I was referring to all cause mortality and I actually sited three separate studies ( none of which I claim as my own
)
But returning to your point it is important to note that the expected relative risk factor for the 4- 4,4/ group was 0,92 compared to the reference group's of 5-5,4% of 1 I.e. An expected improvement of 0.08 for a non diabetic. Within the 95% confidence level this still allows a range of relative risk factor for the lower group of 0,5 to 1,62.
In order to get to the 4-4,4% group from a 5- 5,4% starting point one has to pass through the 4,5-4,9% band and this does have a worse expected risk factor at 1.06 than the reference 5-5,4% band ( 95% confidence range for the 4,5-4,9% band 0,78-1,42)
Out of the 14099 original participants only 469 (3,3%) fell in the 4-4,4% band. The study explicitly states that a higher proportion of hepatitis C sufferers fell in this band. So as a proportion of the total sample in model 4 which excludes hepatitis C, they must represent an even smaller percentage.
So by the above:
If I was a non diabetic and somehow managed to lower my hba1c from 5-5,4% all the way to 4-4,4% I would expect to lower my expected mortality risk excluding hepatitis c by 8 per cent. But if I missed my target and ended up in the 4,5-4,9 range I would expect to increase the same risk by 6%.
But I am a diabetic.
Only a tiny percentage of non diabetics manage 4-4,4%. Is it realistic to expect me or any other diabetic to match them.
Even the reference 5-5,4% is a tough ask for most diabetics. In order to go even lower would require greatly increased medication or dietary restriction. Any additional risk associated by such increased medication or dietary restrictions is not captured by the above studies which concentrates on non diabetics.