Transcript Common Reference Intervals (and everything you wanted to

Reference Intervals for New
Methods
Dr Graham Jones
Department of Chemical Pathology
St Vincent’s Hospital, Sydney
www.sydpath.stvincents.com.au
Reference Intervals for new Methods
Subtitle: “Reference
Intervals”
Contents
• Introduction to reference intervals
• Reference intervals for the new method
– Derive de-novo
– Transfer from old method
– Literature
– Other Laboratories
• Conclusions
Defining Reference Intervals
• Central 95% of results from a reference
population
– IFCC/NCCLS definition
• Excludes 2.5% above and below interval
• For healthy population are “Health-associated
Reference intervals”
• Can be any population, but must be defined
– eg, pregnant, premature, hospitalised, treated.
Other forms:
• Other statistical cuttoffs
– Troponin: 99th centile of healthy population
– Apo (a): 80th centile of total population
• Recommended interval (decision point)
– Impaired fasting glucose (6.1 - 6.9 mmol/L)
– Target LDL concentration (<2.0 mmol/L)
• Therapeutic Interval
– Drugs, INR, APTT, TSH
Current Paradigm
• Based on recommendations from the NCCLS and
the IFCC
• Repeated in Product Information from most reagent
suppliers
• Encoded in the NATA summary of ISO/IEC guide
17025.
– laboratories may perform their own detailed
reference interval studies
or
– may validate reference intervals published
elsewhere for their own methods and populations
Generating a new reference interval
•
•
•
•
•
•
Define and select reference population*
Define collection conditions and numbers
Collect samples
Analyse samples
Perform statistical evaluation*
Put into practice
• Tietz Textbook covers standard approach very
well (HE Solberg)
Define Reference Population
• Source
– eg blood bank, lab volunteers, students
• Numbers
• Exclusions
• Likely Partitioning
– Age
– Sex
– Other
• Difficult to get extremes of age and high
numbers
Study Imprecision
• Estimates of reference limits has limitations
• Expressed as the confidence interval of the
Reference Limits, eg 90% CI of the upper and
lower reference limits
• Confidence intervals decrease as the number of
people in the study increases.
Large n
Small n
Non-parametric statistics
• Lowest number where error envelope can be
calculated is 120
• For n=120
– 2.5th centile is 4th lowest result
– 90% confidence limit for LRL is lowest sample
and 7th lowest sample
• These values often very scattered giving wide
intervals
NORIP STUDY
ALT (U/L)
Female ALT (n=1220)
1
2
Percent:
1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0
+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
1 |•
.
.
.
.
11 2 |•••••••••
.
.
.
.
3 |••••••••••••••••••••••••••••••••••••
.
.
19 4 |••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••>
5 |••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••>
27 6 |•••••••••••••••••••••••••••••••••••••••••••••••••••••
.
7 |•••••••••••••••••••••••••••••.
.
.
33 8 |••••••••••••••.
.
.
.
9 |••••••••
.
.
.
.
.
.
.
.
41 10 |•••
11 |•••
.
.
.
.
.
.
.
.
49 12 |••
13 |•
.
.
.
.
.
.
.
.
57 14 |•
15 |•
.
.
.
.
16 |•
.
.
.
.
17 |
.
.
.
.
18 |
.
.
.
.
19 |
.
.
.
.
+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
Female Upper Reference Limit: 45.6 (90% CI 42.5 – 49.3, n=1220)
Male Upper Reference Limit: 68 (90% CI 63.4 – 73.6, n=1080)
Generating Intervals
• Is hard to do well
• Requires time and effort and money
• But any local data may be very useful for
validation of other intervals
“Impossible” Intervals
• Some reference intervals are essentially
impossible to produce from local studies:
–
–
–
–
Paediatric intervals
Stages of pregnancy (eg hCG in 5th week)
Stages of menstrual cycle
Nutritional parameters
• Reflects local diet
• May normalise deficiency state
Transfer Intervals from previous
method
• Implies previous intervals are good
– Check source and validity
• Transfer requires good correlation
• Advantage is clinical acceptance
– Note: much the following data related to
introduction of a Bayer Centaur for Vitamin
B12.
