Calculating the probability of change between two consecutive laboratory test results
Re: Explaining laboratory test results to patients: what the clinician needs to know. Maurice John O’Kane and Berenice Lopez. 351:doi 10.1136/bmj.h5552
With great interest we read the article by O’Kane and Lopez: “Explaining laboratory test results to patients: what the clinician needs to know”. The authors provide a comprehensive overview and discuss the possibilities and limitations of laboratory tests for diagnosis, monitoring, risk stratification and screening. Of particular interest, they discuss the variability of test results over time due to (pre)analytical and biological variation. Awareness of day to day variability within an individual, is particularly important when monitoring disease progression or response to treatment. Importantly, reference values have limited value in the setting of monitoring as they are based on variation between individuals, whereas the interpretation of serial measurements requires information on the day to day variation within an individual. O’Kane and Lopez recommend laboratory test results to be considered as 'ranges' instead of single numbers and to determine whether the change between subsequent laboratory test results is real by looking at the degree of overlap of the ranges. [1] This is what experienced clinicians intuitively do when assessing whether serial laboratory results reflect a true change, or rather reflect expected variability. There is however a more formal way to objectively distinguish between ‘true changes’ and random variation between serial test results. For a change between consecutive measurements to become significant, the difference must be larger than the change that can reasonably be expected due to normal biological and analytical variation. This threshold value is termed the Reference Change Value (RCV). The RCV can be calculated for each laboratory test and depends on the biological within-person variability (CVI) and the analytical variability (CVA). [2]
RCV= √2 * Zscore* √(CVa^2 + CVi^2 )
In this formula, the Z-score represents the number of standard deviations and correspond to the desired probability. Commonly used Z-scores are 1.96 and 2.56.These Z-scores calculate the percentage increase or decrease that is required to become statistically significant, with a false positive rate of 5%, (p <0.05) and 1% (p <0.01) respectively.
For easy calculation of the RCV we developed Labtracker+ [3], a free, CE certified, medical smartphone app for iOS. The RCV principle is used to calculate the probability of a true change between serial laboratory results. Over 100 laboratory parameters are currently available in Labtracker+. Decision support by Labtracker, using the RCV principle, may be a useful addition to clinical intuition.
References:
[1] O’Kane and Lopez, Explaining laboratory test results to patients: what the clinician needs to know. BMJ 2015; 351: h5552
[2] Fraser, C.G. and Harris, C.G., Generation and application of data on biological variation in clinical chemistry. Crit Rev Clin Lab Sci, 1989. 27(5): p. 409-37
[3] https://appsto.re/nl/T3lr8.i
Competing interests:
No competing interests
22 December 2015
Judith M. Hilderink
PhD student
Richard. P Koopmans, Roger J.M.W. Rennenberg, Marja P. van Dieijen-Visser, Steven J.R. Meex
Central Diagnostic Laboratory, Maastricht University Medical Center, Maastricht, the Netherlands
P. Debeyelaan 25, 6229 HX Maastricht, the Netherlands
Rapid Response:
Calculating the probability of change between two consecutive laboratory test results
Re: Explaining laboratory test results to patients: what the clinician needs to know. Maurice John O’Kane and Berenice Lopez. 351:doi 10.1136/bmj.h5552
With great interest we read the article by O’Kane and Lopez: “Explaining laboratory test results to patients: what the clinician needs to know”. The authors provide a comprehensive overview and discuss the possibilities and limitations of laboratory tests for diagnosis, monitoring, risk stratification and screening. Of particular interest, they discuss the variability of test results over time due to (pre)analytical and biological variation. Awareness of day to day variability within an individual, is particularly important when monitoring disease progression or response to treatment. Importantly, reference values have limited value in the setting of monitoring as they are based on variation between individuals, whereas the interpretation of serial measurements requires information on the day to day variation within an individual. O’Kane and Lopez recommend laboratory test results to be considered as 'ranges' instead of single numbers and to determine whether the change between subsequent laboratory test results is real by looking at the degree of overlap of the ranges. [1] This is what experienced clinicians intuitively do when assessing whether serial laboratory results reflect a true change, or rather reflect expected variability. There is however a more formal way to objectively distinguish between ‘true changes’ and random variation between serial test results. For a change between consecutive measurements to become significant, the difference must be larger than the change that can reasonably be expected due to normal biological and analytical variation. This threshold value is termed the Reference Change Value (RCV). The RCV can be calculated for each laboratory test and depends on the biological within-person variability (CVI) and the analytical variability (CVA). [2]
RCV= √2 * Zscore* √(CVa^2 + CVi^2 )
In this formula, the Z-score represents the number of standard deviations and correspond to the desired probability. Commonly used Z-scores are 1.96 and 2.56.These Z-scores calculate the percentage increase or decrease that is required to become statistically significant, with a false positive rate of 5%, (p <0.05) and 1% (p <0.01) respectively.
For easy calculation of the RCV we developed Labtracker+ [3], a free, CE certified, medical smartphone app for iOS. The RCV principle is used to calculate the probability of a true change between serial laboratory results. Over 100 laboratory parameters are currently available in Labtracker+. Decision support by Labtracker, using the RCV principle, may be a useful addition to clinical intuition.
References:
[1] O’Kane and Lopez, Explaining laboratory test results to patients: what the clinician needs to know. BMJ 2015; 351: h5552
[2] Fraser, C.G. and Harris, C.G., Generation and application of data on biological variation in clinical chemistry. Crit Rev Clin Lab Sci, 1989. 27(5): p. 409-37
[3] https://appsto.re/nl/T3lr8.i
Competing interests: No competing interests