- Email: [email protected]
Measurement of Fibrinogen in Frozen Plasma
P
e
r
g
a
m
o
THROMBOSIS RESEARCH
n
Thrombosis Research 88 (1997) 481484
REGULAR ARTICLE
Measurementof Fibrinogenin Frozen Plasma Susan E. McNerlan, Vivienne L. S. Crawford and Robert W. Stout Department of Geriatric Medicine, The Queen’s University of Belfast, Whitla Medical Building, Belfast, Northern Ireland. (Received 29 August 1997 by Editor P.J. Gaffney; revised/accepted 4 November 1997)
Abstract Several large studies have compared fibrinogen measurements determined over a particular time interval. These assays are subject to difficultiesencountered by all laboratories on tests carried out over a period of time such as assaydrift.To avoidthis problem, plasma can be stored frozen and fibrinogen determined in a large number of samples simultaneously.However, a thorough comparisonof measurements carried out in fresh and frozen plasma has not yet been performed. Fibrinogen concentration was therefore determined in fresh plasma samples and then at a later date in the same samples after storage at –70”C. A good correlation was observed between the two measurements, however, bias increased at the higher fibrinogen levels which are most criticalin the determination of thrombotic risk. An increase in measurement error as a result of freezingwas also observed.These effectsmay,therefore, be important considerations in future studies of this nature. @1998 Elsevier Science Ltd.
dering its accurate measurement using inexpensive and reproducible assays of utmost importance. A common cause of inaccuracy in measurement of various proteins is assay drift. This problem may arise when comparisons are to be made between samplesanalysedover a period of time. The possibility of drift can be reduced by using the same batch of reagents for all samples and by incorporating a quality control standard into every test. To exclude drift completely,plasma can be stored at –70”C and all measurements then performed in large batches over a short period of time. However, a thorough investigation of fibrinogen levels measured in both fresh and frozen plasma samples has not been performed. We have, therefore, measured the fibrinogen concentration in fresh plasma samples, stored the plasma at –70°C and then determined the fibrinogen concentration again at a later date to assesshow the two values compare. 1. Materials and Methods
Key Words: Fibrinogen; Clauss
s
1.1. Samples
everal studieshave recently confirmedthe role of fibrinogen as an independent risk factor for cardiovasculardisease [1–3],thereby ren-
Abbreviations:SPSS, SuperiorPerformanceStatisticalSoftware;
CV, coefficientof variation. Correspondingaufhor:Dr. Susan McNerlan, Department of Geriatric Medicine, The Queen’s University of Belfast, Whitla Medical Building, 97 Lisburn Road, Belfast BT9 7AB, United Kingdom. Tel: (01232) 272034; Fax: (01232) 263870;E-mail: (s.mcnerlan@ qub.ac.uk).
004g-sg4g/gT $17.00 + .00@1998Elsevier PII S0049-3848(97)00283-1
As part of a continuing study on seasonal variation of hemostatic factors in elderly people, venous blood samples were collected on several occasions from 59 healthy subjects using a 21 gauge needle and dispensed into bottles containing 3.8’XOtrisodium citrate (9 parts blood to 1 part anticoagulant). Plasma was separated from the cells within 3 h of collection by centrifugation at 2000g for 20 min. Some fresh plasma was used immediately for
Science Ltd. Printed in the USA. All rights reserved.
482
S.E. McNerlanet al./Thrombosis Research88 (1997)481484
fibrinogen determination and the remainder was aliquotted and frozen at –70”C until required but for a maximum of 16 months. The research was approved by the Research Ethics Committee of The Queen’s University of Belfast. . ..
