Professional Documents
Culture Documents
The synchronous blind comparison test of TSH Rapid Quantitative Test (fluorescence
immunochromatography) of Guangzhou Wondfo Biotech Co., Ltd. and Roche thyroid
stimulating hormone assay kit (electrochemiluminescence) is conducted to investigate the
equivalence between this reagent and the same product that has been approved for listing.
Result:
(1) The content and process of this study meet the experimental design.
(2) The concentration range of the selected samples of the center meets the study
requirements.
(3) The qualitative analysis of the results of serum samples obtained from 220serum
sample evaluation reagents and reference reagents is conducted and a total of 210
samples with the same results as the evaluation reagent and the reference reagent.
Of which 86 cases are detected with abnormalities, 124 cases are detected normal
and 10 cases with inconsistent results. The two methods total compliance rate:
95.45%, 95%CI is [91.80%~97.80%], positive compliance rate: 93.48%, 95%CI
is [86.34%~97.57%], negative compliance rate: 96.88%,95%CI is [ 92.19%~
97.80% ], kappa consistency coefficient k = 0.906, p = 0.000 <0.05. After 11
samples outside the linear range are excluded, the results of 209 serum samples are
statistically analyzed. The Spearman's correlation coefficient between the reagent
and the reference reagent is 0.982 and p=0. Linear regression is used to estimate
the results. The results showed a good linear relationship between the two methods.
The regression coefficient obtained is 1.051, the 95% confidence interval of the
regression coefficient is [1.023~1.079], and the intercept obtained is -0.111. The
95% confidence interval for the intercept is [-0.629~0.408], and the linear
regression equation is y=1.051x-0.111; When the medical decision level is
0.1mIU/L, estimated expected bias is -0.106, the expected bias is 95% confidence
interval [-0.595 ~ 0.383], and when the medical decision level is 10mIU/L,
estimated expected bias is 0.399, and the expected bias is 95% confidence interval
[-0.089~0.888]. The bias at the medical decision level is outside the 12.5%
tolerance. For the Bland-Altman analysis, the mean difference is -0.292, the
standard deviation difference is 3.534, the 95% cutoff is [-7.218 ~ 6.634],
out-of-bounds point is 10 and the ratio of out-of-bounds points is 4.78%.
(4) A total of 116 samples of homologous serum and plasma samples are tested by
evaluation reagents. After 8 samples outside the linear range are excluded, the
results of 108 homologous samples are statistically analyzed. The Spearman
correlation coefficient of the statistical results is 0.994, p = 0; Linear regression is
used to estimate the results and its results showed a good linear relationship
between the two types of samples. The regression coefficient obtained is 1.123, the
95% confidence interval of the regression coefficient is [1.099~1.148], and the
intercepts obtained at the same time for -0.233, the 95% confidence interval for the
intercept is [-0.396~-0.070] and the linear regression equation is y=1.123x-0.233.
When the medical decision level is0.1mIU/L, the estimated expected bias is -0.221,
the expected bias is 95% confidence interval[-0.425~-0.016], and when the
medical decision level is 10mIU/L, estimated expected bias is 0.997, the expected
bias. 95% confidence interval [0.792~1.202]. The bias at the medical decision
level is within 12.5% of the allowable error. The results are analyzed by
Bland-Altman, with a mean difference of -0.138, a standard deviation of 1.064, a
95% cutoff value of [-2.223~1.946], out-of-bounds point is 7 and the ratio of
out-of-bounds points is 6.48%.
(5) A total of 120samples of homologous serum and whole blood samples are tested
by evaluation reagents. After 9 samples outside the linear range are excluded, the
results of 111 homologous samples are statistically analyzed. The Spearman
correlation coefficient of the statistical results is 0.989, p = 0; Linear regression is
used to estimate the results and its results showed a good linear relationship
between the two types of samples. The regression coefficient obtained is 1.120, the
95% confidence interval of the regression coefficient is [1.088~1.152], and the
intercepts obtained at the same time for -0.320, the 95% confidence interval for the
intercept is [-0.554~-0.086] and the linear regression equation is y=1.120x-0.320.
