Rising mortality from motor neurone disease in Sweden 1961-1990: the relative role of increased population life expectancy and environmental factors

Rising mortality from motor neurone disease in Sweden 1961-1990: the relative role of increased population life expectancy and environmental factors

Neilson S, Gunnarsson L-G, Robinson I . Rising mortality from motor neurone disease in Sweden 1961-1990: the relative role of increased population life expectancy and environmental factors. Acta Ncurol Scand 1994: 90: 150-159. 0 Munksgaard 1994.

Recent studies of mortality from motor neurone disease (MND) in Sweden have demonstrated rising levels of mortality from the disease, especially amongst older age groups. Case-control investigations have suggested that certain environmental factors are significantly related to variations in mortality from the disease, and are associated with a probable individual susccptibility to MND. This study applies an innovative epidemiological technique to longitudinal and cohort analysis of Swedish mortality from MND during the period 1961 to 1990. Survival modelling shows that a subpopulation susceptible to MND exists in Sweden, as has been demonstrated in other countries. The increased life expectancy of the Swedish population since 1961 has resulted in more of that susceptible population living to the ages at which MND is expressed, explaining the majority of the increase in mortality from the disease. However, environmental factors may play a role in accelerating the course of M N D and may affect the timing of death within the susceptible sub-population.

In the last 10 years national and regional mortality from M N D in Sweden has been extensively studied ( l ) , particularly in relation to its association with occupation (2, 3 ) , and occupational exposure to a range of potentially neurotoxic substances (4, 5) . These analyses have been set in the context of a demonstrable and major rise in mortality from the disease over the last three decades. In common with studies in other industrialised countries (6), the ex- amination of Swedish mortality from M N D has re- vealed an increasing preponderance of deaths in the older age groups, with the mean age at death in- creasing over the study period. It has been suggested (7) that much of the worldwide rise in M N D mor- tality is an artifact of improved diagnosis or record- ing and it is possible that some of the increase in recorded mortality from the disease in Sweden is due to more accurate diagnosis and recording of existing cases. However, the considerable size of the rise and its age distribution suggest that other factors are at least as important. Indeed the analysis of death cer- tification of M N D in Sweden suggests that the mor-

S. Neilson I , L.-G. GunnarssonL, I. Robinson ' ' The John Bevan MND Research Unit. Brunel. The University of West London, Uxbridge. England, * Department of Neurology. Orebro Medical Center Hospital, Sweden

Key words: amyotrophic lateral sclerosis; Sweden; mortality; Gompertzian analysis; survival model.

Stuart Neilson. The John Bevan MND Research Unit, Brunel, The University of West London, Uxbridge. Middx UB8 3PH, England.

Accepted for publication December 23. 1993

tality data are a reliable guide to the distribution of diagnosed cases of the disease (2).

An apparently more logical explanation for the pattern of increased mortality is the presence, or perhaps the increased potency, of one or more exo- genous factors acting on those who subsequently succumb to the disease (8-12). The role of environ- mental factors in appearing to explain some regional variations in mortality from MND, as well as some of the distinctive differences in case-controlled stud- ies has been a special feature of recent research ( 13- 18). With regard to Sweden, case-control studies have shown that exposures related to agricultural work might partly explain the regional variation found ( 3 ) , and that those exposed to welding or im- pregnating agents were more statistically liable to develop the disease (5 ) . The combination of a history of solvent exposure, a family history of neurodegen- erative or thyroid disease, and being male produced the highest risk ratio ( 5 ) . This interaction between environmental factors and heredity is of major ini- portance and the combination of endogenous and

150

Rising Swedish MND mortality: 1961-90

exogenous factors is a key to understanding the aeti- ology of the disease.

At a general level, therefore, if Swedish mortality data can be interrogated by techniques which allow an estimation of the relationship between endoge- nous and exogenous factors in mortality from MND, then further advances can be made in understand- ing the disease and its genesis. Certain mathemati- cal models of survival in human populations have properties which can discriminate between these factors. Of special significance for the investigation of the demographic structure and survival charac- teristics of conditions with the late age of onset and mortality characteristic of M N D is the relationship posited by the English actuary Benjamin Gompertz. In 1825 Gompertz ( 1 9) proposed a law of human mortality such that mortality rates rise by equal pro- portions in equal intervals of time. The mortality rate R , at age x is dependent on a (theoretical) rate at age zero, R,,, and the exponential slope a. This can be expressed as the exponential relationship shown in Eqn. 1, or(by taking natural logarithms ofboth sides) as the linear relationship shown in Eqn. 2:

R , = R,, e'" (1)

In R , = In R,, + ax (2)