1.2. Fibrinogen Assay Fibrinogen was measured in both the fresh and frozen samples using the Clauss method [4] on a KC1O coagulometer. The frozen plasma samples were thawed rapidly at 37°C for 10 min before analysis. Clauss reagents with the same batch number, including thrombin, fibrinogen standard, and imidazole buffer with a shelf-life of approximately 18 months were purchased from Diagen. Clotting times of the fibrinogen standard were determined at various dilutions ranging from 1/3 to 1/20representing concentrations of 0.88-O.13g/l.Test plasmas were diluted 1/10before measurement, duplicate clotting times recorded, and the corresponding fibrinogenconcentrationscalculatedfrom the fibrinogen standard curve.Tests were standardised by the inclusion of a reference plasma (Irnmuno) in each assay, again the same batch number of plasma was used throughout. The same instrumentation and operator were also used for all tests. 1.3. Statistics Three hundred and sixty-sixpairs of values (fresh and frozen) were entered into the analysis.A paired samples t test was performed to determine the magnitude and significanceif any of the difference between the fresh and the frozen samples. A Pearson correlation coefficient was calculated to quantify the relationship between the two variables (fresh/frozen) and an estimate of bias was obtained from linear regression of the difference between each pair (fresh–frozen) upon the mean of individual pairs ((fresh+ frozen)/2) [5]. All analyses were carried out using Superior Performance Statistical Software (SPSS).
. .*.: . . .. .
. .
.
. ..
2
The coefficient of variation (CV) of the clotting times of the reference plasma at a 1/10 dilution included in every assay of fresh and frozen plasma
5
s
7
Fresh Fibrinogen(g/l)
Fig.1. Plot of fibrinogenmeasuredin fresh plasmavs. frozenplasma.Correlationcoefficient: r=0.953;p<0.0005. Equationof the line: frozen fibrinogen=O.044+0.998x freshfibrinogen. was 2.8Y0.A good correlation (r= 0.953;p<0.0005) was observed between fibrinogen measured in fresh and frozen plasma (Figure 1).However, on comparison of paired samples the means were significantly different, the frozen samples being on average 0.04g/lhigher than the fresh samples (Table 1). Since 366 plasma samples were taken from only 59 individuals,we investigated whether a particular subject showed the same bias, either positive or negative, for all samples. The bias was tested by analysis of variance and found to be random for each sample from the same individual (F= O.82; P=0.82). Therefore, each pair could be treated as individual measures. A graph of pair difference (fresh–frozen) versus mean pair (fresh +frozen/2) showed that the scatter of the differences increased as the concentration Table 1. Table showing means and standard deviations (SD) for fresh and frozen samples arfd results of the paired samples t test Plasma condition
2. Results
4
3
Fresh Frozen
N
Mean (g/l)
SD
366 366
3.5056 3.5413
0.706 0.739
A–O.0357. to -0.013). (95% CI,-0.059 df=365; P=O.002. r=–3.06;
S.E. McNerlan et al./Thrombosis Research 88 (1997) 481-484
4 = ‘1
.
% N
o t 2“
J-
~
-2-
“
.
.
Mean + 2SD (0.1191) . ........... .......................... . “t ... . .. . . .. . . .“ . . . . ...Q:f .%.. t -. . “. ● t, * t..., .O . .. ... . . :.O”. Mean (-0.0089) .. ./. ,.#y”j.. :.. . . . ... t ..
+ a g u, .5 z c
. .% ~. ,:” &@-”+..y2!”;*.:i*” ; “: “.. #”.”.:”...“,: “#.’ .. .“ . “ “ . . ““ ................................ ................. ...................... .
●
.
.
.
Mean- 2SD (:0.1369)
M al
c1
cg
g 0
-.4-
.
-.a,~~
2.0
Mean Pair (Fresh+Frozen/2)
Fig. 2. PIot of mean pair (fresh+ frozen/2) vs. pair difference (fresh–frozen). All values are log transformed.
of fibrinogen increased (~= –0.046; 95Y0CI, –0.08 to –0.01; p= O.004).For example, for fibrinogen values of less than 4.5g/l the calculated bias was –0.0317 (95Y0CI, –0.054 to –0.009) whereas for fibrinogen concentrations greater than 4.5 g/1the bias rose to –0.0778 (95% CI, –0.196 to +0.040). Values were, therefore, log transformed (Figure 2). As shown in Figure 2, the mean difference is –0.0089 and the limits of the agreement are 0.1191 and –0.1369.The antilogsof these limits,therefore, show that for 95Y0of cases the frozen reading will be between 0.872and 1.126times the fresh reading. Thus the frozen measure may differ from the fresh measure by 13Y0above (1–0.872) and 13Y0below (1–1.126). This means, for example, that a fresh plasma fibrinogen concentration of 3g/1may after freezing give a fibrinogen reading of 3.39 g/1 or 2.61 g/l; however, it is more likely to be the higher value, since we have found that frozen samples give higher values than fresh. There was no evidence to suggest that the length of storage time affected the fibrinogen measurement in any way. 3. Discussion Although in routine laboratories fibrinogen is measured in fresh plasma, determination in frozen samples has some advantage, particularly in studies where comparisons are to be drawn between samples taken over a period of time.
