,

ALGORITHM OF RESULT VALIDATION

https://drive.google.com/open?id=0B9hN7lPwYiEhSnZxM3NVdi1JUTg

EXAMPLE LINEARITY STUDY

https://drive.google.com/open?id=0B9hN7lPwYiEhdElhZnhZeFoxNlE

EXAMPLE METHOD COMPARISON

https://drive.google.com/open?id=0B9hN7lPwYiEhVFVYOFZWVTV0VjA

EXAMPLE PRELIMINARY STUDY

https://drive.google.com/open?id=0B9hN7lPwYiEhZWoyVlV4eHo2dnM

TEMPLATE PRELIMINARY STUSY

https://drive.google.com/open?id=0B9hN7lPwYiEhVVFERmU5SlpzeDQ

Preliminary Study: Total Precision

https://drive.google.com/open?id=0B9hN7lPwYiEhMTZxaWF5cnl4SlU

METHOD VERIFICATION

PRELIMINARY STUDY: WITHIN RUN PRECISION

Method verification is a one- time process to determine performance characteristics before a

test system is utilized for patient testing. The experimental confirmation of performance

usually include preliminary study (precision check), linearity, comparison study and 

reference range. 

Preliminary Study

The preliminary study is applied only to newly installed instruments or new measurement 

procedure in the laboratory. This is very important to verify the precision performance of

method in the insert kit is fulfilled before the new method is utilize for patient sample testing.

The basic statistics calculation as below will used to verified the precision of the new

method:

· Mean  = ( Σ x)/ n , n= number of observation

· Variance, σ2 =  Σ (Xi – Mean )2 / (n-1)

· Standard Deviation, SD=  √ σ2 

· Coefficient Variation ( CV ) = (SD/mean) x 100.

The main objective of preliminary study is to verify the precision performances in the insert

kit are fulfilled before we use the new method in our laboratory. These include  CV within 

run precision and CV intermediate or total precision.

We will use control material at 2 or 3 levels as recommended by the manufacturer to asses precision within run and total precision for verification of manufacturer’s claim. We will

accept the manufacturer precision within run and total precision if precision experiment 

within run and total precision is less than manufacturer claim.

The experiment procedures for the preliminary study is to run triplicate per day within two 

hours for five days using control material as recommended by the manufacturer. We

calculate the daily mean and variance (σ2) analytical. From daily mean and variance

analytical, we calculate overall mean, standard deviation (SD) and coefficient variation (CV) analytical for within runs, between days and total CV. Next, we compare CV analytical withinrun and total CV with manufacturer claims. If CV analytical within run and total CV analytical is less than CV within run and total precision manufacturer claim, we will accept the

precision manufacturer claims. If more, we will continue to calculate SD verification (SDV)

from CV manufacturer, chai- distribution and degree of freedom. We only accept if SD

analytical less than SDV.

Preliminary Study: Within Run 

The procedure of the preliminary study is run all the quality control (QC ) material as 

recommended by manufacturer triplicate within two hour for five day and record all the data. 

Daily mean and variance (σ2) were calculate.

Example : Albumin, the total number of level QC are two. 

The calculation for this preliminary study, only for level 2 QC. The mean level 2 is 38 g/l, CV within run is 0.5% and Total CV is 1.1%.In real situation, we must perform precision study

for level 1 and level 2. The data triplicate for five days with mean and variance as below:

Day	Run 1	Run 2	Run 3	Mean	Variance
1	37.7	37.5	37.8	37.73	0.0633
2	37.3	37.5	37.7	37.50	0.0400
3	38.4	37.6	38.1	38.03	0.1633
4	38	37.3	38.2	37.83	0.2233
5	38.1	37.8	38	37.97	0.0233

Note: formula for mean and variance

Mean  = ( Σ x)/ n , n= number of observation or number of run

 Variance, σ2 =  Σ (Xi – Mean )2 / (n-1)

Overall mean and variance

  Overall mean = (37.73 + 37.50+38.03+37.83+37.97)/5 

   = 37.81 g/l

  Overall variance = (0.0633+0.0400+0.1633+0.2233+0.0233)/5

  = 0.1027

Standard deviation (SD) and coefficient variation CV)

  SD  = √variance 

        = √0.1027

        = 0.3204       SD analytical

  CV  =(SD/overall mean) x 100

     =( 0.3204/37.81) x 100

     = 0.8474%      CV analytical

CV manufacturer is 0.5% and CV analytical is 0.8474%, That means analytical CV is more 

than manufacturer CV . Next step is we calculate the SD verification (SDV)

Standard deviation verification (SDV)

SDV = {(CV manufacturer/100 x overall mean) x √C} / √V 

C=  Percentage point ( Ӽ2 (0.05/L, V), L= number of QC level (L=2)

V=  Degree of freedom ( L*n), L= number of QC level, n = number of day (n=5)

SDV = {(1.1/100 x 37.8) x √ 20.48} /√ 10

     = 0.12706

SD analytical is 0.3204, therefore the SD analytical > SDV  means the precision for within

run manufacturer is not acceptable. The manufacturer must perform troubleshooting and

preliminary study for within run and re-evaluate.