When the medical decision level is 0.1mIU/L, the estimated expected bias is
-0.308, the expected bias is 95% confidence interval[-0.565~-0.051], and when
the medical decision level is 10mIU/L, the estimated expected bias is 0.880, the
expected bias. 95% confidence interval [0.623~1.137]. The bias at the medical
decision level is within 12.5% of the allowable error. The results are analyzed by
Bland-Altman, with a mean difference of -0.076, a standard deviation of 1.354, a
95% cutoff value of [-2.730~2.578], out-of-bounds point is 3 and the ratio of
out-of-bounds points is 2.70%.
Conclusion:
Guangzhou Wondfo Biotech Co., Ltd.'s TSH Rapid Quantitative Test (fluorescent
immunochromatography) does not have a statistically significant difference from the test
results of the same product that has been approved for marketing, and has a high degree of
consistency, that is clinical application performance is equivalent. In addition, the
evaluation reagent also has equivalence in detecting homologous serum and plasma, whole
blood samples.
1. Establishment of reference reagent
Reference reagent: Roche thyroid stimulating hormone assay kit
(electrochemiluminescence).
Sample Type: Serum/Plasma.
Registration No.: National Food and Drug Administration (Jin) 2014 No. 2404873.
The Roche thyroid stimulating hormone assay kit (Electrochemiluminescence) has a
higher market share and a better market evaluation. Therefore, this product is selected as a
reference reagent.
N ^
(Yi Yi ) 2
i 1
B̂c
Calculate the estimated expected bias ( ) and its 95% confidence interval at the
clinical cutoff ( X C ).
Bˆ c a (b 1) X c
B̂c ^
1 ( X X )2
95% confidence interval = B C
t / 2, N 2 S y x N C
( X i X )2
N
i 1
Sample Type
Selected serum
Center Excluding samples Effective serum samples
samples
102 0 102
120 2 ( Repeated sample and 118
operational errors result in
insufficient sample size)
Total 222 2 220
Conformance evaluation of effective samples
4.1.2 Qualitative analysis
According to the decision thresholds recommended in the kit instructions for the two
methods, as for the Roche thyroid stimulating hormone assay kit
(electrochemiluminescence) kit, 0.27mIU/L<sample concentration <4.2mIU/L judged
negative and the sample concentration outside its range is judged to be positive. As for T4
Rapid Quantitative Test of Guangzhou Wondfo Biotech Co.,Ltd., 0.3mIU/L<sample
concentration <4.2mIU/L judged negative and the sample concentration outside its range is
judged to be positive.
Wondfo+ 86 4 90
Table7 Coefficienta
B̂c
estimate ( ) and its 95% confidence interval (The confidence interval of the expected
bias includes the lower limit of the confidence interval that allows the bias or allows the
bias to be less than the expected bias or the upper limit of the confidence interval that
allows the bias to be greater than the expected bias.). By comparing the confidence interval
of the expected bias of the medical decision level with the limit of the allowable error, it is
less than the allowable error, and the equivalent of the evaluation reagent and the reference
reagent can be judged. If it is greater than the allowable error, further clinical verification is
required.
B̂c Bˆ c a (b 1) X c
Expected bias estimate ( ):
B̂c ^
1 ( X X )2
95% confidence interval = B C
t / 2, N 2 S y x N C
( X i X )2
N
i 1
B̂c
bias estimate ( )are calculated respectively, and compared with the tolerance error of half
Medical decision Regression The estimated value of the expected 10% allowable Expected bias
level(Xc)nmol/L equation y=bx+a bias of the medical decision-making error 95%
level; a+(b-1)Xc confidence
interval
Lower Upper Lower Upper
limit limit limit limit
X1=0.1mIU/L y=1.051x-0.111 -0.106 -0.0125 0.0125 -0.595 0.383
X2=10mIU/L 0.399 -1.250 1.250 -0.089 0.888
Table 9 Sample information with inconsistent qualitative results for both methods
Roche Wondfo
Sample
Hospital Diagnose symptoms concentration concentration
No.
