CHANGES WITH AGING 



42' 



CUTTING POINT AT \)=j<r 

 THE PROBABILITY OF MISCLASSIFICATION 



.25 



.20 



o 

 .o 

 o 



10 



.00 



Figure 3. Probability ot mis- 

 classifying an individual on 

 the basis of a single measure- 

 ment of cholesterol. 



1/8 1/4 1/2 I 



Ratio v/cr 



V > population standard deviation 

 (J-- individuol ii ii 



Framingham Study ( Dawber, Moore, and Mann, 1957 ) chose a cutting 

 point at 260 mg. cholesterol per 100 ml., and the follow-up showed that 

 above this cutting point the risk was 2.9 times that below this level. In 

 the Albany Study (Doyle ei al, 1957, 1959) the cutting point was 275 

 mg. per 100 ml., and the risk above that level proved to be 3.4 times 

 that of the men with cholesterol values below 275. 



Elsewhere (Keys and Fidanza, I960; Keys and Kihlberg, to be 

 published ) we have shown that these estimates of the relative risks are 

 underestimates of the importance of the relationship between the true 

 mean blood-cholesterol level and the subsequent appearance of the 

 disease. This follows from the fact that in each case the classification of 

 the men was made on the basis of a single measurement. Since there is 

 a considerable degree of variability in an individuaFs cholesterol level 

 at different times (Keys, Anderson, and Grande, 1959), any classifica- 

 tion based on single blood samples will result in some misclassifications 

 in regard to the true mean values of the individuals concerned. Figure 

 3 shows the probability of misclassification associated with different 

 ratios of inter-individual standard deviations when single cholesterol 

 measurements are used to assign indi\'iduals above and below a par- 

 ticular cutting point. In this case the cutting point is the population 

 mean plus half the average inter-individual standard deviation. From 

 available information on the average intra-indi\'idual variabiHty of 



