COAGULATION MEDICINE Original Article
Proficiency Testing of Platelet Counting in Ontario REINHARD C. LOHMANN, M.D.,1 LINDA N. CRAWFORD, M.T.(ASCP),2 AND DONALD E. WOOD, M.D.2
with other components of hematologic cytometry. A major contributor to this shortfall is the method used by the laboratory. It was not a surprise that manual counting had a high coefficient of variation. Unexpectedly, the semiautomated methods used in the mid-1980s also displayed a high coefficient of variation. (Key words: Platelet counting; Proficiency testing; Quality control standards) Am J Clin Pathol 1992; 98:231-236
Accurate platelet counting is more difficult than other blood cellular elements because platelets are small and tend to aggregate or adhere to foreign surfaces. Many methods have been described since Fonio1 first reported indirect platelet counting. Between 1981 and 1990, major changes in methodology and technology of platelet counting have taken place. It is the purpose of this review to examine how these changes have reflected on the results of platelet counting in Ontario as portrayed by proficiency testing. In the Province of Ontario, legislation enables the Ministry of Health to require mandatory testing and evaluation of proficiency of all licensed clinical laboratories. Licensing and inspection is carried out by the Ministry of Health. Through an agreement between the Ontario Medical Association and the Ministry of Health, the Ontario Medical Association is identified as an agent for the purpose of proficiency testing. The Laboratory Proficiency Testing Program is the unit within the Ontario Medical Association that performs mandatory testing and proficiency evaluation in Ontario. Its terms of reference have
been described elsewhere.2-3 The Laboratory Proficiency Testing Program has tested laboratories in Ontario in hematology and other disciplines since 1975 and has concluded that the overall results in hemoglobinometry and white blood cell counting were satisfactory.2'3 The Laboratory Proficiency Testing Program's platelet count testing started as a pilot project in 1979 and became part of routine testing in 1981. MATERIALS AND METHODS The Testing Protocol Testing material for platelet counting is distributed four times per year to all laboratories in Ontario that are licensed to perform this test. The testing dates for each testing period (survey) are specified by instructions accompanying the testing material. Participating laboratories return the results on a standardized form—the Analysis Worksheet—by a stipulated date. The Testing Material
The testing material consists of mammalian platelets From the ^Hematology Laboratory, Victoria Hospital Corporation. London, Ontario, and the 2Laboratory Proficiency Testing Program, To- by glyceraldehyde. From 1979 to 1989, the stabilized ronto, Ontario, Canada. platelets were distributed in a plasma suspension. The testing material was first supplied by BPH of West Chester, Received August 5, 1991; received revised manuscript and accepted for publication January 9, 1992. Pennsylvania (1979 to 1982) and subsequently by Fisher Address reprint requests to Dr. Wood: Director, Laboratory Proficiency Diagnostics of Orangeburg, New York (1983 to 1989). Testing Program, 250 Bloor Street East, Suite 501, Toronto, Ontario, Commencing in 1989, a whole blood product was supplied M4W 1E6. 231
Downloaded from http://ajcp.oxfordjournals.org/ by guest on June 5, 2016
Proficiency testing results provide data for interlaboratory comparisons. The Laboratory Proficiency Testing Program of Ontario, Canada, has tested platelet counting for a decade and other hematology parameters for 15 years. Despite the fact that all testing programs are compromised by the type of testing material used, for platelet count testing, conclusions are possible. The data indicate that most laboratories in Ontario have greater difficulty performing reproducible platelet counting as compared
232
COAGULATION MEDICINE Original Article
by R & D Systems Inc. of Minneapolis, Minnesota, permitting the simultaneous testing of red blood cell counting, white blood cell counting, automated differential counting, hemoglobin, packed cell volume, and red cell indices as well as platelet counting on most types of equipment used at this time in Ontario. Testing material target values for each constituent are selected to represent both normal and abnormal patient populations. Target platelet count values ranged from 35 to 505 X 109/L. The Analysis of Results
Laboratories with results greater than 3 "SD units" from the all-methods' mean are considered separately for committee action. Individual results are manually compared with the peer-method mean and assessed against any particular bias. Special letters or requests for clarification or explanation may be mailed to the Laboratory Director. Additionally, an on-site visit may be arranged if the individual survey results for platelet counting and other hematologic tests are of sufficient concern or if there is a pattern of poor performance over time. If, after the onsite visit, the laboratory's performance does not improve, the laboratory is reported through senior committees to the Ministry of Health. Suspension or withdrawal of the laboratory's license can be the ultimate consequence.
