J Clin Epidemiol Vol. 50, No. 6, pp. 719-723, Copyright 0 1997 Elsevier Science Inc.
0895.4356/97/$17.00 PI1 SOS95+4356(97)00021-8
1997
ELSEVIER
Cohort
Study of Ethnic Group and Cardiovascular and Total Mortality Over 15 Years Aage Tverdal*
RESEARCH
DEPARTMENT,
NATIONAL
HEALTH
SCREENING
SERVICE,
N-0033
OSLO,
NORWAY
ABSTRACT.
who attended two cardiovascular screenings in 1974Altogether, 3365 men and 3266 women 1975 and 1977-1978, have been followed with respect to death for an average of 15 years. The mortality from coronary heart disease was lower in men who reported being of Lappish origin than in men who reported Norse origin at both screenings (Rate ratio: 0.38 [0.20-0.71]), w h en other major risk factors were taken into account. Among men without history of cardiovascular disease or symptoms of angina pectoris the rate ratio became 0.24 (0.09-0.63). With cardiovascular death as the endpoint, the corresponding rate ratios were 0.42 (0.25-0.73) and 0.31 (0.15-0.67). In women, a non-significant lower mortality was seen in the Lappish group than in the Norse group. This study is suggestive of some protection from coronary heart disease in middle-aged men of Lappish origin. J CLIN EPIDEMIOL 50;6:719-723, 1997. 0 1997 Elsevier Science Inc.
KEY WORDS:
Coronary
heart
disease,
mortality,
ethnicity,
INTRODUCTION A cardiovascular the
disease
northernmost
county
study
was carried
in Norway,
out in
in Finnmark,
1974-1975.
The
included questions of whether one or more grandparents were of Finnish or Lappish ancestry. An interesting result from the cross-sectional data was a much lower prevalence of a history of myocardial infarction among the Lapps than could be expected from the level of the coronary risk factors [I]. The purpose of this study was to assess whether this finding of a low myocardial infarction prevalence among the Lapps was reflected in a low coronary heart questionnaire
disease
mortality
MATERIALS
in
a follow-up
study.
AND METHODS
All male and female participants, who were aged 35 to 49 years, and a 10% random sample of people aged 20-34 were invited to a cardiovascular screening in Finnmark, the northernmost county in Norway in 1974-1975. This screening was repeated 3 years later in 1977-1978. Al1 people aged 35-52 years were invited. Of those invited to both screenings, 81% of the men and 87% of the women attended both times. A non-fasting blood sample was taken and the serum was analyzed
for
cholesterol,
triglyceride,
and
glucose
levels.
‘Address for correspondence: Aage Tverdal, Research Department, National Health Screenmg Service, P.O. Box 8155 Dep, N-0033 Oslo, Norway. Accepted for publication on 14 February 1997.
height,
Lappish,
Norse
The same laboratory was used both times and the accuracy and the precision of the analytical results have been reported by Bjartveit er al. [2]. HDL cholesterol was determined only at the second screening enzymatically from the cholesterol remaining in the supematant after precipitation of low density (LDL) and very low density (VLDL) particles by heparin-mangane reagent [3]. The samples were stored at -20°C before HDL determinations were made. The mean level in frozen sera has been shown to be 0.12 mmol/l lower than in fresh sera and, accordingly, 0.12 mmol/l was added to the values determined from the frozen sera [4]. The blood pressures were recorded twice with a sphygmomanometer, 4 mm apart. Diastolic pressure was recorded at phase 5, or phase 4 if phase 5 was lacking. The last of the two pairs of recordings was used in this study. Height and weight were measured to the nearest centimeter and half kilogram. Underwear, trousers, and stockings were allowed, but not shoes. A questionnaire was filled in at home and checked by a nurse at the screening site. The questions were about history of cardiovascular disease or diabetes, symptoms of angina pectoris or atherosclerosis obliterans, physical activity during leisure time, smoking habits, and history of heart infarction or angina pectoris in one or more ofparents or sisters or brothers. A previous cohort study revealed that persons who answered “don’t know” to this last question carried the same excess coronary risk as those who answered “yes” [5]. Therefore, these two groups have been combined as “yes.” The two questions that were asked about ethnicity read:
