Weight loss is good, right up until it isn’t: The obesity paradox.

Obesity is bad. I don’t have a citation for that, not because I couldn’t find one, but because, I mean, come on. OK fine — 36.5% of adults suffer from obesity in the US, and it is a key contributor to the pathophysiology of multiple chronic diseases. But obesity is not irreversible, and in welcome news weight loss of 5-10% of body weight can reduce insulin resistance, reduce blood pressure, and generally undo the damage caused by obesity.

So given this knowledge, we have long wrestled with the observation that obesity seems to be inversely associated with mortality in population-level studies. That’s right — when stratifying by BMI categories, patients with a BMI > 30 have in multiple observational studies been less likely to die. But a recent analysis by Edward Yu examines the data in a new way, and can teach us to use our common experience to learn to avoid confounding.

If You’re Only Going to Read One Paragraph

Obesity is still bad. It may look protective in observational studies, but that’s because most people who lose enough weight to move down a BMI category aren’t doing so intentionally, but rather are cachectic related to the inflammatory components of their chronic disease. You should still encourage patients to lose weight, and should in general be skeptical of any observational findings that don’t make sense.

So You’re Saying I Can Eat These Donuts

This obesity paradox was first described in hemodialysis patients in 1999. This investigation was inspired by the observation that hemodialysis (HD) patients in Mississippi (where the investigators were based) died at half the rate of the national average. Mississippi also has a 35.6% obesity rate. Coincidence? Who’s to say based on one observational study, but a one-unit increase in BMI (i.e. from 31 to 32) was associated with a 6% reduction in mortality, P < 0.0001. A 2003 review of this effect in HD patients coined the term “reverse epidemiology” (which we are instead going to call bizarro because I like it better), and found inverse association with mortality not only for obesity, but for hypertension and dyslipidemia as well. Granted one could make the argument that these things are simply tracking obesity given that the associations are drawn from separate studies, but still. Bizarro. This effect has been described in a diverse array of diseases ince then, from stroke to heart failure to acute coronary syndrome.

However, as astute readers I’m sure you have already been making up your own explanation for this as you go. Older people naturally lose weight, so BMI category may simply track age (a known risk factor for mortality, and I can say that because I’m about to turn 30). Inflammatory disease from cancer to advanced heart disease result in weight loss as well — show me an obese patient with NYHA 3 heart failure and I’ll show you a picture of him much larger two years ago.

Ok, but why can’t I Eat the Donut?

Yu and colleagues took data from three fairly famous cohort studies, the Nurses’ Health Study I and II and Health Professionals Follow-Up Study. These are the same giant cohorts that gave us the old data on aspirin for primary prevention in men and a tremendous number of epidemiological studies in women. The sample has known limitations, mostly predictable from the fact that it’s made up of health care professionals — men in the cohort were only 15% as likely to have ACS compared to age-matched controls, for example. The sample is also predominantly white.

This gave the authors a massive sample size, with 225,072 men and women followed for a median of 12.3 years. As a test of the theory above, that weight loss may more often be more an acute change related to end stage chronic disease than an intentional change resulting in risk reduction the authors used both a traditional stratification approach with an initial baseline BMI, as well as stratifying instead by peak BMI over the study period. In multivariate analysis, they adjusted for age, smoking status, and sex.

The results?

Yu et al., Ann Intern Med 2017 166:613-20

Using a normal BMI as the reference, every other category (unsurprisingly including underweight) increases your risk of death. Even overweight, which has been found to be protective even in studies that found obesity per se to be risk factors, is an increased risk for mortality when viewed this way. In short, by taking maximum BMI rather than enrollment BMI, we avoid confounding those who have avoided obesity over their lifetimes with those who have moved out of the obese category due to disease-related cachexia.

The study isn’t perfect — for one thing, both of these cohorts utilized survey data rather than measured weights, so self-reporting of BMI may have mis-estimated true weight (however if so it is likely to have done so in a uniform or random way). BMI itself is a limited instrument given its failure to take into account body composition (this is the very reason that weight loss due to chronic illness looks the same in cohort studies as weight loss due to diet and exercise). However, on the whole, this is an excellent and novel analysis, and offers not only an explanation of bizarro epidemiology but also a positive analysis for the epidemiological risk of increased maximum body weight.

Just Tell Me I Can Eat the Donut

Jeez, go ahead, fine. What’s with you and this donut anyway? It’s like I’m talking to my own id.

Thanks, I Was Super Hungry. So You’re Saying That Obesity Is Bad, But Weight Loss Is Also Bad?

Pretty much? I think what I’m saying is weight loss is hard. It’s well-established that people lose weight quickly in response to lifestyle interventions, then plateau and slowly regain it. So, if you are seeing true, sustained weight loss, you should consider whether it is truly intentional or rather might represent an undiagnosed chronic disease. And you should still be encouraging all of your obese patients to lose at least 5-10% of their body weight to aid in management of their comorbidities. Now you can recommend this without fear that you’re somehow dooming them to increased rates of death if they do move down a BMI category — when analyzed pragmatically, BMI remains an independent risk factor for mortality.

Tomorrow on IM HEAT

Mental health day, as I’m starting on a new service and may be too busy to write (hence the shorter piece today). Check back in on Friday for a piece by budding nephrologist Dr. John Nawn on novel approaches to prevent AKI in ICU patients! I learned something reading it, and I bet you will too!

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s