Analyses of general mortality in many human and animal populations have indicated that the Gomp- ertzian relationship approaches the status of what has been described by some as a biological law. The exponential rise of mortality with age has also been observed in the case of mortality from certain dis- eases such as stroke (20), Parkinson's disease (21), and some cancers (22). However the application of the same technique to some other conditions such as Huntington's chorea (23) has demonstrated an ap- parent paradox, that is a peak of mortality followed by a relatively steep decline, rather than the expected continuing exponential increase with advancing age observed in general mortality. Huntington's chorea is known to be a genetically determined condition, and it can be shown that the observed pattern of mortality is typical of conditions arising in a sus- ceptible subpopulation. Due to the burden of Huntington's chorea mortality, in addition to all other causes of death, the susceptible subpopulation declines in size faster than the general population leading to an apparent decline in mortality rates at advanced ages. Elimination of the competing risk of general mortality (24) transforms the observed rates of mortality into underlying (i.e.: within subpopula- tion) rates, demonstrating clear Gompertzian dy- namics at all ages in Huntington's chorea and ad- ditionally providing an estimate of the size of the susceptible subpopulation. M N D mortality rates

exhibit remarkably similar patterns, indicative of a similarly predisposing factor, with a peak mortality rate in the sixth or seventh decade and declining rates thereafter (25). The mathematical model de- veloped for this analysis has been applied previously to M N D mortality in England and Wales both na- tionally (26) and regionally (27), as well as in Japan (28) and the United States (29), with the conclusion that life expectancy is the primary factor affecting observed M N D mortality rates (30).

The determination of the nature and extent of a subpopulation (genetically) susceptible to M N D would confirm the conclusions of a previous Swed- ish study ( 5 ) that predisposition is an important as- pect of the disease. However, it is improbable that genetic susceptibility to the disease could have in- creased sufficiently to account for the substantial rise in mortality that has occurred over a relatively short period, even if the origin of susceptibility is spontaneous mutation. Thus other possibilities, such as the increased expression of underlying and pre- existing susceptibilities, must be explored in order to sustain the earlier endogenous-exogenous model of the causation of MND. The key hypothesis in this respect lies in the fact that the median age of nior- tality from M N D occurs close to the mean life ex- pectancy of the general population in industrialised countries. I f overall life expectancy, and that in the susceptible subpopulation, rises by even a small margin, mortality from M N D would be expected to rise dramatically through the differential survival of the predisposed proportion of the population. By developing Gonipertzian techniques and applying them to Swedish mortality from M N D it is possible to estimate the nature and size of the susceptible subpopulation, to assess the relative role of chang- ing life expectancy in the rise in M N D mortality, and evaluate the contribution of environmental factors to the longitudinal pattern of mortality from the dis- ease.

Material and methods

A total of 4738 deaths in Sweden were coded with M N D as the underlying cause of death according to

and ICD9-335.2 (1987-90) (33) during the period 1961 to 1990. Deaths were categorised in the 5-year age-groups from 0-4, 5-9, etc., up to 85 and above. The base population in the same agc and sex group- ings was used to calculate age-specific M N D nior- tality rates for each year in the study. Mortality rates were calculated for each year from 1961 to 1990 for men, women and the total population. Mortality rates were also calculated by year of birth, in five- year intervals from 1880 to 1940, and subjected to the same analysis as the cross-sectional data.

ICD7-356 (1961-68) (3 l), ICD8-348 (1969-86) (32)

151

Neilson et al.

0.01 -

Table 1 Age-specific rates of MND by sex, total number of deaths and estimated size of the susceptible subpopulation (per 1000,000 persons born) in Sweden during the inter- vals 1961-70. 1971-80 and 1981-90

1

I I I I 1 I I 1

Rate (men) Rate (women) Deaths (totall

Age 1961-70 1971-80 1981-90 1961-70 1971-80 1981-90 196 1-70 1971-80 198 1-90

0-4 5-9

10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 8 5 t

All ages 0-54 5 5 t Subpop.

0.49 0.12 0.10 0.13 0.06 0.12 0.13 0.34 0.67 1.40 2.30 4.34 5.36 7.71 8.80 7.48 6.29 5.69

1.71 0.51 6.15

257.72

0.13 0.12 0.09 0.04 0.09 0.00 0.03 0.09 0.21 0.10 0.08 0.11 0.13 0.25 0.33 0.37 0.53 0.66 1.22 0.92 2.23 2.57 4.07 3.65 7.68 6.58 9.45 9.59

12.68 14.12 14.14 15.70 15.65 14.47 7.17 12.75

2.28 2.47 0.4 1 0.42 8.73 9.21

379.44 410.3

0.63 0.03 0.1 1 0.00 0.04 0.05 0.03 0.21 0.45 0.79 1.42 2.71 3.92 4.55 5.90 4.96 3.60 1.47

