivariate analysis was performed on the 122 drug-free specimens using the variables race and gender and a P-value adjusted for multiple comparisons. EPI aggregation in CL-PRP and BD-PRP and REL in CL-PRP were significantly reduced in males (P<0.0001) (Figure 2). Blacks had significantly more aggregation than Whites using ADP in CL-WB (P=0.0001) and significantly less REL using collagen and thrombin in PRP and WB and AA in WB (P<0.0001). Ristocetin agglutination in Blacks was significantly lower than in Whites in CL-PRP (P<0.0001) (Figure 3) but not in BD-PRP or either WB system. No significant gender or race differences were seen using MP-WB.
Ristocetin responsePlatelet response to ristocetin varied by instrument (Figure 1A–D, Figure 3), with BD-PRP showing similar medians at all concentrations. Because of these results, a further study was performed using additional concentrations on a new panel of 7 healthy control subjects (Figure 4). In CL-PRP, median agglutination dropped to near zero at 1.00 mg mL−1 but in BD-PRP was strong at 0.75 mg mL−1. At 0.50 mg mL−1, a concentration often used to detect von Willebrand disease Type 2B (VWD2B), CL-PRP showed little response (range 0–2%), while BD-PRP continued to show a median agglutination of 10% (range 0–90%).
Intra-individual variation47gures 5–7 show variation in individual test results over time by subject. To assess intra-individual variation, a coefficient of variation (CV) for each subject with four or more specimens was calculated by agonist and method (Table VI). For PRP aggregation, median CVs were below 15% with collagen, AA, ADP, and EPI using both instruments. For WB aggregation, median CVs were between 15 and 30%. REL showed the greatest intra-individual variation. Eighteen subjects had large CVs due to extremely low results on individual specimens. When these diverged from results on the subject’s other specimens, as illustrated by open circles in Figures 5–7, they were termed “aberrant” specimens.
Effect of food exposuresThe histories of all subjects were examined for reported illness or intake of flavonoid-rich foods or alcohol prior to each specimen. Of 75 specimens drawn after flavonoid-rich food exposures, 24 (32.0%) had aberrant results, compared to 4 of 47 specimens (8.5%) without such exposures (P=0.0035). The distribution of exposures was significantly different between the aberrant and non-aberrant specimens (P<0.0001) (Table VII). Of 28 specimens with aberrant results, 24 (85.7%) were drawn after flavonoid-rich food intake within 18 hours, compared to 51/94 (54.3%) with consistent results (P=0.0035). Exposure within 1–6 hours with or without earlier exposure characterized 19/94 specimens (20.2%) with non-aberrant results and 15/28 (53.6%) with aberrant results (P=0.0004). No other alcohol intake or illnesses were reported in the aberrant group. The tests affected and the specific foods consumed are shown in Table VIII.
DiscussionGuidelines for platelet function testing (Christie et al, 2008; Hayward et al, 2010; Cattaneo et al, 2013) stress the importance of insuring that the patient and normal controls have not taken drugs influencing platelet function. Clinical and Laboratory Standards Institute guidelines also recommend that the patient be fasting (Christie et al, 2008). The characteristics of the normal subjects for reference range definition and daily controls are not otherwise defined. Most clinical laboratories rely on a group of proven donors who have demonstrated normal platelet function on previous occasions. Repeated study of 5 such subjects with LTA showed good reproducibility over a 2 year period (Hevelow et al, 2007). Repeated WBA in 7 subjects over 5 years, however, showed abnormal results on one-third of 86 studies (Refaii et al, 2010). We tested 64 unselected healthy subjects, finding striking differences that must be considered in interpretation of patient results.
Abnormal results in a control subject are often ascribed to the subject having forgotten that they took medication. To improve on simple recall, we prospectively collected exposure data. All subjects had drug exposure on at least one specimen; more than 60% of specimens followed some drug exposure. Although only 17% of exposures were to drugs known to affect platelet function, we excluded all in order to focus on other sources of variability. To mimic the clinical situation, we used the first drug-free specimen from each subject to calculate reference ranges by non-parametric methods, including 95% of the normal population, as recommended (Hayward et al, 2008), and used the 81 remaining drug-free specimens as a test group. Diagnosis of platelet function disorders involves a panel of tests, ranging from at least 5 tests when measuring aggregation alone to a total of 10 when both aggregation and release are measured. The chance that a single test result will fall below a 2.5 percentile cut-off is 12.5% for an aggregation profile using 5 agonists and approaches 25% for a profile including both aggregation and release. Our findings that 6–21% of aggregation profiles and 28–30% of profiles measuring both aggregation and release were abnormal are close to that expectation. The strategy suggested to reduce the number of false positive profiles by requiring abnormality with >1 agonist to classify a profile as abnormal (Hayward et al, 2009) is effective; however, it would exclude patients with a true bleeding disorder affecting response to a single agonist. In spite of the statistical expectation of a large number of false positive tests, the frequency of specimens with minimal aggregation and absent release among healthy subjects is concerning.