Number of Laboratories The number of laboratories providing platelet counting decreased from 392 to 345 between 1981 and 1990 because of surrender or withdrawal of license for platelet counting or the cessation of the entire laboratory operation. In Ontario, new licenses or augmentation of licenses are difficult to obtain. Type of Equipment Counting methods may be clumped into three groupings: manual methods, semiautomated machines, and fully automated machines. Manual counting procedures consist of microscopic hemacytometer counts performed after a manual dilution of whole blood through bulk dilution or Unopette techniques (Becton-Dickinson, Franklin Lakes, NJ). Semiautomated counting equipment is based on impedance counting after individual dilution of whole blood performed either manually or by standalone, automated diluters, or on preparations of plateletrich plasma. Fully automated counting equipment is based largely on impedance counting on whole blood aspirated and diluted directly by the instrument without manual steps other than positioning the tube of blood. There has been a major shift in platelet counting methodology during the observation span. In 1981, the prevailing method was manual light microscopic examination, with or without phase contrast. The number of laboratories using manual methods has decreased from 311
Manual Semi-Automated 300
Automated
200
Number and Frequency of Surveys From 1981 to 1990, a total of 41 surveys were conducted with a total of 122 vials of platelet count testing material being sent to each laboratory. Of these, two surveys (each with six vials of testing material) had to be invalidated because of material problems, usually platelet clumping. Survey frequency and number of challenges per survey have changed over the years. Five vials are distributed to each laboratory four times a year, one vial
1
1981
1
1
1
1
1
1
1
1982 1983 1984 198S 1986 1987 1988 1989 1990
Year FIG. 1. Number of laboratories using three platelet counting methods, 1981-1990.
A.J.C.P. • August 1992
Downloaded from http://ajcp.oxfordjournals.org/ by guest on June 5, 2016
Results are compiled on a Hewlett Packard 3000 computer. An all-methods' all-laboratory mean is computed for each vial and standard deviation (SD) calculation applied. This is denned as an "SD unit." Individual method means and biases are also calculated. The data are then manually reviewed by the committee in light of the performance of the entire cohort. If the results appear to be satisfactory, they are distributed to the participating laboratories on a Survey Report, which displays the participant's results alongside the all-methods' mean and the calculated SD. In addition, the Survey Report contains two histograms: one showing the participant's result in "SD units" as a deviation from the reference mean, and the other showing all results plotted in actual counts on which the participant's own result is highlighted. Outlying results are brought to the attention of the respective laboratory by a special identifier on the Survey Report.
to be tested each day on five consecutive days. Periodically, two or more vials from the same lot number with identical target values are issued for precision studies. No testing material has been mailed between November and February, to avoid samples being exposed to very cold temperatures.
233
LOHMANN, CRAWFORD, AND WOOD Platelet Counting in Ontario TABLE 1. COMPARATIVE COEFFICIENTS OF VARIATION (CV) FOR WBC AND PLATELET COUNTING RESULTS IN REPRESENTATIVE SURVEYS 1985-1990 WBC
Platelet
All Methods (mean X 109/L)
1990 1989 1988 1987 1986
cv%
11.2 5.9 10.6 15.5 11.0
3.1 4.8 4.2 3.6 4.5
All Methods (mean X 10* IL) 243 188 199 188 50
•
16 e»
14 -
> y
12 -
«
10 •
1 CV %
T
s
•s fe «
5.6 8.4 9.6 10.6 18.1
e Coe
Year
18 •
• • • .• • ':
• •
• • • • •
8-
•• / • •
B •
•
• •
42 •
WBC = white blood cell count.
n I
, —
i—i—«—i—
50
100
150
200
i — •
250
i
•
300
i
350
• i 400
• — i — > — i » 450 500
550
All-methods' Mean x lO'/L 2. Platelet count versus measured coefficient of variation, 19871990.
FIG.
As expected, the coefficients of variation varied inversely to the level of the target value. Figure 2 shows the range of platelet counts during the period 1987 to 1990, with the corresponding coefficients of variation. The coefficients of variation lowered as the platelet count increased. Survey results for platelet counting in both the low and the normal ranges were compared over the observation period by selecting vials with the worst coefficients of variation. Figure 3 shows that the platelet counting results at both ranges displayed a modest improvement that is most pronounced at lower levels. Each surveyed laboratory is coded by the method used for platelet counting, and by the type of equipment and the manufacturer. For 1990, there were 21 different codes for platelet counting (Table 3). For the purpose of this review, these codes were grouped into three categories: manual methods; semiautomated methods; automated equipment. Figure 4 compares the three categories of counting methods for survey results with means more than 100 X 109/L between 1981 and 1990 expressed in coefficients of variation.