720
TABLE
A. Tverdal
1. Ethnic
origin
Origin
Group definition
Norse Finnish Lappish
of group
members
of two
or more
Lappish
grandparents Finnish
No No or don’t know Yes
No Yes No or don’t know
“Are two or more of your grandparents of Lappish (or Finnish) origin?” Three groups were compared in this study, as defined in Table 1. Of 11,998 men and women aged 35+ years at second screening, 7293 reported being of Norse, Finnish, or Lappish origin at one or both screenings. Of these, 663 1 (91%) gave identical answers, this is the study population. Of those excluded, 4056 did not know their ethnic origin and 649 reported to be of both Finnish and Lappish origin at one or both screenings. An additional questionnaire was handed out at the second screening that was intended to be filled in at home and then returned by mail. Of the study population, 86% returned this questionnaire. Among the questions were: (1) “Do you drink wine or liquor during one week?” (2) Do you drink beer during one week?” Also, a question about the number of cups of coffee drank was asked with the following categories: no coffee or less than 1 cup per day; l-2 cups; 34 cups; 5-6 cups; 7-8 cups; and 9+ cups. These categories were coded as 1 to 6. The type of coffee was not asked for.
TABLE
2. Characteristics
of study
Information about education and income from the population census in 1970 was linked to the file. Education was coded in nine categories (0 to 8) with 0 representing the lowest level (pre-school) and 8 representing the highest level (research) [6]. Income was defined as taxable income plus special deductions. People contributed person-time from date of screening in 1977-1978 to date of death, date of emigration, or 31 December 1992. All Norwegian death certificates are coded in the Central Bureau of Statistics according to a Norwegian adaptation of the International Classification (ICD) and strict rules for selecting the underlying cause. During 1977-1985, the eighth revision of ICD [7] was in use; during 1986-1992 the ninth revision was in use [8]. Deaths with the following codes for underlying cause have been used here: ICD-8: 410-411, 412.0-412.3, 413 and ICD-9: 410-413, 414.0414.1, 414.3, 414.9. Cardiovascular disease is defined as ICD-8: 390-458 and ICD-9: 390-459 [7,8]. Adjustments for confounders have been done with the Cox proportional hazards model [9]. The confounders were defined as the mean of values from the two screenings where appropriate. Otherwise, the values from the second screening have been used (Table 2).
RESULTS
The Finns had the highest mean level of blood pressure, total cholesterol, and cigarette smoking, and also the high-
population Men Norse (II = 2100)
Mean of two screenings Systolic blood pressure (mmHg) Diastolic blood pressure (mmHg) Total cholesterol (mmol/l) Triglycerides (mmol/l) No. of cigarettes/day BMI (g/cm’)” Height (cm)’ Physical activity (1 = low, 2, 3, 4 = high) Second screening Age (years) History of cardiovascular disease or symptoms (%) Family history (%) HDL cholesterol (mmol/l)b No. of cups of coffee/day‘ Beer last week (%)’ Wine last week (%)’ Information from census in 1970 Income (100 NOK) Education (0 = low, 1 thru 8 = high)
135 86 6.6 1.9 8.3 2.5 175 2.0
Finnish (n = 714)
139
Women Lappish (n = 551)
Norse (II = 2000)
130 81 6.5
2.0 9.0 2.6 173 2.1
134 83 6.9 2.0 8.5 2.5 166 1.9
45 11.3 40 1.3 4.3 26.2 36.8
44 13.0 48 1.3 4.6 29.4 35.3
338 2.8
277 2.4
87 7.0
aMen: n = 1958, 661, 480; Women: n = 1868, 677, 486 (Norse, Finnish, Lappish). bMen: II = 1963, 655, 481; Women: n = 1855, 658, 492 (Norse, Finnish, Lappish). ‘Men: n = 1816, 603, 442; Women: n = 1752, 622, 466 (Norse, Finnish, Lappish).
