close

Вход

Забыли?

вход по аккаунту

?

Epidemiology of multiple sclerosis in U.S. veterans V. Ancestry and the risk of multiple sclerosis

код для вставкиСкачать
Epidemiology. of Multiple Sclerosis
in US. Veterans: V. Ancestry and the b s k
of Multiple Sclerosis
William F. Page, PhD," John F. Kurtzke, MD,? Frances M. Murphy, MD,P and James E. Norman, Jr, PhD-f
Self-reported ancestry data for the U.S. population from the 1980 decennial census and multiple sclerosis (MS) risk
data derived from a large series of World War I1 white male veterans with MS and matched controls were aggregated
on a state level and analyzed to determine the relationship between ancestry and MS risk. A significant portion of the
state-by-state variation in MS risk is explainable statistically by differences in ancestry among state populations, even
when geographic latitude is included in analyses. In the main, Swedish and other Scandinavian ancestry is most
consistently associated with places with increased MS risk. In some analyses, Italian, French, and (to a lesser extent)
Scottish ancestries are also associated with increased risk, whereas English and Dutch ancestries are each associated
with decreased risk, but most of these non-Scandinavian correlations may reflect predominantly geography per se.
These findings provide evidence that ancestry of the resident population, a confounded measure of genetic susceptibility and cultural environment, is part of the complicated picture of MS as a disease of place.
Page WF, Kurtzke JF, Murphy FM, Norman JE Jr. Epidemiology of multiple sclerosis in U.S. veterans:
V. Ancestry and the risk of multiple sclerosis. Ann Neurol 1993;33:632-639
Perhaps the most intriguing feature about multiple
sclerosis (MS) is that its occurrence varies with geography, i.e., generally, most people of European origin are
at high risk both in their homelands and in their New
World settlements. The reasons for this variation remain unknown and still raise the basic issue of nature
versus nurture [l]. In three previous articles of this
series, we have explored this subject, examining the
effects of race, sex, and geography { 2 ) ; latitude and
climate 131; and migration r4} on the risk of MS in an
unusually large cohort of patients with MS and
matched controls, comprising U.S. veterans of World
War I1 (WW 11) and the Korean Conflict. In this article, we examine the effect of another factor, population
ancestry, on the risk of M S .
Materials and Methods
The entire series, described in detail earlier E21, consists of
5,305 US. veterans of WW I1 or the Korean Conflict who
were judged by the then Veterans Administration (now the
Department of Veterans Affairs) as "service-connected" for
MS. Such a decision for this disease required definitive evidence of clinical signs attributable to MS during or within 7
years after military service. Each patient with MS was
marched with a military control by using year of birth, date
of entry into and branch of service, and survival of the war.
From the *Medical Follow-up Agency, Institute of MedicineiNational Academy of Sciences, and ?Neurology Service, Veterans Affairs Medical Center, Washington, DC; and *National Heart, Lung,
and Blood Institute, Bethesda, MD.
A random sample of 80 patients was reviewed for diagnosis
(J. F. K.), and 96% (77 patients; 95% confidence interval,
89.3-99.2%) of these patients met all criteria of the Schumacher Panel for definite MS 151.
For each patient with MS and control, residence was defined by his or her place (state and county) of birth and
place of residence at entry into active duty (EAD). To avoid
confounding by sex, race, and time, tabulation of MS risk by
individual state in this report is limited to the largest subset,
that of WW I1 white male patients and controls, 8,019 veterans.
Some anomalies in the distribution of MS risk by state (Fig
1) in fact suggested that latitude, per se, might not be the
sole determining factor of risk; for example, the northern
tier of latitude contains states with very high, high, and moderate MS risk. In casting about for an explanation for such
variation, a hypothesis was formed that ancestry might be a
contributing factor. The interest in this hypothesis was
strengthened when it was also noted, although it is not very
apparent in Figure 1, that not all states in the southern tier
show the same risk; in particular, Louisiana, Arizona, and
southern California do not have as low a risk as other southern tier states. This assessment was undertaken because it was
believed that these differences in risk might be attributableto
differences in ancestry, the three southern states noted having relatively large proportions of residents with, respectively, French and Hispanic ancestry (W. F. P.). The working
hypothesis was thus formed that MS risk might be better
Received Nov 9, 1992. Accepted for publicarion Jan 21, 1993.
