“We need less research, better research, and research done for the right reasons”

– Doug Altman (1)

In their POINT article “Current Evidence and Future Perspectives,” Carbone et al. summarize observational studies of people with an obesity-related comorbidity that find improved outcomes in those who had obesity as opposed to those who did not have obesity. Since obesity is a known cause of obesity-related comorbidities, the authors find the results of these studies to be paradoxical. I do not, as the results are easily explainable due to systematic errors in the research methods. Unfortunately, the opposing viewpoint provides no attempt to explain the rationale producing the nonintuitive results and solve the so-called paradox. Carbone et al. acknowledge that “hypotheses have been proposed and previously described to explain the obesity paradox.” However, they cite only 4 papers, all of which were written by proponents of the so-called obesity paradox. They did not cite the long list of articles describing how biased research methods can create nonintuitive and invalid results, such as finding improved outcomes in those with unhealthy characteristics such as excessive adiposity or the behavior of cigarette smoking (215). Future studies using the same methods will not help us to find the truth or advance healthcare. It is time to put the logic back into epidemiologic research.

The results in the studies cited by the proponents of the so-called obesity paradox are expected because there are biases from both reverse causation (which can be thought of as confounding) (see COUNTERPOINT article) and a form of selection bias known as collider stratification (5) or index event bias (13). Observing people until death will show that weight loss is harmful because dying people lose weight through the dying process (1618). Consequently, worsening illness confounds the assessment of the effect of body weight on longevity. The studies also fit the textbook definition of structural selection bias (a bias that occurs when conditioning on a variable affected by the exposure of interest) (12). In each study claiming a paradox, subjects are followed only after contracting an obesity-caused comorbidity, e.g., heart failure, hypertension, coronary heart disease, pulmonary arterial hypertension, and atrial fibrillation, and obesity status is measured after disease status is ascertained. Consider the section by Carbone et al. on hypertension (HTN) in their POINT article where they state: “Obesity, particularly visceral adiposity, is also associated with increased activation of the sympathetic nervous system (SNS), which is, in turn, linked to HTN,” (19) and then, “However, overweight and obesity have shown to confer some degrees of protection once HTN was diagnosed [read conditioned upon]. . .” (20). It is incorrect to state that the obesity conferred protection, as it implies that obesity itself caused a benefit, which is not true. Rather, the association of obesity with better outcomes was due to a noncausal relationship between the variables that developed due to the study design as shown in Figure 1 (this is similar to figure 3 in “Harmful exposures can appear protective, repeatedly, when there is bias”) (see COUNTERPOINT article). Conditioning upon the outcome of HTN creates an inverse association between obesity (a cause of HTN) and other causes of HTN (genetics, kidney disease, etc.).

FIGURE 1.

A structural diagram showing the causes of mortality from hypertension (HTN). The box around HTN diagnosis signifies that the analysis is restricted to only those with HTN. The dotted line signifies that this restriction induces a negative association between the causes of HTN. The red lines signify that this is the more dangerous path to HTN. “Other things” = kidney disease, genetics, etc.

FIGURE 1.

A structural diagram showing the causes of mortality from hypertension (HTN). The box around HTN diagnosis signifies that the analysis is restricted to only those with HTN. The dotted line signifies that this restriction induces a negative association between the causes of HTN. The red lines signify that this is the more dangerous path to HTN. “Other things” = kidney disease, genetics, etc.

Close modal

Best evidence about the effect of something (e.g., obesity) comes when research compares 2 groups with an equal chance of an outcome except for that “something,” and then follows those people over time (21,22). In the case of studies about the effects of obesity, the ideal study would be if the people with obesity would have had the same risk for bad outcomes as the people without obesity had the obese people not been obese. However, in the observational studies cited by the proponents of the so-called obesity paradox, the non-obese people all had another reason for their illness, and the analyses do not attempt to correct this flaw. The reason that the results appear contrary to what one may believe about obesity is because they are wrong. The comparisons are invalid, and repeating the comparisons over and over again is not helpful. When the methods are biased, the truth will not emerge even if more studies of similar design and analysis are conducted.