(n
Finnish = 714)
Lappish (n = 552)
2.4 162 1.9
133 83 6.7 1.5 5.2 2.5 160 1.9
131 81 6.6 1.4 3.2 2.7 154 1.7
45 9.4 34 1.4 4.6 30.1 29.6
45 11.4 40 1.5 4.0 7.0 14.9
45 14.1 49 1.5 4.1 5.0 12.4
44 12.5 40 1.5 4.1 3.2 5.6
194 2.2
a2 2.5
61 2.2
35 2.1
:::
Cohort Study
est prevalence of a history of cardiovascular disease, diabetes, or symptoms of angina pectoris (Table 2). Among men, the Lapps had the lowest prevalence of such disorders. The proportion who consumed wine or liquor during one week was lowest in the Lapps. The lowest proportion who consumed beer in a week was seen in Lappish women, whereas Lappish men reported the highest proportion. The Norse had the highest level of education and income. There was a huge difference in mean income between men and women due to the fact that in 1970 many women did not have work outside the household. In men, the lowest coronary heart disease mortality was seen in those of Lappish origin (Table 3). Adjustment for several possible confounders strengthened this pattern. Additional adjustment for use of beer, wine or liquor, cups of coffee, and HDL cholesterol did not substantially change the rate ratios in men (rate ratios: Finnish 1.07 [0.68-1.691 Lappish 0.49 [0.23-1.041 for total material, and Finnish 0.97 [OS-1.721, Lappish 0.29 [O.lO-0.851 for men without history of cardiovascular disease or symptoms). One should keep in mind that with these additional covariates the study population is approximately 10% smaller (Table 2). In women there were few coronary deaths with wide confidence intervals for rate ratios which all encompassed 1.0 (Table 3). The age-adjusted rate ratios of coronary or cardiovascular death were above 1.0 and became lower than 1.0 when the whole battery of confounders were included as covariates in the model. To examine whether the rate ratio of coronary death between the Norse and the Lappish was the same in men and women, two Cox models were run: with and without an interaction term of Lappish and sex. The two models did not differ significantly (p = 0.16). The body height was on the average much lower in the Lappish group (Table 2). One could imagine that low height expressed genetic purity. If there were a protection against coronary death in being Lappish, one might expect a decreasing mortality with decreasing height. Table 4 shows the association between height and coronary death, specifically for each ethnic group. An inverse relationship is seen in all three groups, and the rate ratios do not differ beyond chance. It is noted that the height gradient in the Norse became much weaker when adjusting for the set of risk factors, while the inverse height relationship was strengthened in the Lapps.
DISCUSSION This follow-up study revealed a lower coronary heart disease mortality in Lappish men than in men who reported to be of Norse or Finnish origin. In women the pattern was similar, but the estimates were more imprecise with confidence intervals encompassing a null relationship. The study population had attended two screenings three years apart. The requirement of a consistent answer at two independent occasions to the question of ethnicity has
made misclassification less likely. Nevertheless, as the Lapps is a minority group, one might expect them to be more prone to misclassify themselves. Anyhow, there is no reason to believe that any misclassification should be dependent on the outcome and, as such, one would expect the estimates in this study to be biased towards the null. The rate ratio (Lappish versus Norse) became smaller when taking several covariates into the model. This suggests that the lower coronary mortality cannot be ascribed to residual confounding. It should be borne in mind that most of the covariates were based on two independent measurements. We do not know the attendance rate in each ethnic group, as information about ethnicity was