RESULTS Proficiency testing results for platelet counting consistently showed much larger variances than comparable cell counting throughout the observation period. Table 1 compares white blood cell counting results with platelet counting results for representative surveys during the last 5 years. It can be seen that the coefficients of variation at the levels depicted were in the 3% to 5% range for white blood cell counts and in the 5% to 18% range for platelet counts. Within hematologic cytometry, only reticulocyte counting produced worse results. Table 2 shows the corresponding results of three precision studies for white blood cell and platelet counting. The percentage difference normalized the actual difference between white blood cell and platelet counting results, and allowed a comparison of the data. There was a relative lack of platelet count precision, as expressed in higher percentage differences.
TABLE 2. PRECISION STUDY RESULTS FOR WHITE BLOOD CELL COUNT AND PLATELET COUNTING IN THREE SURVEYS Platelet
WBC Precision Study Survey Vials
Actual Result All Methods (mean X 10"IL)
Result Difference* (X109/L)
9002 C vs. E 8908 C vs. E 8808 A vs. C
20.0 2.0 14.4
0.5 0.2 0.4
% Difference^
Actual Result All Methods (mean X 109/L)
Result Difference* (X109/L)
% Difference^
2.5 10.2 2.8
499 54 41
31 12 8
6.2 22.2 19.4
* Result difference = actual difference between vial results achieved by 90% to 94% of all laboratories.
t % Difference = result difference divided by the mean value (X100).
Vol. 98 • No. 2
Downloaded from http://ajcp.oxfordjournals.org/ by guest on June 5, 2016
in 1981 to fewer than 70 in 1990. Although semiautomated machines were popular in the early 1980s (Fig. 1), the trend today is toward fully automated counting equipment. Although newer semiautomated machines are now available that have improved solid-state electronics, it is clear that the data from this review cannot be used to comment on these newer machines.
All-methods' Mean
Proficiency Testing of Platelet Counting in Ontario REINHARD C. LOHMANN, M.D.,1 LINDA N. CRAWFORD, M.T.(ASCP),2 AND DONALD E. WOOD, M.D.2
with other components of hematologic cytometry. A major contributor to this shortfall is the method used by the laboratory. It was not a surprise that manual counting had a high coefficient of variation. Unexpectedly, the semiautomated methods used in the mid-1980s also displayed a high coefficient of variation. (Key words: Platelet counting; Proficiency testing; Quality control standards) Am J Clin Pathol 1992; 98:231-236
Accurate platelet counting is more difficult than other blood cellular elements because platelets are small and tend to aggregate or adhere to foreign surfaces. Many methods have been described since Fonio1 first reported indirect platelet counting. Between 1981 and 1990, major changes in methodology and technology of platelet counting have taken place. It is the purpose of this review to examine how these changes have reflected on the results of platelet counting in Ontario as portrayed by proficiency testing. In the Province of Ontario, legislation enables the Ministry of Health to require mandatory testing and evaluation of proficiency of all licensed clinical laboratories. Licensing and inspection is carried out by the Ministry of Health. Through an agreement between the Ontario Medical Association and the Ministry of Health, the Ontario Medical Association is identified as an agent for the purpose of proficiency testing. The Laboratory Proficiency Testing Program is the unit within the Ontario Medical Association that performs mandatory testing and proficiency evaluation in Ontario. Its terms of reference have
been described elsewhere.2-3 The Laboratory Proficiency Testing Program has tested laboratories in Ontario in hematology and other disciplines since 1975 and has concluded that the overall results in hemoglobinometry and white blood cell counting were satisfactory.2'3 The Laboratory Proficiency Testing Program's platelet count testing started as a pilot project in 1979 and became part of routine testing in 1981. MATERIALS AND METHODS The Testing Protocol Testing material for platelet counting is distributed four times per year to all laboratories in Ontario that are licensed to perform this test. The testing dates for each testing period (survey) are specified by instructions accompanying the testing material. Participating laboratories return the results on a standardized form—the Analysis Worksheet—by a stipulated date. The Testing Material
