As usual, I appreciate your thoughtful approach to explaining statistics and study outcomes in science articles that have to do with issues that are controversial in our current situation. We do need thoughtful ways to prevent and treat type 2 diabetes. And because some of what makes it a challenging health care problem have to do with social and psychological aspects of our humanity, like some people eating a lot of sugary treats in part to deal with depression, it isn't 100% solvable by physical sciences.
I read the whole piece and was about to call you out on the title (which felt fairly click-baity). Until you yourself pointed out that it was indeed biased and an example of how any one stat can be tweaked to tell any story you want.
My bigger concern today is how many of the policies being put in place and cuts being made are going to disproportionately impact higher risk populations. I anticipate a lot of even the positive-looking stats to turn negative in the coming years as a result of it. 😞
I am sort of stuck on the call out that obesity causes diabetes. I know it's a risk factor, but I know individuals who aren't obese at all and still have type 2 diabetes.
I came to ask a similar question. There is an association, but a cause? I don't know any proof of that. One could argue that obesity is a *symptom* of diabetes in some people....
There is some debate here, but after the DiRECT results - which I linked to - I think we can safely say that obesity directly causes diabetes. It is not the *only* cause, because diabetes is complex and multifactorial. Age is a big cause, and genetics play a major role. However, if you get people who are obese and have diabetes to lose 10-15% of their bodyweight, they no longer meet the diagnostic criteria for diabetes. It's not clear whether this is a cure, as I've discussed in another comment, but it does mean that at least in those people the obesity was causing the diabetes.
Perhaps you're looking past another way (besides dying as you suggest) an individual can remove themselves from the prevalence data...controlling their A1C and blood glucose through lifestyle and diet changes. That works pretty well as an alternative to dying.
I did discuss diabetes remission. As I said, it's contentious whether this results in a cure or just a temporary reprieve, and whether it has any impact on the underlying pancreatic damage. I'm addition, remission does not usually remove a person from these datasets for several reasons.
In particular, thanks for highlighting the fundamental difference between incidence and prevalence. And, the fact that increasing prevalence can result from people with the disease living longer, even if incidence is flat or decreasing.
But - adjusting incidence for age runs a risk of confusing age with birth year. There is inherently an age-period-cohort problem. Period is the year of diagnosis and cohort is year of birth. Age period cohort problems are inherently unsolvable unless you make strong assumptions. (Strictly speaking, the solution is unidentified.) For example, assume that birth year has no effect. It might be impossible to validate the assumption.
The CDC’s appendix with detailed methods for diabetes doesn’t actually say how they adjusted incidence for age, even though it has a section on that.
If you have further insights on that I’d be interested.
I studied a very similar problem with autism in depth. It’s difficult, but possible, to get solid answers.
I have to admit, I am not certain how incidence was adjusted for age in these analyses. Usually I would assume it is reported age when the survey was done, and in general the CDC stratifies by age group, but I cannot see any information on this either. I would generally at this point email the CDC but I have not had much luck getting responses from them since the new administration gutted the organisation!
I am not sure if the age-period cohort problem applies here (and to be transparent, I am not well versed in it).
Your example of autism is apt as the discrepancy between age and birth year can skew data as the “time interval” for the incidence of autism is pretty narrow. This may also apply to Type 1 diabetes.
Since the time interval to develop Type 2 diabetes is pretty broad, the difference between year of birth and age should not affect results.
I didn't see age of diagnosis discussed. We know that Type II diabetes diagnosed at an earlier age is usually more aggressive in terms of complications, such as kidney disease, eye complications, etc. and is associated with higher mortality.
As usual, I appreciate your thoughtful approach to explaining statistics and study outcomes in science articles that have to do with issues that are controversial in our current situation. We do need thoughtful ways to prevent and treat type 2 diabetes. And because some of what makes it a challenging health care problem have to do with social and psychological aspects of our humanity, like some people eating a lot of sugary treats in part to deal with depression, it isn't 100% solvable by physical sciences.
I read the whole piece and was about to call you out on the title (which felt fairly click-baity). Until you yourself pointed out that it was indeed biased and an example of how any one stat can be tweaked to tell any story you want.
My bigger concern today is how many of the policies being put in place and cuts being made are going to disproportionately impact higher risk populations. I anticipate a lot of even the positive-looking stats to turn negative in the coming years as a result of it. 😞
I am sort of stuck on the call out that obesity causes diabetes. I know it's a risk factor, but I know individuals who aren't obese at all and still have type 2 diabetes.
I came to ask a similar question. There is an association, but a cause? I don't know any proof of that. One could argue that obesity is a *symptom* of diabetes in some people....
There is some debate here, but after the DiRECT results - which I linked to - I think we can safely say that obesity directly causes diabetes. It is not the *only* cause, because diabetes is complex and multifactorial. Age is a big cause, and genetics play a major role. However, if you get people who are obese and have diabetes to lose 10-15% of their bodyweight, they no longer meet the diagnostic criteria for diabetes. It's not clear whether this is a cure, as I've discussed in another comment, but it does mean that at least in those people the obesity was causing the diabetes.
Perhaps you're looking past another way (besides dying as you suggest) an individual can remove themselves from the prevalence data...controlling their A1C and blood glucose through lifestyle and diet changes. That works pretty well as an alternative to dying.
I did discuss diabetes remission. As I said, it's contentious whether this results in a cure or just a temporary reprieve, and whether it has any impact on the underlying pancreatic damage. I'm addition, remission does not usually remove a person from these datasets for several reasons.
Depletion of susceptibles? All those who could have gotten diabetes have it and the people remaining are less likely to get it.
This is interesting, thank you.
In particular, thanks for highlighting the fundamental difference between incidence and prevalence. And, the fact that increasing prevalence can result from people with the disease living longer, even if incidence is flat or decreasing.
But - adjusting incidence for age runs a risk of confusing age with birth year. There is inherently an age-period-cohort problem. Period is the year of diagnosis and cohort is year of birth. Age period cohort problems are inherently unsolvable unless you make strong assumptions. (Strictly speaking, the solution is unidentified.) For example, assume that birth year has no effect. It might be impossible to validate the assumption.
The CDC’s appendix with detailed methods for diabetes doesn’t actually say how they adjusted incidence for age, even though it has a section on that.
If you have further insights on that I’d be interested.
I studied a very similar problem with autism in depth. It’s difficult, but possible, to get solid answers.
I have to admit, I am not certain how incidence was adjusted for age in these analyses. Usually I would assume it is reported age when the survey was done, and in general the CDC stratifies by age group, but I cannot see any information on this either. I would generally at this point email the CDC but I have not had much luck getting responses from them since the new administration gutted the organisation!
I am not sure if the age-period cohort problem applies here (and to be transparent, I am not well versed in it).
Your example of autism is apt as the discrepancy between age and birth year can skew data as the “time interval” for the incidence of autism is pretty narrow. This may also apply to Type 1 diabetes.
Since the time interval to develop Type 2 diabetes is pretty broad, the difference between year of birth and age should not affect results.
I didn't see age of diagnosis discussed. We know that Type II diabetes diagnosed at an earlier age is usually more aggressive in terms of complications, such as kidney disease, eye complications, etc. and is associated with higher mortality.