Note: The value C is obtain from chi-square table or from window excel 

( Ӽ2 (0.025,10).

https://drive.google.com/open?id=0B9hN7lPwYiEhbTRva1ZpaGdCSEU

LINEARITY STUDY

https://drive.google.com/open?id=1Fwi-mQmgZbAw_sPHRvz1eziv5q8HkFDPrFRQVuVVuNQ

PRELIMINARY STUDY: WITHIN RUN PRECISION

https://drive.google.com/open?id=0B9hN7lPwYiEhUWRXTk1sQ3ZDems

APPLICATION OF MEASUREMENT UNCERTAINTY IN CLINICAL LABORATORY

Sairi Satari

Chemical Pathology Unit

Penang Hospital

The uncertainty of measurement is a fundamental concept indicating the range within the “true value of a measurement should lie”. Although not commonly reported with the result, the calculation of measurement uncertainty (MU) has become common in the clinical laboratory routine. MU estimates are usually used in analytical troubleshooting and interpretation of pathology results. In analytical troubleshooting, the expanded uncertainty will be obtained by multiplying MU with 1.96, In the interpretation of numerical pathology results, the patients’ results are compared with the reference interval or medical decision limit and the expanded uncertainty will be obtained by multiplying MU with 1.65. If the interpretation is done by comparing with previous result from the same patient, the expanded uncertainty will be obtained by multiplying MU with 2.77. In patients’ management, MU is significant during validation of the patients’ results for accuracy and reliability. 

https://drive.google.com/open?id=0B9hN7lPwYiEhRWZCejlUa1cwQjg

https://drive.google.com/open?id=0B9hN7lPwYiEhYzZBbllpRGM0bWc

INTERNAL QUALITY CONTROL (IQC) DAN PENYELESAIAN MASALAH

SAIRI SATARI

Pegawai Sains Kimiahayat

BAHAGIAN 2

TARGET MEAN, SD DAN CV

Target mean, SD dan CV yang diabadikan dalam graf Levey Jennings akan digunapakai selagi lot QC yang sama digunakan sebagai material untuk memantau kualiti ujian makmal. Oleh itu, pengiraan  target mean, SD dan CV perlu memenuhi kriteria berikut;

·         Minima data 20 dari lot QC yang hendak digunapakai dan data membentuk taburan normal,

·         Analisa sampel QC perlu melibatkan beberapa lot kalibrator dan pertukaran lot reagent,

·         Penglibatan beberapa operator dalam penyediaan larutan kalibrator, QC dan analisisa,

·         Penggunaan glass pipette yang sesuai semasa menyediakan larutan kalibrator dan QC.

Kriteria di atas perlu dipenuhi kerana target mean, SD dan CV membawa maksud tersurat iaitu ralat sistematik dan ralat random. Ralat-ralat ini dijelmakan sebagai target mean dan target SD dan diabadikan dalam graf Levey Jennings sebagai mean ±2SD. 2SD. Ia adalah nilai 95% confident limit. Ralat yang wujud ini perlu konsisten selagi kita menggunakan lot QC yang sama. Pertambahan ralat atau kewujudan ralat baru menyebabkan perubahan target mean dan target SD. Oleh itu kita perlu mengira mean, SD dan CV setiap bulan. Nilai mean setiap bulan perlu sama dengan target mean secara statistik dan CV setiap bulan perlu kurang dari target CV. Perubahan lot QC memerlukan kita mengira semula target mean dan SD.

Taburan normal dapat dikenalpasti apabila set data memberi nilai mean, median dan mode yang sama (mean=median =mode)

Taburan normal

Mean +/- 2 SD

Setiap kali kita menganalisa sampel QC, keputusan QC perlu berada dalam julat mean + 2SD atau mean -2SD dalam keadaan seimbang. Oleh kerana kita sukar untuk menentukan keputusan QC berada dalam julat mean + 2SD atau mean - 2SD yang seimbang, maka graf Levey Jennings digunapakai dalam memantau QC harian. Westgard rule pula digunakan untuk memantauan ralat sistematik dan ralat random semasa proses analitikal.

Graf Levey Jenning

Multi Westgard rule

Penyediaan larutan QC dan kalibrator perlu menggunakan glass pipette yang sesuai. Jika kita ingin menyediakan larutan 5 ml, maka kita perlu menggunakan pipette 5ml dan bukannya pipette 10 ml. begitu juga untuk menyediakan larutan 1ml, kita perlu menggunakan glass pipette 1 ml. Setiap pipette mempunyai uncertainty. Pipette yang mempunyai isipadu yang besar mempunyai uncertainty yang besar, begitu juga sebaliknya.