1.35 0.34 4.05

154.35

0.19 0.03 0.04 0.03 0.00 0.00 0.15 0.16 0.31 0.86 1.37 2.87 4.24 6.19 7.79 7.34 4.97 5.64

1.73 0.26 5.40

210.92

0.13 0.00 0.02 0.03 0.02 0.10 0.05 0.17 0.45 0.67 1.48 2.22 5.29 6.99 8.51

10.13 7.69 5.18

2.04 0.25 6.38

245.6

38 5 6 4 2 4 3

14 30 56 93

175 205 121 191 1 1 1 48 14

1,211 255 956 193.97

9 4 4 2 6 2 9

12 20 5 1 91

171 262 32 1 323 218 107 48

1,660 210

1,450 273.48

6 1 1 4 5 7 8

16 29 37 94

128 267 327 371 328 164 74

1,867 208

1,659 305.76

t ix . I

152

i; Observed rate of MND mortali t in Sweden in the 1960's, 70's and 0's

1961-1970 - 1971-1980 - 1981-1990 1 _ _ _ _ _ I 1

s)garithni of the obscrvcd rate of mortality from MND in Sweden during the intervals 1961-70, 1971-80 and 1981-90

Rising Swedish MND mortality: 1961-90

Recorded mortality from M N D has been shown to be extremely accurate (34, 35) and a recent study in Sweden demonstrated false positive reporting rates of only 4", and false-negative rates of only 5 % (2). The observed rates of M N D mortality and absolute numbers of deaths are shown in Table 1 for each of the three decades studied.

The survival function for Gompertzian mortality is given by Eqn. 3 (27), in which S , represents the proportion of the general population susceptible to M N D at any age s and S,, represents the initial susceptible proportion at birth. The observed rate of mortality from MND, * R , at age s, will therefore be determined both by the underlying rate of mortality within the subpopulation, R , , and by the proportion of susceptible individuals remaining in the popula- tion, S , , as shown in Eqn. 4 below.

S , = S,, exp [ R,,/a ( 1 - enx)] (3)

* R , = S , R , (4)

In practice the value of S,, is readily estimated by the integral of observed mortality across the lifespan, given by the area under the M N D mortality rate curve (Eqn. 5) . This estimate is independent of the assumption of a Gompertzian mortality rate distri- bution. When integrated from a lower age bound, the integral also estimates the number of susceptible individuals alive as a proportion of the total popu- lation alive at any age s (Eqn. 6):

By inserting Eqn. 6 into Eqn. 4 it is possible to calculate the underlying rate of M N D mortality di- rectly from division of the observed rate of mortal- ity by the estimated proportion of the remaining population susceptible to MND, as shown in Eqn. 7:

R , = *R,/I ,.'.*R, dx (7)

Therefore, if the mortality rate from M N D in Swe- den is Gompertzian and if M N D is restricted to an inherently susceptible subpopulation, then a plot of the logarithm of the underlying mortality R , against age will result in a straight line, which can be tested by linear regression. The underlying mortality rates from M N D were calculated for each year and re- gression was performed on the logarithm of the un- derlying rate against age over the age range 40-85 years.

The Strehler-Mildvan modification of the Gomp- ertz model ( 3 6 ) provides the possibility of a more

comprehensive analysis of the factors influencing ALS mortality. They propose a negative linear rela- tionship between the rate of exponential increase in mortality (TX) and the logarithm of the extrapolated mortality rate at age zero ( R J . This implies that a high initial mortality rate will be accompanied by a low rate of increase with age, and that mortality rate curves, even when drawn from populations with dif- fering life expectancies, will intersect at an age point determined by the rate of ageing ( B ) . Mortality rates at this age point will therefore be static and inde- pendent of variations in life expectancy. This rela- tionship was tested by linear regression of the (3 , log R,) pairs calculated from ALS mortality rates in each year studied using Eqn. 8.

a = B (log K - log R,) (8)

Table 2. Results of linear regression of the calculated inderlying rate of MNO mortality among men and women in Sweden, 1961-1990

Men Women

Year 109 Ro 1 R 2 1% Ro 1 R2

1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990

1961-70 1971-80 1981-90

Minimum Mean Maximum

0.680 0.770 1.154 1.01 1 0.801 0.326 0.529 0.550 0.727 0.166 0.258

-0.065 -0.383 -0.431 -0.016

0.348 0.149

-0.052 0.107 0.073 0.302

-0.083 -0.069 -0.485 -0.161 -0.015

0.190 0.276 0.369 0.372

0.757 0.06 1 0.113

-0.485 0.252 1.154

0.044 0.952 0.040 0.954 0.035 0.961 0.036 0.964 0.041 0.986 0.049 0.960 0.047 0.972 0.047 0.962 0.045 0.979 0.053 0.992 0.052 0.992 0.054 0.976 0.058 0.978 0.058 0.973 0.053 0.985 0.046 0.978 0.049 0.975 0.052 0.977 0.051 0.993 0.051 0.968 0.047 0.976 0.052 0.961 0.052 0.973 0.058 0.934 0.054 0.981 0.052 0.999 0.049 0.994 0.047 0.992 0.046 0.994 0.046 0.977