collected at the screening site. However, as the attendance rate was high in the study population, it seems unlikely that the more favorable mortality in the Lapps can be due to a more “healthy selection” of attendees among the Lapps. This is supported by the mean levels of risk factors (Table 2). Furthermore, the attendance rate in the municipalities where the majority of the population is Lapps was not different from the attendance rate in the other municipalities (data not shown). A large proportion of the screenees was excluded, primarily due to a statement of unknown ethnicity, but also due to discordant answers. However, in men the coronary and cardiovascular mortality was similar in the excluded group and the study population. Also, a complete crosstabulation of the ethnicity answers at the first and second screenings, given in [lo], showed a lower proportion of “Don’t know” answers to the Lappish than the Finnish question. Of those who answered “Yes” to the Lappish question at the first screening 84% answered “Yes” the second time as well. The comparable percentage for the Finnish question was 78. It therefore appears unlikely that the proportion of Lapps who answered “Don’t know” or gave inconsistent answers was high enough, and with a high enough cardiovascular mortality, to seriously influence the results. The majority of Lapps lives in a few municipalities, and we cannot prove that the diagnosis on the death certificate is as valid as in other municipalities. There is, however, no evidence to the contrary. Also, the total mortality rate is lower in Lapps, even if the deficit is lower than for cardiovascular mortality. The difference is largely made up by higher rates of violent and pulmonary deaths in Lapps. Mortality from sudden deaths, was similar in men of Norse and Lappish origin. The baseline characteristics are generally more unfavorable in Lappish than in Norse men, so if knowledge of the risk factor level has influenced the statement on cause of death, one would expect the Lapp rate ratio estimates to be biased toward the null. Finnmark county, where this study was carried out, has the highest coronary heart disease mortality in Norway [2]. The mortality varies between the counties by a factor of 2. However, most of this difference can be explained by differences in total serum cholesterol levels [S]. The differ-
Abbreviations: “Systolic hWithout
1.00 1.69 1.31
40 65 48 30 22 41
12 7 4
8 2 3
1.oo 0.75 1.54
1.00 1.09 0.33
246 254 80
67 23 6
1.00 1.19 0.50
354 398 171
57s
Ageadj.
(95% Cl)
for age and risk facto&
ratio
and total
1.00 0.27 (0.03-2.24) 0.75 (0.12-4.80)
1.00 0.80 (0.27-2.39) 0.88 (0.21-3.77)
1.00 0.88 (0.53-1.46) 0.24 (0.09-0.63)
1.00 0.91 (0.61-1.35) 0.38 (0.20-0.71)
Adj. other
Rate
death
cardiovascular,
107 41 14
Deaths
Rate/ 100,000
Coronary
of coronary, 1974-1978
phys~al
19 5 8
28 14 10
83 32 11
131 53 21
among
actiwty,
72 54 109
94 130 120
305 353 147
433 515 256
P-v
family
history
1.oo 0.77 1.64
1.00 1.42 1.35
1.oo 1.21 0.48
1.25 0.61
1 .oo
Ageadj.
Cardiovascular
group
Rate/ 100,000
by ethnic
Deaths
death
adj. = adjustment; Cl = confidence interval; p-yrs = person-years. blood pressure, total cholesterol, triglycerides, cigarettes/day, BMI, height, history of cardiovascular disease or symptoms.
Mean Total material Norse (30228) Finnish (10299) Lappish (8201) Men Without historyh Norse (27242) Finnish (9067) Lappish (7503) Women Total material Norse (29805) Finnish (10751) Lappish (8344) Women Without historyh Norse (26443) Finnish (9259) Lappish (7315)
Group h-w-)
TABLE 3. Rates and rate ratios in Finnmark county, Norway,
(95% Cl)
infarction
1.oo 0.39 (0.13-1.19) 0.65 (0.21-2.08)
1.oo 0.76 (0.33-1.55) 0.59 (0.22-1.60)
1.oo 0.92 (0.59-1.43) 0.31 (0.15-0.67)
of heart
income,
318 259 355
336 391 395
672 838 560
education
1.00 0.85 1.25
1.21 1.29
1 .oo
1.30 0.85
1 .oo
1.00 1.28 0.87
Ageadj.