Evidence assessment would be improved if researchers explicitly stated their assumptions about causes and effects (ideally with figures to minimize confusion), instead of couching all of their observations in terms of nonspecific associations (23). The other viewpoint detailed many of the harmful effects of obesity (although not listing obesity explicitly as a cause of the effects) and then how obesity is associated with better outcomes when studies select only those with obesity-related comorbidities. However, they did not attempt to explain the causes of the associations. If they worked with methodologists to diagram the causes, then they would see that the association of obesity and “better outcomes” is not due to obesity causing better outcomes (see Figure 1). Since they do not draw out their causes and effects, they continue to be deceived. To quote Judea Pearl, PhD, in a book chapter (“Paradoxes Galore”) detailing why people become deceived by these apparently paradoxical findings: “The correlations[s] observed [are], in the purest and more literal sense, an illusion. Or perhaps even a delusion: that is, an illusion we brought upon ourselves by choosing which events to include in our data set and which to ignore” (7).

I am not certain of the reasons why the proponents of the so-called obesity paradox continue to unnecessarily complicate the evidence assessment. For example, the other viewpoint included a long discussion about the discrepancies between “obesity” defined as excess adiposity and the clinical use of body mass index (BMI), an imperfect measure of obesity. It remains a “red herring.” The mislabeling of some people as obese who do not have excessive body fat is not why there is misinterpretation of the results as paradoxical.

In addition to the biases mentioned above, there remains the possibility that, once people with obesity acquire a cardiovascular comorbidity, then a higher body weight is beneficial. This does not appear to be the case (24), but it is theoretically possible, and it could be studied. If the promoters of the so-called obesity paradox believe that extra weight is beneficial in people with obesity and a comorbidity, then they should propose a study that would randomize obese people with obesity-related comorbidities to either weight gain or control (14). If higher levels of obesity are beneficial, then those randomized to weight gain would have better outcomes. There may be skeptics on the grant review committee or the institutional review board, but the promoters of the so-called obesity paradox could cite all of their previous observational data. However, I am unaware of any such attempts to study weight gain in the obese. On the other hand, there are numerous randomized clinical trials studying weight loss in obese individuals with obesity-related comorbidities, and these consistently show that intentional weight loss causes improvements in health (2531).

Research studies should specify their questions, and if one's question is, “Among people with CVD, who, on average, lives longer: people with a BMI ≥ 30 or those with a BMI < 30?” the answer is clearly those with a BMI ≥ 30, partly because the comparison group (those with a BMI < 30) must have had a cause other than obesity for their CVD (see Figure 1). As such, a clinician could say to a patient with CVD who also has obesity, “Great news. Studies show that you will live longer than those unfortunate folks who have CVD and who are not obese.” However, clinical treatment decisions should be based upon causal effects, and the clinical treatment question is not, “Will you live longer than another person who contracted a similar disease but through a different mechanism?” Rather, the clinical treatment question is, “What is the best treatment for you?” Here, for the person with obesity who has an obesity-caused morbidity, randomized controlled trials provide valid evidence that treatment recommendations should include weight loss (2531). Achieving weight loss is difficult and complicated, and future research should be devoted to improving effectiveness of treatment options (4).

Due to the rising number of avenues for publication, many with suspect peer review and others with a simple lack of awareness of the biases, there are now several articles per week reporting paradoxical findings related to obesity (32). If your goal is more publications, then you too can follow the same strategy, whether it be with obesity-related comorbidities or other situations; simply choose those with a condition, compare the causes of that condition, and one of the causes will falsely appear to be beneficial (33). Since the results will seem shocking, you can call for more research, and so it goes. However, then we will see more publications reporting invalid results, and the public will be further confused. For the sake of personal and public health, journal reviewers and editors must recognize the effect of structural biases in studies. A recent editorial pleaded, “Journals should no longer accept ‘obesity paradox’ articles” (8). Evidence-based medicine calls on us to incorporate the best evidence into clinical treatment decisions. This must entail the best methods for evidence assessment. It is time to take heed of the wisdom of Doug Altman: “We need less research, better research, and research done for the right reasons” (1).

