Articles
www.thelancet.com Published online November 28, 2012 http://dx.doi.org/10.1016/S0140-6736(12)61426-3 1
Published Online
November 28, 2012
http://dx.doi.org/10.1016/
S0140-6736(12)61426-3
See Online/Articles
http://dx.doi.org/10.1016/
S0140-6736(12)61804-2
Cardiology Department,
Veterans Affairs Medical
Center, Washington DC, USA
(Prof P F Kokkinos PhD,
C Faselis MD,
D Panagiotakos PhD,
M Doumas MD); Georgetown
University School of Medicine,
Washington DC, USA
(Prof P F Kokkinos); George
Washington University School
of Medicine, Washington DC,
USA (Prof P F Kokkinos,
C Faselis); Cardiology Division,
Veterans Affairs Palo Alto
Health Care System, Palo Alto,
CA, USA (Prof J Myers PhD); and
Stanford University, Stanford,
CA, USA (Prof J Myers)
Correspondence to:
Prof Peter F Kokkinos,
Veterans Affairs Medical Center,
Cardiology Division,
50 Irving Street NW,
Washington DC, 20422, USA
peter.kokkinos@va.gov
Interactive effects of fitness and statin treatment on mortality
risk in veterans with dyslipidaemia: a cohort study
Peter F Kokkinos, Charles Faselis, Jonathan Myers, Demosthenes Panagiotakos, Michael Doumas
Summary
Background Statins are commonly prescribed for management of dyslipidaemia and cardiovascular disease. Increased
fitness is also associated with low mortality and is recommended as an essential part of promoting health. However, little
information exists about the combined effects of fitness and statin treatment on all-cause mortality. We assessed the
combined effects of statin treatment and fitness on all-cause mortality risk.
Methods In this prospective cohort study, we included dyslipidaemic veterans from Veterans Affairs Medical Centers in
Palo Alto, CA, and Washington DC, USA, who had had an exercise tolerance test between 1986, and 2011. We assigned
participants to one of four fitness categories based on peak metabolic equivalents (MET) achieved during exercise test
and eight categories based on fitness status and statin treatment. The primary endpoint was all-cause mortality adjusted
for age, body-mass index, ethnic origin, sex, history of cardiovascular disease, cardiovascular drugs, and cardiovascular
risk factors. We assessed mortality from Veteran’s Affairs’ records on Dec 31, 2011. We compared groups with Cox
proportional hazard model.
Findings We assessed 10 043 participants (mean age 58·8 years, SD 10·9 years). During a median follow-up of 10·0 years
(IQR 6·0–14·2), 2318 patients died, with an average yearly mortality rate of 22 deaths per 1000 person-years. Mortality
risk was 18·5% (935/5046) in people taking statins versus 27·7% (1386/4997) in those not taking statins (p<0·0001). In
patients who took statins, mortality risk decreased as fitness increased; for highly fit individuals (>9 MET; n=694), the
hazard ratio (HR) was 0·30 (95% CI 0·21–0·41; p<0·0001) compared with least fit (≤5 METs) patients (HR 1; n=1060).
For those not treated with statins, the HR for least fit participants (n=1024) was 1·35 (95% CI 1·17–1·54; p<0·0001) and
progressively decreased to 0·53 (95% CI 0·44–0·65; p<0·0001) for those in the highest fitness category (n=1498).
Interpretation Statin treatment and increased fitness are independently associated with low mortality among
dyslipidaemic individuals. The combination of statin treatment and increased fitness resulted in substantially lower
mortality risk than either alone, reinforcing the importance of physical activity for individuals with dyslipidaemia.
Funding None.
