In this issue, Kelley et al1 apply meta-analysis to assess the impact of aerobic exercise on lipid and lipoproteins. During the past 30 years, there have been many reports on exercise's effects on lipoproteins, particularly high-density lipoprotein (HDL) cholesterol, and it is a welcome contribution when these are summarized systematically and quantitatively. Most original research include discussions of earlier articles, but these are frequently written to support the necessity for publishing the authors' own findings, where studies consistent with the authors' results lend credibility to their findings, and those that differ provide the rationale for publication. Well-crafted discussions often include a balance of both but make no pretense of being comprehensive or definitive, that being the domain of reviews.

Reviews may take the form of sagacious opinions of senior researchers or meta-analysis by the more quantitatively inclined. To Kelley et al's credit, the authorship of their article includes both. There are strengths and weaknesses to classic reviews and meta-analyses. More traditional reviews tend to rely on the larger, better known studies. However, most studies, even the large ones, have only marginally adequate sample sizes. Kelley et al's analyses include HDL-cholesterol on 637 subjects from 6 studies, which is 3-fold as many subjects as the largest single study they cite, yet this still yielded less than 90% power to detect a significant exercise effect. The meta-analysis enhances statistical power by pooling over multiple studies. However, notwithstanding analytic evidence to the contrary, we must assume that the published record is a biased representation of the relevant research. Only in the case of large, expensive, randomized controlled clinical trials will the author's enthusiasm to write, the reviewer's enthusiasm to recommend, and the journal's enthusiasm to publish not be influenced by the significance or insignificance of the finding.

Training studies are special because they prove that exercise causes beneficial lipoprotein changes. Unlike cross-sectional reports, training studies are not affected by self-selection bias. Specifically, although many cross-sectional reports show that vigorously active men and women have higher HDL-cholesterol and lower triglycerides than do their more sedentary counterparts, they cannot distinguish the effects of exercise on lipoproteins from the effects of lipoproteins on exercise. It is now clear that cross-sectional studies overestimate the effects of vigorous exercise on HDL-cholesterol because HDL is predictive of exercise. Two separate studies show that sedentary men with initially higher baseline HDL-cholesterol run further after 1 year of training than those starting with lower baseline HDL-cholesterol.2,3 High HDL-cholesterol concentrations have also been reported in sedentary men whose only associations with exercise were their having identical twins who ran an average of 35 miles per week.4 The twin pairs were selected to be discordant for exercise. The active male twins had high HDL-cholesterol, averaging 57.4 mg/dL, whereas their sedentary brothers also had a relatively high average HDL-cholesterol of 52 mg/dL, which falls within the top tertile of adult men.5 Thus, solely the genetic predilection to run, as demonstrated by their more active brothers, conferred high HDL-cholesterol. High HDL may identify individuals genetically endowed with a high proportion of slow-twitch red muscle fibers.3 These fibers are more adaptive to endurance exercise and are relatively enriched with lipoprotein lipase, an enzyme that promotes higher HDL.3

The importance of training studies is evident from the magnitude of self-selection bias that affects cross-sectional comparisons. In the twin study described above,4 the sedentary twins can be viewed as the genetic controls for the active twins and can be used to estimate the effect of running on HDL-cholesterol while eliminating self-selection associated with genes. Dividing the average 5.5 mg/dL difference in the twins' HDL-cholesterol by their 35.7 mile per week difference in mean running distance gives 0.15 mg/dL as the per mile running effect after eliminating genetic differences. Elsewhere, we have shown that HDL-cholesterol increases (mathematically) 0.23 mg/dL per mile run when high- and low-mileage runners are compared.6 This suggests that 36% of the HDL-cholesterol difference between high- and low-mileage runners may be caused by their genetic differences rather than by exercise. However, most exercise studies compare active with totally sedentary men and women, whose genetic differences may be greater still. The presumably sedentary men of the Third National Health and Nutrition Examination Survey III had an HDL-cholesterol of 45.7 mg/dL.7 Dividing the HDL-cholesterol difference between these men and the running twins by 35.7 miles per week would erroneously suggest an even greater exercise effect of 0.30 mg/dL per mile run, which is twice the twins' estimate. Estimates of the exercise effect from training studies are also within the framework of a consistent genetic background. In addition, training studies control for environmental factors that may differ within identical twin pairs while remaining constant within individuals before and after training. Thus, unbiased estimates of exercise-induced increases in HDL-cholesterol derived from training studies will probably be only one-half as large as the corresponding biased cross-sectional estimates.

The principle of HDL elevation with exercise is well established and is endorsed by the Third National Cholesterol Education Program for the treatment of metabolic syndrome.8 Evidence for the benefit of exercise for treating dislipoproteinemia is largely inferred from studies in generally healthy individuals. We now know that exercise-induced increases in HDL-cholesterol are greatest in men with high baseline HDL and least in men with low baseline HDL. One year of exercise training produced increased HDL-cholesterol levels that were 3-fold greater in men who started with normal-to-high HDL-cholesterol (7.0 mg/dL) and more than 2-fold greater in men with intermediate HDL-cholesterol (4.9mg/dL) than in men with initially low HDL-cholesterol (2.3 mg/dL).3 Moreover, there is a tendency for runners who have an identical sedentary twin with low HDL-cholesterol to show less of an HDL-cholesterol difference from their sedentary twin than those with a sedentary twin with high HDL-cholesterol.4 Kelley et al's meta-analysis focuses on individuals with preexisting cardiovascular disease. Their low baseline HDL-cholesterol is not unexpected, given their condition, and proof that their HDL-cholesterol levels increase with vigorous exercise is an important contribution.

In their discussion, the authors correctly acknowledge that modest reductions in low-density lipoprotein (LDL)-cholesterol may belie more substantial alterations in LDL subclasses. Concomitant increases in HDL-cholesterol and decreases in plasma triglycerides are indicative of a shift in the distribution of LDL from smaller, denser particles to larger, more buoyant particles.9 Persons with a preponderance of smaller, denser particles are at substantially greater risk for cardiovascular disease than are those with predominantly larger, more buoyant particles.9 Long-distance runners have lower plasma concentrations of the smaller, denser particles than do sedentary men.10 Other studies show significant mean reductions in the plasma concentration of small LDL in sedentary men randomized into long-term exercise training when compared with mean changes in sedentary controls.11 In contrast, plasma levels of larger, more buoyant LDL do not appear to differ between runners and sedentary men10 and do not appear to change significantly during exercise training.11

The lipoprotein changes induced by exercise are significant, albeit modest, relative to pharmacological treatment. However, these are in addition to other known benefits of exercise, including lowered blood pressure, decreased insulin resistance, and increased cardiovascular fitness. The authors' report of a strong correlation between changes in body mass index and changes in HDL-cholesterol is reminiscent of an earlier debate on whether exercise-induced changes in lipoproteins were caused by the metabolic processes associated with fat loss or by enhanced musculature.11 Although this remains an important unresolved topic, it may not be particularly germane to the practical treatment of dyslipoproteinemia in high-risk individuals. The findings of Kelley, Kelley, and Franklin strengthen the evidence pointing to the prescription of exercise as part of the clinical treatment of patients with cardiovascular disease.

References