The testing material consists of mammalian platelets From the ^Hematology Laboratory, Victoria Hospital Corporation. London, Ontario, and the 2Laboratory Proficiency Testing Program, To- by glyceraldehyde. From 1979 to 1989, the stabilized ronto, Ontario, Canada. platelets were distributed in a plasma suspension. The testing material was first supplied by BPH of West Chester, Received August 5, 1991; received revised manuscript and accepted for publication January 9, 1992. Pennsylvania (1979 to 1982) and subsequently by Fisher Address reprint requests to Dr. Wood: Director, Laboratory Proficiency Diagnostics of Orangeburg, New York (1983 to 1989). Testing Program, 250 Bloor Street East, Suite 501, Toronto, Ontario, Commencing in 1989, a whole blood product was supplied M4W 1E6. 231
Downloaded from http://ajcp.oxfordjournals.org/ by guest on June 5, 2016
Proficiency testing results provide data for interlaboratory comparisons. The Laboratory Proficiency Testing Program of Ontario, Canada, has tested platelet counting for a decade and other hematology parameters for 15 years. Despite the fact that all testing programs are compromised by the type of testing material used, for platelet count testing, conclusions are possible. The data indicate that most laboratories in Ontario have greater difficulty performing reproducible platelet counting as compared
232
COAGULATION MEDICINE Original Article
by R & D Systems Inc. of Minneapolis, Minnesota, permitting the simultaneous testing of red blood cell counting, white blood cell counting, automated differential counting, hemoglobin, packed cell volume, and red cell indices as well as platelet counting on most types of equipment used at this time in Ontario. Testing material target values for each constituent are selected to represent both normal and abnormal patient populations. Target platelet count values ranged from 35 to 505 X 109/L. The Analysis of Results
Laboratories with results greater than 3 "SD units" from the all-methods' mean are considered separately for committee action. Individual results are manually compared with the peer-method mean and assessed against any particular bias. Special letters or requests for clarification or explanation may be mailed to the Laboratory Director. Additionally, an on-site visit may be arranged if the individual survey results for platelet counting and other hematologic tests are of sufficient concern or if there is a pattern of poor performance over time. If, after the onsite visit, the laboratory's performance does not improve, the laboratory is reported through senior committees to the Ministry of Health. Suspension or withdrawal of the laboratory's license can be the ultimate consequence.
Number of Laboratories The number of laboratories providing platelet counting decreased from 392 to 345 between 1981 and 1990 because of surrender or withdrawal of license for platelet counting or the cessation of the entire laboratory operation. In Ontario, new licenses or augmentation of licenses are difficult to obtain. Type of Equipment Counting methods may be clumped into three groupings: manual methods, semiautomated machines, and fully automated machines. Manual counting procedures consist of microscopic hemacytometer counts performed after a manual dilution of whole blood through bulk dilution or Unopette techniques (Becton-Dickinson, Franklin Lakes, NJ). Semiautomated counting equipment is based on impedance counting after individual dilution of whole blood performed either manually or by standalone, automated diluters, or on preparations of plateletrich plasma. Fully automated counting equipment is based largely on impedance counting on whole blood aspirated and diluted directly by the instrument without manual steps other than positioning the tube of blood. There has been a major shift in platelet counting methodology during the observation span. In 1981, the prevailing method was manual light microscopic examination, with or without phase contrast. The number of laboratories using manual methods has decreased from 311
Manual Semi-Automated 300
Automated
200
Number and Frequency of Surveys From 1981 to 1990, a total of 41 surveys were conducted with a total of 122 vials of platelet count testing material being sent to each laboratory. Of these, two surveys (each with six vials of testing material) had to be invalidated because of material problems, usually platelet clumping. Survey frequency and number of challenges per survey have changed over the years. Five vials are distributed to each laboratory four times a year, one vial
1
1981
1
1
1
1
1
1
1
1982 1983 1984 198S 1986 1987 1988 1989 1990
Year FIG. 1. Number of laboratories using three platelet counting methods, 1981-1990.
