1:
2: \documentstyle[aps,preprint]{revtex}
3: \begin{document}
4:
5: \draft
6: \title{Phenomenological Theory of Survival}
7:
8: \author{Mark Ya. Azbel'}
9:
10: \address{School of Physics and Astronomy,Tel Aviv University,\\
11: Ramat Aviv, 69978 Tel Aviv, Israel}
12: \date{\today}
13: \maketitle
14:
15: \begin{abstract}
16: Theoretical analysis proves that human survivability is dominated by an
17: unusual physical, rather than biological, mechanism, which yields an exact law.
18: The law agrees with all experimental data, but, contrary to existing theories,
19: it is the same for an entire species, i.e., it is independent of the population,
20: its phenotypes, environment and history. The law implies that the survivability
21: changes with environment via phase transitions, which are simultaneous for all
22: generations. They
23: allow for a rapid (within few percent of the life span) and significant
24: increase in the life expectancy even above its value at a much earlier age.
25:
26: \end{abstract}
27: \newpage
28: \narrowtext
29: Mortality is one of the most universal and important phenomena in biology.
30: Human mortality is extensively studied over two centuries \cite{1,2,3,4,5} and is arguably
31: the best statistically quantified biological phenomenon. There exist several
32: evolutionary theories of aging \cite{6} (the first one \cite{7} is only 50 years old).
33: Yet, C. Franceschi (in ref.\cite{8}) notes: "Longevity is a trait with some
34: peculiarities because of the unnecessary nature of aging". This paper
35: proves that, in contrast to all existing theories, human mortality is dominated
36: by an unusual physical (rather than biological) mechanism. Presumably, the same
37: mechanism dominates mortality of laboratory fly populations.
38:
39: I study a large
40: and diverse amount of over 150,000 data points from 1528 period life tables,
41: for 3 races on 4 continents in 16 countries during over a century of their
42: history, 36 cases total (here and on ''a case'' denotes the population of a
43: given sex in a given country at any time in its history; in the USA white,
44: black and total populations are considered as separate cases). Each period
45: life table presents in any given case the probability $l_x$ to survive to any
46: (yearly) age x in a given calendar year. To escape any arbitrary scaling or
47: adjustment, I use life table variables only.
48:
49: The survivability $l_{x}$ depends on age $x$, calendar year $t$ and a complete
50: set $A$ of all parameters, comprehensively describing a considered population,
51: i.e. its phenotypes and their heterogeneous social, medical, dietary, etc. conditions during the
52: previous $x$ years (from $t$ - $x$ till $t$). So,$l_{x}=l_{x}(t,A)$. Mortality rate,
53: and thus its statistics and accuracy, are low in early age. The number of
54: survivors, and thus accuracy, decrease in old age. So, middle age
55: survivability, e.g., $l _{40}$, is determined more accurately. Present
56: $l_{x}(t,A)$ and $l_{40}(t,A)$ from the same life table as the ordinate and
57: abscissa of a point. According to Fig. 1, for different races, countries, continents,
58: histories $l_x$ is dominated
59: by the function $l_x(l_{40})$, which depends only on $l_{40}$, i.e. invariant to $A$ and
60: is therefore the same for an entire species. The deviation of $l_x$ from this
61: function depends on $A$ (and thus on population phenotypes, living conditions,
62: history). It is relatively small. From now on I denote $l_x(l_{40})$ as "the invariant
63: survivability" and, unless specifically stated otherwise, consider this (dominant)
64: fraction of survivability only.
65:
66: By Fig. 1, at any age and in every case, $l _{40}$
67: significantly changes with living conditions. Thus, since living conditions
68: of different groups in a given case are different, these groups may have
69: different probabilities $l_{x}^{\prime }$ to survive to $x$ years (the value
70: of $l_{x}$ in the population is the average $l_{x}^{\prime }$; $0\leq
71: l_{x},l_{x}^{\prime }\leq 1$). Suppose the probability density of a given $%
72: l_{40}^{\prime }$\ in the population is c($l _{40}$,$l_{40}^{\prime }$%
73: ), i.e.
