women, mortality improvements are likely to remain slower than males for the next decade or so. Similarly, life expectancy in Japan can be expected to improve less rapidly than it otherwise would because of the rapid increase in the prevalence of smoking in that country. On the other hand, the United States has been in the vanguard of a global obesity epidemic, and obesity also appears to be an important contributor to the shortfall in life expectancy in the United States. If the obesity trend in the United States continues, it may offset the longevity improvements expected from reductions in smoking. However, recent data on obesity for the United States suggest that its prevalence has leveled off, and some studies indicate that the mortality risk associated with obesity has declined. The interplay between obesity levels and obesity risks bears watching as an important factor in future longevity trends in the United States.
While the panel believes it made progress in identifying some of the main factors that have been driving differences in life expectancy among wealthy countries, it also identified many research gaps. With respect to behavioral risk factors, a reliable marker of the damage from smoking exists—mortality rates from lung cancer. No such clear-cut and widely available marker has been identified for obesity, physical inactivity, stress, lack of social integration, or the other risks considered in this report. Furthermore, evaluation of the importance of these risks is based primarily on observational studies that follow forward people with differing levels of exposure. These studies are subject to many biases, especially those associated with omitted variables, self-selection into categories, and reverse causation. Without randomized controlled trials, it is difficult to overcome these problems that plague observational studies. While it is sometimes difficult, expensive, and ethically challenging to alter individual behavior, there is no perfect substitute for such trials. On the other hand, studies that take advantage of natural experiments, such as increased cigarette taxes or a dramatic change in the use of hormone therapy, can sometimes serve as valuable supplements to randomized controlled trials.
The panel concluded that a history of heavy smoking and current levels of obesity are playing a substantial role in the relatively poor longevity performance of the United States. Yet these behaviors are products of a broad social and economic context encompassing, for example, a level of affluence that supports large numbers of automobiles, low taxes on gasoline, and dispersed residences and workplaces that encourage driving; a climate and soil in part of the country that are conducive to growing tobacco; a productive agricultural sector that produces inexpensive foods; and a public health system that is highly dispersed and thus heavily dependent on regional rather