The major medical journals of the world receive far more papers than they can ever publish, and so it is rather surprising when dull, trivial or bad work appears in them. This must mean either that the editors of the journals, like Homer, sometimes nod, or that the general standard of the work submitted for publication is lower than one might hope or suppose.
A recent paper in the New England Journal of Medicine, entitled “Glucose Levels and Risk of Dementia,” by no fewer than fourteen authors, is a case in point. They repeatedly measured the blood glucose levels of 2067 people aged on average 76 at the start of the study, followed them up for a median length of 6.8 years, and correlated the levels with the patient’s chances of developing dementia.
It was already known that diabetics are at increased risk of developing dementia, not surprisingly in view of the damage that diabetes does to small blood vessels in the brain. But the authors of the paper put forward the hypothesis that higher levels of glucose even in non-diabetics would increase the risk of developing dementia.
They indeed found that non-diabetic patients with a blood sugar level of 115 milligrams per decilitre were more likely to develop dementia than those with a level of 100 milligrams. However, the extra chance, 1.18 times, though statistically significant, was so small that its significance in any other sense must be doubted. Generally speaking, epidemiological surveys which find such small differences are not of much value from the point of view of elucidation of the causation of diseases. If you trawled through a hundred factors – coffee consumption, number of begonias in the garden, subscription to a newspaper, etc. – you would probably find five such factors with odds rations as large (or small).
More than this, however, some of the data given by the authors do not fit with their hypothesis and indeed are mysterious and inexplicable. According to a graph in the paper, the blood glucose level among diabetics does not correlate in a linear fashion with the chance of developing dementia. Instead, the graph is a flattish, j-shaped one, in which, over a third of it, higher levels of blood glucose correlate with a lower chance of developing dementia. The increased risk of someone developing dementia with a blood glucose level of 150 compared with someone with a blood level of 170 was actually higher than that quoted for non-diabetics with levels of 115 and 100, respectively.
This is absurd. Since diabetes is defined simply by the level of glucose in the blood, there is no reason why the relationship between blood glucose levels and the development of dementia should take the form described in this paper. Elephantine labor has thus given birth to a scientific mouse.
What is the average clinician supposed to conclude from this paper? He already knew that diabetes was bad for his patients and would advise against those habits that gave rise to it. Is he likely to put the fear of God into his patients with average blood glucose levels of 115 by saying, “Look, you have a 1.18 times chance of developing dementia in 6 years’ time compared with people with a level of 100, so alter your diet”? Is he likely to say to his patients with a blood sugar level of 150 “I think you should try to get it up to 170 to lessen your chances of getting dementia”?
Not all that appears in the world’s most distinguished medical journals is distinguished. Laborious hackwork exists in all fields of human endeavor; indeed, perhaps it is the norm.