EURURO Vol. 72 No. 6

941 1030

proxies of healthfulness (eg, stage, age, comorbidities) will
also mitigate this bias.
Second, assessing prediagnosis physical activity—or
any prediagnosis exposure—in a case-only survival analy-
sis introduces selection bias. This selection bias structure
has been discussed primarily in the context of prediag-
nosis obesity and chronic disease; obesity is often
associated with lower mortality among those with chronic
disease (the “obesity paradox”). However, such findings do
not imply that obesity before diagnosis will improve
survival after disease diagnosis . Because obesity
increases the risk of chronic disease, nonobese patients
with chronic disease must have other risk factors
(eg, genetic predispositions) that may also be associated
with worse survival after diagnosis. Related to this
analysis, evidence suggests that physical activity
decreases the risk of many cancers and chronic diseases,
including prostate cancer . Thus, when Wang et al
examine prediagnosis physical activity and survival after
prostate cancer diagnosis among cases only, patients who
were diagnosed with prostate cancer despite high physical
activity were more likely to have other risk factors
(eg, family history of prostate cancer) that may be associated
with more aggressive prostate cancer. As a result, the
observed inverse association between prediagnosis physical
activity and survival after diagnosis is probably under-
estimated here. In this specific situation, this selection bias
structure does not have very detrimental implications; it is
only underestimating the benefits of physical activity.
However, it should still be avoided. When assessing
unhealthy prediagnosis exposures such as obesity among
patients with chronic disease, these exposures can decep-
tively appear to improve survival. Thus, associations
between prediagnosis exposures and disease-specific mor-
tality are typically better analyzed within the whole
population rather than among cases only.
In their examination of postdiagnosis physical activity
and prostate cancer survival, the authors took important
steps to mitigate reverse causation. However, in doing so,
they probably also concealed some of the true beneficial
effects of physical activity on prostate cancer progression.
Reverse causation, whereby progressing disease influences
risk factors (eg, disease often causes individuals to lose
weight, eat less, exercise less, change medications) is a
critical issue in many epidemiological analyses, and
analytical measures must be taken to address this. Here,
the authors applied physical activity exposure from 4 –6 yr
previous to the current exposure period in case men became
less active because they were sick. However, this prevents
investigators from analyzing the impact of more recent
physical activity, which may also have true benefits
independent of reverse causation. Lagging also excludes
anyone who died within 4–6 yr after diagnosis, so the
authors were less able to assess the effect of postdiagnosis
physical activity on men with more aggressive disease.
In summary, it is likely that many analyses, including the
current one by Wang et al, arrive at underestimates of
the inverse relationship between physical activity and
prostate cancer. This further highlights one of the many
health benefits of physical activity. Physical activity may
often achieve prostate cancer benefits of similar magnitude
to many prostate cancer drugs and treatments.
Conflicts of interest: The author has nothing to disclose.
Acknowledgments: This research was funded by the Harvard T.H. Chan
School of Public Health Nutrition Department.
References
[1]
[2]
.
[3]
.
E U R O P E A N U R O L O GY 7 2 ( 2 0 17 ) 9 4 0– 9 41 941