Females have been observed to have greater aggregation than males in some studies (Johnson et al, 1975; Meade et al, 1985; Silver et al, 1993; Becker et al, 2006) but not in others (Beyan et al, 2006; Seyfert et al, 2007; Rubak et al, 2012). Here, only epinephrine showed a gender difference. The number of abnormalities seen in healthy subjects using a relatively high concentration of epinephrine (10 μM) in two different instruments makes one question its utility for diagnosis. Although it is included on recommended agonist panels (Hayward et al, 2009; Dawood et al, 2012; Cattaneo et al, 2013), we believe that its use in the initial screening panel may lead to misdiagnosis, particularly in males.
Ristocetin response showed a significant race difference. A single nucleotide polymorphism in von Willebrand factor, D1472H, which is common in Blacks, interferes with ristocetin binding to von Willebrand factor and falsely lowers ristocetin cofactor (Flood et al, 2010). DNA was not available on our subjects to investigate this possibility. The race difference was evident only in CL-PRP. BD-PRP and the WB systems did not show this difference. Black subjects also showed decreased release with collagen and thrombin in PRP and WB and with AA in WB. These differences might also be due to common genetic variants.
Subjects who consumed flavonoid-rich foods prior to testing were significantly more likely to show results differing from their usual pattern than those without such consumption. Studies on the effects of foods on platelet function have focused primarily on their potential long-term use in the prevention of cardiovascular disease and have been conducted after exposure for 1–4 weeks (Pearson et al, 2005; Holt et al, 2005). Flavonoids, components of such common foods as chocolate, tea, red wine, and beer, have been most extensively studied. Black tea (Wolfram et al, 2002) and chocolate (Innes et al, 2003) have been demonstrated to inhibit LTA. Chocolate (Murphy et al, 2003) and grape products, including red wine and grape juice (Keevil et al, 2000; Pignatelli et al, 2002), have shown an effect on WBA. Few studies have assessed how rapidly these effects occur and their persistence. Chocolate prolonged the Platelet Function Analyzer-100 closure time within 4–6 hours (Murphy et al, 2003). Red wine retained an effect at 12 hours post consumption (Pignatelli et al, 2002). In this study, more flavonoid-rich foods were consumed in the evening, but those with aberrant results most often had both morning and evening exposures. These data suggest that a morning fast prior to blood draw is not sufficient to eliminate the effect of inhibitory foods; an abstention from specific foods of 24 hours may be required to eliminate their effect. Other studies have shown no effect of eating breakfast (Stegnar et al, 2010) or lunch (Silver et al, 1993) prior to LTA, but neither considered the intake of specific foods. Our study was limited by the large number of specimens excluded due to drug exposure and the complexity of the dietary information collected. Flavonoid-rich foods are common and often consumed together. Longitudinal studies will be required to determine the timing and magnitude of food effects on clinical platelet function testing.
The test systems used vary in basic mechanism, type of specimen, and the specific proprietary reagents provided as agonists. The BD-PRP system showed the least intra-individual variation over time. Both WB systems showed more variability than PRP, perhaps due to standardization of the platelet count in PRP. The MP-WB system produced the fewest abnormal results in these healthy subjects. CL-PRP and CL-WB results, using the same instrument and reagents, were similar; however, in the CL-PRP system, no ATP release was detected on 6.6% of AA, 8.2% of ADP, and 13.1% of EPI tests, preventing calculation of a lower limit of the reference range. Exclusion of such low results from reference range calculations and yet considering them to be diagnostic in patients is not an acceptable practice. Overall, addition of ATP release measurements to aggregation profiles increased the number of abnormal findings in healthy subjects.
The panels of agonists and concentrations used (Table II) were those recommended by the manufacturers at the time of the study. They are similar but not identical to the recommendations published more recently (Dawood et al, 2012; Cattaneo et al, 2013). Reagent formulations vary by manufacturer, and this may contribute to the differences observed among the systems tested. BD-PRP showed less variability than CL-PRP with different ristocetin concentrations. From this study, it could not be determined if the BD system can detect clinically important differences in ristocetin response, such as that seen in VWD2B. In VWD2B, aggregation with low ristocetin concentrations is expected to be increased over the minimal response usually seen in normal individuals; however, normal subjects had up to 90% aggregation with 0.50 mg/mL ristocetin in the BD system, suggesting that it might be difficult to identify VWD2B with this system.
We conclude that gender, race, diet, and test system affect results of platelet function testing in healthy subjects and that these differences should be considered when interpreting results in patients referred for evaluation of bleeding. Epinephrine and ristocetin are particularly problematic; abnormal results seen with either of these agonists alone may reflect population variation and should be interpreted cautiously. Exclusion of flavonoid-rich foods from the diet for 24 hours prior to testing may decrease false positive results. While it might be presumed that the systems with the least intra-individual variability and the fewest false-positive results would be the best choice for patient testing, increased specificity often results in decreased sensitivity. A similar method comparison using patients with known platelet function defects would be required to determine whether the methods showing the least variability retain sufficient sensitivity for detection of mild platelet defects and whether release measurements, which decrease specificity, add appreciably to diagnostic efficacy. Because of the high rate of abnormal results observed in normal individuals, confirmation of all abnormal results by demonstrating the reproducibility of the defects in a different specimen, along with specific receptor, granule, or DNA studies are needed to ensure accurate diagnosis and avoid incorrectly labeling patients as having an abnormality of platelet function.