G. D. Maynard 
277 
inclusions in a group must be set against the accidental exclusions. For instance, 
is it reasonable to consider that in two cities of the same country, during the same 
period, and under similar Registration Laws, rates of 1184 and 440 per million 
are merely fortuitous or due to errors of registration ? Such a suggestion means 
that either more than half the cases are overlooked in the one town or more than 
100°/^ too many are recorded in the other, unless the error is to be divided 
between them. Before dismissing all differences in rates as being due to errors in 
diagnosis or imperfections of record it is well to realise exactly what this suggestion 
entails. 
Not only is it often said that better registration accounts largely for the varying 
rates in different places, but that much of the increase in the cancer death-rate, 
as also that occurring in the diabetes death-rate, is due to better diagnosis. The 
statistics show that both these diseases are very markedly on the increase. As 
a rule before death occurs both diseases give a very definite clinical picture, 
and run a more or less protracted course. When dealing with death returns, 
difficulties in early diagnosis are of no consequence. These factors are in favour of 
a correct diagnosis being made, and probably fewer mistakes are made in the 
diagnosis of these groups than in almost any others. In the last 30 years there 
has been very little added to our knowledge, that would aid the general practitioner 
to a more accurate diagnosis, in the final stages of these diseases. Doubtful 
tumours are more often submitted to the microscope for confirmation of diagnosis 
than formerly, but the effect of this on the returns would rather be to exclude 
from the cancer group cases that would otherwise have been found there ; for 
a fatal tumour of doubtful origin would most probably have been classed as cancer 
in the absence of definite knowledge to the contrary. There does not seem 
therefore, sufficient ground for believing that the increase in the rates is in any 
great part due to improved medical knowledge, or that this will account for the 
growth in the cancer and diabetes death-rates. 
If we refer to the cancer death-rates for " occupation groups " we find that 
those consisting of men who by their social status can best afford highly skilled 
medical care, or those who have State medical attendance provided, are the groups 
showing the lowest death-rates, while the labouring and agricultural classes show 
distinctly higher rates. This is the reverse of what one would be led to expect if 
the explanation that better diagnosis accounts for the increased rate — is the correct 
one. 
There is still a further reason for thinking that the group returned as cancer is 
fairly homogeneous in character. It will be seen from Fig. 1, how good a fit a 
Pearson Curve gives for the male death-rates, tabulated in five-yearly periods. 
The dotted polygons at ages 92-5 and 97-5 are the original observations and the 
constants of this curve are given below. Sheppard's adjustments were not used as 
I did not expect to find high contact at both ends. Finding however that there 
was high contact I recalculated the curve, using Sheppard's adjustments and 
Piometrika vii 3(; 