1.
Altman
DG.
The scandal of poor medical research
.
BMJ
.
1994
;
308
:
283
4
.
2.
Banack
HR,
Kaufman
JS.
Does selection bias explain the obesity paradox among individuals with cardiovascular disease?
Ann Epidemiol.
2015
;
25
:
342
9
.
doi:10.1016/j.annepidem.2015.02.008.
3.
Banack
HR,
Stokes
A.
The ‘obesity paradox’ may not be a paradox at all
.
Int J Obes
.
2017
;
41
:
1162
3
.
doi:10.1038/ijo.2017.99.
4.
Shrier
I.
The “obesity paradox” is not a paradox: time to focus on effective treatments
.
JACC Hear Fail
.
2016
;
4
:
234
.
doi:10.1016/j.jchf.2015.11.002.
5.
Cole
SR,
Platt
RW,
Schisterman
EF,
Chu
H,
Westreich
D,
Richardson
D,
Poole
C.
Illustrating bias due to conditioning on a collider
.
Int J Epidemiol
.
2010
;
39
:
417
20
.
doi:10.1093/ije/dyp334.
6.
Greenland
S.
Quantifying biases in causal models: classical confounding vs collider-stratification bias
.
Epidemiology
.
2003
;
14
:
300
6
.
7.
Pearl
J,
MacKenzie
D.
The book of why: the new science of cause and effect
.
New York
:
Hachette Book Group
;
2018
.
8.
Peeters
A.
Journals should no longer accept ‘obesity paradox’ articles
.
Int J Obes (Lond)
.
2018
;
42
:
584
5
.
doi:10.1038/ijo.2017.259.
9.
Schooling
CM,
Au Yeung
SL.
‘Selection bias by death’ and other ways collider bias may cause the obesity paradox
.
Epidemiology
.
2017
;
28
:
e16
7
.
doi:10.1097/EDE.0000000000000591.
10.
Hernández-Díaz
S,
Schisterman
EF,
Hernán
MA.
The birth weight ‘paradox’ uncovered?
Am J Epidemiol.
2006
;
164
:
1115
20
.
doi:10.1093/aje/kwj275.
11.
Weuve
J,
Tchetgen Tchetgen
EJ,
Glymour
MM,
Beck
TL,
Aggarwal
NT,
Wilson
RS,
Evans
DA,
Mendes de Leon
CF.
Accounting for bias due to selective attrition
.
Epidemiology
.
2012
;
23
:
119
28
.
doi:10.1097/EDE.0b013e318230e861.
12.
Hernán
MA,
Hernández-Díaz
S,
Robins
JM.
A structural approach to selection bias
.
Epidemiology
.
2004
;
15
:
615
25
.
doi:10.1097/01.ede.0000135174.63482.43.
13.
Dahabreh
IJ,
Kent
DM.
Index event bias as an explanation for the paradoxes of recurrence risk research
.
JAMA
.
2011
;
305
:
822
3
.
doi:10.1001/jama.2011.163.
14.
Lajous
M,
Banack
HR,
Kaufman
JS,
Hernan
MA.
Should patients with chronic disease be told to gain weight? The obesity paradox and selection bias
.
Am J Med
.
2015
;
128
:
334
6
.
doi:10.1016/j.amjmed.2014.10.043.
15.
Porta
M,
Vineis
P,
Bolúmar
F.
The current deconstruction of paradoxes: one sign of the ongoing methodological “revolution.”
Eur J Epidemiol
.
2015
;
30
:
1079
87
.
16.
Katz
AM,
Katz
PB.
Diseases of the heart in the works of Hippocrates
.
Br Heart J
.
1962
;
24
:
257
64
.
17.
Doehner
W,
Anker
SD.
Cardiac cachexia in early literature: a review of research prior to Medline
.
Int J Cardiol
.
2002
;
85
:
7
14
.
18.
Alley
DE,
Metter
EJ,
Griswold
ME,
Harris
TB,
Simonsick
EM,
Longo
DL,
Ferrucci
L.
Changes in weight at the end of life: characterizing weight loss by time to death in a cohort study of older men
.
Am J Epidemiol
.
2010
;
172
:
558
65
.
doi:10.1093/aje/kwq168.
19.
Hall
JE,
do Carmo
JM,
da Silva
AA,
Wang
Z,
Hall
ME.
Obesity-induced hypertension: interaction of neurohumoral and renal mechanisms
.
Circ Res
.
2015
;
116
:
991
1006
.
doi:10.1161/CIRCRESAHA.116.305697.
20.
Stamler
R,
Ford
CE,
Stamler
J.
Why do lean hypertensives have higher mortality rates than other hypertensives? Findings of the hypertension detection and follow-up program