Introduction
Results of several clinical trials have shown that statin
treatment substantially reduces morbidity and mortality of
individuals with coronary heart disease.1–3 On the basis on
these findings, the Adult Treatment Panel 3 and other
expert panels have issued guidelines4 for statin treatment
of patients with established coronary heart disease.5 Trials
also suggest that statin treatment provides health benefits
for individuals with high risk of cardiovascular disease
who do not have coronary heart disease.6–9
Expert panels on management of lipids have also
emphasised the importance of lifestyle changes
for reduction of cardiovascular risk.4,5 These recom
men dations are based on evidence from large
epidemiological studies. Data from these studies show
inverse, graded, independent, and robust associations
between physical activity (fitness) and mortality risk in
apparently healthy participants10–16 and in patients with
cardiovascular disease, irrespective of age, sex, or
comorbidities.14,17–20 Mortality risk is highest for patients
with low fitness; risk decreases as fitness increases
irrespective of sex, presence of other risk factors, or
age.14,15,17,19–21
Although a healthy lifestyle—including physical activity
and fitness—is promoted as an essential component for
prevention and management of coronary heart disease,
little data are available regarding the combined health
benefits of fitness and statin treatment. Furthermore, for
dyslipidaemic patients who cannot take statins, whether
increased mortality risk can be abated by increased fitness
is unclear. We assessed the separate and combined effects
of statin treatment and exercise capacity on allcause
mortality risk in veterans with dyslipidaemia.
Methods
Study design and patients
This prospective cohort study included patients from the
Veterans Affairs Medical Centers in Washington, DC,
USA and Palo Alto, CA, USA. The cohort was taken from
a database of more than 20 000 veterans who had dys lipi
daemia (defined by the International Classifica tion of
Diseases) and who had a symptomlimited exercise toler
ance test between 1986, and 2011. The test was adminis
tered either as part of a routine assessment or to assess
exerciseinduced ischaemia. This information, along with
the patient’s medical history, was electronically stored.
Articles
2 www.thelancet.com Published online November 28, 2012 http://dx.doi.org/10.1016/S0140-6736(12)61426-3
Exclusion criteria were: history of an implanted pace
maker; development of left bundle branch block during
the test; inability to complete the test because of
musculoskeletal pain or impairments; exercise capacity
less than 2 metabolic equivalents (MET); instability or
need for emergency intervention; bodymass index less
than 15·5 kg/m²; impaired chronotropic response to
exercise; HIV/AIDS; and missing data.
The study was approved by the institutional review board
at each institution, and all participants gave written
informed consent.
Procedures
Clinical characteristics and demographic and drug
information were obtained from the patients’ electronic
medical record before the exercise tolerance test. Each
individual was asked to verify the information, including
history of chronic disease, current treatments, and
smoking habits. Bodyweight and height were assessed by a
standardised scale and recorded before the test.
Duration of statin treatment was based on the first and
last date that statins were prescribed for each patient.
Individuals were judged to be on statin treatment if these
two dates were more than 3 months apart. The most recent
assessments of lipid and lipoprotein con centrations
represent poststatin treatment values. For patients not
taking statins, the lipid assessment values before diagnosis
of dyslipidaemia represent the first assessment and the
most recent lipid assessment represents the final value.
Exercise capacity was assessed by a standard treadmill
test with the Bruce protocol at the centre in Washington
DC, and by an individualised ramp protocol22 at the Palo
Alto centre. Peak exercise time was recorded in minutes.
Peak exercise capacity (in MET) was estimated with
standardised equations on the basis of peak speed and
grade for the ramp protocol22 and peak exercise time for
the Bruce protocol.23 One MET was defined as the energy
expended at rest, which is roughly equivalent to an
oxygen consumption of 3·5 mL per kg of bodyweight
per min. Participants were encouraged to exercise until
volitional fatigue in the absence of symptoms or other
indications for stopping.24 Use of handrails was allowed
only if necessary for balance and safety. Agepredicted
peak exercise heart rate was evaluated with a population
specific equation.25 Drug regimens were not altered
before testing.