Address correspondence to Dr Page, Medical Follow-up Agency,
2101 Constitution Avenue, NW,
DC 20418,
632 Copyright 0 1993 by the American Neurological Association
Multiple Sclerosis Risk (actual)
rl 86 - 115
r1401
1.41 8 above
(based on EAD state)
Fig 1. Multiple sclerosis risk (actual). Patient-to-controlratio
ranges for state of residence at entry into active duty for white
males of World War I1 service. (Modij5edfromK2lrt.de et al,
Neurology 1979;29:1228-1235.)
explained by differences in ancestry rather than merely by
differences in geography. The lack of ancestry data for the
individual patients with MS and controls in the cohort (see
later discussion) required the analysis of population-based
information.
Ancestry data for this analysis were sought from the U.S.
decennial census, but the census did not include data on
ancestry until 1980; before then, the census collected information only on parents’ birthplace. Although earlier material
would have been preferable, the 1980 data present some
notable advantages, being readily available in tabulated form
[6] and quite comprehensive, as some 188 million persons
reported their ancestry in 1980. W e should emphasize that
these nationwide 1980 census data are not ancestry data collected from our individual patients with MS and controls.
The census material on ancestry exhibits some of the usual
problems connected with self-reported data, despite the fact
thar field tests of potential items were undertaken in advance
of the actual census. There are, for example, clearly identifiable instances of misreporting, such as an unusually large
proportion of persons living in the state of Georgia whose
responses caused them to be (mis)classified as having Russian
(i.e., “Georgian”) ancestry. There are other limitations as
well. In particular, two possible meanings of the term
“Scotch-Irish are confounded, i.e., the first, a person of
Northern Irish (Scotch-Irish) ancestry, cannot always be distinguished from the second, a person who simultaneously
reports both Scottish and Irish ancestry.
In the 1980 census, data are tabulated separately for respondents who reported “at least one specific ancestry group”
(total response) and for those who reported only “a single
ancestry group” (single response); the difference of these two
figures is the number of respondents giving a multiple response, such as “German-English.” With the exception of
17 unique triple-ancestry categories (e.g., English-GermanIrish), all other multiple responses were coded only to the
first and second reported ancestry groups. Because single
responses might be considered to represent “purer” ancestry
data, with the inference that both parents share that stock,
our analyses were performed for the single responses (e.g.,
Scottish o r Irish alone) as well as for total responses.
Analyses here included only the 10 largest self-reported
ancestry groups (English, German, Irish, French, Italian,
Scottish, Polish, Dutch, Swedish, and Norwegian) as well
as two additional groups of substantive interest, Danish and
“Scandinavian, not otherwise specified.” In this report, the
unit of analysis is the individual state of the 48 coterminous
states (and the District of Columbia) with 10 o r more observations in the veteran series (this excluded only Nevada). MS
risk was calculated twice, once using the patients’ and controls’ states of birth and once using their states of EAD.
Latitude (midstate latitude measured to the nearest 0.1 degree) and ancestry data (number of respondents in a given
state reporting a given ancestry divided by the total number
of respondents in that state) were then appended to MS risk
values for each state.
Both ancestry and geography yield two kinds of information, i.e., ancestry data includes both single and total responses and geography, both birth and EAD state. Analysis
tables, therefore, show four sets of figures, one set for each
combination of ancestry and geography. In these analyses,
MS risk, more accurately, a relative odds ratio rather than a
true relative risk, is defined as the number of patients with
MS divided by the number of controls in a given state.
The correlational and regression analyses were performed
using the Statistical Analysis System (SAS) computer software [7}. Both sets of analyses used weighted data, which
means that each state’s observation is weighted so that the
larger the state, the larger its effect in the analyses. Weights
were calculated by dividing the number of patients and controls in a given state by the total number (8,019) of patients
and controls. Regression models were fit in a stepwise manner, using ancestry categories and geography (latitude) as independent variables and MS risk (by state) as the dependent
variable. Because of the relatively high intercorrelations
among ancestry groups and geography, there is the potential
for producing unstable estimates of regression coefficients.