.
Hypertension
.
1991
;
17
:
553
64
.
doi:10.1161/01.HYP.17.4.553.
21.
Greenland
S,
Robins
JM.
Identifiability, exchangeability, and epidemiological confounding
.
Int J Epidemiol
.
1986
;
15
:
413
9
.
22.
Greenland
S,
Robins
JM.
Identifiability, exchangeability and confounding revisited
.
Epidemiol Perspect Innov
.
2009
;
6
:
4
.
doi:10.1186/1742-5573-6-4.
23.
Hernán
MA.
The C-word: scientific euphemisms do not improve causal inference from observational data
.
Am J Public Health
.
2018
;
108
:
616
9
.
doi:10.2105/AJPH.2018.304337.
24.
Banack
HR,
Kaufman
JS.
The obesity paradox: understanding the effect of obesity on mortality among individuals with cardiovascular disease
.
Prev Med (Baltim)
.
2014
;
62
:
96
102
.
doi:10.1016/j.ypmed.2014.02.003.
25.
Stamler
R,
Stamler
J,
Grimm
R,
Gosch
FC,
Elmer
P,
Dyer
A,
Berman
R,
Fishman
J,
Van Heel
N,
Civinelli
J.
Nutritional therapy for high blood pressure. Final report of a four-year randomized controlled trial—the Hypertension Control Program
.
JAMA
.
1987
;
257
:
1484
91
.
26.
Whelton
PK,
Appel
LJ,
Espeland
MA,
Applegate
WB,
Ettinger
WH,
Jr,
Kostis
JB,
Kumanyika
S,
Lacy
CR,
Johnson
KC,
Folmar
S,
Cutler
JA.
Sodium reduction and weight loss in the treatment of hypertension in older persons: a randomized controlled trial of nonpharmacologic interventions in the elderly (TONE). TONE Collaborative Research Group
.
JAMA
.
1998
;
279
:
839
46
.
27.
Mertens
IL,
Van
Gaal LF.
Overweight, obesity, and blood pressure: the effects of modest weight reduction
.
Obes Res
.
2000
;
8
:
270
8
.
doi:10.1038/oby.2000.32.
28.
Pack
QR,
Rodriguez-Escudero
JP,
Thomas
RJ,
Ades
PA,
West
CP,
Somers
VK,
Lopez-Jimenez
F.
The prognostic importance of weight loss in coronary artery disease: a systematic review and meta-analysis
.
Mayo Clin Proc
.
2014
;
89
:
1368
77
.
doi:10.1016/j.mayocp.2014.04.033.
29.
Kitzman
DW,
Brubaker
P,
Morgan
T,
Haykowsky
M,
Hundley
G,
Kraus
WE,
Eggebeen
J,
Nicklas
BJ.
Effect of caloric restriction or aerobic exercise training on peak oxygen consumption and quality of life in obese older patients with heart failure with preserved ejection fraction
.
JAMA
.
2016
;
315
:
36
.
doi:10.1001/jama.2015.17346.
30.
Kritchevsky
SB,
Beavers
KM,
Miller
ME,
Shea
MK,
Houston
DK,
Kitzman
DW,
Nicklas
BJ.
Intentional weight loss and all-cause mortality: a meta-analysis of randomized clinical trials
.
PLoS One
.
2015
;
10
:
e0121993. doi:10.1371/journal.pone.0121993.
31.
Pathak
RK,
Middeldorp
ME,
Meredith
M,
Mehta
AB,
Mahajan
R,
Wong
CX,
Twomey
D,
Elliott
AD,
Kalman
JM,
Abhayaratna
WP,
Lau
DH,
Sanders
P.
Long-term effect of goal-directed weight management in an atrial fibrillation cohort
.
J Am Coll Cardiol
.
2015
;
65
:
2159
69
.
doi:10.1016/j.jacc.2015.03.002.
32.
Stovitz
SD,
Banack
HR,
Kaufman
JS.
Structural bias in studies of cardiovascular disease: let's not be fooled by the “obesity paradox”
.
Can J Cardiol
.
2018
;
34
:
540
2
.
doi:10.1016/j.cjca.2017.10.025.
33.
Stovitz
SD,
Banack
HR,
Kaufman
JS.
Paediatric obesity appears to lower the risk of diabetes if selection bias is ignored
.
J Epidemiol Community Health
.
2018
;
72
:
302
8
.
doi:10.1136/jech-2017-209985.

Conflicts of Interest and Source of Funding: None.

Author notes

1University of Minnesota, Department of Family Medicine and Community Health, Minneapolis, MN 55455 USA