Entire cohort
(n=10 043)
Taking statins
(n=5033)
Not taking statins
(n=5010)
p value
Age (years) 58·8 (10·9) 59·4 (10·4) 57·6 (11·3) <0·0001
Weight (kg) 90·5 (17·7) 92·4 (17·8) 88·5 (17·3) <0·0001
Body-mass index (kg/m²) 29·2 (5·2) 29·7 (5·3) 28·7 (5·2) <0·0001
Resting heart rate (beats per min) 71 (13) 71 (13) 71 (13) 0·76
Resting systolic blood pressure (mm Hg) 131 (20) 130 (20) 131 (21) 0·11
Resting diastolic blood pressure (mm Hg) 79 (12) 79 (12) 79 (12) 0·25
Peak MET 7·4 (2·6) 6·9 (1·9) 7·8 (3·0) <0·0001
Ethnic origin
African-American 5192 (52%) 3346 (66%) 1846 (37%) <0·0001
White 4425 (44%) 1624 (32%) 2801 (56%) <0·0001
Other 426 (4%) 63 (1%) 363 (7%) <0·0001
History of cardiovascular disease 4198 (42%) 2980 (59%) 1218 (24%) <0·0001
Family history of cardiovascular disease 1687 (17%) 609 (12%) 1078 (22%) <0·0001
Current smoker 2222 (22%) 1266 (25%) 956 (19%) <0·0001
Hypertension 5500 (55%) 3032 (60%) 2468 (49%) <0·0001
Type 2 diabetes mellitus 3775 (38%) 2075 (41%) 1700 (34%) <0·0001
Drug treatments
β blocker 1489 (15%) 860 (17%) 629 (13%) <0·0001
Calcium channel blockers 1773 (18%) 963 (19%) 810 (16%) <0·0001
Angiotensin-converting enzyme inhibitors 1869 (19%) 1249 (25%) 620 (12%) <0·0001
Angiotensin receptor blockers 232 (2%) 200 (4%) 32 (1%) <0·0001
Angiotensin-converting enzyme inhibitors or
angiotensin receptor blockers
1973 (20%) 1336 (26%) 637 (13%) <0·0001
Diuretics 1271 (13%) 1025 (20%) 246 (5%) <0·0001
Nitrates or vasodilators 546 (5%) 185 (4%) 361 (7%) <0·0001
Aspirin 668 (7%) 480 (10%) 188 (4%) <0·0001
Data are mean (SD) or n (%). p value calculated by Z test for ethnic origin, by χ² test for drug treatments and medical history (cardiovasular disease, hypertension, and
diabetes mellitus), and by t test for of the remaining variables. MET=metabolic equivalents.
Table 1: Baseline characteristics
Articles
www.thelancet.com Published online November 28, 2012 http://dx.doi.org/10.1016/S0140-6736(12)61426-3 3
We created four fitness categories on the basis of peak
MET. Patients with a peak MET value of 5·0 or less (the
lowest 25th percentile of the MET values acheived) were
classed as least fit; those with a peak MET value of 5·1–7·0
(26th–50th percentile) were classed as moderately fit; those
with a peak MET value of 7·1–9·0 (51st–75th percentile)
were classed as fit; and those with a peak MET value of
more than 9·0 (>75th percentile) were classed as highly fit.
We formed two groups (treated with statins and not treated
with statins) within each fitness category, giving eight
categories in total.
The primary endpoint was death from any cause. Dates
of death were first assessed from the database and verified
from the Veterans Affairs Benefi ciary Identification and
the Record Locator System File. This system is used to
calculate benefits for relatives of deceased veterans and is
95% complete and accurate.26 Mortality was assessed on 31
Dec 31, 2011.
Statistical analysis
Followup time is presented as median with IQRs. We
calculated mortality rate as the ratio of events to person
years of followup. Continuous variables are presented as
means and SDs and categorical variables as relative
frequencies. We tested baseline associations between
categorical variables with χ² or Z tests. We did oneway
ANOVA and t tests to evaluate mean differences of
normally distributed variables between individuals taking
and not taking statins. We tested the assumption of
equality of variances between groups by Levene’s test, and
the assumption of normality with probability–probability
plots. We calculated hazard ratios (HRs) for allcause
mortality for the four fitness categories, the two statin
treatment categories, and the eight fitness–statin categories
with Cox proportional hazard models. We also assessed
first order interactions between statin treatment and
fitness groups. For the four fitness categories, the least fit
category was the reference group. For the statin treatment
categories, no statin treatment was the refer ence group.
For the eight fitness–statin categories, the least fit, treated
with statins category was the reference group. We also
used Cox proportional hazard models to assess risk in the
eight fitness–statin categories. We adjusted analyses for
age, bodymass index, ethnic origin, sex, history of
cardiovascular disease, cardio vascular medications
(angiotensinconverting enzyme inhibitors, angiotensin
receptor blockers, β blockers, calcium channel blocker,
diuretics, vasodilators), and cardio vascular disease risk
factors (hypertension, type 2 diabetes mellitus, smoking).