74:
75: \begin{equation}
76: \int\limits_{0}^{1}c(l_{40},l_{40}^{\prime })dl_{40}^{\prime }=1,\qquad
77: l_{x}(l_{40})=\int\limits_{0}^{1}l_{x}^{\prime }\cdot
78: c(l_{40},l_{40}^{\prime })dl_{40}^{\prime } \label{1}
79: \end{equation}
80:
81: By Eq. \ref{1}, $0\leq \min (l_{40}^{\prime })\leq l_{40}\leq \max
82: l_{40}^{\prime }$.So, $l_{40}^{\prime }=0$ when $l_{40}^{{}}$ = 0 and $%
83: l_{40}^{\prime }=1$ when $l_{40}=1$. Thus, invariance, which allows one to
84: introduce c and which yields Eq. \ref{1}, implies that any population is
85: homogeneous in $l_{40}^{\prime }$ at the distribution boundaries. In virtue
86: of invariance, the dependence of $l_{x}^{\prime }$\ on $l_{40}^{\prime }$\
87: is the same as the dependence $l_{x}(l_{40})$, i.e., $l_{x}^{\prime
88: }=l_{x}(l_{40}^{\prime }).$ This, by Eq. \ref{1}, yields a remarkable
89: symmetry of the invariant survivability to the transformations, specified by
90: the function $c$ of two variables:
91:
92: \begin{equation}
93: l_{x}\left[ \int\limits_{0}^{1}l_{40}^{\prime }\cdot c(l_{40},l_{40}^{\prime
94: })dl_{40}^{\prime }\right] =\int\limits_{0}^{1}l_{x}(l_{40}^{\prime })\cdot
95: c(l_{40},l_{40}^{\prime })dl_{40}^{\prime } \label{2}
96: \end{equation}
97: Such invariance is sufficiently restrictive to mathematically accurately
98: determine $l_x(l_{40})$ (as an invariant of survivability dynamics with respect
99: to any population and environmental changes). Indeed,
100: in a general case the function $l_{x}(l_{40})$, which yields Eq. \ref{2},
101: must be linear. A special case of $c(l_{40},l_{40}^{\prime })=c^{\prime }\cdot \delta
102: (l_{40}^{\prime }-l^{\prime })+c^{"}\cdot \delta (l_{40}^{\prime }-l^{"})$%
103: yields $c^{\prime }+c^{"}=1$ by Eq.\ref{1}, and $l_{x}(l_{40})=l_{x}(c^{%
104: \prime }\cdot l^{\prime }+c^{"}\cdot l^{"})=c^{\prime \cdot }l_{x}(l^{\prime
105: })+c^{"}\cdot l_{x}(l^{"})$ by Eq. \ref{2}. The latter equation implies
106: $dl_{x}/dl_{40}=dl_{x}/dl_{x}^{\prime }=$const. On the other hand, a linear
107: $l_{x}(l_{40})$ yields Eq. \ref{2} with any arbitrary $c$ in virtue of Eq. \ref{1}.
108: Similarly, if an entire population is distributed in a certain interval of $%
109: l_{40}$, then $l_{x}(l_{40})$ is linear in this interval, and the population
110: is homogeneous at its ends (as it was in the previous case).
111: Correspondingly, if any population is homogeneous at certain $l_{40}$
112: points only, then an entire population is distributed within one of the
113: intervals $\lambda _{s}<l_{40}<\lambda _{s+1}$ (s = 0,1 is its ordinal
114: number):
115: \begin{equation}
116: \bigskip \int\limits_{\lambda _{s}}^{\lambda _{s+1}}c(l_{40},l_{40}^{\prime
117: })dl_{40}^{\prime }=1;\qquad l_{40}=\int\limits_{\lambda _{s}}^{\lambda
118: _{s+1}}l_{40}^{\prime }\cdot c(l_{40},l_{40}^{\prime })dl_{40}^{\prime }
119: \label{3}
120: \end{equation}
121: and $l_{x}(l_{40})$ is linear within these intervals. When $l_{40}=\lambda
122: _{s+1}$, an entire population of all ages moves to the next interval.