A.J.C.P. • August 1992
Downloaded from http://ajcp.oxfordjournals.org/ by guest on June 5, 2016
Results are compiled on a Hewlett Packard 3000 computer. An all-methods' all-laboratory mean is computed for each vial and standard deviation (SD) calculation applied. This is denned as an "SD unit." Individual method means and biases are also calculated. The data are then manually reviewed by the committee in light of the performance of the entire cohort. If the results appear to be satisfactory, they are distributed to the participating laboratories on a Survey Report, which displays the participant's results alongside the all-methods' mean and the calculated SD. In addition, the Survey Report contains two histograms: one showing the participant's result in "SD units" as a deviation from the reference mean, and the other showing all results plotted in actual counts on which the participant's own result is highlighted. Outlying results are brought to the attention of the respective laboratory by a special identifier on the Survey Report.
to be tested each day on five consecutive days. Periodically, two or more vials from the same lot number with identical target values are issued for precision studies. No testing material has been mailed between November and February, to avoid samples being exposed to very cold temperatures.
233
LOHMANN, CRAWFORD, AND WOOD Platelet Counting in Ontario TABLE 1. COMPARATIVE COEFFICIENTS OF VARIATION (CV) FOR WBC AND PLATELET COUNTING RESULTS IN REPRESENTATIVE SURVEYS 1985-1990 WBC
Platelet
All Methods (mean X 109/L)
1990 1989 1988 1987 1986
cv%
11.2 5.9 10.6 15.5 11.0
3.1 4.8 4.2 3.6 4.5
All Methods (mean X 10* IL) 243 188 199 188 50
•
16 e»
14 -
> y
12 -
«
10 •
1 CV %
T
s
•s fe «
5.6 8.4 9.6 10.6 18.1
e Coe
Year
18 •
• • • .• • ':
• •
• • • • •
8-
•• / • •
B •
•
• •
42 •
WBC = white blood cell count.
n I
, —
i—i—«—i—
50
100
150
200
i — •
250
i
•
300
i
350
• i 400
• — i — > — i » 450 500
550
All-methods' Mean x lO'/L 2. Platelet count versus measured coefficient of variation, 19871990.
FIG.
As expected, the coefficients of variation varied inversely to the level of the target value. Figure 2 shows the range of platelet counts during the period 1987 to 1990, with the corresponding coefficients of variation. The coefficients of variation lowered as the platelet count increased. Survey results for platelet counting in both the low and the normal ranges were compared over the observation period by selecting vials with the worst coefficients of variation. Figure 3 shows that the platelet counting results at both ranges displayed a modest improvement that is most pronounced at lower levels. Each surveyed laboratory is coded by the method used for platelet counting, and by the type of equipment and the manufacturer. For 1990, there were 21 different codes for platelet counting (Table 3). For the purpose of this review, these codes were grouped into three categories: manual methods; semiautomated methods; automated equipment. Figure 4 compares the three categories of counting methods for survey results with means more than 100 X 109/L between 1981 and 1990 expressed in coefficients of variation.
RESULTS Proficiency testing results for platelet counting consistently showed much larger variances than comparable cell counting throughout the observation period. Table 1 compares white blood cell counting results with platelet counting results for representative surveys during the last 5 years. It can be seen that the coefficients of variation at the levels depicted were in the 3% to 5% range for white blood cell counts and in the 5% to 18% range for platelet counts. Within hematologic cytometry, only reticulocyte counting produced worse results. Table 2 shows the corresponding results of three precision studies for white blood cell and platelet counting. The percentage difference normalized the actual difference between white blood cell and platelet counting results, and allowed a comparison of the data. There was a relative lack of platelet count precision, as expressed in higher percentage differences.
TABLE 2. PRECISION STUDY RESULTS FOR WHITE BLOOD CELL COUNT AND PLATELET COUNTING IN THREE SURVEYS Platelet
WBC Precision Study Survey Vials
Actual Result All Methods (mean X 10"IL)
Result Difference* (X109/L)
9002 C vs. E 8908 C vs. E 8808 A vs. C
20.0 2.0 14.4
0.5 0.2 0.4
% Difference^
Actual Result All Methods (mean X 109/L)
Result Difference* (X109/L)
% Difference^
2.5 10.2 2.8
499 54 41
31 12 8
6.2 22.2 19.4
* Result difference = actual difference between vial results achieved by 90% to 94% of all laboratories.
t % Difference = result difference divided by the mean value (X100).
Vol. 98 • No. 2
Downloaded from http://ajcp.oxfordjournals.org/ by guest on June 5, 2016
in 1981 to fewer than 70 in 1990. Although semiautomated machines were popular in the early 1980s (Fig. 1), the trend today is toward fully automated counting equipment. Although newer semiautomated machines are now available that have improved solid-state electronics, it is clear that the data from this review cannot be used to comment on these newer machines.
All-methods' Mean