For that reason, we have also presented tables with only
the sign and not the size of regression coefficients of each
statistically significant predictive variable but with the proportion of variance in MS risk explained thereby, which is a
measure of the strength of association.
Results
Table 1 shows the frequency distribution of reported
ancestry in 15 classes for the entire United States in
1980. Single ancestry rates for these range from a high
of 20.03% for English to a low of 0.20% for Scandinavian, not otherwise specified; total ancestry rates range
from 26.33 to 0.25%. A comparison of single versus
total ancestry reporting shows some ancestry groups
with similar values, English, Italian, and Portuguese,
for example, and others with rather different values,
Irish and Scottish, for example.
Table 2 shows four sets of correlations between MS
risk and ancestry, one set for each combination of ancestry (single and total) and geography (birth and EAD
state) for all ancestries among the 12 we analyzed.
Most striking are the highly significant positive correlations of Scandinavian ancestry, especially Swedish ancestry, with MS risk; these are statistically significant
Page et al: Ancestry and the Risk of MS 633
Table 2. Cowelation of Ancestry (Single and Total) with
Multiple Sclerosis Risk (Tabulated by Birth and EAD State)
Table I. Percentage Distribution of Ancestry" Groupj
for U.S. Population by Single and Total Ancestryb
English (%)
German (%)
Irish (%)
French (96)
Italian (%)
Scottish (%)
Polish (%)
Dutch (%)
Swedish (%)
Norwegian (%)
Hungarian (%)
Welsh (%)
Danish (%)
Portuguese (%)
Scandinavian' (%)
Respondents (in
millions)
Single
Ancestry
Total
Ancestry
20.03
15.13
8.72
2.58
5.81
0.99
3.21
1.18
1.09
1.06
0.61
0.26
0.36
0.52
0.20
26.33
26.14
21.33
6.85
6.47
5.34
4.37
3.35
2.31
1.83
0.94
0.88
0.81
0.54
0.25
118.60
188.30
'Categorization of self-reported ancestry from the 1980 Census of
Population.
bSingle and multiple responses, respectively, to the 1980 Census
question on ancestry.
'Not otherwise specified.
(Source: Table 2, 1780 Census of Population, Ancestry of the Population by State: 1980, Supplementary Report PC80-SI-10, 1983,
Bureau of the Census, Washington, DC.)
for all combinations of single or total ancestry and birth
or EAD state. Thus, in a given state, the higher the
proportion of persons reporting Scandinavian ancestry
on the 1980 census, the higher the risk of MS reported
among WW I1 white male veterans who were born in
or entered military service from that state.
Polish ancestry also shows statistically significant
positive correlations (except for single ancestry by
EAD state) and both German and French total ancestry
are significantly correlated with MS risk for EAD state.
All of these are, however, notably smaller than the
Scandinavian correlations. On the other hand, English
ancestry, either single or total, has a statistically significant negative correlation with MS risk for birth or
EAD state, which means that the higher the proportion
of a state's population reporting English ancestry, the
lower the associated risk of MS. In general, the pattern
of correlations of MS risk across birth and EAD state
and single or total state ancestry is similar.
Given this preliminary evidence of a statistical association between MS risk and ancestry, the next step
was to quantify this association more precisely. Thus,
weighted stepwise multiple regression models were fit
by using all 12 ancestry categories to predict MS risk.
Table 3 shows the sign of the regression coefficient
634 Annals of Neurology Vol 33 No 6 June 1993
Pearson Correlation Coefficients
Birth State
English
German
Irish
French
Italian
Scottish
Polish
Dutch
Swedish
Norwegian
Danish
Scandinavian
EAD State
Single
Ancestry
Total
Single
Ancestry
Ancestry
Total
Ancestry
-0.565"
0.250
-0.136
0.152
0.212
0.156
0.296'
0.077
0.545"
0.40Ib
0.306'
0.489"
-0.425b
0.276
-0.149
0.276
0.238
-0.077
0.352'
-0.045
0.545"
0.445b
0.345'
0.509"
-0.565"
0.260
-0.114
0.131
0.207
0.107
0.283
0.101
0.631"
0.444b
0.263
0.554"
-0.409b
0.309'
-0.066
0.294'
0.228
0.151
0.349'
0.023
0.634"
0.497"
0.375b
0.566"
(NOS)
< 0.001; ' p < 0.01; ' p < 0.05.
