Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
The following is a correction to the paper by Loukolaet al (J Med Genet 1999;36:819-22). An algorithm was introduced by Wijnen et al 1to predict the probability of finding a disease causingMLH1 or MSH2mutation in patients with HNPCC or possible HNPCC. The variables of this formula are the mean age of colorectal cancer (CRC) diagnosis in the family, fulfilment of the so called Amsterdam criteria, and the presence of endometrial cancer in the family. In addition to this basic formula, an alternative formula to be used in small families was introduced. The variables of this alternative formula are the mean age of CRC diagnosis in the family, the number of patients with CRC in the family, and the number of patients with endometrial cancer in the family. Applying these algorithms, a probability (p1 for the basic formula and p2 for the alternative formula) of 20% or higher for a germline mutation was proposed to justify mutation analyses. We tested these formulae in a series of 509 consecutive colorectal adenocarcinoma samples.2 MSI analysis and genomic sequencing of MLH1 andMSH2 for the MSI positive samples had been previously performed.3
Unfortunately, we used an inaccurate factor in the alternative formula. We used L=1.4 + (−0.09)V1 + 0.27 V2 + 0.75 V3, when the first factor should have been 1.8. Thus, the results obtained using the alternative formula were not accurate. We proposed that the alternative formula was able to identify only three out of 10 mutation positive patients when first degree pedigrees (on average eight family members) were used and six out of 10 when extensive pedigrees (on average 39 family members) were used. The correct results are: the alternative formula was able to detect six out of 10 (with first degree pedigrees) and eight out of 10 (with extensive pedigrees) mutation carriers.
To test the algorithm of Wijnen et al 1 further, we have now analysed 535 additional consecutive colorectal adenocarcinoma samples. The subjects ranged in age from 29 to 91 years with a mean age of 67 years. MSI analysis and genomic sequencing of MLH1 andMSH2 for MSI positive samples were performed previously (Salovaara et al, submitted); 66 out of 535 (12%) samples were MSI positive and 18 of these patients had an MLH1 orMSH2 germline mutation. We then combined the results with the previous set of samples and describe here the results from 1044 consecutive colorectal cancer patients. The results of all 28 mutation positive samples are presented in table 1. The basic formula identified five out of 28 (18%) mutation carriers when first degree pedigrees (on average eight family members) were used, and 18 out of 28 (64%) when extensive pedigrees (on average 39 family members) were used. The corresponding results for the alternative formula are 13 out of 28 (46%) and 21 out of 28 (75%). In addition, these formulae identified 11 patients in whom no MLH1 orMSH2 mutations could be found. They have all been tested and found negative for the two most common Finnish founder mutations. One of them was found to be a FAP patient and another one had juvenile polyposis. Three were sequenced forMLH1 and MSH2mutations with negative results. The remaining six patients were all MSI negative, diagnosed before the age of 40, and had no family history of cancer.
Extensive pedigree data are difficult to obtain in clinical practice. When relying on first degree pedigrees, which are generally easily obtained during patient interview, the mathematical algorithms proposed by Wijnen et al 1 were able to detect 18% (p1) and 46% (p2) of mutation carriers. Even with extensive pedigree information, the formulae were able to detect only 61% (p1) and 75% (p2) of carriers, meaning that every third or every fourth were missed. Based on the analysis of 1044 colorectal cancer probands, we conclude that these mathematical formulae alone are of limited value and use of additional tools such as MSI screening is warranted.