0.042 0.982 0.052 0.992 0.050 0.992

0.035 0.934 0.049 0.976 0.058 0.999

0.680 0.829 0.946 0.776 0.621 0.503 0.461 0.06 1 0.242 0.519 0.910 0.853 0.558 0.652 0.36 1 0.025 0.025

-0.364 -0.012

0.425 0.406 0.516 0.126

-0.085 -0.227 -0.426 -0.134

0.028 0.273

-0.137

0.581 0.336 0.105

-0.426 0.316 0.946

0.044 0.042 0.040 0.043 0.045 0.048 0.048 0.054 0.05 1 0.047 0.040 0.040 0.044 0.043 0.047 0.05 1 0.051 0.058 0.053 0.046 0.046 0.044 0.050 0.053 0.056 0.059 0.055 0.052 0.049 0.054

0.046 0.048 0.051

0.040 0.048 0.059

0.922 0.946 0.971 0.984 0.982 0.982 0.984 0.971 0.989 0.98 1 0.943 0.852 0.934 0.983 0.976 0.957 0.957 0.953 0.982 0.974 0.954 0.977 0.992 0.948 0.980 0.983 0.977 0.993 0.97 1 0.903

0.993 0.973 0.992

0.852 0.965 0.993

153

Neilson et al.

This modification of the Gompertz model allows the identification of an aetiopathogenic factor ( K ) , which can be considered to be the sum of all envi- ronmental causative influences upon ALS mortality, and which affects mortality rates at all ages by the same proportion. This is in contrast to the competi- tive effects of changing life expectancy, which de- crease early mortality and shift the age distribution of deaths to later ages. Thus, in an unchanging caus- ative environment and with a constant rate of age- ing, mortality rates at all ages are determined solely by Ro,

Results

The age-specific mortality rates from M N D in Swe- den among men and women in each decade are given in Table 1. Mortality rates in the oldest age groups have approximately doubled over the period under study, with evidence of a decline in the rates for those aged below 55 years. The summary groups show that whilst mortality has increased by 48% overall, the increase is accounted for entirely by a

rise of 54% in those aged 55 years and over whilst there has been a decrease of 21 % in the mortality rate amongst those aged 0-54 years.

The logarithm of the observed rate of M N D nior- tality in Sweden, shown in Fig. 1 for each decade, is nearly linear over the age range 30-69 years, declin- ing from a linear increase thereafter. Linear regres- sion of these three curves yielded a squared corre- lation coefficient of R'> 0.95 over this age range. I t can also be seen that the observed mortality rate is constant at an age of approximately 55 years, as reported in earlier studies, and confirming the obser- vation above that M N D mortality rates are rising in the older group whilst falling in those aged under 55 years. However, the upper bound of 69 years re- stricts this analysis to only 56"; of all M N D deaths in Sweden and results in an artificially low intersect point and gradient, due to the terminal curvature.

The integral of the observed rates (Eqn. 5 ) are reported in Table 1 and estimate the size of the sus- ceptible subpopulation as approximately 0.38",, of all men and 0.21 7 , of all women, or about 0.27",, of the total population. The size of the subpopulation

Logarithm of the underlyin rate of MND in Sweden in the 1960 s, f 0's and 80's

5.0

4.5 s 8- 4.0

r 3.5 5 g 3.0 2, y 2.5 e 3 2.0

8 1.5

1 .0 z

0.5

0 0

r n

._ - rj u

0 10 20 30 40 50 60 70 80 Age (years)

198 1 - I990 1961-1970 - 1971-1980 - - _ _ _ _

I

F;g. 2. 1.ogarithni of the calculated underlying rate of mortality from M N D in Sweden during the intervals 1961-70. 197 1-80 mid 1% 1-90.

154

Rising Swedish MND mortality: 1961-90

0.05

susceptible to M N D in Sweden, estimated from the cross-sectional data, was seen to increase over the 30 years studied by approximately 60% in both sexes.

The underlying mortality rate from M N D was calculated using Eqn. 7 and linear regression was performed on the logarithm of the underlying rate against age for ages 40-84 years (nine data points) encompassing 93% of all M N D deaths in Sweden. This yielded the (log R,, a ) pairs shown in Table 2 for all years 1961-1990. The squared cor- relation coefficient approached unity for both men (R'> 0.934, mean 0.976) and women (R'> 0.852, mean 0.965), indicating that underlying mortality in- creases exponentially throughout the entire life span. The regression results show that the initial rate of mortality log R,, has fallen from about 1.0 in 1961 towards 0.0 in 1990 in both sexes, compensated by an increase in the exponential gradient a due to in- creasing mortality in the eldest age groups. The loga- rithm of the calculated underlying rate of mortality from M N D is shown in Fig. 2, in which it can be seen that the curves are indeed linear in both sexes to the tabulated upper age limit.