EAD = entry into active duty; NOS
=
not otherwise specified.
(Source: 1980 Census of Population, Ancestry of the Population by
State: 1980, Supplementary Report PC80-S1-10, 1983, Bureau of
the Census, Washington, DC.)
(indicating positive or negative association) of each significant predictive variable, along with the proportion
of variance in MS risk explained, a measure of the
strength of association.
Perhaps the single most important finding in Table
3 is that these few ancestry variables together explain
45 to 60% of the state-by-state variance in MS risk,
whether one considers birth or EAD state and whether
one looks at single or total ancestry. Moreover, except
for one model, i.e., birth state MS risk by single ancestry, all other models contain only two significant terms,
one for the proportion of Swedish ancestry and one
for Italian ancestry. In both cases there is a positive
association.
In the birth state, single ancestry model, Scandinavian ancestry, not otherwise specified, replaces Swedish ancestry, and Scottish ancestry appears for the only
time as a positive factor. English ancestry, on the other
hand, is strongly associated negatively with MS risk.
This negative association, however, could be at least
partly attributable to statistical artifact. The only model
in which English ancestry has a significant negative effect among these regression models is in that involving
single ancestry, in which responses are by definition
mutually exclusive. Thus, the larger the proportion of
English reported as a single ancestry, the necessarily
lower proportion of other single ancestries, such as
Swedish or Itahan, in a given state; this might produce
an artificially lower predicted risk for English ancestry.
Table 3. Proportion of Variance in Multiple Sclerosis Risk
Explained bj AncestTya (Single and Total Responses)
Birth State
Variable and
Direction of
Associationb
(96)
EAD State ($25)
Table 4. Proportion of Variance in Multiple Sclerosis Risk
Explained by Ancestry" (Single and Total Responses) and
Geography (Birth and EAD State)
Birth State
Single
Ancestry
Total
Ancestry
Single
Ancestry
Total
Ancestry
Swedish(positive)
Italian (positive)
Scandinavian'
(positive)
Scottish (positive)
English (negative)
...
29.7
39.8
40.2
...
16.2
15.2
17.6
12.3
...
...
.. .
16.5
...
...
...
31.9
...
...
...
Total model
60.7
45.9
55.0
57.8
'Based on 1980 decennial census data aggregated by state.
bPositive means that multiple sclerosis (MS) risk increases with increasing values of the variable; negative means that MS risk decreases
with increasing values of the variable.
CNototherwise specified.
Variable and
Direction of
Associationb
(5%)
EAD State (%)
Single
Ancestry
Total
Single
Ancestry
Total
Ancestry
Latitude
(positive)
Swedish
(positive)
French (positive)
Dutch (negative)
60.1
60.1
66.5
66.5
4.1
3.0
9.0
8.5
5.3
5.4
3.9
4.4
...
...
2.3
...
Total model
69.5
68.5
81.7
79.4
Ancestry
"Based on 1980 decennial census data aggregated by state.
bPositive means that multiple sclerosis (MS) risk increases with increasing values of the variable; negative means that MS risk decreases.
EAD = entry into active duty.
EAD = entry into active duty.
However, recall that in the univariate analyses of Table
2 , each English ancestry comparison was strongly and
significantly correlated negatively with MS risk.
In Table 4, state latitude is added to the ancestry
and MS models. We may note first that latitude is
somewhat associated with self-reported ancestry simply
due to historic patterns of settlement, and thus it and
ancestry may be somewhat confounded in these analyses, and second, that latitude is very strongly associated
with MS risk [2-47. Although latitude is seen to be
the strongest predictive factor in Table 4, there remain
statistically significant associations of ancestry with MS
risk, even after latitude has been considered. In all
these models, Swedish and French ancestry are significantly associated with a higher risk of MS, and Dutch
ancestry appears in one model as a significant negative
factor. As with English ancestry in Table 3, the Dutch
association occurs in the analysis of single ancestry data,
and may thus be at least partially an artifact.