 BERSAMBUNG

 INTERNAL QUALITY CONTROL (IQC) AND PROBLEM SOLVING

SAIRI  SATARI

Biochemist
Part 1

RECTIFICATION IS ESSENTIAL, ALBEIT ROUTINE PROCEDURES

Internal Quality Control (IQC) is a process or system to monitor the quality of laboratory testing in terms of accuracy and precision. It includes pre-analytical and analytical phase. IQC should be able to detect the changing of systematic error and random error during the analytical process. Therefore, when we determine the target mean and SD or CV, the systematic error and random error should be taken into account. Drastic changes of the systematic error and random error will be affect the mean and SD or CV.

Target Mean, SD and CV will be change

Levey Jenning’s chart and Westgard rule are the common procedures used in the laboratories to determine the systematic error and random error.  The IQC must be compliance with the Wesgard rule before the patient results are released.

Usually, when we determine the target mean, SD or CV, we will collect 20 data point from selected level of quality control (QC) within 20 working days.  We calculate the mean, SD, and CV from the data. If the calculated mean is within the manufacturer’s mean range, then the mean and SD will be used as a target mean, and target SD. The target mean and SD will be stated in the Levey Jennings chart as mean ±2 SD (2 SD is 95% confident limit based on the normal distribution). The acceptability of our daily QC will be determine by Westgard rule.

The Levey Jenning’s chart and the Wesgard rule are based on the normal distribution. Thus, the setting of the target mean and SD or CV based on 20 data points only are not sufficient. The setting of target mean and SD or CV should take into account that the data collected should be in a normal distribution. Therefore, we must determine whether 20 data points will form a normal distribution or not. If not, we need to add more data until it forms a normal distribution.

The target mean and SD form manufacturer recommendation are wrong in term of setting procedure of the target mean and SD in the laboratory. We must calculate our target mean and SD based on our condition in the laboratory.

Levey Jenning Chart

When we perform repeatable analysis to the same QC lots, the data collected will generally be in a normal distribution but, we still have to prove that the data is in the normal distribution by ensuring that the mean, median and mode generate the same value. The target mean will reflect the systematic error and SD or CV will reflects the random error. Both of these errors will have to be consistent as long as we are using the same lot of QC in our analysis. Thus, the setting of the target mean and SD QC need to be based on the normal distribution at 95% confident limit or Mean ± 2SD. If the mean setting and SD is not based on the normal distribution, then we will have difficulty in solving QC non-compliance during analysis.

The Levey Jenning’s Chart and the Westgard Rule are based on the Normal Distribution and it can be help us to solve the problem for non-conformant daily QC.

TO BE CONTINUED

INTERNAL QUALITY CONTROL (IQC) DAN PENYELESAIAN MASALAH

SAIRI SATARI

BAHAGIAN 1

“YANG BIASA TIDAK SEMESTINYA BETUL. PERBETULKAN YANG BIASA”

Internal Quality Control (IQC) adalah suatu proses atau sistem untuk memantau kualiti ujian makmal dari aspek ketepatan (accuracy) dan kejituan (precision). Ia merangkumi fasa pre-analitikal dan analitikal. !QC sewajarnya dapat mengesan perubahan ralat sistematik dan ralat random semasa proses analitikal. Sehubungan itu, apabila kita menentukan target mean dan SD atau CV, penyumbang ralat sistematik dan random perlu diambil kira. Perubahan yang drastik terhadap ralat sistematik dan ralat random akan mengubah mean dan SD atau CV.

Jitu dan tepat

Jitu tetapi tidak tepat

Graf Levey Jennings dan Westgard rule adalah prosedur biasa yang digunapakai di makmal dalam menentukan kepatuhan IQC sebelun keputusan makmal dikeluarkan. Adalah menjadi kebiasaan, apabila kita menentukan target mean, SD atau CV, kita hanya mengumpulkan 20 data dalam jangka masa 20 hari bekerja dari lot IQC yang hendak digunakan. Dari 20 data ini kita kira mean, SD dan CV. Apabila mean yang dikira terkandung dalam julat manufacturer, maka mean dan SD tersebut digunapakai sebagai target mean dan target SD dan diabadikan dalam graf Levey Jennings. Wesgard rule digunakan bagi menentukan IQC harian kita sama ada dapat diterima atau sebaliknya.    