123: Clearly, such piecewise $l_{x}(l_{40})$ satisfies Eq.\ref{2}. When $\lambda
124: _{0}=0,\lambda _{1}=1,$ this case reduces to the previous one. If an
125: infinite number of edge points in Eq.\ref{3} condenses and forms a
126: continuous interval (where any population is homogeneous), then the
127: invariant $l_{x}(l_{40})$ in this interval remains undetermined.
128: Of course, the possibility to distinguish this case from the case of
129: a large number of short linear segments is limited by the disregarded
130: non-invariant contributions to survivability, but demographic studies
131: prove that survivability is heterogeneous and exclude the former case
132: (of a population which is homogeneous in a finite interval). Thus, in a
133: general case
134: \begin{equation}
135: l_{x}=R_{x}^{(s)}l_{40}+a_{x}^{(s)},\qquad \lambda _{s}\leq l_{40}\leq
136: \lambda _{s+1}.
137: \end{equation}
138: So, except for the number of linear segments, the
139: very existence of the invariant survivability allows one to establish its exact
140: law, without any experiments, approximations, and assumptions.\\
141: Equation (4) is the implication of invariance only. Such invariance was also
142: demonstrated \cite{9} for medfly and fruitfly
143: families whose different populations were extensively studied in different
144: conditions. Fly statistics is rather low, thus their $l_{x}$ was studied as
145: a function of the life expectancy at birth (which averages $l _{x}$
146: over different ages and is therefore more smooth). The study was based on
147: fly populations hatched the same day. However, since in most cases fruitflies
148: were kept in stationary conditions, it suggests that the invariant
149: survivability law is general for humans and flies.
150:
151: Piecewise linear law (4) and its age independent invariant intersections are
152: the main predictions of this paper. They are verified with all experimental
153: data - see, e.g., $l_{1}$, $l_{60}$,$l_{80}$ vs $l_{40}$ in Fig.2.
154: Piecewise linear law agrees with (but has never been suggested in)
155: demographic approximations \cite{3}. Slope jumps in Eq.(4) and Fig.2 are consistent with
156: ref \cite{5}. However, in contrast to the qualitative observations in ref.\cite{5}, accurate
157: Eq.(4) yield quantitative predictions. Since all $l_{x}$ belong to the same
158: calendar year, their slope jumps are simultaneous for all generations.
159: Indeed, although the intersections in Fig.2 are reached at different times
160: in different countries for different sexes, they are simultaneous (to the
161: invariance accuracy) at different ages, and (in $l_{40}$) for different populations.
162: The invariant law allows for the
163: prediction of intersections. For instance, the extrapolation of $l_{80}$
164: beyond $\lambda _{1}=0.95$ in Fig. 1 yields $l_{40}=1$ (while $l _{80}$
165: is finite), i.e. implies no deceased until $40y$. This is hardly possible, and
166: suggests a crossover, which is indeed seen in Fig.2 at $\lambda _{2}=0.97$
167: The extrapolations of $l_{60}$ and $ l_{80}$ to $l_{40}<0.94$ yield
168: $l_{60}=0$ and $l_{80}=0$\ , i.e., no survivors beyond $60y$ (when $l_{40}=0.27$)
169: and $80y$ (when $l_{40}=0.42$). This is never true, and suggests an
170: intersection at $0.4<l_{40}<0.7$.
171:
172: To elucidate the nature of the invariant law, present Eq. (4) in a different
173: form:
174:
175: \begin{equation}
176: \bigskip l_{x}=c_{x}l_{x}^{(s)}+(1-c_{s})l_{x}^{(s+1)}~\quad when~\quad
177: l_{x}^{(s)}\leq l_{x}\leq l_{x}^{(s+1)}
178: \end{equation}
179:
180: Here $l_{40}^{(s)}=\lambda _{s};l_{x}^{(s)}=l_{x}(\lambda _{s})$ \ from
181: Eq.(4); $c_s$ reduces to $l_{x}^{(s)}$\ and $l_{40}$ [thus, by Eq. (3), to c].