Italian ancestry, a factor in three of the four models
in Table 3, does not appear at all in Table 4; that is,
only when latitude is not included in the model, as
seen in Table 3, is Italian ancestry a significant predictive factor. And it never attained significance in the
univariate analyses (see Table 2). Intercorrelations between the proportion of persons with Italian ancestry
and all other ancestries (data not shown) are significant
and negative between Italian ancestry and Scandinavian
ancestry; more important, though, Italian ancestry is
significantly and positively correlated with latitude. Because Italian ancestry is a significant factor only in the
absence of latitude, with which it is significantly corre-
lated, it may be that it is acting as a proxy for latitude
in the Table 3 models. If so, then the presence of
Italian ancestry in those models may be considered a
statistical artifact.
Conversely, French ancestry, a significant factor in
all four models in Table 4 , does not appear at all in
Table 3. In contrast to Italian ancestry, however,
French ancestry does not have a significant correlation
with latitude. However, the univariate analyses of Table 2 show only one weak correlation of significance.
In summary, latitude and ancestry together explain 70
to 80% of the state-by-state variability in MS risk, and
self-reported ancestry of the population remains a statistically significant predictor of MS risk, even when
latitude is considered.
Although the larger a proportion of variance a model
explains the better, it is also important to examine the
discrepancies between a model's predicted values and
the actual data. Figure 1 shows actual MS risk by state,
calculated using EAD state. Figure 2 displays the corresponding predicted MS risk by EAD state, calculated
from the model using single-response ancestry data
without latitude. In comparing the actual values in Figure 1 with the predicted values in Figure 2 , there is
fairly good agreement between the maps, except for
the very high risk states. In fact, the actual and predicted values have a (Spearman) rank-order correlation
of 0.924. Figure 3A plots the same actual and predicted
MS risks together on a common scatterplot. In this
case, the diagonal line represents the theoretically ideal
situation where predicted and actual risk are identical;
the larger the number of points lying close to this line,
the better the fit of the model. Figure 3B shows the
same data except that latitude has been added to the
Page et al: Ancestry and the Risk of MS
635
Multiple Sclerosis Risk (predicted)
(based on EAD state)
Fig 2. Multiple sclerosis risk (predicted),Predicted patient-tocontrol ratio ranges for state of residence at entry into active
duty for white males of World War II service; predictions based
on single-response ancestry data shown in Table 3 and excluding latitude.
model. The reader will note how fit of the model is
improved. Figure 3C shows data similar to Figure 3B
(latitude and single ancestry model results) for MS risk
based on birth residence. Overall, the model for EAD
(R2 = 81.7%) fits the actual data somewhat better than
the birth model (Rz = 69.5%).
Discussion
We have shown that self-reported ancestry by itself
explains nearly as much of the state-by-state variation
in MS risk (45-6096) as does geography (latitude) per
se (60-67%’0). Even when ancestry is combined with
geography in a joint model, it remains a statistically
significant and independent predictor of MS risk. Although the specific ancestry groups significantly predicting MS risk vary somewhat by the choice of single
or total reporting of ancestry and birth or EAD state,
Swedish (or, in one case, Scandinavian) ancestry is always a significant positive predictor of MS risk, i.e., the
higher the proportion of persons reporting Swedish
ancestry in a given state, the higher the risk of MS.
In the regressions without geographic latitude, Italian
ancestry is significantly and positively associated with
MS risk in three of four models; in all four latitude plus
ancestry models, French is significantly and positively
associated with MS risk. In the birth state analysis without latitude, Scottish ancestry is a positive risk factor
and English ancestry a negative one, and Dutch ancestry appears as a negative factor in the single ancestry
model including latitude.