We tested assumption of proportionality for all Cox pro
portional hazard analyses graphically, by plotting the
cumulative hazards of the logarithms of the covariates; the
proportionality assumption was fulfilled for each model.
All hypotheses were two sided and p less than 0·05 was
deemed statistically significant. We did all statistical
analyses with SPSS (version 19.0).
Role of the funding source
There was no funding source for this study. The corres
ponding author had full access to all data in the study and
had final responsibility to submit it for publication.
Patients treated with statins Patients not treated with statins
Before statin
treatment (n=2959)
After statin
treatment (n=2959)
Mean difference
(95% CI)
p value Initial (n=1433) At follow-up
(n=1433)
Mean diffrence
(95% CI)
p value
Total cholesterol (mmol/L) 6·1 (0·8) 4·4 (1·2) 1·7 (1·6 to 1·7) <0·0001 6·0 (0·8) 5·1 (1·1) 0·9 (0·8 to 0·9) <0·0001
Triglycerides (mmol/L) 1·6 (0·9) 1·5 (1·0) 0·2 (0·1 to 0·2) <0·0001 1·5 (0·9) 1·5 (0·8) 0·1 (0·0 to 0·2) 0·03
HDL cholesterol (mmol/L) 1·2 (0·3) 1·2 (0·3) 0·1 (0·0 to 0·1) <0·0001 1·2 (0·4) 1·2 (0·4) 0·0 (0·0 to 0·0) 0·22
LDL cholesterol (mmol/L) 4·2 (0·7) 2·6 (1·0) 1·6 (1·5 to 1·6) <0·0001 4·0 (0·8) 3·6 (0·9) 0·4 (0·4 to 0·5) <0·0001
Data are mean (SD) unless stated otherwise. p values calculated by paired t test.
Table 2: Lipid and lipoprotein concentrations
Patients
(n)
Deaths
(n; %)
Unadjusted HR
(95% CI)
Age-adjusted
HR (95% CI)
Fully adjusted
HR (95% CI)*
p value†
Whole cohort (n=10 043)
≤5·0 MET 2084 920 (44%) 1 1 1 ··
5·1–7·0 MET 2727 726 (27%) 0·61 (0·55–0·67) 0·68 (0·62–0·75) 0·70 (0·63–0·77) <0·0001
7·1–9·0 MET 3040 443 (15%) 0·34 (0·31–0·38) 0·48 (0·43–0·54) 0·50 (0·45–0·57) <0·0001
>9·0 MET 2192 229 (10%) 0·22 (0·19–0·25) 0·35 (0·30–0·40) 0·38 (0·32–0·44) <0·0001
Patients taking statins (n=5032)
≤5·0 MET 1060 389 (37%) 1 1 1 ··
5·1–7·0 MET 1573 329 (21%) 0·56 (0·49–0·65) 0·64 (0·55–0·74) 0·64 (0·55–0·74) <0·0001
7·1–9·0 MET 1705 173 (10%) 0·28 (0·23–0·33) 0·40 (0·33–0·48) 0·41 (0·34–0·49) <0·0001
>9·0 MET 694 39 (6%) 0·17 (0·12–0·23) 0·29 (0·21–0·41) 0·32 (0·23–0·45) <0·0001
Patients not taking statins (n=5011)
≤5·0 MET 1024 531 (52%) 1 1 1 ··
5·1–7·0 MET 1154 397 (34%) 0·65 (0·57–0·74) 0·72 (0·64–0·82) 0·74 (0·65–0·84) <0·0001
7·1–9·0 MET 1335 270 (20%) 0·40 (0·34–0·46) 0·56 (0·48–0·65) 0·57 (0·49–0·66) <0·0001
>9·0 MET 1498 190 (13%) 0·22 (0·18–0·25) 0·35 (0·29–0·41) 0·37 (0·31–0·44) <0·0001
Excluding people in the least fit category (≤5·0 MET) who died within 2 years of follow-up (n=9797)
≤5·0 MET 1968 804 (41%) 1 1 1 ··
5·1–7·0 MET 2662 661 (25%) 0·64 (0·57–0·70) 0·70 (0·44–0·56) 0·70 (0·63–0·77) <0·0001
7·1–9·0 MET 2997 400 (13%) 0·35 (0·31–0·40) 0·50 (0·44–0·56) 0·50 (0·45–0·57) <0·0001
>9·0 MET 2170 207 (10%) 0·22 (0·19–0·26) 0·35 (0·30–0·41) 0·38 (0·32–0·44) <0·0001
HR=hazard ratio. MET=metabolic equivalents. *Adjusted for age, body-mass index, ethnic origin, sex, β blockers,
calcium channel blockers, angiotensin-converting enzymes, angiotensin receptor blockers, aspirin, diuretics, history of
smoking, history of cardiovascular disease, hypertension, and diabetes mellitus. †For fully adjusted HR.
