Page 50 - IJPS-1-1

Earlier and more rapid ageing: Does nutrition contribute?

over males because women were comparatively less affected by the sweeping social
changes that were taking place in those years (notably in smoking and alcohol consump-
tion) and by war. Our analyses showed that ageing started earlier and was more rapid in
recent cohorts in all the seven countries. We do not yet know why this happened. However,
we noted that this rather counter-intuitive result is consistent with the outcomes of about
30 years of experiments on calorie restriction (CR) imposed on animals and we argued that
the changing dietary regime of humans in the past century or so may have played a part in
this evolution.
2. Data and Methods

The first studies on mortality acceleration assumed that it would begin around sexual ma-
turity, i.e., at about the age of 12 (Olshansky and Carnes, 1997), in line with the most im-
portant macro-theories of ageing which considered senescence a consequence of the pro-
gressive weakening of natural selection after puberty (Hamilton, 1966). On the other hand,
empirical findings pointed in another direction. In hunter-gatherer populations for instance,
the onset of mortality acceleration was estimated to occur between 40 and 50 years (Gur-
ven and Kaplan, 2007; Gurven, Kaplan and Supa, 2007; Hill, Hurtado and Walker, 2007).
Salinari and De Santis (2015) also identified a late ageing onset between 30 and 50 years
for cohorts born in the nineteenth and twentieth century in fourteen different countries
(earlier in more recent cohorts).
These findings seem to corroborate the idea that the (unobservable) physiological
process of ageing does not need to translate into an immediate rise in the (observable)
death rate with age because cells can absorb and partly repair molecular damage as long as
it remains below a given threshold (Franceschi et al., 2007; Kirkwood and Austad, 2000).
Therefore, a possible way of indirectly evaluating the pace of ageing is to observe if and
how its onset evolved over time, in which an earlier occurrence, i.e., an earlier arrival at
the “dysfunctional threshold” may signal more rapid ageing.To estimate the onset of mor-
tality acceleration, we employed the methodology proposed by Bai (2010) for the identifi-
cation of common breaks in panel data. For the sake of simplicity, we assumed that the
mortality of a group of C homogeneous cohorts (for instance the ten Swedish cohorts born
between 1890 and 1899) is observed between 25 and 75 years and that ageing starts within
this interval, at an unknown age k 0, which means that the force of mortality µ x is approx-
imately constant up to age x ≤ k 0 and increases thereafter. We also assumed that all the in-
dividuals of a given cohort share the same age at the onset of mortality acceleration. In a
more realistic but complex scenario, this acceleration is admitted to start at different ages
but even in this case, the same methodology can be applied using the mean age at the be-
ginning of the process (Salinari and De Santis, 2015). This can be formalized as follows:
µ , c x α = c ε + , c x x = 25,26, ,k (1)
0
µ , c x = α c e β x + ε , c x x = k + 0 1, ,75
where c denotes a cohort in the group, x stands for age, α is the mortality experienced
c
before the onset of mortality acceleration, β is the rate of ageing and ε , c x is the error
term. Bai’s technique is insensitive to the distribution of errors, which we therefore disre-
garded in the following. In all cases, it is possible to show that the error terms in equations
(1)–(3) are normally distributed (Brillinger, 1986; Horiuchi and Wilmoth, 1998; Salinari
and De Santis, 2014; 2015).
After logarithmic transformation (Figure 1(A)) we arrived at:
ln µ , c x lnα = c ν + , c x x = 25,26, ,k (2)

0
International Journal of Population Studies | 2015, Volume 1, Issue 1 44