483
Assay drift can sometimes account for observed changes but can be monitored by the incorporation of standards into each assay performed. However, drift can be completely excluded by storing samples and then testing large batches within a period of a few days or weeks. The results of the current study indicate that a good correlation exists between fibrinogen concentrations, measured in fresh plasma, and in the same plasma at a later date after storage at –70°C (r= 0.953;p<0.0005), although frozen plasma determinationswere significantlyhigher than in the fresh plasma. The actual difference between the means, however, was small (0.04g/1).Such a smallvariation in measurement would be negligible,for example, in the determination of cardiovascular risk in individual patients. In clinical terms this difference is also small.For example, a difference as high as 0.78 gilbetween winter and summer measures of fibrinogen has been reported [6] and exercise is reported to reduce levels by 0.46 g/1[7]. It may, however, be important to consider the random bias occurring within individuals reported in these results when carrying out certain epidemiologicalstudies. It suggeststhat there is an increase in measurement error as a result of freezing; freezing imparts an added noise. This effect might be important when looking at summer/winter differences. A real increase in winter may be dampened by freezing, whereas a decline in summer may be masked, resulting in a weakened seasonal rhythm. The converse may also be true where a stronger rhythm could artificially result. The reason why the fresh plasma yielded a lower result than the frozen samples is uncertain, in fact the oppositewouldperhaps be expectedsincefreeze/ thawing can sometimescause the formation of cryoprecipitate of which fibrinogen is the main protein component. The removal of fibrinogen in this precipitate would therefore result in a lower value determined in the plasma. One possible reason could be that evaporation occurs upon thawing, resulting in a higher fibrinogen concentration in the plasma. However, the difference between values is perhaps too small to warrant an explanation. On analysis of the deviation from the mean, the bias increased at higher fibrinogen values. Plasma samples containing low levels of fibrinogen therefore gave a better comparison indicating that, in conjunction with the incorporation of an intema-
484
S.E. McNerlan et al./Thrombosis Research 88 (1997) 481484
tional standard, values obtained at different laboratory sitescouldbe compared whether measurements were made in fresh or frozen plasma samples.However, higher amounts of fibrinogen,which are most critical in the determination of increased risk, presented most variation and are perhaps more difficult to compare. In conclusion, this study indicates that determinations made in fresh plasma are comparable to those made in frozen, although freezing may introduce increased error in measurements. WearegratefidtoAnneBrown,researchnurse,ChrisC.Patterson, statistician,the staff of the Hematology Departmentat Belfmt CityHospital,andall volunteerswho took part in thestudy.This project wasfundedby a grant,fromthe Departmentof Health.
References 1. Ernst E. Fibrinogen, an independent cardio-
vascular risk factor. J Int Med 1990;227:365–72.
2. Meade TW, North WRS. Population-based distributions of hemostatic variables. Br Med Bull 1977;33:283-88. 3. Stone MC, Thorp JM. Plasma fibrinogen— a major coronary risk factor. J R Coil Gen Prac 1985;35:565-69. 4. Clauss A. Gerinnungsphysiologische schnellmethode zur bestimmung des fibrinogen. Acta Haematol 1957;17:23746. 5. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;i:307–10. 6. Stout RW, Crawford VLS. Seasonal variations in fibrinogenconcentrationsamong elderly people. Lancet 1991;338:9-13. 7. Elwood PC, Yarnell JWG, Pickering J, Fehily AM, O’Brian JR. Exercise, fibrinogen and other risk factors for ischaemic heart disease. Caerpherhilly prospective heart disease study. Br Heart J 1993;69:183-87.