=m 1

A negative linear relationship between log R,, and a was determined by applying linear regression to the results in Table 2. Applying Eqn. 8 to the (log R,,, ( 1 )

pairs the following relationships were determined for men, women and the total population respectively:

" '~=0.01387 (3.768 - hg"'R,,) (R2 = 0.937)

'"a = 0.01334 (3.95 1 - lag "'R,,) (R' = 0.956)

"U = 0.01303 (4.032 - log"R,,) (R' = 0.950)

0.05 .c ct a 0.05 -a

0.04-

0.04

0.04

0.04

0.04

This determines a series of mortality curves for men which intersect at a mortality rate of lo3.'"' or 5900 deaths per 100000 men at an age of 1/0.01387 or 72.1 years. The other two points are 9000 deaths per 100000 women at age 75.0 years and 10828 deaths per 100000 persons at age 76.8 years. In other words the underlying mortality rates at these ages have remained unchanged over the past 30 years in Sweden, whilst mortality rates at all younger ages have declined by a substantial proportion. Men age at a rate of nearly 1.49" per annuni and women at 1.3% per annum. The strong linear relationship be- tween a and log R,, can be seen in Fig. 3.

I I --II

= I

I

I I I I

I

I

I I I I I I

Plot of alpha against log Ro for MND mortality in Sweden from 1961 to 1990

0.06 I I

0.05

0.05 =I --

Fix. 3. Plot o f c, against log R,, for M N D mortality in Sweden froni 1961 to 1990.

155

Neilson et at.

Table 3. Results of non-linear regression of the observed rates of MND mortality by year of birth from 1880 to 1940 and the proportion of variation explained (equivalent R2)

log R, 7. S.S.R. S.D. %Error RZ

1880 1885 1890 1895 1900 1905 1910 1915 1920 1925 1930 1935 1940

1.89 0.03 0.51 0.32 33% 0.67 1.82 0.03 4.21 0.92 41% 0.59 1.46 0.03 17.69 1.88 43% 0.57 0.84 0.04 39.18 2.80 44% 0.56 0.50 0.05 9.51 1.38 19% 0.81 0.17 0.05 2.36 0.69 9% 0.91 0.08 0.05 1.22 0.49 7% 0.93 0.53 0.05 3.52 0.84 15% 0.85 0.44 0.05 0.82 0.41 12% 0.88 0.35 0.05 0.67 0.37 17% 0.83 0.20 0.05 0.19 0.19 17% 0.83 0.08 0.05 0.16 0.18 32% 0.68 0.01 0.05 0.01 0.03 10% 0.90

The results of the cross-sectional analysis were confirnied by non-linear regression analysis of mor- tality rates tabulated by year of birth from 1880 to 1940. The observed mortality rate function (Eqn. 4) was fitted to each cohort mortality rate curve by minimisation of the sum of the squared residuals

(23), determining the best fitting value of log R,, in each birth-year. The parameters a and K were de- termined by the results of the cross-sectional analy- sis reported above, and the subpopulation was ini- tially set at 0.3%. The full results of the non-linear regression analysis are shown in Table 3. The stan- dard deviation of the fitted data lay between 7% and 44"/, of the observed values, with a mean of 23'2,. This is equivalent to a squared correlation coefficient of 0.56 < R'< 0.93, with a mean of 0.77. The (theo- retical) initial rate of mortality, log R,, fell from 1.89 for those born in 1880 to 0.0 1 for those born in 1940, whilst the exponential gradient a rose from 0.028 to 0.052. The size of the susceptible subpopulation was re-estimated by minimisation of the sum of squared residuals for all cohorts resulting in an optimum value of 0.00279, or 0.28% of each birth cohort. The observed cohort mortality data and fitted curves are shown for all cohorts in Fig. 4.

Discussion

The results of this study, using a complementary range of epidemiological techniques, represent

Rates of mortalit in Sweden by year of birth, 188 d (a) to 1940 (m)

0.001 I I I I I 1 I I I

0 10 20 30 40 50 60 70 80 ' Age at death (years)

I

Fig. 4 , Observed and calculated underlying rates of mortality from M N D in Sweden by year of birth from 1880 (a) to 1940 (111).