(mIU/L) (mIU/L)
Eighth People's Hospital of Primary
6 3.93 4.38
Guangzhou hyperthyroidism
Eighth People's Hospital of
92 Hepatitis B virus 0.092 0.77
Guangzhou
Eighth People's Hospital of
101 Diabetes NOS 4.38 2.96
Guangzhou
Eighth People's Hospital of
102 Medical examination 4.72 4.02
Guangzhou
Eighth People's Hospital of Normal pregnancy
106 4.37 3.85
Guangzhou supervision
Eighth People's Hospital of
110 Thyroid nodules 5.07 3.5
Guangzhou
The Second Affiliated
Hospital of Guangzhou 7 stomach ache 3.90 5.04
Medical University
The Second Affiliated
46 Lung malignancy 0.394 0.28
Hospital of Guangzhou
Medical University
The Second Affiliated
Hospital of Guangzhou 54 Early pregnancy 0.289 0.26
Medical University
The Second Affiliated
Hospital of Guangzhou 99 4.69 3.93
Medical University
The 220 effective serum samples are tested with the evaluation reagent and the
reference reagent. There are 10 samples inconsistent with a ratio of 4.55%. The inconsistent
sample values of the two methods are all close to the reference range and are acceptable.
4.2 Test results and analysis of homologous serum and plasma samples
4.2.1A total of 116 serum samples are selected from the two centers, of which 116 are valid
serum samples (100%). See Table 1 for detailed selection results.
Table10 Selected sample basic situation
B̂c
estimate ( ) and its 95% confidence interval (The confidence interval of the expected
bias includes the lower limit of the confidence interval that allows the bias or allows the
bias to be less than the expected bias or the upper limit of the confidence interval that
allows the bias to be greater than the expected bias.). By comparing the confidence interval
of the expected bias of the medical decision level with the limit of the allowable error, it is
less than the allowable error, and the equivalent of the evaluation reagent and the reference
reagent can be judged. If it is greater than the allowable error, further clinical verification is
required.
B̂c Bˆ c a (b 1) X c
Expected bias estimate ( ):
B̂c ^
1 ( X X )2
95% confidence interval = B C
t / 2, N 2 S y x N C
( X i X )2
N
i 1
B̂c
estimate ( )are calculated respectively, and compared with the tolerance error of half of
Medical decision Regression The estimated value of the expected 10% allowable Expected bias
level(Xc)nmol/L equation y=bx+a bias of the medical decision-making error 95%
level; a+(b-1)Xc confidence
interval
Lower Upper Lower Upper
limit limit limit limit
X1=0.1mIU/L y=1.123x-0.233 -0.221 -0.0125 0.0125 -0.425 -0.016
X2=10mIU/L 0.997 -1.250 1.250 0.792 1.202
Table 18 Coefficienta
B̂c
estimate ( ) and its 95% confidence interval (The confidence interval of the expected
bias includes the lower limit of the confidence interval that allows the bias or allows the
bias to be less than the expected bias or the upper limit of the confidence interval that
allows the bias to be greater than the expected bias.). By comparing the confidence interval
of the expected bias of the medical decision level with the limit of the allowable error, it is
less than the allowable error, and the equivalent of the evaluation reagent and the reference
reagent can be judged. If it is greater than the allowable error, further clinical verification is
required.
B̂c Bˆ c a (b 1) X c
Expected bias estimate ( ):
B̂c ^
1 ( X X )2
95% confidence interval = B C
t / 2, N 2 S y x N C
( X i X )2
N
i 1
B̂c
estimate ( )are calculated respectively, and compared with the tolerance error of half of
Medical decision Regression The estimated value of the expected 10% allowable Expected bias
level(Xc)nmol/L equation y=bx+a bias of the medical decision-making error 95%
level; a+(b-1)Xc confidence
interval
Lower Upper Lower Upper
limit limit limit limit
X1=0.1mIU/L y=1.120x-0.320 -0.308 -0.0125 0.0125 -0.565 -0.051
X2=10mIU/L 0.880 -1.250 1.250 0.623 1.137
4.3.6Absolute bias analysis
As shown in the following figure, the Bland-Altman analysis of the mean value of the
difference between the evaluation reagent and the reference reagent yields an average of
1.354 between the two methods (absolute bias) with a standard deviation of 19.397, 95%
The boundary value is [-2.730~2.578], the out-of-bounds point is 3 and the out-of-bounds
point ratio is 2.70%.
