The contribution by Dabelea et al. (1) to the growing debate on the accelerator hypothesis is an important one, but I wonder if there is a confounder that has not been accounted for in the reasoning. The report revolves principally around Fig. 2, which shows, after appropriate adjustments, a clear inverse relationship between age at diagnosis and BMI (the acceleration predicted) among those whose fasting C-peptide (FCP) levels lay below the median, but none among those whose FCP lay above. The difference is interpreted to mean that any relationship to insulin resistance applies only to a subset of type 1 diabetic children with low β-cell reserve.
The accelerator hypothesis argues that “type 1 and type 2 diabetes are the same disorder of insulin resistance, set against different genetic backgrounds” (2). It predicts a general inverse relationship between BMI (surrogate for insulin resistance) and age at diagnosis and identifies three accelerators that determine the rate at which the β-cell mass declines during life: constitution (genes/gestation), insulin resistance (lipotoxity and antigenicity), and immune response (HLA) genotype (response to insulin resistance–induced antigenicity).
The one adjustment that was not made to the regressions in Fig. 2 of Dabelea et al.’s report may be the crucial one: the HLA genotype. Those children who fulfilled the prediction were younger than those who did not. The younger the type 1 diabetic child at diagnosis, the more likely he/she is to carry high-susceptibility HLA genes (3), and so it is possible, even probable, that the proportion of children carrying high-susceptibility HLA genes was greater in the younger group than in the older group. Reactive HLA genes are the third accelerator of β-cell loss, and presenting the data in this way may serve to support the hypothesis for the entire group, rather than qualify it according to FCP.
The only difference between the behavior of the two groups in Fig. 2 may be one of tempo, and while the age range studied was sufficient to demonstrate the predicted inverse relationship in the younger group carrying more intensely reactive HLA genes, it may not have been wide enough to demonstrate the corresponding relationship for the older group carrying less reactive genes. An age range spanning many decades may be needed where the tempo is slower. DNA is available to the SEARCH study, and it will be interesting to learn in due course whether the distribution of HLA genotypes was indeed different between the two groups.
The accelerator hypothesis has now been subject to the scrutiny of several independent cohort studies. The Birmingham study mentioned by Dabelea et al. was small and of mixed race. (The relationship between BMI and insulin resistance is different between children of Asian and European descent [4].) The racial distribution of the children in Dabelea et al.’s study is not detailed but may be important. Two other U.K. studies of almost exclusively white children (5,6) and a large European study involving many thousands of predominantly white children (7) have all shown the predicted inverse relationship between age at onset and BMI on simple univariate regression. Furthermore, a German study involving 920 children with type 1 diabetes shows the same (8), and a study of pre–type 1 diabetic children from Australia suggests that the more insulin resistant the child, the more rapidly he/she progresses to type 1 diabetes (9). Longitudinal studies will be important to further elucidate the accelerator hypothesis, as Dabelea et al. suggest, but the real test will be a randomized controlled trial to reduce insulin resistance in at-risk children of type 1 diabetes.