e
r
g
a
m
o
THROMBOSIS RESEARCH
n
Thrombosis Research 88 (1997) 481484
REGULAR ARTICLE
Measurementof Fibrinogenin Frozen Plasma Susan E. McNerlan, Vivienne L. S. Crawford and Robert W. Stout Department of Geriatric Medicine, The Queen’s University of Belfast, Whitla Medical Building, Belfast, Northern Ireland. (Received 29 August 1997 by Editor P.J. Gaffney; revised/accepted 4 November 1997)
Abstract Several large studies have compared fibrinogen measurements determined over a particular time interval. These assays are subject to difficultiesencountered by all laboratories on tests carried out over a period of time such as assaydrift.To avoidthis problem, plasma can be stored frozen and fibrinogen determined in a large number of samples simultaneously.However, a thorough comparisonof measurements carried out in fresh and frozen plasma has not yet been performed. Fibrinogen concentration was therefore determined in fresh plasma samples and then at a later date in the same samples after storage at –70”C. A good correlation was observed between the two measurements, however, bias increased at the higher fibrinogen levels which are most criticalin the determination of thrombotic risk. An increase in measurement error as a result of freezingwas also observed.These effectsmay,therefore, be important considerations in future studies of this nature. @1998 Elsevier Science Ltd.
dering its accurate measurement using inexpensive and reproducible assays of utmost importance. A common cause of inaccuracy in measurement of various proteins is assay drift. This problem may arise when comparisons are to be made between samplesanalysedover a period of time. The possibility of drift can be reduced by using the same batch of reagents for all samples and by incorporating a quality control standard into every test. To exclude drift completely,plasma can be stored at –70”C and all measurements then performed in large batches over a short period of time. However, a thorough investigation of fibrinogen levels measured in both fresh and frozen plasma samples has not been performed. We have, therefore, measured the fibrinogen concentration in fresh plasma samples, stored the plasma at –70°C and then determined the fibrinogen concentration again at a later date to assesshow the two values compare. 1. Materials and Methods
Key Words: Fibrinogen; Clauss
s
1.1. Samples
everal studieshave recently confirmedthe role of fibrinogen as an independent risk factor for cardiovasculardisease [1–3],thereby ren-
Abbreviations:SPSS, SuperiorPerformanceStatisticalSoftware;
CV, coefficientof variation. Correspondingaufhor:Dr. Susan McNerlan, Department of Geriatric Medicine, The Queen’s University of Belfast, Whitla Medical Building, 97 Lisburn Road, Belfast BT9 7AB, United Kingdom. Tel: (01232) 272034; Fax: (01232) 263870;E-mail: (s.mcnerlan@ qub.ac.uk).
004g-sg4g/gT $17.00 + .00@1998Elsevier PII S0049-3848(97)00283-1
As part of a continuing study on seasonal variation of hemostatic factors in elderly people, venous blood samples were collected on several occasions from 59 healthy subjects using a 21 gauge needle and dispensed into bottles containing 3.8’XOtrisodium citrate (9 parts blood to 1 part anticoagulant). Plasma was separated from the cells within 3 h of collection by centrifugation at 2000g for 20 min. Some fresh plasma was used immediately for
Science Ltd. Printed in the USA. All rights reserved.
482
S.E. McNerlanet al./Thrombosis Research88 (1997)481484
fibrinogen determination and the remainder was aliquotted and frozen at –70”C until required but for a maximum of 16 months. The research was approved by the Research Ethics Committee of The Queen’s University of Belfast. . ..
