Long-term outcome of the Malmö Preventive Project: mortality and cardiovascular morbidity

Authors


Göran Berglund, MD, PhD, Department of Medicine, University of Lund, University Hospital, S-205 02 Malmö, Sweden (fax: + 46 40 92 32 72; e-mail: Goran.Berglund@medforsk.mas.lu.se).

Abstract

Abstract. Berglund G, Nilsson P, Eriksson K-F, Nilsson J-Å, Hedblad B, Kristenson H, Lindgärde F (University Hospital, Malmö, Sweden). Long-term outcome of the Malmö Preventive Project: mortality and cardiovascular morbidity. J Intern Med 2000; 247: 19–29.

Objectives. To analyse the effects on mortality and cardiovascular morbidity in a population-based sample, invited to an intervention programme incorporating a baseline screening examination and treatment programmes for subjects with cardiovascular risk factors, high alcohol intake and, in women, suspicion of breast cancer on mammography.

Setting. Section of Preventive Medicine, Department of Medicine, University Hospital, Malmö, Sweden.

Subjects. Birth cohorts (aged 32–51 years) invited to screening examination (men = 9.923; women = 4.422) were compared to birth cohorts not invited (men = 6.655; women = 4.290). Mean participation rate in the invited cohorts was 71% (range 64–78%).

Screening examination. Between 1974 and 1992 a baseline screening including a physical examination, blood pressure, a questionnaire regarding, e.g. family history, lifestyle, and socio-economic factors, laboratory tests of serum cholesterol, triglycerides, gamma-glutamyl-transferase, blood glucose before and after an oral glucose load, as well as a mammography examination in women, was performed.

Interventions. Subjects with hypertension; hyperlipidaemia; diabetes or glucose intolerance; high alcohol intake; or, in women, suspicion of breast cancer were referred to special outpatient clinics.

Main outcome measures. Total and cause-specific mortality, nonfatal myocardial infarction, and stroke, from the screening examination until the end of 1995, was followed in both the intervention and control groups, using national and/or local registries.

Results. Total mortality did not differ significantly between the intervention group and control group. Cause-specific deaths were also similar except for ‘other’ deaths amongst men being significantly lower in the intervention group, mainly due to a lower mortality from ‘other’ causes (suicide, alcohol related deaths) in men under 40 years of age at baseline. Women under 40 years of age had a significantly lower mortality from cancer in the intervention group than in the control group. Nonfatal myocardial infarction and stroke did not differ between intervention and control group in either sex. Within the invited birth cohorts, nonparticipants had a higher total and cause-specific mortality.

Conclusions. Risk factor screening for major diseases such as cardiovascular disease, alcohol abuse, diabetes mellitus and breast cancer, and subsequent treatment of the detected risk factors/diseases – The Malmö Preventive Project – did not reduce total mortality in the intervention group as a whole. In subjects under 40 years of age at entry, total mortality was lower in the intervention group than in the control group. In men, this seemed to be due to a reduction of alcohol-related deaths, whilst in women death from cancer was reduced.

Introduction

The benefits of screening and case-finding leading to interventive actions against detected risk factors, with the aim of preventing cardiovascular disease, has been a debated issue for many years [12]. Most controlled studies within this field have not been able to show any substantial long-term intervention effects on cardiovascular mortality or morbidity, despite huge practical efforts and large costs. These studies with a multifactorial risk factor intervention approach can be divided into community-based intervention studies, e.g. the North Karelia Study in Finland [3] and the Minnesota Heart Health Study in the USA [4], and studies based on individual intervention and follow-up, e.g. the Multiple Risk Factor Intervention Trial (MRFIT) [5] in the USA, and the Gothenburg Primary Preventive Trial (GPPT) in Sweden [6].

The lack of health-protecting effects in these studies may have been due to either the application of ineffective methods for prevention, or of concomitant beneficial time–trend effects in control areas or control populations. Another aspect has been the difficulty of reaching less privileged social groups with health promotion for the prevention of coronary heart disease [7]. Previous population-based studies have mostly used a control area study design, thus being hampered by the time–trend changes in the whole population as well as variable losses of subjects to follow up [3–7]. The Swedish Council on Technology Assessment in Health Care concluded, in a report to the Swedish government in 1997, that community-based cardiovascular prevention programmes have so far not provided enough positive evidence to support further large-scale intervention studies within this field [8].

In Malmö, the third largest city of Sweden (250 000 inhabitants), a large-scale screening programme for cardiovascular risk factors and health promotion was started in 1974. Men and women from the local population were recruited for risk factor screening, lifestyle modification and medical intervention in high-risk individuals, as well as a follow-up survey [9]. Subjects were invited because they were born in prespecified years (range 1921–49). No randomized controls were chosen at the time of screening. We therefore chose to compare long-term outcome in the invited birth cohorts with that of noninvited cohorts of subjects born close to the years of birth of the invited subjects. Thus, a combined group of baseline examination participants and nonparticipants of the invited group, was compared with that of noninvited subjects. The existence of national registers of good quality in Sweden [10] has been pivotal in this process.