death
for age and risk factors (95% Cl)
ratio
screenings
1.00 0.82 (0.50-1.34) 1.02 (0.57-1.82)
1.oo 1.05 (0.71-1.56) 1.00 (0.60-1.67)
1.00 1.07 (0.80-1.43) 0.63 (0.41-0.96)
1.01 (0.78-1.30) 0.65 (0.45-0.93)
1 .oo
Adj. other
Rate
cardiovascular
Total
two
820 1000 695
P-Yrs
Rate/ 100,000
attended
pectorls,
84 24 26
100 42 33
183 76 42
248 103 57
Deaths
who
or angina
years
for age and risk factors”
ratio
35-52
1.00 0.93 (0.65-1.33) 0.42 (0.25-0.73)
Adj. other
Rate
death
persons
Cohort
Study
723
TABLE 4. Rate ratio of coronary death ethnic groups among men 35-52 years cardiovascular screedings in Finnmark 1974-1978
Rate
ratio
by height in three who attended two county, Norway, (per Adj. other
Ethnic
group
Deaths
Men Norse Finnish Lappish
99 39 13
Abbreviations: “Systolic
Age.
adj. blood
= adjustment; pressure, total
RMI height, physical activity, pectorw, income, educanon.
adj.
0.70 0.76 0.80
0.92 0.75 0.60
10 cm) for age and risk factors’ (95% Cl)
(0.65-1.29) (0.43-1.33) (0.18-1.97)
Cl = confidence interval. cholesterol, triglycerides, cigarettes/day,
family
hisrory
of heart infarction
or angina
ence in coronary mortality between ethnic groups within one county, on the other hand, cannot be explained by the well-known risk factors. It is hard to think of some lifestyle factors that are not accounted for in this study that could explain the differences in coronary mortality between ethnic groups living in the same county. So it is reasonable to think along the lines of some genetic differences that are related to the lower coronary mortality in Lappish men. Although there is a suggestive protective effect also in Lappish women, the estimates are imprecise and our data are also compatible with no protective effect in Lappish women. During the follow-up period, the women will have been at risk both in the pre-menopausal and the menopausal state. It might be that a protective effect only is present in the menopausal state. In conclusion, this study provides evidence that middle-
aged men of Lappish origin are more protected against coronary death than men of Finnish and Norse origin. References 1. National Mass Radiography Service, Health Services of Finnmark County, Central Laboratory, Ullevdl Hospital, Faculty of Medicine, University of Tromsti. The cardiovascular study in Finnmark 1974-75. Nordic Council for Arctic Medical Research, Report 25. 1979; 1-195. 2. Bjartveit K, Foss OP, Gjervig T, Lund-Larsen PG. The cardiovascular disease study in Norwegian counties. Background and organization. Acta Med Stand 1979; (Suppl. 634): l-70. 3. Burstein M, Scholnick H, Morfin R. Rapid method for the isolation of lipoproteins from human sera by precipitation with polyions. J Lipid Res 1970; 11: 583-585. DS, Ferde OH, Arnesen E. Distribution of high-den4. Thelle sity lipoprotein cholesterol according to age, sex, and ethnic origin: Cardiovascular disease study in Finnmark 1977. J Epidemiol Comm Health 1982; 36: 243-247. 5. Tverdal A. A Mortality Follow-Up of Persons Invited to a Cardiovascular Disease Study in Five Areas in Norway. Oslo: National Health Screening Service; 1989. 6. Vassenden K. Folke-og boligtellingene 1960, 1970 og 1980. Dokumentasjon av de sammenlignbare filene. Rapport 87/ 2. Oslo-Kongsvinger: Statistisk sentralbyrd; 1986. 7. Sratistisk sentralhyrd. Standard for gruppering av sykdommer-skader-d@dsCsaker. Oslo: Statistisk sentralbyri; 1973. 8. Statistisk sentralbyrd. Klassifikasjon av sykdommer, skader og dedsirsaker. Oslo-Kongsvinger: Statistisk sentralbyrd; 1990. 9. Norusis MJ/SPSS Inc. SPSS for Windows. Advanced Statistics. Release 6.0. Chicago: SPSS, Inc.; 1993. 10. National Health Screening Service, Health Services of Finnmark, Sogn og Fjordane, and Oppland counties, Ullevdl Hospital, Central Laboratory, Oslo. The Cardiovascular Disease Study in Norwegian Counties. Results from Second Screening. Oslo: National Health Screening Service; 1988.