Comorbidity as a clue to pathogenesis

Based on some 84 million person-years of data from the Danish Psychiatric Central Research Register, Plana-Ripoll et al. (2019) have shown convincingly that comorbidity is pervasive across all diagnostic classes, is temporally patterned yet persistent and bidirectional. Some pairs carry strikingly high absolute risks, but all occur much above chance. In no instance is the association negative, whereby one condition decreases the probability of another. The authors make no assumption that the disorder following the first is causally related to the prior disorder. The analyses to date have examined only pairs of disorders rather than comorbidity across three or more. This report is of exceptional significance because of its implications for aetiology. The findings make one ask what is going on for such comorbidity to come about? The authors see the pattern that has emerged as supporting the hypothesis that some disorders share their risk architecture. In an accompanying Editorial, Hyman (2019) looks upon the findings as ‘additional evidence that shared factors underlie the pathogenesis of mental disorders’. What these factors are and how they contribute to pathogenesis now await explication.

Caspi and Moffitt (2018) have proposed something closely congruent with the Danish data: that all mental disorders have a common vulnerability. They call this the p-factor. It is seen as ‘a quantitatively distributed, stable, generalized liability to develop any and all forms of psychopathology across the life course …’ The Danish findings and this hypothesis now need to be taken further. It is proposed here that for those pairs of disorders with high hazard ratios, a final common pathway in pathogenesis must lead to only one outcome if both disorders are to develop: this must be at the level of brain structure and function. Many factors, including environmental exposures, are assumed to lead to this. Plana-Ripoll et al. have not yet indicated what mediates such wide comorbidity and in what domains. A strategy is now needed to determine where to look. It seems appropriate to focus initially on those disorders that co-occur with the highest hazard ratios. If their pathogenesis is similar, any similarity in aetiological variables and any similarity in neural pathology are likely to be the most apparent.

Here, we propose that the shared risk architecture may lie in any of the following domains: vulnerability to the two disorders as conferred by common genes; similar environmental exposures, either biological or social; similar developmental trajectories; and similar response to the same treatments; and similar developmental changes in actual brain structure or function over time. Throughout, attention needs to be accorded to the temporal pattern of association, in which the probability of the second disorder emerging has been found to decline sharply after the first 12 months. While this may be attributable to the diagnostic activity of clinicians, it could also reflect temporal changes in the intensity of pathogenic variables.

There is already some evidence that the affective disorders and schizophrenia have shared genes and similar effective treatments. The same environmental exposures are already known to be associated with a diversity of diagnostic groups, childhood abuse being one, though how this wide effect is mediated is far from clear. The association between intellectual disability and schizophrenia or other disorders is well established. For evidence of shared brain pathology, Caspi and Moffitt cite five recent publications, all in recent years. In one of these, reduced grey matter volumes in the dorsal anterior cingulate cortex and both left and right anterior insulae were found in schizophrenia, bipolar disorder, obsessive–compulsive disorder, depression, anxiety and substance addiction (Barch, 2017). But such findings bring their own puzzlement: these six disorders each have distinct differences in their phenomenology, their epidemiology and their course over time, yet they are found to have similar brain changes. A wide range of diagnoses sharing similar brain changes does little to illuminate the pathogenesis of any one of them. One explanation may be that current methods of examining brain structure and function are still too coarse to identify differentiating features. For that, demonstration of morphological changes needs to be complemented by better technical methods for examining functional interaction between brain areas.

As a contribution to clinical epidemiology, it would be hard to overestimate the scientific significance of the Danish data. They may carry information of far-reaching value if adequately interpreted. That may require knowledge of brain function which is not yet available. While the necessary advances in technology are awaited, further investigation may throw light on mechanisms whereby two diagnoses occur much above chance in the same individual. In addition to these scientific endeavours, the Danish findings also offer exciting possibilities for entirely new strategies in treatment and prevention. Any contribution to these would be most timely.