Previous reports from the Malmö Preventive Project have shown beneficial long-term treatment effects of intervention in selected individuals with impaired glucose tolerance [11], or alcohol abuse [12–16]. Another observational study has focused on risk factors for ischaemic heart disease and long-term outcome in males [17]. No data on population-based controls were, however, included in these early reports.

The aim of this study was to investigate long-term disease outcomes in a large urban population, selected by year of birth and invited to a screening examination, compared to that of corresponding, noninvited birth cohorts of subjects from the same area. A secondary objective was to compare disease outcomes in participants and nonparticipants of the invited cohort, and to describe socio-economic factors in these two groups that might influence outcome.

Materials and methods

Subjects

In Malmö, Sweden a preventive case-finding programme for cardiovascular risk factors, alcohol abuse and breast cancer, the Malmö Preventive Project, started at the Section of Preventive Medicine, Department of Medicine, University Hospital in 1974. The aim of the project at the start was to invite and examine large strata of the adult population, mostly middle-aged males and females, born in prespecified years and at the time living in the city, in order to find high-risk individuals for preventive intervention on cardiovascular risk factors, alcohol abuse, impaired glucose tolerance, and breast cancer [911–17].

Subjects were invited to participate in a comprehensive risk factor screening, including a physical examination, a panel of laboratory tests, a glucose intolerance test, and a mammography for women [9]. Additionally, every participant filled in a self-administered questionnaire (see below). Between 1974 and 1992, a total of 21 911 males and 8676 females attended the screening programme, with an overall attendance rate of 71.2% (range 64–78%), somewhat differing between years. Males were mostly screened in the first half of the period (1974–82) and females in the latter half (1981–92), thus implying different follow-up time periods for males and females.

When the intervention programme was launched in 1974, the design did not incorporate a noninvited control group. However, it was soon realized that a control group would be useful, hence only every second age cohort was invited, leaving every other age cohort as a control group.

Subjects of the invited male birth cohorts of 1927, 1928, 1929, 1944, 1946, and 1948 (n = 9923 ) were thus compared to noninvited (control) subjects of the male birth cohorts of 1925, 1943, 1945, and 1947 (n = 6655 ). Correspondingly, invited female birth cohorts of 1928, 1930 and 1938 (n = 4422 ) were compared to noninvited control female birth cohorts of 1927, 1929 and 1939 (n = 4290 ). The created groups of invited and noninvited birth cohorts were, thus, close in age at screening examination. The small age differences were adjusted for in the statistical analyses.

Participants were defined as those of the invited subjects who attended the screening examination. It has not been possible to establish full participation in, and degree of adherence to the various intervention programmes.

Baseline examination

Physical examination

All subjects were examined for height (cm), and weight (kg) in light indoor clothing. The body mass index (BMI) was calculated (kg/m2). Blood pressure (mmHg) and pulse rate (beats min–1) were measured twice after 10-minutes’ rest in the supine position by use of appropriate technical equipment (a sphygmomanometer with an appropriate cuff width and a mercury manometer), and a mean figure was recorded.

Questionnaire

The subjects answered about 260 questions using a computer. The questionnaire centred on family history of cardiovascular disease, hypertension and diabetes; smoking habits and signs of high alcohol consumption (Mm-Mast instrument ) [16]; physical activity at work and during leisure time; dietary habits and weight gain, presence of symptoms and signs of cardiovascular disease, alcohol abuse and history of malignancies.

Laboratory investigation

Subjects were invited to come to the screening for a blood test after an overnight fast. The following variables were analysed, using routine methods at the Department of Clinical Chemistry, University Hospital, Malmö; serum total cholesterol, serum triglycerides, serum gamma-glutamyl-transferase (GGT), and blood glucose.

Mammography

Screening for breast cancer was offered to all women above 45 years of age. They were reinvited every fourth year. The examination was performed with standard mammography equipment by staff from the Department of Radiology at University Hospital.

Intervention programmes (Fig. 1)

The design and interventions have previously been described in detail [9]. In short, the Section of Preventive Medicine was an integrated part of the Department of Medicine, Malmö University Hospital, and occupied facilities close to the hospital for screening and for diagnosis and treatment for those subjects with diseases and/or risk factors detected at screening. The intervention programme was thus an individually orientated, high-risk approach with resources set aside for both screening and subsequent treatment within the budget of the Departments of Medicine, Radiology and Surgery.

The criteria for intervention and numbers of subjects intervened on were described in 1983 [9] whilst this is a short summary and up-date of the intervention performed.

Hypertension was defined as repeated blood pressures above 160 and/or 100 mmHg or current antihypertensive therapy at the screening examination. Those subjects were referred to their ordinary primary care physician if they had one. If not, they were cared for at a specially created outpatient hypertension clinic. Close to 8% were diagnosed as hypertensives and cared for at this outpatient clinic throughout most of the follow-up.