182: Equation (5) accurately separates ''nature'' and ''nurture'' in survival.
183: ''Nature'' reduces to the fixed set of the intersection survivabilities .
184: The set depends only on age and is invariant, i.e., independent of
185: phenotypes (and thus of a specific DNA sequence at least in an entire
186: species), their living conditions and life history. The dependence of the
187: set on age is not determined by invariance, but invariance implies that it
188: is the same for at least an entire species. (Moreover, it scales onto the
189: same functions for species as remote as humans and flies \cite{9}). Thus, it
190: does not change at least as long as the species does not evolve into a
191: different species (demographic data in Fig.2 verify it for 100-150 years).
192: So, the set must be inheritable. Such set reminds of the body temperature
193: (which in any living conditions is the same with few percent accuracy) of an
194: entire class of birds and a subclass of placental mammals. Presumably, both
195: the set and the body temperature are genetically determined, and independent
196: of a specific DNA sequence. But the set, unlike the body temperature,
197: strongly depends on age. ''Nurture'' distributes the survivability $l_{x}$
198: at a given age between two adjacent intersection survivabilities. The
199: concentration $c_{s}(0\leq c_{s}\leq 1)$ is age independent (and may be
200: related to, e.g., $l_{40}$). Thus, survivability follows environment (in
201: particular, a new intersection survivability emerges) simultaneously for all
202: generations.
203:
204: Equation (5) relates $c_{s}$, and thus $l_{x},$ to $l_{1}$. Since $l_{x}$ is the survival
205: probability in the same calendar year as $l_{1}$, so, by Eq. (5), the
206: survivability accurately and rapidly follows the change in
207: environment according to the value of $l_{1}$ (i.e., the infant mortality $%
208: q_{0}=1-l_{1}$), which is established in less than two years. Indeed, whatever the difference in
209: environmental factors is, close values of infant mortality imply very close
210: fractions of deceased at any given age in 1885 Swedish and 1947 Japanese
211: females, despite of their different races, continents, countries, and 62
212: year gap in their different history. Since $l_{1}$ depends on the
213: environment in a given year only, the invariant survivability $l_{x}$ is
214: reversible (an entire survival curve comes back when $q_{0}$ changes
215: non-monotonically and returns to its previous value), and statistically
216: independent of the life history during its $x$ years, despite country
217: specific, highly and irregularly changing, non-monotonic and non-stationary
218: living conditions. World wars and epidemics, e.g., flu in 1918 Europe,
219: significantly decrease $l_{x}$. (For instance, in 1915 the probability for a
220: French male to survive to $80y$ was 5 times less than in 1913, twice less than
221: in 1917, and 3.5 times less than for a 1861 Swedish male). Yet, they just
222: slightly shift the plots (mostly vertically, and relatively little) in
223: Fig.2. In a couple of years (which estimate the relaxation time at few
224: percents of the life span) all plots restore their invariant dependence,
225: i.e., the memory of the previous life history is erased. Accurate
226: reversibility of the invariant survivability does not decrease with aging,
227: even in old age. Unless such reversibility is related to some perfect
228: biological rehabilitation (which is hardly possible), it implies an
229: adiabatic change in a certain thermodynamic equilibrium. This is consistent
230: with its relaxation time (rapid compared to the life span, but enormous on a
231: microscopic scale). Equation (5) reduces $l_{x}$ to the fixed set of $%
232: l_{x}^{(s)}$. Since $l_{x}^{(s)}$ reversibly change into each other, they
233: are related to different equilibrium thermodynamic states of the same
234: system, i.e. to different phases. Thus, Eq. (5) relates the invariant
235: survivability to a certain phase equilibrium, and the jumps in Fig.1 slopes
236: to the emergence of a new phase. The phase concentration $c_{s}$, which
237: Eq.(5) reduces to, e.g., $l_{1}$, is independent of age. This suggests that
238: the age dependence of a \ $l_{x}^{(s)}$ is related to the difference in
239: survivabilities provided by the same phase at different ages. Quasi-equilibrium
240: phases in a living homeostatic being might be
241: related to a meso- (e.g., a cell) or microscopic (e.g., DNA
242: configuration) scale.
