Response to Lee et al.

To the Editor: We thank the correspondents, Professors Lee, MacDonald, and Gusella, for their comments¹ and also thank the editors for allowing us to respond briefly to some of the points they have raised.

The original claim was that huntingtin CAG length does not affect the duration of survival after disease onset in Huntington disease (HD).² This claim was based on lack of significant association in two datasets using statistical methodology that we argue was inappropriate because it ignored data censoring. Our article now under discussion³ analyzed a superset of one of these two data (Registry HD) and addressed only the previous claim of no association, which had also been put forth by others.⁴ Using statistical methodology that accounted for various censoring mechanisms, we found a highly statistically significant association between CAG length and duration of survival after so-called age of onset, especially when the reported age of onset was controlled within the data models. Consequently, one could argue that the now-rejected null hypothesis does not serve as evidence to reinforce biological conjectures. This does not imply that the conjectures are incorrect, and the correspondents now introduce other evidence in support of some of their ideas. This also in no way implies that CAG length has the strongest (or even the strongest genetic) effect on mortality. That is an entirely different question to which the available data could not speak.

HD survival is not explained entirely by CAG repeat length of the huntingtin gene. Our article concluded with a reminder that CAG length does not equal destiny. While other genetic factors may play a leading role, one can also reasonably posit that eventual mortality at the end of this decades-long disease is largely due to an intertwining of myriad and often subtle biological and non-biological influences.

The correspondents propose alternatives to the idea that longer CAG length drives more “aggressive” disease. Commenting upon the empirically demonstrated association, we did offer this idea as a post hoc hypothesis, without further explanation of what exactly was meant by “aggressive” or what intermediate biological mechanisms might be involved. The statement was intentionally vague, and we are quite pleased if it serves as a catalyst to more detailed scientific debate. We will only offer the following logical and semantic points for this further discussion.

We must disagree with the assertion that equal age of onset implies equally aggressive disease. A still photograph of two cars alongside each other does not speak to whether one is in the process of passing the other. There seems no contradiction in hypothesizing that two people with the same illness onset age may have differing progression rates of underlying disease. (Many other diseases could be cited in which this is the case.)

It is argued that the observed CAG association may have been confounded by other unmeasured causal factors. However, inherent to the statistical definition of confounding, this would only be the case if those factors and CAG length exhibited statistical dependency. In the absence of dependency, which we are tempted to assume by default, these other factors would not affect the CAG association with survival. If there were a dependency, then we agree that the observed CAG association with survival (the only thing claimed with confidence) may not be due to CAG length per se. The cause of such statistical dependence would require further examination.

We believe there is a need for clarifying discussion about what is meant by CAG length’s “influencing” disease. The issue of somatic expansion is a case in point. A leading theory, discussed in Lee et al.’s Letter to the Editor, is that HD progression is caused primarily by the rate of CAG somatic expansion within neurons rather than by original CAG length. However, expansion rate is strongly, inherently influenced by original CAG length.⁵ As demonstrated by the correspondents⁶ and others,⁷ somatic expansion is also influenced by genetic variation in DNA mismatch repair genes. However, the connection between baseline CAG length and the extent of somatic expansion remains, regardless of other influences.⁵ We would therefore classify the effect of somatic expansion as an intermediate factor in the biological influence of CAG length upon disease. Others, perhaps focusing on only the most proximal mechanisms, would not. This inconsistency has already proven grounds for debates that could have been avoided by clarification of terminology.