156

Rising Swedish MND mortality: 1961-90

further confirmation of the findings of a previous study in Sweden ( 5 ) which suggest that mortality from M N D is directly triggered by the interaction of endogenous and exogenous factors. The role of en- dogenous factors is demonstrated epidemiologically by the considerable body of evidence that mortality from the disease arises only in those in a discrete susceptible subpopulation. The size of this subpop- ulation in Sweden is estimated at approximately 0.38",, of men and 0.21Y0 of women, which com- pares well with the estimates for England and Wales as a whole (26), the counties of England and Wales separately (27) and Japan (28), indicating that the relative size of susceptible populations appear to be globally uniform, even though the age distributions of mortality may substantially differ.

The rise in the size of the susceptible subpopula- tion in Sweden over the study period is of particu- lar interest. It is not likely that the underlying size of the subpopulation has changed significantly over 30 years, particularly if the origin of the susceptibility is considered to be genetic in some form. More plau- sible explanations of the increase relate to two groups of factors. Whilst it is possible that the apparent rise in recorded mortality is an artifact of the increased accuracy of death certification of a constant size of underlying susceptible subpopulation, the short du- ration of the disease and generally high reliability of death certification in M N D discounts such an ex- planation. However, as the subpopulation is calcu- lated using cross-sectional M N D mortality and rel- evant population data, secular trends in mortality will clearly affect the estimated size of the subpop- ulation. The analysis of cohort data appears to con- firm a stable subpopulation size in remarkably close agreement with our previous estimates.

In addition to the identification of the existence and approximate size of subpopulation susceptible to M N D in Sweden, this study also demonstrates that the rate of ageing amongst that subpopulation is greater than that for the population as a whole. The rate of ageing (sometimes described as "loss of vitality" in other analyses using survival models) is essentially an endogenous, probably genetically based process, which is calculated independently of the effects of exogenous factors on mortality. The rate of ageing of individuals susceptible to M N D has been calculated in previous studies as about 1.3 per annuni, and the data for Sweden indicates a similar figure. Thus, the rate of ageing in M N D is greater than that calculated for general populations, estimated at around 0.97 "b in various investigations (30, 36, 37). Therefore M N D can be considered to be a disease associated with accelerated ageing.

The role of exogenous factors in triggering the expression of M N D in susceptible individuals is more difficult to unravel, largely because of the com-

plex consequences of increased life expectancy in Sweden and most other industrialised countries. In- creased life expectancy over the past three decades is associated with considerable demographic changes including variations in general and M N D mortality rates. This process may account for the finding that R , has declined from 1961-1990 in Sweden. This observation, consistent with that observed elsewhere, combined with the observation that K is static and nearly equal in all countries studied, shows that while the annual risk of an environmental M N D trigger factor is constant, the average magnitude of the im- pact of the hazards has actually declined since 1961 in Sweden. This either means that increasing popu- lation health has led to greater resistance to a con- stant hazard, or that the magnitude of environmental hazards has declined. The former possibility seems more likely in so far as increasing life expectancy in both the general population and subpopulations has occurred, based on lower levels of earlier mortality. The latter possibility may have played a contributory role since labour safety and working conditions have greatly improved in Sweden during the last three decades. However, assuming an induction period of at least ten to twenty years (38), aetiologically rel- evant exposures with regard to the M N D mortality under study would have occurred some decades ear- lier, rendering the analysis of associations to par- ticular substances more difficult.

The Strehler-Mildvan modification of the Gomp- ertz model of ageing states that vitality falls linearly with age and that as vitality falls the number of potentially fatal challenges to the individual rises (36). From this observation it can be seen that chal- lenges sufficient to trigger the onset of M N D in a young individual will necessarily be larger (and hence rarer) than those triggering onset in an older indi- vidual. Using this framework, it can be argued that chemical exposures may represent infrequent, large challenges sufficient to generate responses in younger susceptible individuals whilst onset at older ages is additionally triggered by the effects of increasingly common but minor exposures in the susceptible sub- population. This general model accords well with the results of a previous Swedish study ( 5 ) .

It may be hypothesised that whilst factors such as solvents, implicated as statistically significant in mortality from M N D in Sweden in earlier studies, could in various ways accelerate the expression of MND, they are in fact acting on healthier (longer lived) populations - which includes those suscep- tible to the disease. Thus, the paradox may be that it is increased health (expressed in greater longevity), rather than decreased health which has led to more of the susceptible subpopulations in Sweden reach- ing those ages at which MND, rather than other conditions, causes death. In summary this observa-

157

Neilson et al.

tion implies that exogenous factors may still be im- portant in triggering MND, and the subsequent ex- pression of the disease in Sweden, but their cumulative effects, as measured by the extent and timing of M N D mortality, has in recent years been muted through acting on an increasingly robust (long lived) population.

