Back in the Fight Aging! archives, you'll find a post on breaking out historical data on increases in human longevity into two components: firstly an increase in the average years lived, and secondly a reduction in early mortality - that more people are reaching ever closer to the average. This second statistical behavior is often presented as compression of morbidity, with the goal being to reduce the time spent in ill health at the end of life.
There is some debate over whether compression of morbidity is in any way a realistic or even useful goal for medical science, as opposed to aiming for increased human longevity through repair and reduction in the ongoing damage that causes aging. If you consider aging in terms of reliability theory, for example, it seems dubious that one could engineer the machineries of human life to last a set time and then fall apart very rapidly at the end - at least not without deliberately making it fall apart at the end. If all you are doing is consistently removing damage, then you extend the length of life, but don't do much about the time taken to fall apart when you stop repairing damage.
In any case, here is a recent paper that revisits this structural decomposition of increased longevity. The researchers here suggest that it is longevity, not compression of morbidity, that is the important factor.
In low-mortality countries, life expectancy is increasing steadily. This increase can be disentangled into two separate components: the delayed incidence of death (i.e. the rectangularization of the survival curve) and the shift of maximal age at death to the right (i.e. the extension of longevity).
We studied the secular increase of life expectancy at age 50 in nine European countries between 1922 and 2006. The respective contributions of rectangularization and longevity to increasing life expectancy are quantified with a specific tool.
For men, an acceleration of rectangularization was observed in the 1980s in all nine countries, whereas a deceleration occurred among women in six countries in the 1960s. These diverging trends are likely to reflect the gender-specific trends in smoking. As for longevity, the extension was steady from 1922 in both genders in almost all countries. The gain of years due to longevity extension exceeded the gain due to rectangularization. This predominance over rectangularization was still observed during the most recent decades.
Disentangling life expectancy into components offers new insights into the underlying mechanisms and possible determinants. Rectangularization mainly reflects the secular changes of the known determinants of early mortality, including smoking. Explaining the increase of maximal age at death is a more complex challenge. It might be related to slow and lifelong changes in the socio-economic environment and lifestyles as well as population composition. The still increasing longevity does not suggest that we are approaching any upper limit of human longevity.
There are two ways to think of upper limits on human longevity. One is that there may exist a distinct process of damage accumulation that eventually claims even the most hardy survivors, and which is not yet greatly affected by modern medicine. One candidate is TTR amyloidosis, based on autopsies of supercentenarians.
The second way of looking at limits on longevity is that there are no limits. Or rather, the only limits are those imposed by the lack of biotechnologies to circumvent them. Heart disease used to be a death sentence, for example, and now it is not. The same goes for all of the low-level mechanisms that drive aging and create conditions like heart disease - we are limited because, while we know how to go about repairing these forms of damage, we do not yet have the means in hand to do so.