Transferring Intervals
Total Protein
140
y = 1.0007x - 0.5037
R2 = 0.9957
120
AU2700
100
80
60
40
20
0
0
20
40
60
80
100
120
140
Modular <P>
Wide range of results, assayed over several days, excellent correlation
And linearity. Transfer with no problem
Transferring Intervals – more difficult
Vitamin B12
y = 0.82x + 38
500
450
156 (bottom
of normal)
400
350
Centaur
BAYER
181 (top of
deficient)
300
250
200
150
100
50
0
0
100
200
300
400
500
Access
BECKMAN
107 (bottom of
normal)
126
SydPath
133 (top of
deficient)
95% Confidence Limits
Slope: 0.78 – 0.86
Intercept: 28 - 48
Correlation Data
• Patient samples
• Focus on results near limits
• Beware effect of extreme values on statistics
– Passing and Bablock preferred to linear
regression
• Use correlation data from several days and
calibrations
• Review source of previous Intervals
Validation of reference intervals
• NCCLS protocol
• Measure 20 samples appropriate for reference
interval on new method
• Exclude outliers
• If 2 or fewer are outside proposed inetrvals
– Accept intervals
• If >2 are outside proposed intervals
– Measure another 20
– If 2 or fewer are outside – accept intervals
• Cannot detect overly wide intervals
Review Previous Method
• Previous method may have significant amounts
of data (information)
• For many assays many of the results will be on
“normal” patients
• For all assays will allow assessment of
previous reference intervals
• Methods:
– Inspection
– Frequency histograms (all data, some data)
– Formal methods (Bhattacharya)
Access
6
66
126
186
246
306
366
426
486
546
606
666
726
786
846
906
966
1026
1086
1146
1206
350
300
250
200
150
100
50
0
126 pmol/L
5.8% rate of “low”
results
Access B12 (actual)
Centaur - predicted
500
400
180 pmol/L cuttoff
16% positive rate
300
200
100
Centaur B12 (predicted)
1020
960
840
900
780
720
600
660
540
420
480
360
300
180
240
120
60
0
Assess effect of possible
Assay change
Data Mining old results
• Bhattacharya, LG. Journal of the Biometric Society.
1967;23:115-135.
• Example data: Frequency Distribution of the forkal length of
the Porgy caught by pair-trawl fishery in the East China Sea.
Bhattacharya
• Assumes Gaussian (or Log Gaussian)
distributions
• Assumes a significant proportion of requests
are on unaffected individuals
4500
4000
3500
3000
2500
2000
1500
1000
500
0
patient values
Bhattacharya
0
0.05
Creatinine
0.1
0.15
0.2
Data Mining
• Bhattacharya ignores effects of outliers and
samples not part of majority distribution.
• Reference intervals based on majority.
4500
4000
3500
3000
2500
2000
1500
1000
500
0
patient values
Bhattacharya
0
0.05
0.1
0.15
Creatinine (mmol/L)
0.2
Literature
• Look for same method
• Equivalent population
• Sources
– Peer-reviewed publications
– Gray Literature
• Abstracts (eg AACB, AACC, ACB)
– Company literature
• Product information (PI)
• Other
Literature sources
• Vital where population reference intervals may
be of limited use
• Dietary factors
• Special groups
– Eg paediatrics
• Numbers are prohibitive
– eg 99th centile for troponins
• Following examples taken from SydPath data for
creatinine (Roche) and Vitamin B12 (Centaur)
Combining data
• Local and blood bank (M 101, F 110)
M: 62 – 105 umol/L F: 51 – 82 umol/L.
• Literature:
• South Australia (M 293, F 269) Mazzachi BC et al,
Clin Lab. 2000;46:53-55
M: 62 – 106 umol/L F: 44 – 80 umol/L
• Germany (M 127; F125) Junge et al. Clin Chem
Acta. 2004;344:137-148
M: 63 – 103 umol/L F: 48 – 85 umol/L
• Values rounded out as follows:
M: 60 – 110 mmol/L F: 40 – 90 mmol/L
Literature Sources - distribution
180 pmol/L
150
300
450
600
750
Vitamin B12 – ACS:180 Klee 2000 (pmol/L)
Homocysteine and Methylmalonic acid
relative to serum B12 (Centaur)
Homocysteine
MMA
1562 people, age >65. MMA and Homocysteine.