Graf Levey Jennings dan Wesgard rule adalah berasaskan taburan normal. Oleh itu, Penetapan target mean dan SD atau CV hanya berdasarkan 20 data semata-mata adalah tidak mencukupi.Penetapan target mean dan SD atau CV perlu mengambil kira bahawa, data yang dikumpulkan perlu berada dalam taburan normal. Oleh itu, kita perlu menentukan samada 20 data yang dikumpulkan itu membentuk taburan normal atau pun tidak. Jika tidak, kita perlu menambah data sehingga membentuk taburan normal.

Amalan menggunakan target mean dan SD dari manufacturer adalah salah dari segi prosedur penetapan target mean dan SD IQC di makmal.m Justeru, kita perlu mengira target mean dan SD makmal kita.

Graf Levey Jenning

Apabila kita melakukan analisa berulang kali kepada IQC yang sama lot, data yang dikumpulkan secara umumnya akan membentuk taburan normal. Walaupun begitu, kita perlu membuktikan bahawa data tersebut membentuk taburan normal dengan memastikan bahawa mean ,median dan mod memberi nilai yang sama. Target mean adalah mengambarkan ralat sistematik dan SD atau CV mengambarkan ralat random. Kedua-dua ralat ini perlu konsisten sepanjang masa selagi kita menggunakan lot yang sama dalam analisis kita. Justeru, penetapan standard mean dan SD IQC perlu berdasarkan taburan normal data pada 95% confident limit atau Mean ± 2SD. Jika penetapan mean dan SD tidak berasaskan taburan normal, maka kita akan mengalami kesukaran dalam menyelesaikan masalan ketidakpatuhan IQC semasa menjalankan analisa.

95% confident limit (tburan normal)

Berdasarkan taburan normal ini maka Graf Levey Jennings dan Westgard rule dapat membantu dalam menyelesaikan masalah ketidakpatuhan IQC harian. 

BERSAMBUNG…

LONG TERM IMPRECISION: HOW TO CALCULATE

Introduction

            The monitoring of long term laboratory imprecision is very important for patient management, especially in the follow up for the response to treatment based on the laboratory data. Currently, the long term imprecision is based on the combination of the intermediate CV as CV= √ (CV₁²+CV₂²+ CV₃² +….). This formula provides bigger CV  compared to the generated CV by the analyzer for the same duration.

            The aim of this study is to determine alternative method in calculation for the long term imprecision compared to the generated CV in the same duration.

Method

            The Albumin quality control data, Level 1 and Level 2 on the Architect c8000 analyzer for six months ( July, August, September, October, November and December  2012) were used in this study. The analyzer generated the mean, SD and CV for the last six months duration and on monthly basis (July to December respectively). The monthly mean, SD and CV were entered in the excel work sheet for calculation of the long term imprecision in two methods:

i.                    Based on the combination of monthly CV and

ii.                  Based on the combination of the monthly variance.

The value of the total CV obtained in the above method was compared to the generated CV for the last six month (July to December).    

Results

            The results from the experiment are as the  following :

1.                 Automated generated  Total CV by the analyzer for the duration of July to December

 

DURATION	LEVEL 1				LEVEL 2
	N	MEAN (g/l)	SD	% CV	N	MEAN ( g/l)	SD	% CV
July to Dec	231	40.54	0.827	2.04	234	24.7	0.61	2.46

 2.         Monthly data (July to December)

MONTH	N	LEVEL 1			LEVEL 2
		MEAN (g/l)	SD	CV	MEAN	SD	CV
JULY	2	41	0	0.00%	24.5	0.707	2.89%
AUGUST	57	40.54	0.569	1.40%	24.745	0.440	1.78%
SEPT	57	39.72	0.526	1.32%	24.18	0.563	2.33%
OCT	54	40.333	0.476	1.18%	24.91	0.446	1.79%
NOV	52	41.481	0.504	1.22%	25.245	0.434	1.72%
DEC	9	41.556	0.527	1.27%	25.111	0.333	1.33%

 3.         Imprecision calculation

       

Imprecision calculation (six month)	Level  1 QC	Level 2 QC
Based on the monthly CV	2.86%	4.99%
Based on the monthly Variance	2.04%	2.46%
Automated generated by analyzer (six month)	2.04%	2.46%

      

The results showed that CV generated by automated analyser based on the QC data July to December were 2.04% for level 1 and 2.46% for level 2. The CV calculated based on the combination of monthly CV from July to December on the same QC data were 2.86% for level 1 and 4.99% for level 2. On the other hand, when the CV is calculated based on the combination of monthly variance on the same QC data, it was found that CV for level 1 was 2.04% and level 2 was 2.46%.

Conclusion

            Based on this study, it was found that the long term CV calculated based on the combination monthly variance gave  comparable result to that of  automatically generated CV  by the analyser where as combination of the monthly CV gave much bigger CV. Therefore, it can be concluded that the CV calculated by monthly variance is more reliable compared to the combination of the monthly CV.