243:
244: While the existence of survivability phases is accurately proven, their
245: microscopic nature, as well as that of the parameter which determines their
246: concentrations, remains unknown (as ''units of heredity'' were to Mendel).
247: However, any phase equilibrium may be reversibly
248: manipulated. This means that human life expectancy may be rapidly (within
249: few years) reversed to its value at a much earlier age. (Note that in just 8
250: years from 1947 till 1955 the life expectancy of Japanese females increased
251: 26\% at birth, 15\% at $60y$, and 20\% at $80y$).
252:
253: Presumably, the change in phase concentrations may affect other than
254: survivability characteristics, in particular, aging and disease resistance.
255: Indeed, Eq.(5) is not violated even in Japan prior to 1949 (see Fig.1) and
256: in Finland from 1890 till 1940 (see Fig.2), although their mortality has a
257: strongly tubercular age pattern during this period.
258:
259: The suggested phase equilibrium nature of survivability implies that
260: non-equilibrium (in particular, sufficiently non-stationary and
261: heterogeneous conditions which depend on age, sex, social mobility,
262: immigration and other factors), may lead to more than two adjacent phases
263: and to hysteresis in the adjustment. The contribution of non-equilibrium
264: phases is most pronounced in old age, when the difference in $l_{x}^{(s)}$
265: is significantly higher. This agrees with Fig.2. Hysteresis yields small
266: jumps in $l_{x}$ and shifts of the intersections, which are indeed present
267: in certain cases. The accuracy of the invariant law estimates maximal
268: concentrations of ''extra'' phases. They are lower for females, presumably
269: suggesting lower heterogeneity of female populations.
270:
271: Survivability independence of life history implies no correlation between
272: early and old age invariant mortalities in a given cohort (born the same
273: year). This disagrees with evolutionary theories of aging \cite{6}.
274: Stochastic mutation accumulation \cite{7} theory is inconsistent with reversibility
275: and rapid accurate survivability change with environment. Optimal allocation
276: of metabolic resources (pleiotropy and
277: disposable soma theories) implies strong
278: correlation between survival in young and old age in populations born the same year. This
279: is inconsistent with the survivability independence of life history.
280: Reversibility is also inconsistent with mortality theories (thelomers,
281: oxygen consumption, free radicals, somatic mutations). Although natural
282: mortality in the wild is mostly due to extrinsic hazards, invariant
283: mortality, which dominates in species as remote as humans and flies (in
284: certain protected environment), and which rapidly, accurately and
285: simultaneously in all generations changes with environment, calls for
286: biological, evolutionary and microscopic physical theories.\\
287: These conclusions, as well as an unusual mechanism of mortality, are accurate
288: implications of the exact law, which agrees with all demographic approximations
289: and studies \cite{3,5}. For instance, the infant mortality is widely appreciated by
290: demographers as a sensitive barometer of environmental conditions. However,
291: demographic approximations are developed primarily as a useful tool of maximally
292: accurate estimation and forecast. They are often country and time specific, and
293: approximate specific demographic data better than Eq. (4), but they do not consider
294: nor care about the underlying general law and its mechanism. In contrast, my goal
295: is the exact law, albeit of the invariant mortality only (to discover inertia, one
296: must disregard friction!) and its mechanism, which yield new biological insights.
297:
298:
299: \begin{description}
300: \item[Acknowledgement]
301: \end{description}
302:
303: \vspace{0in}This paper was stimulated by enlightening discussions with
304: Profs. Y. Aharonov, S.M. Jazwinski, S. Horiuchi, D.S. Thaler, A. Libchaber,and
305: B. Tsirelson. I am grateful to Prof. S.
306: Horiuchi, S. Smallwood, A. Hanika, K. Anderson, S. Helgadottir, C. Capocci,
307: E. Waeyetens, N. White, U. Wegmueller, V. Kannisto, M. Nieminen, K. Deaves
308: for life tables; to I. Kolodnaya for crucial technical
309: assistance; A. Kobi for computer simulations, J. \& R. Meyerhoff chair for
310: financial support.