Table 3: Mortality risk by fitness
Articles
4 www.thelancet.com Published online November 28, 2012 http://dx.doi.org/10.1016/S0140-6736(12)61426-3
Results
We assessed 20 023 people for eligibility. We enrolled
10 043 veterans (9700 men and 343 women). 5192 were
AfricanAmerican (mean age 57·8 years, SD 10·7),
4425 were white (mean age 59·5 years, SD 10·9), and
426 were other (mean age 57·6 years, SD 11). Median
followup was 10·0 years (IQR 6·0–14·2); providing
105 334 personyears. 2318 (23·1%) patients died (no data
were missing), with an average yearly mortality of
22 deaths per 1000 personyears (95% CI 13–31). Patients
not treated with statins had significantly higher mortality
than did patients treated with statins (1386/4997 vs
935/5046 died; 27·7% vs 18·5%; p<0·0001). No interaction
existed between site and METs (p=0·66) or ethnic origin
and METs (p=0·40).
Individuals taking statins tended to be older and had a
higher bodymass index and lower exercise capacity than
those not taking statins (table 1). The prevalences of
cardiovascular disease, smoking, hypertension, type 2
diabetes mellitus, and use of β blockers, calcium chan
nel blockers, angiotensinconverting enzyme inhibitors,
angiotensin receptor blockers, diuretics, and aspirin
were also higher in patients taking statins than in those
not taking statins. Family history of cardiovascular
disease, and use of nitrates or vasodilators was higher in
patients not taking statins than in those taking statins.
Median duration of statin treatment was 70 months.
For patients not treated with statins, median duration
between the first and final lipid assessments was
51 months. Patients treated with statins had a significantly
greater reduction in total cholesterol, LDLcholesterol,
HDLcholesterol, and triglycerides than did those not
taking statins (table 2).
Significant predictors of allcause mortality were: age
(HR 1·05, 95% CI 1·04–1·05; p<0·0001), hypertension
(1·22, 1·12–1·34; p<0·0001), smoking (1·26, 1·15–1·39;
p<0·0001); type 2 diabetes (1·39, 1·27–1·51; p<0·0001),
bodymass index (0·98, 0·97–0·99; p=0·001), and exercise
capacity. For each 1 MET increase in exercise capacity,
adjusted mortality decreased (HR 0·88, 95% CI 0·86–0·89;
p<0·0001). Adjusted mortality risk was also significantly
lower for patients taking statins than those not taking
statins (0·65, 0·59–0·71; p<0·0001).
The effect of exercise capacity was stronger in the statin
treatment group. For each 1 MET increase in exercise
capacity, the HR for adjusted mortality was 0·83 (95% CI
0·80–0·87; p<0·0001) for those taking statins compared
with 0·89 (95% CI 0·87–0·91; p<0·0001) for those not
taking statins. The interaction between fitness categories
and statin treatment also significantly affected mortality
risk (p=0·007).
Table 3 and the figure show mortality risk across fitness
categories. For the entire cohort in the final adjusted
model, mortality risk progressively decreased as exercise
capacity increased (table 3). Similarly, the adjusted risks for
individuals treated with statins were progressively lower
with increased exercise capacity.
To account for the possibility that the higher mortality
rates in the low fitness categories were caused by
underlying diseases (such as cachexia or musculoskeletal
or peripheral vascular conditions) and not low fitness per
se (reverse causality), patients who died within the first
2 years of followup (n=246) were excluded and the analyses
were repeated. The association between exercise capacity
and mortality risk remained (p<0·0001 for all compa