Thus, the relationship of exogenous and endoge- nous factors in the development of M N D and sub- sequent mortality from the disease requires consid- erable re-examination following the focus in earlier studies which appear to have been undertaken as- suming a virtually unequivocal role for environmen- tal agents in the genesis of the condition. The iden- tification of a significant endogenous factor in this investigation, through the location of a susceptible subpopulation with a higher rate of ageing, indicates that the role of exogenous factors must be consid- ered only in an endogenous context. The role of increased life expectancy and its association with mortality from M N D further suggests that an undue emphasis on environmental agents alone may not be a profitable way forward in the understanding of mortality from the disease.

Conclusions

This study has indicated that mortality from M N D has indeed risen substantially over the period from 1961-1990 in Sweden. This rise is for the most part a real increase, and cannot be attributed to increased accuracy in diagnosis or the recording of death. A subpopulation susceptible to M N D has been iden- tified and its approximate size calculated. More of this subpopulation are now living to the ages at which M N D is expressed and therefore leading to increased mortality from the disease. This study supports the growing evidence of a genetic involvement in spo- radic M N D (39-41) and it seems likely that initial susceptibility is conferred through a genetic (al- though not necessarily inherited) defect. Overall the analysis suggests that environmental agents d o not cause MND, but niay still play a role in precipitat- ing the symptoms of the condition. The annual risk of environmental trigger factors in the disease has not increased over the study period and it appears that, extrapolating from the increased longevity of the susceptible subpopulation, the impact of such factors niay have actually declined since 1961.

References

1. GUNNARSSON L-G. LINDBERG G . [!]SODERFELT B, AX- ELSON 0. The mortality of niotor neuron disease in Sweden. Arch Neurol 1990: 47: 42-46.

2. GUNNARSSON L-G. PALM R. Motor neuron disease and heavy ni;inuaI labour: an epidemiologic survey of Varrnland County. Sweden. Neuroepidcniiology 1984: 3: 195-206.

3 . GUNNARSSON L-G. LINDBERG G. [!]SODERFELDT B. AX- ELSON 0. Amyotrophic lateral sclerosis in Slveden in relation to occupation. Acta Neurol Scand 1991: 83: 394-398.

4. GUNNARSSON L-G. LINDBERG G. Amyotrophic lateral scle- rosis in Sweden 1970-83 and solvent exposure. Lancet 19x9: 335: 958.

5. GUNNARSSON L-G. BODIN L. SODERFELDT B. AXELSON 0. A case-control study of motor-neuron disease: its relation t o heritability. and occupational exposures. particulorly t o solvents. Br J Ind Med 1992: 49: 791-798.

6. WILLIAMS DB. WINDEBANK AJ. Motor neuron disease (amyotrophic lateral sclerosis). Mayo Clin Proc I99 I : 06:

7. SWASH M. SCHWARTZ MS. LI T-M. Trends in mortality from niotoneuron disease. Lancet 1989: 335: 958.

8. KURTZKE JF. BEEBE G W . Amyotrophic lateral sclerosis: I . A case-controlled comparison based on ALS deaths. Neu- rology 1980: 30: 453-462.

9. FLATEN TP. Rising mortality from motoneuron disease. Lnn-

10. LILIENFELD DE. CHAN E. EHLAND J et al. Rising mortal- ity from motoneuron disease in thc USA, 1962-XJ. Lancet

11. MARTYN CN. Poliovirus and motor neuron disease. J Ncu- rol 1990: 237: 336-338.

12. KURTZKE JF. Risk factors in amyotrophic lateral sclerosis. Adv Neurol 1991: 56: 245-270.

13. JOKELAINEN M. PALO J. WIKSTROM J . The distribution o f amyotrophic lateral sclerosis in Finland. J Neurol Sci 1975: 25: 473-479.

14. HOLLOWAY S. EMERY AEH. The epidemiology of niotor neuron disease in Scotland. Muscle Nerve 1982: 5: 131-133.

15. BUCKLEY J. WARLOW C. SMITH P. HILTON-JONES D. I R - V I N E S. TEW JR. Motor neuron disease in England and Wales, 1959- 1979. J Neurol Neurosurg Psychiatry 1983: 46: 197-205.

16. LEONE M. CHI^ A. MORTARA P. Rosso MG. SCH~FFER D. Motor neuron disease in the province of T u r i n . Italy. 1971-1980. Acta Neurol Scand 1983: 68: 316-327.

17. BHARUCHA NE. SCHOENBERG BS. RAVEN KH. P ICKLE LW. BYAR DP. MASON T J . Geographic distribution o f niotor neuron disease and correlation with possible etiologic Factors. Neurology 1983: 3 3 : 91 1-915.

18. SIENKO DG. DAVIS JP. TAYLOR JA. BROOKS BK. Aniy- otrophic lateral sclerosis: a case-control study follo\ving de- tection of a cluster in a small Wisconsin community. Arch Neurol 1990: 47: 38-41.