B12 measured on Bayer Centaur
Bin width 50 pmol/L. Red Arrow 200 pmol/L.
- Clarke et al, Am J Clin Nutrit. 2003;77:1241-7.
Dorevich Pathology
Sikaris et al
25,201 B12 measurements
ACS:180 and Bayer Centaur
Central 95% of results with
Normal Hb and MCV:
178 - 741 pmol/L
Vitamin B12 v MCV
MCV
126
160
140
120
100
80
60
40
20
0
0
500
1000
Vitamin B12 (Access, pmol/L)
• SydPath Data (3 months, 1497 results)
• Beckman-Coulter Access
1500
Product Information
VB12, pmol/L
588
441
294
181 pmol/L
156 pmol/L
147
140 pmol/L
Centaur Vitamin B12
• 6 studies using Centaur or ACS:180
•
•
•
•
Product information
3 x refereed publications
1 x AACB abstract
1 x local study (NZ)
• Data combined to make reference interval
– Deficient <120 pmol/L
– Indeterminate 120 – 180 pmol/L
– Replete >180 pmol/L
Product Information…
30
R ecom m ended Interval: 3 - 25
(2.5th to 97.5th Centile)
25
90% CI of URL: 19 - 39
# of Results
20
15
10
5
0
0
2
4
6
8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40
Manufacturer’s Interval: well-defined population, appropriate exclusions
But: Outliers?, bi/trimodal distribution?
Other laboratories
• If someone has done the work, and uses the
same method, review their work and apply the
intervals.
• Need to verify assay bias.
• Collaborative effort between several labs with
the same method may be a powerful method of
setting reference intervals
– Spanish Group
– NORIP:
http://wip.furst.no/norip/
Combining Laboratories
• 13 Spanish laboratories (all Centaurs)
• 11 – 15 samples from each laboratory (tot 150 samples)
• Combined data used for Reference Intervals
– Ferre-Masferrer et al. Clin Chem Lab Med 2001;39:166-169
Other Sources: accuracy base
• In order to share method-specific literature
need to ensure assay accuracy.
• “Is my Bloggs method for X working the same
as everyone else’s Blogg’s method?”
• QAP results
– measure QAP samples
– look up results for method group
• QC material target values
• Shared samples
Comparison with QAP targets
QAP Endocrine program - free T3
SydPath Results
14
y = 0.9382x + 0.4922
R2 = 0.9944
12
10
8
6
4
2
0
0
2
4
6
8
10
12
14
QAP Centaur Median
Method-specific medians (and scatter) available on QAP website
Plea to manufacturers...
• Searching refereed literature by trade names
can be difficult
– ie Abbott, Elecsys, Immulite, Vitros are terms
that are not often searchable in Medline,
pubmed etc
• If companies keep a resource library of
information it would be very useful.
• Note “google Scholar” can be useful
– http://scholar.google.com
Clinical Input
• Previous slides about Vitamin B12 are taken
from a presentation to haematologists at St
Vincent’s Hospital
• Actively seeking their input on decision points
• Allows inclusiveness and practical input
Putting it all together
• Different sources will give (slightly) different
values.
• Judgement is required to combine data
• Other factors include:
–
–
–
–
–
Precision of intervals
Long term precision of assay(s)
Biological variation
Rounding for ease of memory
Partitioning
Implementing
• Recommend temporary footnote
• eg change in method and change intervals, see
lab for further details
• Make further details available if needed
– source document (NATA)
– Handout
– Website
Conclusions
• A new method is a good time to review
reference intervals
• Uncritical transfer of old intervals is bad
practice
• Many sources of information can be used
• Judgement is required for final decision
• Working with other labs may be of great
benefit