1.2. Fibrinogen Assay Fibrinogen was measured in both the fresh and frozen samples using the Clauss method [4] on a KC1O coagulometer. The frozen plasma samples were thawed rapidly at 37°C for 10 min before analysis. Clauss reagents with the same batch number, including thrombin, fibrinogen standard, and imidazole buffer with a shelf-life of approximately 18 months were purchased from Diagen. Clotting times of the fibrinogen standard were determined at various dilutions ranging from 1/3 to 1/20representing concentrations of 0.88-O.13g/l.Test plasmas were diluted 1/10before measurement, duplicate clotting times recorded, and the corresponding fibrinogenconcentrationscalculatedfrom the fibrinogen standard curve.Tests were standardised by the inclusion of a reference plasma (Irnmuno) in each assay, again the same batch number of plasma was used throughout. The same instrumentation and operator were also used for all tests. 1.3. Statistics Three hundred and sixty-sixpairs of values (fresh and frozen) were entered into the analysis.A paired samples t test was performed to determine the magnitude and significanceif any of the difference between the fresh and the frozen samples. A Pearson correlation coefficient was calculated to quantify the relationship between the two variables (fresh/frozen) and an estimate of bias was obtained from linear regression of the difference between each pair (fresh–frozen) upon the mean of individual pairs ((fresh+ frozen)/2) [5]. All analyses were carried out using Superior Performance Statistical Software (SPSS).
. .*.: . . .. .
. .
.
. ..
2
The coefficient of variation (CV) of the clotting times of the reference plasma at a 1/10 dilution included in every assay of fresh and frozen plasma
5
s
7
Fresh Fibrinogen(g/l)
Fig.1. Plot of fibrinogenmeasuredin fresh plasmavs. frozenplasma.Correlationcoefficient: r=0.953;p<0.0005. Equationof the line: frozen fibrinogen=O.044+0.998x freshfibrinogen. was 2.8Y0.A good correlation (r= 0.953;p<0.0005) was observed between fibrinogen measured in fresh and frozen plasma (Figure 1).However, on comparison of paired samples the means were significantly different, the frozen samples being on average 0.04g/lhigher than the fresh samples (Table 1). Since 366 plasma samples were taken from only 59 individuals,we investigated whether a particular subject showed the same bias, either positive or negative, for all samples. The bias was tested by analysis of variance and found to be random for each sample from the same individual (F= O.82; P=0.82). Therefore, each pair could be treated as individual measures. A graph of pair difference (fresh–frozen) versus mean pair (fresh +frozen/2) showed that the scatter of the differences increased as the concentration Table 1. Table showing means and standard deviations (SD) for fresh and frozen samples arfd results of the paired samples t test Plasma condition
2. Results
4
3
Fresh Frozen
N
Mean (g/l)
SD
366 366
3.5056 3.5413
0.706 0.739
A–O.0357. to -0.013). (95% CI,-0.059 df=365; P=O.002. r=–3.06;
S.E. McNerlan et al./Thrombosis Research 88 (1997) 481-484
4 = ‘1
.
% N
o t 2“
J-
~
-2-
“
.
.
Mean + 2SD (0.1191) . ........... .......................... . “t ... . .. . . .. . . .“ . . . . ...Q:f .%.. t -. . “. ● t, * t..., .O . .. ... . . :.O”. Mean (-0.0089) .. ./. ,.#y”j.. :.. . . . ... t ..
+ a g u, .5 z c
. .% ~. ,:” &@-”+..y2!”;*.:i*” ; “: “.. #”.”.:”...“,: “#.’ .. .“ . “ “ . . ““ ................................ ................. ...................... .
●
.
.
.
Mean- 2SD (:0.1369)
M al
c1
cg
g 0
-.4-
.
-.a,~~
2.0
Mean Pair (Fresh+Frozen/2)
Fig. 2. PIot of mean pair (fresh+ frozen/2) vs. pair difference (fresh–frozen). All values are log transformed.
of fibrinogen increased (~= –0.046; 95Y0CI, –0.08 to –0.01; p= O.004).For example, for fibrinogen values of less than 4.5g/l the calculated bias was –0.0317 (95Y0CI, –0.054 to –0.009) whereas for fibrinogen concentrations greater than 4.5 g/1the bias rose to –0.0778 (95% CI, –0.196 to +0.040). Values were, therefore, log transformed (Figure 2). As shown in Figure 2, the mean difference is –0.0089 and the limits of the agreement are 0.1191 and –0.1369.The antilogsof these limits,therefore, show that for 95Y0of cases the frozen reading will be between 0.872and 1.126times the fresh reading. Thus the frozen measure may differ from the fresh measure by 13Y0above (1–0.872) and 13Y0below (1–1.126). This means, for example, that a fresh plasma fibrinogen concentration of 3g/1may after freezing give a fibrinogen reading of 3.39 g/1 or 2.61 g/l; however, it is more likely to be the higher value, since we have found that frozen samples give higher values than fresh. There was no evidence to suggest that the length of storage time affected the fibrinogen measurement in any way. 3. Discussion Although in routine laboratories fibrinogen is measured in fresh plasma, determination in frozen samples has some advantage, particularly in studies where comparisons are to be drawn between samples taken over a period of time.