19. GOMPERTZ B. On the nature of the function expressive of the law of human mortality. Phil Trans Roy Soc I.ondon 1x25:

20. RIGGS JE. The decline of niortality due t o stroke: ;I competi- tive and deterministic perspective. Neurology 1991 : 4 I : 1335- 1338.

2 1. RIGGS JE. Longitudinal Gonipertzian analysis o f Parkin- son’s Disease mortality in the United States. 1955- 1986 - the dramatic increase in overall mortality since 1980 is the natural consequence of deterniinistic mortality dynamics. Mech Age Devel 1990: 55: 221-233.

22. RIGGS JE. Longitudinal Gonipertzian analysis of primary malignant brain tuniour mortality in the United States 1962- 1987. Rising mortality in the elderly is the natural conse- quence of competitive deterministic dynnmics. Mech Age Devel 1991: 60: 225-241.

23. NEILSON S, ROBINSON I. Reinterpreting mortality statistics: some uses of Gonipertzian analysis in epidemiological re- search. J Clin Epidemiol 1993: 46: 1063-1069.

24. DAVID HA. MOESCHBERGER ML. The theory of competing risks. London: Griffin. 1978.

25. NEILSON S. ROBINSON I . CLIFFORD ROSE F. HUNTER M .

54-82.

CCI 1989: 335: 1018-1019.

1989: 335: 710-712.

115: 513-585.

158

Rising Swedish MND mortality: 1961-90

Rising mortality from motor neurone disease: an explanation. Acta Ncurol Scand 1993: 87: 184- 191.

26. NEILSON S. ROBINSON 1. HUNTER M . Longitudinal Gomp- ertzian analysis of ALS mortality in England and Wales, 1963- 1990: estimates of susceptibility in the general popula- tion. Mcch Age Devel 1992: 64: 201-216.

27. NEILSON S. ROBINSON I. Cross sectional Gompertzian analysis: the development of a “Gonipertz mortality ratio” (GMR) and its applicability. Mech Age Devel 1993: 68: 137- 149.

28. NEILSON S. ROBINSON I. KONDO K. A new analysis of mortality from motor neurone disease in Japan, 1950- 1990: rise and fall in the postwar years. J Neurol Sci 1993: 117: 46-53.

29. RIGCS JE. Longitudinal Gompertzian analysis of amyo- trophic lateral sclerosis niortality in the United States, 1977- 1986 - evidence for an inherently susceptible subset. Mech Age Dcvel 1990: 55: 207-220.

30. NEILSON S. ROBINSON I. HUNTER M . Static and dynamic models of interdiscase competition: past and projected nior- tality from amyotrophic lateral sclerosis and multiple sclero- sis. Mcch Age Devcl 1993: 66: 223-241.

3 1, WHO. International classification of diseases. Geneva: World Hcalth Organisation. 1957.

32. WHO. International classification ofdiseases. Geneva: World Health Organisation. 1968.

33. WHO. International classification ofdiseases. Geneva: World Health Organisation, 1989.

34. O’MALLEY F, DEAN G. ELIAN M. Multiple sclerosis and motor neurone disease: survival and how certified after death. J Epidemiol Community Health 1987: 41: 14-17.

35. HOFFMAN PM. BRODY JA. The reliability of death certifi- cate reporting for amyotrophic lateral sclerosis. J Chron Dis

36. STREHLER BL, MILDVAN AS. General theory of mortality and ageing. Science 1960: 132: 14-21.

37. RIGGS JE. Longitudinal Gompertzian analysis of adult mor- tality in the United States, 1900-1986. Mech Age Dcvcl 1990:

38. CALNE DB. MCGEER E. EISEN A. SPENCER P. Alzhcimer’s disease, Parkinson’s disease and motoneurone disease: abio- tropic interaction between ageing and environment? Lancet

39. SIDDIQUE T. FIGLEWICZ DA et al. Linkage of a gene causing familial amyotrophic lateral sclerosis to chromosome 2 1 and evidence of genetic-locus heterogeneity. N Engl J Mcd 1991:

40. KOSEN DR. SIDDIQUE T. PATTERSON D et al. Mutations in Cu/Zn superoxide disniutase gene arc associated with familial amyotrophic lateral sclerosis. Nature 1993: 362: 59-62.

41. JONES CT. BROCK DJH. CHANCELLOR AM. WARLOW CP. SWINGLER KJ. CuiZn superoxide-distiiutase (SODI) mutations and sporadic amyotrophic lateral sclerosis. Lancet

1971: 24: 5-8.

54: 235-247.

1986: 2: 1067-1070.

324: 1381-1384.

1993: 342: 1050-1051.

159