The meaning of all these ancestry factors is not immediately clear, and we suggest that this complicated
picture arises because ancestry is linked both to genetics and to (cultural) environment. And in this specific
context, ancestry is also linked to geography, with
which MS risk is strongly related. The presence of
Swedish (or Scandinavian) ancestry in all the models,
birth or EAD and single or total ancestry, mark it as a
636 Annals of Neurology Vol 33 No 6 June 1993
major underlying risk factor. Coupled with the known
high rates of MS in Scandinavian countries, this suggests that Scandinavian ancestry may play either an environmental role or a genetically determined part in
increasing susceptibility to MS. The high rates of MS in
Scotland likewise suggest such a hypothesis for Scottish
ancestry. The troublesome fact that Scottish ancestry
appears as a statistically significant factor in only one
model could be explained by the fact that Scottish ancestry in the U.S. population is mostly mixed (Scottish
ancestry has the highest ratio of total to single ancestry
in Table 1); thus, only in the single ancestry analysis is
there enough “pure” ancestry data to show an excess
risk of MS, if there is a “real” relationship.
The speculation about the meaning of ancestry takes
another turn when discussing Italian and French ancestry. Italian ancestry, in particular, deserves special mention because it was strongly correlated with latitude
and did not appear as a significant factor in any model
that included latitude. This strongly suggests that the
association of Italian ancestry and MS risk is a statistical
artifact. Although French ancestry is not in itself significantly associated with latitude, the states whose residents report a large proportion of French ancestry (a
10% or larger proportion of total ancestry) are Connecticut, Louisiana, Massachusetts, Maine, Minnesota,
New Hampshire, Rhode Island, and Vermont; with
the exception of Louisiana and Minnesota, all of them
are in the MS high-risk northeastern United States.
Although it may be the case that MS risk is truly higher
in states whose populations contain a higher proportion
of persons with French ancestry, it may also be that
because the historic patterns of settlement have tended
to concentrate persons of French ancestry in the urban
Northeast. Not only geography, but also urban residence has been shown to be a risk factor for MS {8].
Thus, the association between MS risk and French ancestry may be mediated through cultural factors such
as urbanization or others.
In addition, lest the ancestry model results appear
too clear cut, it should also be noted that the choice
of the customary (though strictly speaking, arbitrary)
p-value of 0.05 in the statistical models means that
some possibly strong ancestry factors were not considered “significant.” Scottish ancestry was such a “borderline” factor in the birth risk, total ancestry model, and
also in the corresponding total ancestry model that included latitude. Dutch ancestry was a borderline negative factor in the single ancestry model with latitude.
In the EAD risk models, English was a borderline negative risk factor in the single ancestry model and French
a borderline positive risk factor in the total ancestry
model; Norwegian ancestry was a borderline risk factor
for the single ancestry model with latitude, and Scandinavian ancestry, not otherwise specified, a borderline
risk factor in the total ancestry model with latitude.
2.8
2.6
2.4
1
I
WA
2.4
TI
I
1.o
0.4
0.2
1
c/
/
1
0.2
1.Q
0.8
0.6
0.4
&
0.2
0
1
0.4
1
1
,
1
l
l
I
0.6
0.8
1.0
1.2
1.4
1.6
1.8
,
2.0
,
A
”._
-
2.6
-
2.4
-
I
I
I
I
,
0.2
0.4
0.6
0.8
1.0
,
1.2
,
1.4
,
1.6
,
1.8
,
,
2.0
2.2
I
2.2
Predicted Value of MS Risk
2.8
1.6
gg :
..
/g
k
0
1.8
x z
.%a
. . .>
*./.
.
1.6
0.6
2.0
..
1.8
0.8
2.2
WA
In the latitude plus ancestry models, the proportion
of explained variation in MS risk is somewhat larger
for EAD state ancestry than birth state ancestry. This
is consistent with an EAD preponderance over birthplace in MS risk by geography seen for earlier analyses
of this case series [4],i.e., northern migration between
birth and EAD increased the risk of MS and southern
migration decreased it. More generally, other migration studies confirm that migration between MS risk
areas does alter the risk, and in the appropriate directions 191. Further, because the association of MS risk
with birth state ancestry is not appreciably stronger
C
Predicted Value of MS Risk
Fig 3. (A) Actual versus predicted multiple sclerosis (MS) risk.
Actual patient-to-controlratio ranges for state of residence at ent y into active duty (EAD)for white males of World War I1 service (from Fig I ) plotted against predicted patient-to-control ratios (from Fig 2). (8)Actual versus predicted MS risk. Actual
patient-to-control ratio ranges for state of residence at EAD for
white males of WW I1 service (from Fig 1) plotted against predicted patient-to-control ratios based on single-response ancestry
data shown in Table 4 and including latitude. (C)Actual versus predicted MS risk. Actual patient-to-controlratio ranges for
state of birth for white mules or WW I1 service plotted against
predicted patient-to-control ratios based on single-response ancestry data shown in Table 4 and including latitude.
than the association with EAD, a case cannot be made
for considering these effects of ancestry to be genetically based. Instead, as noted earlier, ancestry is a confounded measure of genetic and environmental factors,
and even in itself may be a proxy measure for other
underlying factors that remain unidentified. Even so,
the mere fact that ancestry has a significant statistical
association with MS risk independent of geography at
least leads one to search for underlying risk factors
such as those possibly related to genetic susceptibility
or to cultural differences.
That ancestry may play a part in the determination
of MS risk is not a new thesis. The very low rates of
MS in American soldiers of Asian ancestry and the low
rate among those of African ancestry in our series 121
has provided support for the statement that MS is “the
white man’s burden” [9]. Ebers and Bulman [lo] had
done a similar analysis of the data in our series and
concluded that “the distribution of MS in the United
States, at least in part, reflects the distribution of genetic susceptibility factors.”
Page et al: Ancestry and the Risk of MS
637
Fig 4. Rates per 1,000 for rejections because of multiple sclerosis
among white drabees of World War I by state of residence at induction. (Reprinted with permission from Davenport CB, Proc
Assoc Res Nerv Ment Dis 1922;2:8-19.)
As our final draft was being prepared, the full article
by B u h a n and Ebers { 1I] appeared. They found “the
highest correlations between MS prevalence (from our
material 12)) and Scandinavian birth/ancestry ([Spearman] Y = 0.73).” Their assessments included foreignborn for 1730 and 1780, as well as the same ancestry
data we used, but limited to single ancestry materials.
They found nearly the same degree of positive relations with the British Isles (including England alone)
as they did for Scandinavia when 1930 data or foreignborn 1980 distributions were used, but their text gives
a -0.40 correlation for “British Isles ancestry” for
1980 (not shown in their table). Bulman and Ebers
[111 state on page 7 1: “The analysis we report here
supports a genetic explanation for the geography of
MS in the U.S.”
A link between ancestry and MS risk has even older
antecedents. Using some of the earliest tabulated data
on geography and MS in America, i.e., those based on
638 Annals of Neurology Vol 3 3 No 6 June 1993
drafted men in WW I, Davenport {I21 noted among
whites a high rate of MS in the Great Lakes region
and the State of Washington (Fig 4). He produced the
hypothesis that “there is some race inhabiting [these
regions) that is especially sub jecr to multiple sclerosis. . . . One thinks of the big Swedes that live in
this country.” However, he continued that “whether or
not . . . multiple sclerosis [is) especially common
among Scandinavians cannot be definitely asserted.”
Finally, as always, some caveats are in order. First,
we must emphasize that the ancestry data used in these
analyses are aggregate data for the nation and not findings specific to our own series. In such a situation there
is always the risk of an “ecological fallacy,” which occurs when differences between groups are attributed
to one characteristic of the groups when they are actually due to some other characteristic. The most conservative interpretation of ecological studies such as the
one presented here is that they provide clues for the
further study of cause 1137. We hope to obtain individual data on ancestry for future analyses of MS risk
within our series.
Second, even when these data are aggregated at the
state level, sample sizes are small and statistical estimates are therefore somewhat unstable. We have considered this in our choice to report the results of stepwise regression in terms of variance explained rather
than coefficient values. The use of weighted regression,
moreover, provides stabler estimates; without weighting the states by size, the effect of including or excluding the data from a single small state would be more
apt to have a material effect on overall results.
Last, one should remember that the ancestry data
used in these analyses come from the 1980 decennial
census and were collected some decades after the veteran patiendcontrol cohort was defined. However, the
overall distribution of MS in this country has changed
little during this century, at least into the 1970s [97.
This temporal difference, though, might underlie some
of the discrepancies between predicted and actual values of MS risk for some states; for example, one discrepant state, Florida, is among the fastest growing and
may have a very different ancestry distribution in 1980
than it had in the 1940s. For California there is also
the methodological problem reflected in the fact that
our earlier analyses divided the state between middle
and southern latitude tiers (and therefore different MS
risk areas). In the analyses in this report, however, ancestry data are available only for the state as a whole,
and thus California’s two disparate MS risk areas have
been combined.
In summation, by using self-reported ancestry data
from the 1980 decennial census and data on MS risk
derived from a large series of WW I1 white male veteran patients with MS and matched controls, it has
been shown that a significant portion of the state-bystate variation in MS risk is explainable statistically by
differences in ancestry of the population of the states.
Even when geographic latitude is considered, ancestry
still explains a statistically significant amount of the
variance in MS risk. Swedish and other Scandinavian
ancestry is most consistently associated with place of
increased MS risk, but there are several significant associations with Italian and French ancestries, and a sin-
gle significant association of Scottish ancestry with increased MS risk and English and Dutch ancestries with
decreased risk. These findings provide evidence that
ancestry of the resident population, a confounded measure of genetic susceptibility and cultural environment,
is part of the complicated picture of MS as a disease
of place.
This work was supported by Department of Veterans Affairs contract no. V688P-1532 (Neuroepidemiology Research Program,
VAMC, Washington, DC) to the National Academy of Sciences.
References
1. Skegg DCG. Multiple sclerosis: nature or nurture? Br Med J
1991;302:247-248
2. Kurtzke JF, Beebe GW, Norman JE. Epidemiology of multiple
sclerosis in U.S. veterans: 1. Race, sex, and geographic variation.
Neurology 1979;29:1228- 1235
3. Norman JE, Kurtzke JF, Beebe GW. Epidemiology of multiple
sclerosis in U.S. veterans: 2. Latitude, climate and the risk of
MS. J Chronic Dis 1983;36:551-559
4. Kurtzke JF, Beebe GW, Norman JE. Epidemiology of multiple
sclerosis in U.S. veterans: 111. Migration and the risk of MS.
Neurology 1985;35:672-678
5. Schumacher GA, Beebe GW, Kibler RF, et al. Problems of
experimental trials of therapy in multiple sclerosis: report by
the panel on the evaluation of experimental trials of therapy in
multiple sclerosis. Ann NY Acad Sci 1965;122:552-568
6. Bureau of the Census. Supplementary report: Ancestry of the
population by state: 1980, PC80-S1-10. Washington, DC: U.S.
Government Printing Office, 1983
7. SAS Institute, Inc. SAS procedures guide, release 6.03 edition.
Cary, NC: SAS Institute Inc, 1988
8. Beebe GW, Kumke JF, Kurland LT, et al. Studies on the natural history of multiple sclerosis. 3. Epidemiologic analysis of the
Army experience in World War 11. Neurology 1967;17:1-17
9. Kurtzke JF. Multiple sclerosis from an epidemiological viewpoint. In: Field EJ, ed. Multiple sclerosis. A critical conspectus.
Lancaster, England: MTP Press Ltd, 1977
10. Ebers GC, Bulman D. The geography of MS reflects a genetic
susceptibility. Neurology 1986;36(suppl):108(Abstract)
11. Bulman DE, Ebers GC. The geography of MS reflects genetic
susceptibility.J Trop Geogr Neurol 1992;2:66-72
12. Davenport CB. Multiple sclerosis from the standpoint of geographic distribution and race. Proc Assoc Res Nerv Ment Dis
1922;2:8- 19
13. Lilienfeld AM, klienfeld DE. Foundations of epidemiology.
New York: Oxford Press, 1980:375
Page et al: Ancestry and the Risk of MS
639
Документ
Категория
Без категории
Просмотров
0
Размер файла
956 Кб
Теги
epidemiology, ancestry, veterans, sclerosis, multiple, risk
1/--страниц
Пожаловаться на содержимое документа