5. Conclusion
The qualitative analysis of the results of serum samples obtained from 220serum
sample evaluation reagents and reference reagents is conducted and a total of 210 samples
with the same results as the evaluation reagent and the reference reagent. Of which 86 cases
are detected with abnormalities, 124 cases are detected normal and 10 cases with
inconsistent results. The two methods total compliance rate: 95.45%, 95%CI is [91.80%~
97.80%], positive compliance rate: 93.48%, 95%CI is [86.34% ~ 97.57%], negative
compliance rate: 96.88%,95%CI is [ 92.19%~97.80% ], kappa consistency coefficient k =
0.906, p = 0.000 <0.05. After 11 samples outside the linear range are excluded, the results
of 209 serum samples are statistically analyzed. The Spearman's correlation coefficient
between the reagent and the reference reagent is 0.982 and p=0. Linear regression is used to
estimate the results. The results showed a good linear relationship between the two methods.
The regression coefficient obtained is 1.051, the 95% confidence interval of the regression
coefficient is [1.023~1.079], and the intercept obtained is -0.111. The 95% confidence
interval for the intercept is [-0.629 ~ 0.408], and the linear regression equation is
y=1.051x-0.111; When the medical decision level is 0.1mIU/L, estimated expected bias is
-0.106, the expected bias is 95% confidence interval [-0.595~0.383], and when the
medical decision level is 10mIU/L, estimated expected bias is 0.399, and the expected bias
is 95% confidence interval [-0.089~0.888]. The bias at the medical decision level is
outside the 12.5% tolerance. For the Bland-Altman analysis, the mean difference is -0.292,
the standard deviation difference is 3.534, the 95% cutoff is[-7.218~6.634], out-of-bounds
point is 10 and the ratio of out-of-bounds points is 4.78%.
A total of 116 samples of homologous serum and plasma samples are tested by
evaluation reagents. After 8 samples outside the linear range are excluded, the results of
108 homologous samples are statistically analyzed. The Spearman correlation coefficient of
the statistical results is 0.994, p = 0; Linear regression is used to estimate the results and its
results showed a good linear relationship between the two types of samples. The regression
coefficient obtained is 1.123, the 95% confidence interval of the regression coefficient is
[1.099~1.148], and the intercepts obtained at the same time for -0.233, the 95% confidence
interval for the intercept is [--0.936~-0.070] and the linear regression equation is
y=1.123x-0.233. When the medical decision level is0.1mIU/L, the estimated expected bias
is -0.221, the expected bias is 95% confidence interval[-0.425~-0.016], and when the
medical decision level is 10mIU/L, estimated expected bias is 0.997, the expected bias.
95% confidence interval [0.792~1.202]. The bias at the medical decision level is within
12.5% of the allowable error. The results are analyzed by Bland-Altman, with a mean
difference of -0.138, a standard deviation of 1.064, a 95% cutoff value of [-2.223~1.946],
out-of-bounds point is 7 and the ratio of out-of-bounds points is 6.48%.
A total of 120samples of homologous serum and whole blood samples are tested by
evaluation reagents. After 9 samples outside the linear range are excluded, the results of 111
homologous samples are statistically analyzed. The Spearman correlation coefficient of the
statistical results is 0.989, p = 0; Linear regression is used to estimate the results and its
results showed a good linear relationship between the two types of samples. The regression
coefficient obtained is 1.120, the 95% confidence interval of the regression coefficient is
[1.088~1.152], and the intercepts obtained at the same time for -0.320, the 95% confidence
interval for the intercept is [-0.554~-0.086] and the linear regression equation is
y=1.120x-0.320. When the medical decision level is 0.1mIU/L, the estimated expected bias
is -0.308, the expected bias is 95% confidence interval[-0.565~-0.051], and when the
medical decision level is 10mIU/L, the estimated expected bias is 0.880, the expected bias.
95% confidence interval [0.623~1.137]. The bias at the medical decision level is within
12.5% of the allowable error. The results are analyzed by Bland-Altman, with a mean
difference of -0.076, a standard deviation of 1.354, a 95% cutoff value of [-2.730~2.578],
out-of-bounds point is 3 and the ratio of out-of-bounds points is 2.70%.
Conclusion:
Guangzhou Wondfo Biotech Co., Ltd.'s TSH Rapid Quantitative Test (fluorescent
immunochromatography) does not have a statistically significant difference from the test
results of the same product that has been approved for marketing, and has a high degree of
consistency, that is clinical application performance is equivalent. In addition, the
evaluation reagent also has equivalence in detecting homologous serum and plasma, whole
blood samples.