311:
312:
313: \begin{references}
314: \bibitem{1} Halley,E., Phil. Trans. Roy. Soc.{\bf 17}, 596 (1693);
315: Euler, L., Histoire de l'Academie Royale des Sciences et Belle- Lettres,
316: 144-164 (1760).
317:
318: \bibitem{2} Gompertz, B., Phil. Trans. Roy. Soc. {\bf A115}, 513 (1825)
319:
320: \bibitem{3} Coale, A.J., Demeny, P., Vaughan, B., Regional Model
321: Life Tables and Stable Populations, 2nd ed. (academic, NY, 1993); Lee, R.D.,
322: Carter, L.R., J. Am. Stat. Assoc. {\bf 87}, 659 (1993); Indirect Techniques for
323: Demographic Estimation, Manual X, pp.12-21, UN publication (1993), and refs.
324: therein.
325:
326:
327: \bibitem{4} Australian life tables (1999), Australian Government Actuary,
328: Canberra; Tables de Mortalit\'{e} (1999), Statistics Belgium, Brussels; Life
329: tables for Japan (1994), Ministry of Health and Welfare, Tokyo; Berkeley
330: Mortality Database, http://demog.berkeley.edu/wilmoth/mortality
331:
332: \bibitem{5} Kannisto, V. (1994), Development of oldest-old mortality,
333: 1950-1990, Odense, Denmark (Odense Univ. Press). Horiuchi, S., Wilmoth, J.R.
334: (1998). Demography {\bf 35} 4, 391-412. Wilmoth, J.R., Horiuchi, S. (1999),
335: Demography {\bf 36} (4), 475-495. Tuljapukar, S., Li, N., Boe, C. , Nature
336: {\bf 405}, 789-792 (2000). Horiuchi, S. in Health and Mortality: Issues of Global
337: Concern, 54 (1999), UN, NY, and refs. therein.
338:
339: \bibitem{6} Kirkwood, T.B.L., Austad, S.N. , Nature {\bf }, 233-238 (2000).
340: \bibitem{7} Medawar, P.B., An Unsolved Problem of Biology, Lewis, London, (1952).
341: \bibitem{8} Koenig, R. Science {\bf} 291, 2074 (2001).
342: \bibitem{9} Azbel, M. Ya., PNAS USA {\bf 96} 3303, 15368 (1999)
343:
344: \end{references}
345: \begin{figure}
346: \caption {Probabilities (vertical axis, upper, middle and lower curves correspondingly)
347: to survive to 1, 60, 80 years vs $l_{40}$ (horizontal axis) for females in 1909-1997
348: Australia, 1880-1998 Belgium, 1891-1996 Japan, 1861-1995 Sweden.
349: Swedish and Japanese females are denoted by full square, open square, and
350: full triangle, open triangle; others by full circle, open circle; open signs denote
351: 1914-1919; 1939-1947. Solid lines are linear regressions, which verify predicted
352: piecewise linearity. }
353: \label{f1}
354: \end{figure}
355:
356:
357: \begin{figure}
358: \caption {Same as in Fig. 1, but for all cases (males and females in 1909-1997
359: Australia, 1880-1998 Austria; 1880-1998 Belgium, 1950-1987 Canada and provinces;
360: 1851-1998 England and Whales; 1881-1998 Finland, 1898-1995 France; 1871-1994 Germany;
361: 1841-1998 Iceland; 1925-1992 Ireland; 1891-1996 Japan, 1846-1998 Norway; 1930-1992 Scotland;
362: 1861-1995 Sweden, 1878-1993 Switzerland; 1900-1997 USA, white, black, and total population.)
363: To amplify invariance and piecewise linear dependence, some of the linear segments are slightly
364: rotated and shifted. (This does not violate piecewise linearity.)}
365: \label{f2}
366: \end{figure}
367:
368: \end{document}
369: