- Email: [email protected]
The epidemiology of multiple sclerosis: the Iceland model onset-adjusted prevalence rate and other methodological considerations
Journal of the Neurological Sciences, 111 (1992) 143-152 © 1992 Elsevier Science Publishers B.V. All rights reserved 0022-510X/92/$05.00
143
JNS 03818
The epidemiology of multiple sclerosis: the Iceland model Onset-adjusted prevalence rate and other methodological considerations Charles M. Poser a, John Benedikz b and Patricia L. Hibberd c Departments of a Neurology and c Medicine, Harvard Medical School, a Neurological Unit, Beth Israel Hospita~ Boston, MA, USA, b Department of Neurology, University of Iceland School of Medicine, and the Neurological Service, Landspitalinn, Reykjavik, Iceland, and c Section of Infectious Diseases, Massachusetts General Hospital, Boston, MA, USA (Received 27 December, 1991) (Revised, received 9 March, 1992) (Accepted 12 March, 1992)
Key words: Epidemiology; Multiple sclerosis; Iceland; Onset adjusted prevalence; Age adjustment; Comparison populations; Ethnic diversity Summary The epidemiology of multiple sclerosis (MS) is characterized by the fact that there is an uneven distribution of the disease throughout the world. The two most commonly used indices of its frequency are the incidence and prevalence rates. The incidence rate reflects, to a great extent, the influence of environmental factors in triggering the clinical manifestations of the disease, if it is based upon the actual date of the initiation of symptoms, rather than the date of diagnosis. The prevalence rate is currently based on the date of diagnosis and includes all MS patients who are alive on a particular date, without regard to their ethnic origin, the site and the duration of residence or any other factors that may have influenced the acquisition of the disease. We propose that in order to make the concept of the prevalence rate possibly more meaningful, the term should refer, retrospectively, to all patients whose symptoms eventually led to the diagnosis of MS, even though the diagnosis was not yet established on an earlier prevalence day. In addition, only patiems of the same ethnic background who have spent their prepuberal years in the geographical area under study should be included. We are calling this measure the onset-adjusted prevalence rate. Another problem encountered in epidemiological studies of MS relates to the common practice of adjusting prevalence data obtained for age and sex in one area to what is referred to as a "standard" American (or world) population, groups of great ethnic and age diversity. It is also curious that in many studies the bases for comparison of populations are census data obtained many years previously. We suggest that age and sex adjustment should be apphed only to similar ethnic groups born and raised under different environmental conditions. We believe that data obtained by calculating an onset-adjusted prevalence rate restricted to a homogeneous group of patients sharing the same environment during the prepuberal years may provide valuable etiological clues.
Introduction Despite the large number of epidemiological studies of multiple sclerosis (MS) that have been published during the past 50 years, many of the conclusions from the data remain or have become controversial, with the
Correspondence to: Charles M. Poser, MD, Neurological Unit, Harvard Medical School, Beth Israel Hospital, 330 Brookline Avenue, Boston, MA 02215, USA. Tel.: (617)735-2063; fax: (617)735 5216.
possible exceptions of the classic studies of Alter et al. (1966) Dean and Kurtzke (1971), and Detels et al. (1978) regarding the significance of age at migration. It is difficult to compare these data because of the lack of uniform diagnostic criteria, uncertainty of the length of time elapsing between the date of symptomatic onset and diagnosis, deficiencies in case finding, and enormous variability in the clinical presentation and course of the disease. In addition, there are problems in applying classical epidemiological methods and procedures to a disease subject to a combination of genetic and environmental factors of debatable etiology and
144 uncertain pathogenesis. This article will address two main areas: that of the calculation of incidence and prevalence, and the traditional epidemiological practice of age-adjustment to what are considered to be standard populations. These considerations were derived, in part, from a recently completed study of the epidemiology of MS in Iceland (Benedikz et al. 1991).
Definitions
The two most commonly used measurements for estimating the frequency of the disease are incidence and prevalence. Schoenberg (1978) wrote that "incidance measures the frequency of addition of new cases of a disease within a specific population; it is calculated for a given time interval and a given place". Kurtzke (1985), another widely respected neuro-epidemiologist, defined incidence as "the number of cases of the disease beginning in a unit of time within the specified population. This is generally given as an annual incidence rate in cases per 1000 population per year". Neither of these definitions specifies whether cases should be counted as "beginning" when the diagnosis of MS was established, (presumably by a neurologist), or retrospectively, as "beginning" when the symptoms of the disease subsequently diagnosed as MS had first been noticed. Therefore, data are published using either date of diagnosis or date of onset. The year of diagnosis is not etiologically relevant and simply represents the random time at which cases happen to be seen by a neurologist who recognizes the clinical diagnostic significance of symptoms and signs. In fact, our Iceland data (Benedikz et al. 1991) clearly demonstrated that case ascertainment is directly related to the availability of neurologists in the area. Prevalence, according to Schoenberg (1978) measures "the frequency of all current cases of disease (existing cases and new ones) within a specific population and is calculated for a given time and given place", while for Kurtzke (1985) "the point prevalence rate refers to the number of people affected [by the disease] at one point in time within the community expressed per unit of population". Thus only living patients are counted on prevalence day. It is curious that texts or articles dealing with the epidemiology of MS have rarely mentioned the possible etiological significance of incidence or prevalence rates. The figures have been used either to measure changes in the frequency of the disease over time, or to compare groups or geographic areas. None the less, incidence and prevalence represent quite different indices: incidence reflects environmental factors that bring out symptoms in individuals already afflicted by either the MS trait (Poser 1986) or the disease itself in an asymptomatic form; prevalence, on the other hand,
calculated according to the traditional definition, simply means the number of cases known at a certain time. It sheds no light on possible etiological or pathogenetic factors. We believe that the results of our epidemiological study of MS in Iceland (Benedikz et al. 1991) may provide a basis for redefining both incidence and prevalence, and indicate the potential significance of prevalence rates in terms of MS predisposition and etiology. The characteristics of the population of Iceland provide ideal material for such a study for the following reasons: the total population of the island, approximately 250 000 people, is extraordinarily homogeneous, being of mixed Norwegian and Irish descent with almost no immigration in the last 1000 years. The health services are excellent and are available free to all inhabitants. The level of education is remarkably high and literacy is 100%. There are 10 neurologists practicing in the country, all of whom have received advanced specialty training in Scandinavia, the United Kingdom or the United States. Many Icelanders have the disease, so that its manifestations are well known to most physicians, the degree of suspicion is high, and referral to neurologists is early and frequent. A MS registry has been in operation for over 30 years. A total of 252 MS patients had been identified in Iceland from the beginning of the century until December 31, 1985. All but four of them were born and raised in Iceland, with their Icelandic ancestry on both sides going back many generations. The four who were excluded from consideration had been born in Denmark, the Faroe Islands, England, and Iran. Approximately half the inhabitants of Iceland live in an urban environment: the metropolitan area, which includes Reykjavik and its suburbs, and Keflavik. The rest of the island is considered rural, since the only town of any size, Akureyri, has approximately 15 000 inhabitants.
Incidence data
Starting in the mid-1930s figures on incidence have been collected. The date of symptomatic onset was determined as accurately as possible, often a very short time after the initial manifestations of the disease, especially after the first neurologist began practicing in Reykjavik in 1936. Fig. I shows that there has been no significant change in the mean annual incidence of MS based on the year of onset from 1936 to the end of 1985, the rate ranging between 1.8 and 2 . 7 / y e a r / 100000. The mean annual incidence based on the year of diagnosis ranges from 2.2 to 3.4/year/100000, with a statistically significant but biologically meaningless
145 Rate and 95% Confidence Limits (Poisson)
5
Rate a n d 95% Confidence Limits (Poisson)
100
4.34 °
~
4
O
..............................................................................................................
81,8
8O .............................................................
s 3
=~
~
°=:!: 2
3.4~
.....~-~§ . . . . . . . . . . . . . . . . . . . . . . .
r[
I
2.89
I
oJ
";;; .............. T ................. I~:i~°
.....
................ . . . . . . . . . . . . . . . . . . . . . . .
. . . . .
62.4
60
~ 4o
,
~ 43.2 ,~-
~ 42.9
~0:2-i~ .~
r~
~.~_i
[
2 10 .....
. . . . . . . . . . .
. . . . .
N 2o
12-31-1955 12-31-1965 12-31-1975 12-31-1985
0 19~6-~945
1946-19s-~
19ss-~gs5
196s-1975
197s-!98s
Year Intervals
Prevalence Day Fig. 3. Iceland: prevalence of multiple sclerosis by year of diagnosis 1955-1985.
Fig. 1. Iceland: mean annual incidence of multiple sclerosis by year of onset 1936-1985.
Rate and 95% Confidence Limits (Poisson)
5
4.26 ..........................................................................
=. °°4
.q29
~
646
.............................
ill
~
i
13.4o
.... .........X.................................................... .........tl .... o
..........
..............
..........
..........
....
0 1~.%-1945
1946.fgss
19s6-196s
tges-197s
197-~-199s
Year Intervals Fig. 2. Iceland: mean annual incidence of multiple sclerosis by year of diagnosis 1936-1985.
difference (Poisson) between the decades 1956-1965 and 1975-1985 (Fig. 2). Because the year of diagnosis may lag a long time after the time of clinical onset, it provides no meaningful information regarding etiological or environmental factors. It is difficult to understand why an unusual number of cases of MS diagnosed at a particular time are referred to as an epidemic, but this is what has
happened in South Africa (Rosman et al. 1985) and in Kenya (Adam 1989). When the latency between clinical onset and the first visit to the neurologist becomes short enough, as it did in Iceland (Benedikz et al. 1991), the incidence rate of MS remains the most accurate method of calculating the frequency of the disease provided the following conditions are met: that it is based on the estimated date of onset rather than the date of diagnosis and that persons of different ethnic background and those who have moved to the study area with already manifested symptoms of MS (even though still undiagnosed) be excluded from the count. Furthermore, the diagnosis must be established by a neurologist, preferably one with extensive experience in MS, according to generally accepted criteria. Only definite and probable cases should be included. The Icelandic study met all these conditions and demonstrated that the rate of incidence based on date of onset remained unchanged for the last 30 years.
Prevalence data Almost all studies evaluating the prevalence of MS have used the Schoenberg-Kurtzke definition, i.e., counting the number of patients who have been diag. nosed as having MS who are alive on prevalence day and calculating the rate on the basis of the most recent
TABLE 1 PREVALENCE OF MS IN ICELAND BY YEAR OF DIAGNOSIS Prev. day
Number of cases
12/31
New cases
Deaths
Cases alive
1955 1965 1975 1985
61 a 30 59 78
8a 18 11 22
53 65 113 169
a New cases and deaths in entire period prior to Dec. 31, 1955.
Population
Prevalence per 100000 (95% confid, lim.: Poisson)
160633 193 758 219033 240 741
33.0 (24.8-43.2) 33.5 (26.0-42.9) 51.6 (42.6-62.4) 70.2 (60.2-81.8)
146 TABLE 2
I00 L%.
~ A •
MULTIPLE SCLEROSIS IN ICELAND: DURATION OF DISEASE MEAN AND MEDIAN BY DECADE
MiIO Moderate Severe
O0
Patients who died before
Number of cases
Mean (years)
Median (years)
Range (years)
Dec Dec Dec Dec
8 18 11 22
21.1 27.8 27.8 34.1
19 29 26 32
2-51 9-51 9-44 5-73
31, 31, 31, 31,
1955 1965 1975 1985
6o ,
~
40
~
20
census data. According to this method, the prevalence data for MS in Iceland are shown in Table 1 and Fig. 3. A statistically significant increase would appear to have taken place between 1955 and 1985. Increases in prevalence have been noted in many areas and are ascribed mainly to the increasing longevity of MS patients, presumably due to improvement in supportive medical care. This has been true in Iceland as shown in Table 2: the mean duration of illness increased from 21 years in 1955 to 34 years in 1985. Another reason may be related to Acheson's (1977) comment that "because MS patients characteristically may have symptoms without confirmatory new rologic signs this may lead to overlooking, at least at onset, patients with a mild form of the disease". The importance of this factor is demonstrated by the changes in the Kurtzke scores of Icelandic MS patients since the beginning of the century (Fig. 4). There is an increase in the proportion of milder cases diagnosed as MS, which may account for the low incidence before the advent of trained neurologists. A very similar situation may have occurred elsewhere, for instance in the Faroe Islands before World War II. The Icelandic data represent an ethnically and geographically homogeneous population, unaffected by immigration or important environmental changes.
Onset-adjusted prevalence This concept arose from the study of two publications on the prevalence of MS in Lanzarote in the
0
1920
,
1930
i940
1950
1960
i970
1900
1990
Year
Fig. 4. Iceland: total number of MS cases at the end of each decade grouped according to Kurtzke Disability Scale: mild = 0-3, moderate = 4-6, severe = 7-9.
Canary Islands (Table 3). In the first survey in 1981, Sosa et al. (1983) reported having found two cases of MS with a prevalence of 3.3/100000. A second survey 6 years later (Garcia et al. 1989) included nine new cases, with a calculated prevalence of 15.0/100000, a highly significant increase. The second publication (Garcia et al. 1989), however, made note of the fact that 6 of the 9 patients "newly discovered" in 1987 had already had definite symptoms of MS, although they had not been diagnosed as such, at the time of the 1981 survey. Adding these 6 to the original 2 for a total of 8 gives a 1981 onset-adjusted prevalence of 13.3/ 100000 (Table 3). At a later date Reyes-Yanez mentioned (personal communication) that the 2 original patients found in 1981 had left the island before the 1987 survey to seek medical care in another one of the Canary Islands but were still alive. Correcting for those 2 additional cases, the 1987 prevalence was 18.3/ 100000, a difference that is not statistically significant from the 1981 prevalence. These studies led to our conviction that prevalence should include all patients affected by MS, i.e. who had symptoms of MS and who would eventually be diagnosed as having the disease, although such diagnosis
TABLE 3 PREVALENCE OF MULTIPLE SCLEROSIS IN LANZAROTE, CANARY ISLANDS Population: 60000; Latitude: 28°N. Year of survey
Number of cases
Prevalence per 100000
Correction
Onset-adjusted prevalence (CL)
1981 e 1987 b
2 9
3.3 15.0
+6 b= 8 + 2 c----11
13.3 (5.7-26.2) 18.3 (9.1-32.8)
a Sosa et aL (1983). b Garcia et al. (1989): 6 of the 9 patients had onset of MS before 1981 survey. c Reyes-Yanez, M., personal communication, 1991: the 2 patients found in 1981 had left before 1987.
147 TABLE 4 ONSET-ADJUSTED PREVALENCE OF MS IN ICELAND 1955-1985 4"
Prey. day
Number of cases
12/31
Clinical onset
Total cases
Deaths
Cases alive
1955 1965 1975 1985
103 a 36 44 65
103 139 183 248
8a 18 11 22
95 113 b 146 189
Population
Prevalence per 100 000 (95% confid, iim.: Poisson)
160633 193 758 219033 240 741
59.1 (48.1-72.7) 58.3 (48.1-70.6) 66.6 (56.3-78.9) 78.5 (67.9-90.7)
a Clinical onsets and deaths during entire period before Dec. 31, 1955. b Number of cases on previous prevalence day, minus deaths° and plus cases with clinical onset during previous decade.
might not have been made on prevalence day. It also became clear that more meaningful data would be obtained when prevalence is based on a well defined, ethnically homogeneous population who had spent the critical years of putative acquisition of the disease (i.e., between the ages of 5 and 15 years, or puberty) in the geographical area under study. This would then exclude all persons who had moved into the area after that age, with or without the symptoms of MS, as well as those of different ethnic or racial origin. Under such conditions, prevalence rates would more precisely define the role of genetic susceptibility as well as suggest clues regarding potentially significant environmental factors affecting the primary acquisition of the illness. Because all these conditions are fulfilled in Iceland, the calculation of onset-adjusted prevalence rate (OAPR) was carried out (Table 4, Fig. 5). It is clear that no statistically significant changes have taken place between 1955 and 1985. We have also calculated that in order for a statistically significant difference to exist between these years, a net increase of 30 cases with onset between January 1, 1976 and December 31, 1985 would have to be discovered. This is most unlikely to occur in view of the fact that the latency between
ICELAND 1955-1985
symptomatic onset and diagnosis has been reduced to less than 5 years during the past 15 years (Benedikz et al. 1991). Only 6 MS patients have had their first symptoms since January 1, 1986. All the MS data based on age adjustment to the general Icelandic populations are identical to those cited here. Theoretically, all symptomatic MS cases should be included in the prevalence calculations, including those who had died. In order to retain those cases, the number of deaths in the general population during the study period should be included in the denominator, but such numbers are difficult to obtain. For this reason, the calculation of the OAPR also subtracts those MS patients who died prior to prevalence day, assuming that these deaths represent the same proportion of deaths in the general population. On the other hand, all MS patients who had met all the conditions outlined above, but who had left the area for whatever reason and are known to be alive should also be counted, as exemplified by the 2 patients who had left Lanzarote for another of the Canary Islands. In order for OAPR to be calculated, a long enough period of time must have elapsed after the selected prevalence day to insure that the vast majority of cases would have had their clinical onset during the study period or prior to it and have eventually been recognized as suffering from MS.
Rate and 95% Confidence Limits (Polsson) 100
90.7 o o o
80
n_gt7
70s •
.7e.9............7e,~ -7
66.6
60
v 40 =
....2
2
-.3
t,u
~ 2o
12-31-1955
12-31-1965
12-31-1975
12-31-1985
Prevalence Day
Fig. 5. Iceland: onset-adjusted prevalence of multiple sclerosis 19551985.
Discussion: prevalence rates
We are aware of only a single published attempt to correct prevalence data obtained by the classical method: in their study of the changes of frequency of MS in Denmark, Koch-Henriksen and Hyllested (1988) applied an ill-defined "notification delay" corrective factor. Because prevalence rates classically include all living patients, difficulties of interpretation arise with regard to immigrants, i.e. individuals with MS who did not spend the critical prepuberal years in the study area. This is illustrated by the Australian study of Hammond et al. (1988). The Australian data comparing preva-
148 that of the 4 articles dealing with the same population of Northeast Scotland. First, but not unique to these studies, is the inclusion of "possible" as well as "probable" MS cases, the latter including what most other classifications would call "definite". Three of the articles report on surveys carried out in 1970 (Shepherd and Downie 1978), 1973 (Shepherd and Downie 1980) and 1980 (Downie and Phadke 1983); the fourth one (Swingler and Compston 1986) included figures from the previous 3 studies in a general review of MS prevalence in the United Kingdom. In all 3 original surveys (1970, 1973 and 1980) 3 diagnostic categories of MS were used: probable, early and latent probable, and possible; all three were included in the calculations. It was possible to calculate OAPR for these surveys from the content of the articles. The details have been reproduced in Table 5. The original 1970 survey reported a prevalence rate of 127/100000 based on 557 cases (all categories) but 105/100000 for probable cases only. In the 1980 publication (Shepherd and Downie 1980) it was noted that 50 cases had already had enough symptoms at the time of the 1970 survey that they should have been counted as cases of probable MS. Therefore, the OAPR for the 1970 survey would be 117/100000. The 1973 survey reported a prevalence rate of 144/100000, which approaches a significant difference from the previous 1970 survey. However, when the cases are restricted to the probable ones, the corrected OAPR is 117, exactly the same as in 1970. The last survey, in 1980 (Downie and Phadke 1983) reported a prevalence of 178/100000 (all categories) with a statistically highly significant increase over the previous two surveys. However, 25 individuals already suffering from the disease had moved into the area after the 1973 survey. Subtracting those and basing the new calculation only on probable cases, the resulting corrected OAPR is 139/100000, which is barely statistically significantly different from the other
lence between 1961 and 1981 include only patients carrying the diagnosis of MS. It was not possible to calculate the OAPR from these data. Although they did calculate rates separately for native-born and foreigu-born patients, and specified how many of their patients already had MS at the time of immigration, they did not indicate at what age the still asymptomatic immigrants had come to Australia. Such data would be of particular interest in view of the fact that the highly significant difference in prevalence between Perth and Newcastle on the one hand, and Hobart on the other, appear in one of the very few recent studies that Support a putative relationship between MS frequency and latitude. The same questions apply to the Wynn et al. (1990) study of the frequency of MS in Olmsted County, Minnesota: the data are based on already diagnosed MS patients and the only exclusion was of patients who had resided in the county for less than one year. No comment was made about the possibility that MS patients, some already diagnosed, some with still undiagnosed symptoms of the disease, had come to the area prior to the exclusion period. The problem of immigration was taken into consideration by Middleton and Dean (1991) in their study of the population of the Greek part of the divided island of Cyprus (latitude 34°N). They counted all diagnosed MS patients of Greek descent who had been born in the area and excluded refugees from the Turkish-occupied region. They included one Englishwoman who had acquired Cypriot nationality but excluded all foreigners. They found a total of 42 patients, including 1 refugee and the Englishwoman. On the basis of 40 MS patients in a population of 92 100 the prevalence was 43.4/100000, once again raising grave doubts about the significance of latitude. Unfortunately, no data were given about time of clinical onset. Another illustration of the difficulties that arise in trying to interpret published epidemiological data is
TABLE 5 PREVALENCEOF MULTIPLESCLEROSISIN NORTH-EASTSCOTLAND. Year of survey Population All cases Probableonly Corrected probable Prevalence rate corrected rate Confidence limits (Poisson)
1970(1) 440200 557
1973(2) 440200 634 464
517 517
514 b 127 117 138
105 96 115
1980(3) 471350 a 839
144 117 107 127
133 156
682
464
517
682
105
117
155
96 115
107 127
143 167
657 c 178
117 107 127
1986(4) 440200
166 191
145 135 156
139 129 147
References: (1) Shepherd and Downie (1978). (2) Shepherd and Downie (1980). (3) Downie and Phadke (1983). (4) Swingler and Compston (1986). Notes: Underlined numbers prevalencegiven by authors. a Populationcalculatedfrom prevalence rate of 1978. b Addition of 50 eases missedat time of 1970survey. c Subtractionof 25 eases who moved into area after 1973surveybut had alreadyacquired MS.
149 two surveys. Finally, in a compilation of published data by Swingier and Compston (1986), only probable cases were used, but no corrections made, so that the prevalence rates that were published gave credence to a marked but questionable increase in prevalence from 1973 to 1980. The major use of this proposed method of OAPR is to determine if significant changes have occurred over a period of time in ethnically homogeneous populations, counting only individuals who have spent their pre-pubertal years in the geographical region under study.
What is true prevalence?
Epidemiologicai studies can deal only with those cases of disease that have produced symptoms and as a result, must ignore what may be a significant number of cases that will never be diagnosed. Some idea of the numbers of unsuspected, completely asymptomatic cases of MS in the population can be gathered from autopsy data. Vost et al. (1964) in Montreal found 5 cases of unsuspected MS in 440 patients most of whom had died of heart disease, an incidence of 1.25% in this selected group. Gilbert and Sadler (1983), in Ontario, found that 0.2% of 2450 autopsies examined ncuropathologically had typical lesions, despite no history suggestive of MS. Engell (1980) in Denmark found evidence of MS in 0.1% o~ cases where the nervous system was examined pathologically because of malignancy or cerebrovascular disease. In addition to the fact that the spinal cord was not studied in those cases, it is not known what fraction of the total autopsies is represented by the 16 000 cases examined, nor is it clear if these cases were representative of the population at risk for MS. On the basis of such data, assuming the Banm and Rothschild (1981) estimate of 123000 cases, or even Scheinberg's (1987) estimate of
TABLE 6 STANDARDPOPULATIONSBY AGE 1966a Age groups (years)
African %
European %
World %
USAb %
O- 9
20
15
22
15
10-19 20-29 30-39 40-49 50-59 60-69 over 70
20 20 20 10 5 3 2
14 14 14 14 13 9 7
18 16 12 12 9 7 4
14 17 17 12 9 9 8
a Modifiedfrom Doll et aL (1966). b Total US population 1988 from Statistical Abstracts of the US 1990.
200000 cases of MS in the United States, i.e. between 0.05 and 0.1% of the total population, this implies that the number of asymptomatic cases is at least equal to that of known instances of MS. This may well be true in other countries of high prevalence. There does not seem to be any method available for correcting for the unavoidable underestimate of MS prevalence.
Adjustment to a standard population
Data on MS incidence and prevalence from various parts of the world are often adjusted to what are considered standard populations in terms of age distribution. This practice makes a great deal of sense when dealing with mortality rates which are clearly influenced by the population's age distribution (Lilienfeld and Ldienfeld 1980). Such adjustment has been used extensively in cancer epidemiology, but the advisability of doing so has been questioned by Doll et al. (1966): "Cancer incidence varies by age, especially in terms of type. Clearly therefore comparisons of total incidence rates uncorrected for age are of little value in the study of etiology. The most suitable comparisons are those made directly between the individual age specific rates. The traditional way of overcoming this difficulty introduced by differences in age distribution of the population is by the use of rates standardized for age; that is, rates which, if calculated would have been observed if the age-specific rates recorded by each registry had occurred in a standard population with a specified age distribution. But this in fact does not happen. In the study of etiology, therefore, standardized rates are little better than crude rates. They reduce the effect of differences in age distribution; but they do not eliminate it, and if their limitations are forgotten they may be grossly misleading". There are two standard populations that are com. monly used for age adjustment. One was devised by the International Union Against Cancer (Table 6). Although it has the redeeming feature of providing specific data for Africa and for Europe, there are important differences in age distribution between those two and the population of the world as well as that of the United States. The other standard population is that of the United States. It is difficult to understand the rationale for this because there are few, if any, populations that are more heterogeneous. Major differences in ethnic composition exist in many regions of the country (Table 7). An illustration of this problem is provided by the study of Wynn et al. (1990) of MS in Rochester and Olmsted County, Minnesota, in which the crude rates were adjusted to the white population of the United States in 1950. Table 8 shows how different the Minnesota White population is from that of the entire United States. The same study also illus-
150 TABLE 7 ETHNIC DIVERSITY OF THE UNITED STATES
Persons who reported a single ancestry group ~ North east (%)
North central (%)
South (%)
West (%)
0.4 1.2 20.0 15.1 8.7 1.1 1.0 1.I 47.5
0.9 0.1 10.9 12.0 11.4 0.3 1.0 0.7 37.3
0.6 2.0 15.5 28.6 7.6 2.5 0.8 1.9 59.5
0.1 0.1 30.8 9.2 8.9 0.2 0.9 0A 50.6
0.8 1.3 17.6 12.4 6.4 1.7 1.3 1.7 43.2
(B) Medium MS susceptibility Czech 0.7 French 2.6 Greek 0.5 Hungarian 0.6 Italian 5.8 Polish 3.2 Portuguese 0.5 Russian 1.2 Total 15.1
0.5 3.9 1.0 1.1 15.9 6.0 1.3 2.6 26.9
1.3 1,8 0,5 0.8 3,3 5.1 0.0 0.7 13.5
0A 2.7 0.3 0.3 1.9 1.0 0.1 0.6 7.3
0.5 1.8 0.4 0.5 3.6 1.3 1.0 1.1 10.2
12.4 0.4 0.7 0.4 0.2 1,6 0.2 0.1 3,5 19,5
15.4 1.3 0.2 0.1 0.2 0,4 0.2 2.1 0.5 20.4
27.2 2.0 0.2 0.8 0.2 1.0 0,1 6,3 0.2 38,0
8.0 3.1 1.8 0.2 1.9 3.3 2.3 17.2 0.3 38.1
Ethnicity
US (%)
(A) High MS susceptibility
Danish Dutch English German irish Norwegian Scottish Swedish Total
(C) Low MS susceptibility
Afro-Amer. Amer. Ind. Chinese Cuban Filipino Hispanic Japanese Mexican Puertorican Total
17.3 1.6 0,6 0,4 0,5 1.4 0.6 5.9 1.1 29,4
a Groups over 400000 population in entire US from 1980 United States Census.
trates another rather curious tradition: the use of the 1950 population as a basis for comparison for data gathered almost 40 years later. Table 9 indicates how the US population has shifted towards the older age groups. One of the most striking examples of the inappropriate application of the custom of age-adjustment to the United States population was the conclusion reached by Leibowitz et al. (1973) that the Israel-born children of both Ashkenazi and Sephardi immigrants had the same prevalence of MS, whereas the "raw" data indicated that MS was 4 times more common among the former, as it had been for their immigrant parents. More recent data (Biton and Abramsky 1983) confirm tile considerably higher prevalence of MS among the Ashkenazis than among Sephardis although all were born and raised in Israel. Because it is now generally accepted that MS is acquired between the ages of 5 and 15, (or puberty), on the basis of migration studies (Alter et al. 1966; Dean
and Kurtzke 1971), age beyond that period of time plays little if any role in considerations of etiology. Adjustment for ethnic origin would appear to be much more important. In terms of clinical onset, there is no question that the disease is more likely to manifest TABLE 8 PERSONS WHO REPORTED AT LEAST ONE ANCESTRY a UNITED STATES CENSUS 1980 (PCg0-sI-10)
European population only. Ancestry
Minnesota (%)
US (%)
Danish Icelandic Norwegian Scandinavian Swedish
1.90 0.04 12.93 1.28 9.58
0.74 0.02 1.67 0.23 2.11
Total
25.73
4.7
a 83.1% of US population.
151 TABLE 9 POPULATION OF THE UNITED STATES BY AGE GROUPS. Age groups
1950 a
1988 b
(years)
Total (%)
White (%)
Black (%)
Total (%)
White (%)
Black (%)
0- 9 10-19 20-29 30-39 40-49 50-59 60-69 over 70
16.1 18.3 17.2 15.0 13.0 9.9 6.5 4.0
15.7 18.0 17.1 15.0 13.1 10.2 6.7 4.1
19.8 20.5 18.2 15.4 11.7 7.4 4.6 2.5
14.8 13.7 16.8 16.7 11.9 8.9 8.5 8.3
14.2 13.5 16.4 16.6 12.1 9.2 8.9 8.7
18.5 17.9 18A 16.1 9.9 7.7 6.2 5.3
a US Census 1950. b Statistical Abstracts of the US 1990.
itself during the 3rd and 4th d e c a d e s o f life, but the value o f age-adjustment must be g a u g e d in terms o f the extreme difficulty o f finding a truly c o m p a r a b l e p o p u lation. It would be theoretically possible and interesting to c o m p a r e the age-specific prevalence and incidence rates o f MS a m o n g individuals o f Scandinavian descent b o r n and raised in M i n n e s o t a with those in Scandinavia, o r a m o n g Israeli Jews a n d Jews in o t h e r parts o f the world.
Conclusion Because MS often b e c o m e s s y m p t o m a t i c in patients long before the diagnosis is m a d e , it is imperative that such patients be included in all calculations o f prevalence. Obviously, MS patients can be identified only after the diagnosis has b e e n established and therefore the calculations must be retrospective. O A P R must also be restricted to patients w h o share similar ethnic backgrounds a n d have resided in the same area during the putative period o f acquisition o f the disease. T o be meaningful for o t h e r populations, age adjusting the prevalence d a t a so obtained m u s t be subject to the same restrictions. It is suggested that O A P R d a t a according t o these criteria m a y provide important clues about the acquisition o f MS. Acknowledgments This project was supported, in part, by grant No. PP0080 from the National Multiple Sclerosis Society (USA). The generous financial contributions made by Hyman Abadi (Bogota, Colombia), Lorne and Luise MacArthur (Waltham, MA), and Paul and Carol Peterson (North Attleboro, MA), as well as the excellent typing skills of Sheila Golden, and the judicious editorial advice of my wife, Joan Crawford, are all most gratefully acknowledged. We are also indebted to Professors Johan Aarli of Bergen, Norway, and Geoffrey Dean of Dublin, Eire, for their constructive comments.
References Acheson, E. (1977) The epidemiology of multiple sclerosis. Br. Med. J., 33: 9-14.
Adam, A. (1989) Multiple sclerosis: epidemic in Kenya? E. Afr. Med. J., 66: 503-5C~i. Alter, M., U. Leibowitz and J. Speer (1966) Risk of multiple sclerosis related to age at immigration to Israel. Arch. Neurol., 15: 234-236. Bantu, H. and B. Rothschild (1981) The incidence and prevalence of reported multiple sclerosis. Ann. Neurol., 10: 420-428. Benedikz, J., H. Magnusson, C. Poser, E. Benedikz, G. Olafsdottir and G. Gudmundsson (1991) Multiple sclerosis in Iceland: 19001985. J. Trop. Gcogr. Neurol., 1: 16-22. Biton, V. and O. Abremsky (1986) Newer study fails to support environmental factors in etiology of multiple sclerosis (abstract). Neurology, 36 (Suppl. 1): 184. Dean, G. and J. Kurtzke (1971) On the risk of multiple sclerosis according to age at immigration to South Africa. Br. Med. J., 3: 725-729. Detels, R., B. Visscher, R. Halle, R. Malmgren, J. Dudley and A. Coulson (1978) Multiple sclerosis and age at migration. Am. J. Epidemiol., 108: 386-393. Doll, R., P. Palme and J. Waterhouse (1966) Cancer incidence in five continents, International Union Against Cancer, Springer-Verlag, Berlin. Downie, A. and J. Phadke (1983) The Chief Scientist reports...: multiple sclerosis in North East Scotland. Health Bull. (Edinburgh), 42: 151-156. Engell, T. (1989) A ciinico-patho-anatomical study of clinically silent multiple sclerosis. Acta Neurol. Scand., 79: 428-430. Gareia, J., S. Rodriguez, M. Sosa et al. (1989) Prevalence of multiple sclerosis in Lanzarote (Canary Islands). Neurology, 39: 265-267. Gilbert, J. and M. Sadler (1983) Unsuspected multiple sclerosis. Arch. Neurol., 40: 533-536. Hammond, S., J. McLeod, K. Millingen et al. (1988) The epidemiology of multiple sclerosis in three Australian cities: Perth, Newcastie and Hobart, Brain, 111: 1'25. Koch-Henriksen, N. and K. Hyllested (1988) Epidemiology of multipie sclerosis: incidence and prevalence rates in Denmark 19481964 based on the Danish Multiple Sclerosis Registry. Acta NeuroL Scand., 78: 369-380. Kurtzke, J. (1985) Epidemiology of multiple sclerosis. In: Koetsier, J. (Ed.), Handbook of Clinical Neurology, vol. 3, Elsevier Science Publ., Amsterdam, pp. 259-287. Leibowitz, U., E. Kahana and M. Alter (1973) The changing frequenoy of multiple sclerosis in Israel. Arch. Neurol., 29: 107-110. Lillienfeld, A. and D. Lillienfeld (1980) Foundations of Epidemiology, 2nd edn., Oxford University Press, New York. Middleton, L. and G. Dean (1991) Multiple sclerosis in Cyprus. J. NeuroL Sci., 103: 29-36. Poser, C. (1986) The pathogenesis of multiple sclerosis: a critical reappraisal. Acta Neuropathol., 71: 1-10.
152 Rosman, K., H. Jacobs and C. Van tier Merwe (1985) A new multiple sclerosis epidemic? S. AfT. Med. J., 68: 162-163. Scheinberg, L. (1987) Multiple Sclerosis: a guide for patients and their fatuities, Raven Press, New York. Schoenberg, B. (1978) Neurologk~.dEpidemiology, Raven Press, New York. Sosa, M., P. Betancor, C. Rosas et al. (1983) La asclerosis multiple en la provincia de las Palmas. Arch. Neurobiol. (Madrid), 46: 161-166. Shepherd, D. and A. Downie (1978) Prevalence of multiple sclerosis in North East Scotland. Br. Med. J., 2: 314-316.
Shepherd, D. and A. Downie (1980) A further prevalence study of multiple sclerosis in North East Scotland. J. Neurol. Neurosurg. Psychiat., 43: 310-315. Swingler, R: and A. Compston (1986) The distribution of multiple sclerosis in the United Kingdom. J. Neurol. Neurosurg. Psychiat., 49: 1115-1124. Vost, A., D. Wolochow and D. Howell (1964) Incidence of infarcts of the brain in heart disease. J. Pathol. Bacteriol., 88: 463-470. Wynn, D., M. Rodriguez, W. O'Fallon and L. Kurland (1990) A reappraisal of the epidemiology of multiple sclerosis in Olmsted County, Minnesota. Neurology, 40: 780-786.
143
JNS 03818
The epidemiology of multiple sclerosis: the Iceland model Onset-adjusted prevalence rate and other methodological considerations Charles M. Poser a, John Benedikz b and Patricia L. Hibberd c Departments of a Neurology and c Medicine, Harvard Medical School, a Neurological Unit, Beth Israel Hospita~ Boston, MA, USA, b Department of Neurology, University of Iceland School of Medicine, and the Neurological Service, Landspitalinn, Reykjavik, Iceland, and c Section of Infectious Diseases, Massachusetts General Hospital, Boston, MA, USA (Received 27 December, 1991) (Revised, received 9 March, 1992) (Accepted 12 March, 1992)
Key words: Epidemiology; Multiple sclerosis; Iceland; Onset adjusted prevalence; Age adjustment; Comparison populations; Ethnic diversity Summary The epidemiology of multiple sclerosis (MS) is characterized by the fact that there is an uneven distribution of the disease throughout the world. The two most commonly used indices of its frequency are the incidence and prevalence rates. The incidence rate reflects, to a great extent, the influence of environmental factors in triggering the clinical manifestations of the disease, if it is based upon the actual date of the initiation of symptoms, rather than the date of diagnosis. The prevalence rate is currently based on the date of diagnosis and includes all MS patients who are alive on a particular date, without regard to their ethnic origin, the site and the duration of residence or any other factors that may have influenced the acquisition of the disease. We propose that in order to make the concept of the prevalence rate possibly more meaningful, the term should refer, retrospectively, to all patients whose symptoms eventually led to the diagnosis of MS, even though the diagnosis was not yet established on an earlier prevalence day. In addition, only patiems of the same ethnic background who have spent their prepuberal years in the geographical area under study should be included. We are calling this measure the onset-adjusted prevalence rate. Another problem encountered in epidemiological studies of MS relates to the common practice of adjusting prevalence data obtained for age and sex in one area to what is referred to as a "standard" American (or world) population, groups of great ethnic and age diversity. It is also curious that in many studies the bases for comparison of populations are census data obtained many years previously. We suggest that age and sex adjustment should be apphed only to similar ethnic groups born and raised under different environmental conditions. We believe that data obtained by calculating an onset-adjusted prevalence rate restricted to a homogeneous group of patients sharing the same environment during the prepuberal years may provide valuable etiological clues.
Introduction Despite the large number of epidemiological studies of multiple sclerosis (MS) that have been published during the past 50 years, many of the conclusions from the data remain or have become controversial, with the
Correspondence to: Charles M. Poser, MD, Neurological Unit, Harvard Medical School, Beth Israel Hospital, 330 Brookline Avenue, Boston, MA 02215, USA. Tel.: (617)735-2063; fax: (617)735 5216.
possible exceptions of the classic studies of Alter et al. (1966) Dean and Kurtzke (1971), and Detels et al. (1978) regarding the significance of age at migration. It is difficult to compare these data because of the lack of uniform diagnostic criteria, uncertainty of the length of time elapsing between the date of symptomatic onset and diagnosis, deficiencies in case finding, and enormous variability in the clinical presentation and course of the disease. In addition, there are problems in applying classical epidemiological methods and procedures to a disease subject to a combination of genetic and environmental factors of debatable etiology and
144 uncertain pathogenesis. This article will address two main areas: that of the calculation of incidence and prevalence, and the traditional epidemiological practice of age-adjustment to what are considered to be standard populations. These considerations were derived, in part, from a recently completed study of the epidemiology of MS in Iceland (Benedikz et al. 1991).
Definitions
The two most commonly used measurements for estimating the frequency of the disease are incidence and prevalence. Schoenberg (1978) wrote that "incidance measures the frequency of addition of new cases of a disease within a specific population; it is calculated for a given time interval and a given place". Kurtzke (1985), another widely respected neuro-epidemiologist, defined incidence as "the number of cases of the disease beginning in a unit of time within the specified population. This is generally given as an annual incidence rate in cases per 1000 population per year". Neither of these definitions specifies whether cases should be counted as "beginning" when the diagnosis of MS was established, (presumably by a neurologist), or retrospectively, as "beginning" when the symptoms of the disease subsequently diagnosed as MS had first been noticed. Therefore, data are published using either date of diagnosis or date of onset. The year of diagnosis is not etiologically relevant and simply represents the random time at which cases happen to be seen by a neurologist who recognizes the clinical diagnostic significance of symptoms and signs. In fact, our Iceland data (Benedikz et al. 1991) clearly demonstrated that case ascertainment is directly related to the availability of neurologists in the area. Prevalence, according to Schoenberg (1978) measures "the frequency of all current cases of disease (existing cases and new ones) within a specific population and is calculated for a given time and given place", while for Kurtzke (1985) "the point prevalence rate refers to the number of people affected [by the disease] at one point in time within the community expressed per unit of population". Thus only living patients are counted on prevalence day. It is curious that texts or articles dealing with the epidemiology of MS have rarely mentioned the possible etiological significance of incidence or prevalence rates. The figures have been used either to measure changes in the frequency of the disease over time, or to compare groups or geographic areas. None the less, incidence and prevalence represent quite different indices: incidence reflects environmental factors that bring out symptoms in individuals already afflicted by either the MS trait (Poser 1986) or the disease itself in an asymptomatic form; prevalence, on the other hand,
calculated according to the traditional definition, simply means the number of cases known at a certain time. It sheds no light on possible etiological or pathogenetic factors. We believe that the results of our epidemiological study of MS in Iceland (Benedikz et al. 1991) may provide a basis for redefining both incidence and prevalence, and indicate the potential significance of prevalence rates in terms of MS predisposition and etiology. The characteristics of the population of Iceland provide ideal material for such a study for the following reasons: the total population of the island, approximately 250 000 people, is extraordinarily homogeneous, being of mixed Norwegian and Irish descent with almost no immigration in the last 1000 years. The health services are excellent and are available free to all inhabitants. The level of education is remarkably high and literacy is 100%. There are 10 neurologists practicing in the country, all of whom have received advanced specialty training in Scandinavia, the United Kingdom or the United States. Many Icelanders have the disease, so that its manifestations are well known to most physicians, the degree of suspicion is high, and referral to neurologists is early and frequent. A MS registry has been in operation for over 30 years. A total of 252 MS patients had been identified in Iceland from the beginning of the century until December 31, 1985. All but four of them were born and raised in Iceland, with their Icelandic ancestry on both sides going back many generations. The four who were excluded from consideration had been born in Denmark, the Faroe Islands, England, and Iran. Approximately half the inhabitants of Iceland live in an urban environment: the metropolitan area, which includes Reykjavik and its suburbs, and Keflavik. The rest of the island is considered rural, since the only town of any size, Akureyri, has approximately 15 000 inhabitants.
Incidence data
Starting in the mid-1930s figures on incidence have been collected. The date of symptomatic onset was determined as accurately as possible, often a very short time after the initial manifestations of the disease, especially after the first neurologist began practicing in Reykjavik in 1936. Fig. I shows that there has been no significant change in the mean annual incidence of MS based on the year of onset from 1936 to the end of 1985, the rate ranging between 1.8 and 2 . 7 / y e a r / 100000. The mean annual incidence based on the year of diagnosis ranges from 2.2 to 3.4/year/100000, with a statistically significant but biologically meaningless
145 Rate and 95% Confidence Limits (Poisson)
5
Rate a n d 95% Confidence Limits (Poisson)
100
4.34 °
~
4
O
..............................................................................................................
81,8
8O .............................................................
s 3
=~
~
°=:!: 2
3.4~
.....~-~§ . . . . . . . . . . . . . . . . . . . . . . .
r[
I
2.89
I
oJ
";;; .............. T ................. I~:i~°
.....
................ . . . . . . . . . . . . . . . . . . . . . . .
. . . . .
62.4
60
~ 4o
,
~ 43.2 ,~-
~ 42.9
~0:2-i~ .~
r~
~.~_i
[
2 10 .....
. . . . . . . . . . .
. . . . .
N 2o
12-31-1955 12-31-1965 12-31-1975 12-31-1985
0 19~6-~945
1946-19s-~
19ss-~gs5
196s-1975
197s-!98s
Year Intervals
Prevalence Day Fig. 3. Iceland: prevalence of multiple sclerosis by year of diagnosis 1955-1985.
Fig. 1. Iceland: mean annual incidence of multiple sclerosis by year of onset 1936-1985.
Rate and 95% Confidence Limits (Poisson)
5
4.26 ..........................................................................
=. °°4
.q29
~
646
.............................
ill
~
i
13.4o
.... .........X.................................................... .........tl .... o
..........
..............
..........
..........
....
0 1~.%-1945
1946.fgss
19s6-196s
tges-197s
197-~-199s
Year Intervals Fig. 2. Iceland: mean annual incidence of multiple sclerosis by year of diagnosis 1936-1985.
difference (Poisson) between the decades 1956-1965 and 1975-1985 (Fig. 2). Because the year of diagnosis may lag a long time after the time of clinical onset, it provides no meaningful information regarding etiological or environmental factors. It is difficult to understand why an unusual number of cases of MS diagnosed at a particular time are referred to as an epidemic, but this is what has
happened in South Africa (Rosman et al. 1985) and in Kenya (Adam 1989). When the latency between clinical onset and the first visit to the neurologist becomes short enough, as it did in Iceland (Benedikz et al. 1991), the incidence rate of MS remains the most accurate method of calculating the frequency of the disease provided the following conditions are met: that it is based on the estimated date of onset rather than the date of diagnosis and that persons of different ethnic background and those who have moved to the study area with already manifested symptoms of MS (even though still undiagnosed) be excluded from the count. Furthermore, the diagnosis must be established by a neurologist, preferably one with extensive experience in MS, according to generally accepted criteria. Only definite and probable cases should be included. The Icelandic study met all these conditions and demonstrated that the rate of incidence based on date of onset remained unchanged for the last 30 years.
Prevalence data Almost all studies evaluating the prevalence of MS have used the Schoenberg-Kurtzke definition, i.e., counting the number of patients who have been diag. nosed as having MS who are alive on prevalence day and calculating the rate on the basis of the most recent
TABLE 1 PREVALENCE OF MS IN ICELAND BY YEAR OF DIAGNOSIS Prev. day
Number of cases
12/31
New cases
Deaths
Cases alive
1955 1965 1975 1985
61 a 30 59 78
8a 18 11 22
53 65 113 169
a New cases and deaths in entire period prior to Dec. 31, 1955.
Population
Prevalence per 100000 (95% confid, lim.: Poisson)
160633 193 758 219033 240 741
33.0 (24.8-43.2) 33.5 (26.0-42.9) 51.6 (42.6-62.4) 70.2 (60.2-81.8)
146 TABLE 2
I00 L%.
~ A •
MULTIPLE SCLEROSIS IN ICELAND: DURATION OF DISEASE MEAN AND MEDIAN BY DECADE
MiIO Moderate Severe
O0
Patients who died before
Number of cases
Mean (years)
Median (years)
Range (years)
Dec Dec Dec Dec
8 18 11 22
21.1 27.8 27.8 34.1
19 29 26 32
2-51 9-51 9-44 5-73
31, 31, 31, 31,
1955 1965 1975 1985
6o ,
~
40
~
20
census data. According to this method, the prevalence data for MS in Iceland are shown in Table 1 and Fig. 3. A statistically significant increase would appear to have taken place between 1955 and 1985. Increases in prevalence have been noted in many areas and are ascribed mainly to the increasing longevity of MS patients, presumably due to improvement in supportive medical care. This has been true in Iceland as shown in Table 2: the mean duration of illness increased from 21 years in 1955 to 34 years in 1985. Another reason may be related to Acheson's (1977) comment that "because MS patients characteristically may have symptoms without confirmatory new rologic signs this may lead to overlooking, at least at onset, patients with a mild form of the disease". The importance of this factor is demonstrated by the changes in the Kurtzke scores of Icelandic MS patients since the beginning of the century (Fig. 4). There is an increase in the proportion of milder cases diagnosed as MS, which may account for the low incidence before the advent of trained neurologists. A very similar situation may have occurred elsewhere, for instance in the Faroe Islands before World War II. The Icelandic data represent an ethnically and geographically homogeneous population, unaffected by immigration or important environmental changes.
Onset-adjusted prevalence This concept arose from the study of two publications on the prevalence of MS in Lanzarote in the
0
1920
,
1930
i940
1950
1960
i970
1900
1990
Year
Fig. 4. Iceland: total number of MS cases at the end of each decade grouped according to Kurtzke Disability Scale: mild = 0-3, moderate = 4-6, severe = 7-9.
Canary Islands (Table 3). In the first survey in 1981, Sosa et al. (1983) reported having found two cases of MS with a prevalence of 3.3/100000. A second survey 6 years later (Garcia et al. 1989) included nine new cases, with a calculated prevalence of 15.0/100000, a highly significant increase. The second publication (Garcia et al. 1989), however, made note of the fact that 6 of the 9 patients "newly discovered" in 1987 had already had definite symptoms of MS, although they had not been diagnosed as such, at the time of the 1981 survey. Adding these 6 to the original 2 for a total of 8 gives a 1981 onset-adjusted prevalence of 13.3/ 100000 (Table 3). At a later date Reyes-Yanez mentioned (personal communication) that the 2 original patients found in 1981 had left the island before the 1987 survey to seek medical care in another one of the Canary Islands but were still alive. Correcting for those 2 additional cases, the 1987 prevalence was 18.3/ 100000, a difference that is not statistically significant from the 1981 prevalence. These studies led to our conviction that prevalence should include all patients affected by MS, i.e. who had symptoms of MS and who would eventually be diagnosed as having the disease, although such diagnosis
TABLE 3 PREVALENCE OF MULTIPLE SCLEROSIS IN LANZAROTE, CANARY ISLANDS Population: 60000; Latitude: 28°N. Year of survey
Number of cases
Prevalence per 100000
Correction
Onset-adjusted prevalence (CL)
1981 e 1987 b
2 9
3.3 15.0
+6 b= 8 + 2 c----11
13.3 (5.7-26.2) 18.3 (9.1-32.8)
a Sosa et aL (1983). b Garcia et al. (1989): 6 of the 9 patients had onset of MS before 1981 survey. c Reyes-Yanez, M., personal communication, 1991: the 2 patients found in 1981 had left before 1987.
147 TABLE 4 ONSET-ADJUSTED PREVALENCE OF MS IN ICELAND 1955-1985 4"
Prey. day
Number of cases
12/31
Clinical onset
Total cases
Deaths
Cases alive
1955 1965 1975 1985
103 a 36 44 65
103 139 183 248
8a 18 11 22
95 113 b 146 189
Population
Prevalence per 100 000 (95% confid, iim.: Poisson)
160633 193 758 219033 240 741
59.1 (48.1-72.7) 58.3 (48.1-70.6) 66.6 (56.3-78.9) 78.5 (67.9-90.7)
a Clinical onsets and deaths during entire period before Dec. 31, 1955. b Number of cases on previous prevalence day, minus deaths° and plus cases with clinical onset during previous decade.
might not have been made on prevalence day. It also became clear that more meaningful data would be obtained when prevalence is based on a well defined, ethnically homogeneous population who had spent the critical years of putative acquisition of the disease (i.e., between the ages of 5 and 15 years, or puberty) in the geographical area under study. This would then exclude all persons who had moved into the area after that age, with or without the symptoms of MS, as well as those of different ethnic or racial origin. Under such conditions, prevalence rates would more precisely define the role of genetic susceptibility as well as suggest clues regarding potentially significant environmental factors affecting the primary acquisition of the illness. Because all these conditions are fulfilled in Iceland, the calculation of onset-adjusted prevalence rate (OAPR) was carried out (Table 4, Fig. 5). It is clear that no statistically significant changes have taken place between 1955 and 1985. We have also calculated that in order for a statistically significant difference to exist between these years, a net increase of 30 cases with onset between January 1, 1976 and December 31, 1985 would have to be discovered. This is most unlikely to occur in view of the fact that the latency between
ICELAND 1955-1985
symptomatic onset and diagnosis has been reduced to less than 5 years during the past 15 years (Benedikz et al. 1991). Only 6 MS patients have had their first symptoms since January 1, 1986. All the MS data based on age adjustment to the general Icelandic populations are identical to those cited here. Theoretically, all symptomatic MS cases should be included in the prevalence calculations, including those who had died. In order to retain those cases, the number of deaths in the general population during the study period should be included in the denominator, but such numbers are difficult to obtain. For this reason, the calculation of the OAPR also subtracts those MS patients who died prior to prevalence day, assuming that these deaths represent the same proportion of deaths in the general population. On the other hand, all MS patients who had met all the conditions outlined above, but who had left the area for whatever reason and are known to be alive should also be counted, as exemplified by the 2 patients who had left Lanzarote for another of the Canary Islands. In order for OAPR to be calculated, a long enough period of time must have elapsed after the selected prevalence day to insure that the vast majority of cases would have had their clinical onset during the study period or prior to it and have eventually been recognized as suffering from MS.
Rate and 95% Confidence Limits (Polsson) 100
90.7 o o o
80
n_gt7
70s •
.7e.9............7e,~ -7
66.6
60
v 40 =
....2
2
-.3
t,u
~ 2o
12-31-1955
12-31-1965
12-31-1975
12-31-1985
Prevalence Day
Fig. 5. Iceland: onset-adjusted prevalence of multiple sclerosis 19551985.
Discussion: prevalence rates
We are aware of only a single published attempt to correct prevalence data obtained by the classical method: in their study of the changes of frequency of MS in Denmark, Koch-Henriksen and Hyllested (1988) applied an ill-defined "notification delay" corrective factor. Because prevalence rates classically include all living patients, difficulties of interpretation arise with regard to immigrants, i.e. individuals with MS who did not spend the critical prepuberal years in the study area. This is illustrated by the Australian study of Hammond et al. (1988). The Australian data comparing preva-
148 that of the 4 articles dealing with the same population of Northeast Scotland. First, but not unique to these studies, is the inclusion of "possible" as well as "probable" MS cases, the latter including what most other classifications would call "definite". Three of the articles report on surveys carried out in 1970 (Shepherd and Downie 1978), 1973 (Shepherd and Downie 1980) and 1980 (Downie and Phadke 1983); the fourth one (Swingler and Compston 1986) included figures from the previous 3 studies in a general review of MS prevalence in the United Kingdom. In all 3 original surveys (1970, 1973 and 1980) 3 diagnostic categories of MS were used: probable, early and latent probable, and possible; all three were included in the calculations. It was possible to calculate OAPR for these surveys from the content of the articles. The details have been reproduced in Table 5. The original 1970 survey reported a prevalence rate of 127/100000 based on 557 cases (all categories) but 105/100000 for probable cases only. In the 1980 publication (Shepherd and Downie 1980) it was noted that 50 cases had already had enough symptoms at the time of the 1970 survey that they should have been counted as cases of probable MS. Therefore, the OAPR for the 1970 survey would be 117/100000. The 1973 survey reported a prevalence rate of 144/100000, which approaches a significant difference from the previous 1970 survey. However, when the cases are restricted to the probable ones, the corrected OAPR is 117, exactly the same as in 1970. The last survey, in 1980 (Downie and Phadke 1983) reported a prevalence of 178/100000 (all categories) with a statistically highly significant increase over the previous two surveys. However, 25 individuals already suffering from the disease had moved into the area after the 1973 survey. Subtracting those and basing the new calculation only on probable cases, the resulting corrected OAPR is 139/100000, which is barely statistically significantly different from the other
lence between 1961 and 1981 include only patients carrying the diagnosis of MS. It was not possible to calculate the OAPR from these data. Although they did calculate rates separately for native-born and foreigu-born patients, and specified how many of their patients already had MS at the time of immigration, they did not indicate at what age the still asymptomatic immigrants had come to Australia. Such data would be of particular interest in view of the fact that the highly significant difference in prevalence between Perth and Newcastle on the one hand, and Hobart on the other, appear in one of the very few recent studies that Support a putative relationship between MS frequency and latitude. The same questions apply to the Wynn et al. (1990) study of the frequency of MS in Olmsted County, Minnesota: the data are based on already diagnosed MS patients and the only exclusion was of patients who had resided in the county for less than one year. No comment was made about the possibility that MS patients, some already diagnosed, some with still undiagnosed symptoms of the disease, had come to the area prior to the exclusion period. The problem of immigration was taken into consideration by Middleton and Dean (1991) in their study of the population of the Greek part of the divided island of Cyprus (latitude 34°N). They counted all diagnosed MS patients of Greek descent who had been born in the area and excluded refugees from the Turkish-occupied region. They included one Englishwoman who had acquired Cypriot nationality but excluded all foreigners. They found a total of 42 patients, including 1 refugee and the Englishwoman. On the basis of 40 MS patients in a population of 92 100 the prevalence was 43.4/100000, once again raising grave doubts about the significance of latitude. Unfortunately, no data were given about time of clinical onset. Another illustration of the difficulties that arise in trying to interpret published epidemiological data is
TABLE 5 PREVALENCEOF MULTIPLESCLEROSISIN NORTH-EASTSCOTLAND. Year of survey Population All cases Probableonly Corrected probable Prevalence rate corrected rate Confidence limits (Poisson)
1970(1) 440200 557
1973(2) 440200 634 464
517 517
514 b 127 117 138
105 96 115
1980(3) 471350 a 839
144 117 107 127
133 156
682
464
517
682
105
117
155
96 115
107 127
143 167
657 c 178
117 107 127
1986(4) 440200
166 191
145 135 156
139 129 147
References: (1) Shepherd and Downie (1978). (2) Shepherd and Downie (1980). (3) Downie and Phadke (1983). (4) Swingler and Compston (1986). Notes: Underlined numbers prevalencegiven by authors. a Populationcalculatedfrom prevalence rate of 1978. b Addition of 50 eases missedat time of 1970survey. c Subtractionof 25 eases who moved into area after 1973surveybut had alreadyacquired MS.
149 two surveys. Finally, in a compilation of published data by Swingier and Compston (1986), only probable cases were used, but no corrections made, so that the prevalence rates that were published gave credence to a marked but questionable increase in prevalence from 1973 to 1980. The major use of this proposed method of OAPR is to determine if significant changes have occurred over a period of time in ethnically homogeneous populations, counting only individuals who have spent their pre-pubertal years in the geographical region under study.
What is true prevalence?
Epidemiologicai studies can deal only with those cases of disease that have produced symptoms and as a result, must ignore what may be a significant number of cases that will never be diagnosed. Some idea of the numbers of unsuspected, completely asymptomatic cases of MS in the population can be gathered from autopsy data. Vost et al. (1964) in Montreal found 5 cases of unsuspected MS in 440 patients most of whom had died of heart disease, an incidence of 1.25% in this selected group. Gilbert and Sadler (1983), in Ontario, found that 0.2% of 2450 autopsies examined ncuropathologically had typical lesions, despite no history suggestive of MS. Engell (1980) in Denmark found evidence of MS in 0.1% o~ cases where the nervous system was examined pathologically because of malignancy or cerebrovascular disease. In addition to the fact that the spinal cord was not studied in those cases, it is not known what fraction of the total autopsies is represented by the 16 000 cases examined, nor is it clear if these cases were representative of the population at risk for MS. On the basis of such data, assuming the Banm and Rothschild (1981) estimate of 123000 cases, or even Scheinberg's (1987) estimate of
TABLE 6 STANDARDPOPULATIONSBY AGE 1966a Age groups (years)
African %
European %
World %
USAb %
O- 9
20
15
22
15
10-19 20-29 30-39 40-49 50-59 60-69 over 70
20 20 20 10 5 3 2
14 14 14 14 13 9 7
18 16 12 12 9 7 4
14 17 17 12 9 9 8
a Modifiedfrom Doll et aL (1966). b Total US population 1988 from Statistical Abstracts of the US 1990.
200000 cases of MS in the United States, i.e. between 0.05 and 0.1% of the total population, this implies that the number of asymptomatic cases is at least equal to that of known instances of MS. This may well be true in other countries of high prevalence. There does not seem to be any method available for correcting for the unavoidable underestimate of MS prevalence.
Adjustment to a standard population
Data on MS incidence and prevalence from various parts of the world are often adjusted to what are considered standard populations in terms of age distribution. This practice makes a great deal of sense when dealing with mortality rates which are clearly influenced by the population's age distribution (Lilienfeld and Ldienfeld 1980). Such adjustment has been used extensively in cancer epidemiology, but the advisability of doing so has been questioned by Doll et al. (1966): "Cancer incidence varies by age, especially in terms of type. Clearly therefore comparisons of total incidence rates uncorrected for age are of little value in the study of etiology. The most suitable comparisons are those made directly between the individual age specific rates. The traditional way of overcoming this difficulty introduced by differences in age distribution of the population is by the use of rates standardized for age; that is, rates which, if calculated would have been observed if the age-specific rates recorded by each registry had occurred in a standard population with a specified age distribution. But this in fact does not happen. In the study of etiology, therefore, standardized rates are little better than crude rates. They reduce the effect of differences in age distribution; but they do not eliminate it, and if their limitations are forgotten they may be grossly misleading". There are two standard populations that are com. monly used for age adjustment. One was devised by the International Union Against Cancer (Table 6). Although it has the redeeming feature of providing specific data for Africa and for Europe, there are important differences in age distribution between those two and the population of the world as well as that of the United States. The other standard population is that of the United States. It is difficult to understand the rationale for this because there are few, if any, populations that are more heterogeneous. Major differences in ethnic composition exist in many regions of the country (Table 7). An illustration of this problem is provided by the study of Wynn et al. (1990) of MS in Rochester and Olmsted County, Minnesota, in which the crude rates were adjusted to the white population of the United States in 1950. Table 8 shows how different the Minnesota White population is from that of the entire United States. The same study also illus-
150 TABLE 7 ETHNIC DIVERSITY OF THE UNITED STATES
Persons who reported a single ancestry group ~ North east (%)
North central (%)
South (%)
West (%)
0.4 1.2 20.0 15.1 8.7 1.1 1.0 1.I 47.5
0.9 0.1 10.9 12.0 11.4 0.3 1.0 0.7 37.3
0.6 2.0 15.5 28.6 7.6 2.5 0.8 1.9 59.5
0.1 0.1 30.8 9.2 8.9 0.2 0.9 0A 50.6
0.8 1.3 17.6 12.4 6.4 1.7 1.3 1.7 43.2
(B) Medium MS susceptibility Czech 0.7 French 2.6 Greek 0.5 Hungarian 0.6 Italian 5.8 Polish 3.2 Portuguese 0.5 Russian 1.2 Total 15.1
0.5 3.9 1.0 1.1 15.9 6.0 1.3 2.6 26.9
1.3 1,8 0,5 0.8 3,3 5.1 0.0 0.7 13.5
0A 2.7 0.3 0.3 1.9 1.0 0.1 0.6 7.3
0.5 1.8 0.4 0.5 3.6 1.3 1.0 1.1 10.2
12.4 0.4 0.7 0.4 0.2 1,6 0.2 0.1 3,5 19,5
15.4 1.3 0.2 0.1 0.2 0,4 0.2 2.1 0.5 20.4
27.2 2.0 0.2 0.8 0.2 1.0 0,1 6,3 0.2 38,0
8.0 3.1 1.8 0.2 1.9 3.3 2.3 17.2 0.3 38.1
Ethnicity
US (%)
(A) High MS susceptibility
Danish Dutch English German irish Norwegian Scottish Swedish Total
(C) Low MS susceptibility
Afro-Amer. Amer. Ind. Chinese Cuban Filipino Hispanic Japanese Mexican Puertorican Total
17.3 1.6 0,6 0,4 0,5 1.4 0.6 5.9 1.1 29,4
a Groups over 400000 population in entire US from 1980 United States Census.
trates another rather curious tradition: the use of the 1950 population as a basis for comparison for data gathered almost 40 years later. Table 9 indicates how the US population has shifted towards the older age groups. One of the most striking examples of the inappropriate application of the custom of age-adjustment to the United States population was the conclusion reached by Leibowitz et al. (1973) that the Israel-born children of both Ashkenazi and Sephardi immigrants had the same prevalence of MS, whereas the "raw" data indicated that MS was 4 times more common among the former, as it had been for their immigrant parents. More recent data (Biton and Abramsky 1983) confirm tile considerably higher prevalence of MS among the Ashkenazis than among Sephardis although all were born and raised in Israel. Because it is now generally accepted that MS is acquired between the ages of 5 and 15, (or puberty), on the basis of migration studies (Alter et al. 1966; Dean
and Kurtzke 1971), age beyond that period of time plays little if any role in considerations of etiology. Adjustment for ethnic origin would appear to be much more important. In terms of clinical onset, there is no question that the disease is more likely to manifest TABLE 8 PERSONS WHO REPORTED AT LEAST ONE ANCESTRY a UNITED STATES CENSUS 1980 (PCg0-sI-10)
European population only. Ancestry
Minnesota (%)
US (%)
Danish Icelandic Norwegian Scandinavian Swedish
1.90 0.04 12.93 1.28 9.58
0.74 0.02 1.67 0.23 2.11
Total
25.73
4.7
a 83.1% of US population.
151 TABLE 9 POPULATION OF THE UNITED STATES BY AGE GROUPS. Age groups
1950 a
1988 b
(years)
Total (%)
White (%)
Black (%)
Total (%)
White (%)
Black (%)
0- 9 10-19 20-29 30-39 40-49 50-59 60-69 over 70
16.1 18.3 17.2 15.0 13.0 9.9 6.5 4.0
15.7 18.0 17.1 15.0 13.1 10.2 6.7 4.1
19.8 20.5 18.2 15.4 11.7 7.4 4.6 2.5
14.8 13.7 16.8 16.7 11.9 8.9 8.5 8.3
14.2 13.5 16.4 16.6 12.1 9.2 8.9 8.7
18.5 17.9 18A 16.1 9.9 7.7 6.2 5.3
a US Census 1950. b Statistical Abstracts of the US 1990.
itself during the 3rd and 4th d e c a d e s o f life, but the value o f age-adjustment must be g a u g e d in terms o f the extreme difficulty o f finding a truly c o m p a r a b l e p o p u lation. It would be theoretically possible and interesting to c o m p a r e the age-specific prevalence and incidence rates o f MS a m o n g individuals o f Scandinavian descent b o r n and raised in M i n n e s o t a with those in Scandinavia, o r a m o n g Israeli Jews a n d Jews in o t h e r parts o f the world.
Conclusion Because MS often b e c o m e s s y m p t o m a t i c in patients long before the diagnosis is m a d e , it is imperative that such patients be included in all calculations o f prevalence. Obviously, MS patients can be identified only after the diagnosis has b e e n established and therefore the calculations must be retrospective. O A P R must also be restricted to patients w h o share similar ethnic backgrounds a n d have resided in the same area during the putative period o f acquisition o f the disease. T o be meaningful for o t h e r populations, age adjusting the prevalence d a t a so obtained m u s t be subject to the same restrictions. It is suggested that O A P R d a t a according t o these criteria m a y provide important clues about the acquisition o f MS. Acknowledgments This project was supported, in part, by grant No. PP0080 from the National Multiple Sclerosis Society (USA). The generous financial contributions made by Hyman Abadi (Bogota, Colombia), Lorne and Luise MacArthur (Waltham, MA), and Paul and Carol Peterson (North Attleboro, MA), as well as the excellent typing skills of Sheila Golden, and the judicious editorial advice of my wife, Joan Crawford, are all most gratefully acknowledged. We are also indebted to Professors Johan Aarli of Bergen, Norway, and Geoffrey Dean of Dublin, Eire, for their constructive comments.
References Acheson, E. (1977) The epidemiology of multiple sclerosis. Br. Med. J., 33: 9-14.
Adam, A. (1989) Multiple sclerosis: epidemic in Kenya? E. Afr. Med. J., 66: 503-5C~i. Alter, M., U. Leibowitz and J. Speer (1966) Risk of multiple sclerosis related to age at immigration to Israel. Arch. Neurol., 15: 234-236. Bantu, H. and B. Rothschild (1981) The incidence and prevalence of reported multiple sclerosis. Ann. Neurol., 10: 420-428. Benedikz, J., H. Magnusson, C. Poser, E. Benedikz, G. Olafsdottir and G. Gudmundsson (1991) Multiple sclerosis in Iceland: 19001985. J. Trop. Gcogr. Neurol., 1: 16-22. Biton, V. and O. Abremsky (1986) Newer study fails to support environmental factors in etiology of multiple sclerosis (abstract). Neurology, 36 (Suppl. 1): 184. Dean, G. and J. Kurtzke (1971) On the risk of multiple sclerosis according to age at immigration to South Africa. Br. Med. J., 3: 725-729. Detels, R., B. Visscher, R. Halle, R. Malmgren, J. Dudley and A. Coulson (1978) Multiple sclerosis and age at migration. Am. J. Epidemiol., 108: 386-393. Doll, R., P. Palme and J. Waterhouse (1966) Cancer incidence in five continents, International Union Against Cancer, Springer-Verlag, Berlin. Downie, A. and J. Phadke (1983) The Chief Scientist reports...: multiple sclerosis in North East Scotland. Health Bull. (Edinburgh), 42: 151-156. Engell, T. (1989) A ciinico-patho-anatomical study of clinically silent multiple sclerosis. Acta Neurol. Scand., 79: 428-430. Gareia, J., S. Rodriguez, M. Sosa et al. (1989) Prevalence of multiple sclerosis in Lanzarote (Canary Islands). Neurology, 39: 265-267. Gilbert, J. and M. Sadler (1983) Unsuspected multiple sclerosis. Arch. Neurol., 40: 533-536. Hammond, S., J. McLeod, K. Millingen et al. (1988) The epidemiology of multiple sclerosis in three Australian cities: Perth, Newcastie and Hobart, Brain, 111: 1'25. Koch-Henriksen, N. and K. Hyllested (1988) Epidemiology of multipie sclerosis: incidence and prevalence rates in Denmark 19481964 based on the Danish Multiple Sclerosis Registry. Acta NeuroL Scand., 78: 369-380. Kurtzke, J. (1985) Epidemiology of multiple sclerosis. In: Koetsier, J. (Ed.), Handbook of Clinical Neurology, vol. 3, Elsevier Science Publ., Amsterdam, pp. 259-287. Leibowitz, U., E. Kahana and M. Alter (1973) The changing frequenoy of multiple sclerosis in Israel. Arch. Neurol., 29: 107-110. Lillienfeld, A. and D. Lillienfeld (1980) Foundations of Epidemiology, 2nd edn., Oxford University Press, New York. Middleton, L. and G. Dean (1991) Multiple sclerosis in Cyprus. J. NeuroL Sci., 103: 29-36. Poser, C. (1986) The pathogenesis of multiple sclerosis: a critical reappraisal. Acta Neuropathol., 71: 1-10.
152 Rosman, K., H. Jacobs and C. Van tier Merwe (1985) A new multiple sclerosis epidemic? S. AfT. Med. J., 68: 162-163. Scheinberg, L. (1987) Multiple Sclerosis: a guide for patients and their fatuities, Raven Press, New York. Schoenberg, B. (1978) Neurologk~.dEpidemiology, Raven Press, New York. Sosa, M., P. Betancor, C. Rosas et al. (1983) La asclerosis multiple en la provincia de las Palmas. Arch. Neurobiol. (Madrid), 46: 161-166. Shepherd, D. and A. Downie (1978) Prevalence of multiple sclerosis in North East Scotland. Br. Med. J., 2: 314-316.
Shepherd, D. and A. Downie (1980) A further prevalence study of multiple sclerosis in North East Scotland. J. Neurol. Neurosurg. Psychiat., 43: 310-315. Swingler, R: and A. Compston (1986) The distribution of multiple sclerosis in the United Kingdom. J. Neurol. Neurosurg. Psychiat., 49: 1115-1124. Vost, A., D. Wolochow and D. Howell (1964) Incidence of infarcts of the brain in heart disease. J. Pathol. Bacteriol., 88: 463-470. Wynn, D., M. Rodriguez, W. O'Fallon and L. Kurland (1990) A reappraisal of the epidemiology of multiple sclerosis in Olmsted County, Minnesota. Neurology, 40: 780-786.