483
Assay drift can sometimes account for observed changes but can be monitored by the incorporation of standards into each assay performed. However, drift can be completely excluded by storing samples and then testing large batches within a period of a few days or weeks. The results of the current study indicate that a good correlation exists between fibrinogen concentrations, measured in fresh plasma, and in the same plasma at a later date after storage at –70°C (r= 0.953;p<0.0005), although frozen plasma determinationswere significantlyhigher than in the fresh plasma. The actual difference between the means, however, was small (0.04g/1).Such a smallvariation in measurement would be negligible,for example, in the determination of cardiovascular risk in individual patients. In clinical terms this difference is also small.For example, a difference as high as 0.78 gilbetween winter and summer measures of fibrinogen has been reported [6] and exercise is reported to reduce levels by 0.46 g/1[7]. It may, however, be important to consider the random bias occurring within individuals reported in these results when carrying out certain epidemiologicalstudies. It suggeststhat there is an increase in measurement error as a result of freezing; freezing imparts an added noise. This effect might be important when looking at summer/winter differences. A real increase in winter may be dampened by freezing, whereas a decline in summer may be masked, resulting in a weakened seasonal rhythm. The converse may also be true where a stronger rhythm could artificially result. The reason why the fresh plasma yielded a lower result than the frozen samples is uncertain, in fact the oppositewouldperhaps be expectedsincefreeze/ thawing can sometimescause the formation of cryoprecipitate of which fibrinogen is the main protein component. The removal of fibrinogen in this precipitate would therefore result in a lower value determined in the plasma. One possible reason could be that evaporation occurs upon thawing, resulting in a higher fibrinogen concentration in the plasma. However, the difference between values is perhaps too small to warrant an explanation. On analysis of the deviation from the mean, the bias increased at higher fibrinogen values. Plasma samples containing low levels of fibrinogen therefore gave a better comparison indicating that, in conjunction with the incorporation of an intema-
484
S.E. McNerlan et al./Thrombosis Research 88 (1997) 481484
tional standard, values obtained at different laboratory sitescouldbe compared whether measurements were made in fresh or frozen plasma samples.However, higher amounts of fibrinogen,which are most critical in the determination of increased risk, presented most variation and are perhaps more difficult to compare. In conclusion, this study indicates that determinations made in fresh plasma are comparable to those made in frozen, although freezing may introduce increased error in measurements. WearegratefidtoAnneBrown,researchnurse,ChrisC.Patterson, statistician,the staff of the Hematology Departmentat Belfmt CityHospital,andall volunteerswho took part in thestudy.This project wasfundedby a grant,fromthe Departmentof Health.
References 1. Ernst E. Fibrinogen, an independent cardio-
vascular risk factor. J Int Med 1990;227:365–72.
2. Meade TW, North WRS. Population-based distributions of hemostatic variables. Br Med Bull 1977;33:283-88. 3. Stone MC, Thorp JM. Plasma fibrinogen— a major coronary risk factor. J R Coil Gen Prac 1985;35:565-69. 4. Clauss A. Gerinnungsphysiologische schnellmethode zur bestimmung des fibrinogen. Acta Haematol 1957;17:23746. 5. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;i:307–10. 6. Stout RW, Crawford VLS. Seasonal variations in fibrinogenconcentrationsamong elderly people. Lancet 1991;338:9-13. 7. Elwood PC, Yarnell JWG, Pickering J, Fehily AM, O’Brian JR. Exercise, fibrinogen and other risk factors for ischaemic heart disease. Caerpherhilly prospective heart disease study. Br Heart J 1993;69:183-87.