The percentages and inter vs intra were way over my head. Can you explain it for me?
Sure. The article examines the reliability of various cancer markers by looking at how values of the markers are distributed in healthy individuals in two different ways. The intra-individual data from a 6 week period (4 samples per volunteer) were used to examine how values for a
single individual are distributed. In addition, each sample was analyzed multiple times to find the variability and consistency of the analyzer. The results were analyzed in terms of both the mean and the coefficient of variability CV (the standard deviation divided by mean times 100, to get a percent). It was found that the machine
itself had a CV = CVa = 5% for CEA measurement, which can be roughly interpreted as saying that individual measurements
from the same sample are quite likely (probability around 30%) to vary by +/- 5% or so. The most important thing the article analyzes is the reference change values RCV, which is the total CV for an individual scaled to provide a limit for change in the value that can be attributed to random variation. The RCV is often interpreted as being
the critical difference necessary between two values for them to be considered significantly different. The RCV for the CEA marker drawn from healthy individuals was computed to be 73%. This means that a
healthy individual typically will have levels that differ by as much as 73%. I suspect that there are sub-populations as well with a much larger RCV, meaning that the RCV of 73% is not nailed down either. This is clear from Figure 1 in the text, which shows the mean and range of CEA readings of all individuals. For example it is not unusual for a subject to have a reading of over 4 even when their mean level is near 2. I think the study would benefit from a much larger population size to clarify this point.
The inter-individual measurements refer to all the data considered together, including the entire sample population. The CV for the CEA marker for the entire population is larger than that of a typical individual by roughly 20%. The authors conclude:
CVi and CVa determine what constitutes a significant difference between sequential results. For follow-up purposes, individuals are their own refer-
ences, and the clinician is mainly interested in increases or decreases of the concentration of tumour marker. Knowledge of within-person variation and
the critical difference (RCV) are important when tumour markers are used for follow-up. In this study, the calculated critical differences of AFP, CEA and
CA 19-9 between serial measurements were 62.62 %, 72.57 % and 64.71 %, respectively. These values are required before one can be sure that the difference between two measurements is significant.
I think the take-home message is that the CEA readings inherently fluctuate a great deal, with larger fluctuations seen in individuals who have a high baseline CEA level. If two readings do not differ by more than around 70%, they should not be considered as different. That suggests that the CEA is a pretty poor marker and patterns rather than levels should be used as a diagnostic tool.
Hope this helps,
Jeremy
Colon cancer dx Feb. 24, 2009, T3/N2/M0
Right Hemicolectomy Feb. 26, 2009
Stage 3C: 4/19 positive nodes
High grade adenocarcinoma with tumor budding
FOLFOX6 April 15 - Oct. 1, 2009
Elective sub-total colectomy July 3, 2012 due to 2 DALMs
Currently NED