Hyperlipidaemia, defined as serum cholesterol > = 6.5 mmol L–1 and/or triglycerides > = 2.3 mmol L–1, or ongoing lipid-lowering drug therapy, were referred to a lipid outpatient clinic [18]. During 1974–92 around 12% of the subjects were referred to this clinic, the majority being cared for at the facility until the end of follow-up.

Diabetes and glucose intolerance, defined as a fasting blood glucose, or blood glucose 2 h after 30 g oral glucose/body surface area (m2), > = 7 mmol L–1, was found in about 5% of the participants who were referred to a special outpatient clinic for nonpharmacological (mainly) and pharmacological treatment during the major part of follow-up. Results from this screening [19] and intervention [1120] have been published.

High alcohol intake, defined from the questionnaire and/or after an interview in individuals with GGT in the highest decentile of the GGT distribution, was, thus, suspected in about 10% of participants [21]. Those were intervened on at an outpatient clinic at the Department of Preventive medicine outside the University Hospital for periods up to several years. The intervention and its results have been described [12–16].

Breast cancer was detected and defined by repeated mammography examinations performed by the Department of Radiology, and suspected cases were referred to the Department of Surgery, for evaluation. Design of the screening procedures [22] and results [23] have been published.

Register follow-up

All invited subjects who attended the screening procedure have been traced in local registers. Additionally, we used registers based on repeated local population censuses to identify nonattendees, born in the same year as the attendees and living in Malmö in the year when written invitation to screening was sent out. Controls were identified from the same register, based on local population censuses and including subjects born in years close to the birth-years of those invited. Specific birth-cohort effects could be expected to be small in persons born only a few years apart. The created control group could thus be regarded as a suitable substitute for a formally randomized control group.

Subjects in the invited and noninvited groups were followed-up by registers for the same time span, from the baseline screening examination until the end of the year 1995. The follow-up started at the same age in both groups, thereby minimizing age differences between subjects in the two groups.

All subjects were followed in local as well as national registers for total mortality, cause-specific mortality, nonfatal ischaemic heart disease (IHD) and nonfatal stroke. The data from national registers have been provided by the Swedish Board on Health and Welfare and by Statistics Sweden.

Statistical analysis

In survival analyses the Cox proportional hazards model was used to test the mortality rates in the invited group compared to the noninvited group (‘intention-to-treat analysis’), first of all in the total invited group and then in subgroups according to age. When we compared three cohorts of invited ‘older’ males (born in 1927, 1928, 1929) with one control cohort (1925), age-adjustment was undertaken for incidence rates and relative risks in the total group of ‘older’ males. When calculating incidence rates and relative risks in the total male groups, figures were weighted in relation to numbers of ‘younger’ and ‘older’ men. A P-value less than 0.05 was considered significant. All tests were two-sided.

Results

Effects on mortality and morbidity

All age groups ( Table 1)

In the overall comparison between the invited group, based on an intention-to-treat analysis, and the noninvited (control) group, we found no differences for either sex in total mortality, cause-specific mortality, nonfatal IHD, or nonfatal stroke. Mortality labelled as ‘Other deaths’ was significantly lower (P < 0.02) in the invited males (18/10 000 person years) compared to the noninvited males (23/10 000 person years), corresponding to a reduced age-adjusted risk ratio of 0.8 (95% CI 0.6–0.9; P < 0.02). In women there were no statistically significant differences between invited and noninvited birth cohorts.

Influence of age – males ( Table 2a)

Younger males (age group 32–37 years; mean age 34.4 years) of the invited group showed a significantly reduced total mortality compared to the control group (24 vs. 30 deaths per 10 000 person years), corresponding to a reduced age-adjusted risk ratio of 0.8 (95% CI 0.7–0.96; P < 0.02) for the invited group. This was due, at least in part, to the lower mortality in ‘Other deaths’ for the invited group compared to the control group (14 vs. 19 deaths/10 000 person years; P < 0.03). No differences were found in nonfatal IHD or stroke. In absolute numbers, 51 fewer deaths seemed to have been postponed in the invited group of younger men.

Older males (47–51 years; mean age 48.9 years) of the invited group did not differ from the control group in total mortality, cause-specific mortality, nonfatal IHD or stroke.

Influence of age – females (Table 2b)

Younger females (age group 38–40 years; mean age 39.1 years) of the invited group also showed a reduced total mortality compared to the control group (22 vs. 33 deaths per 10 000 person years). The risk ratio was 0.7 (95% CI 0.5–0.9; P < 0.02) for the invited group. This was partly due to a decreased mortality in cancer (9 vs. 15 deaths per 10 000 person years; P < 0.05) in the invited group. The risk ratio for cancer death was 0.6 [0.4–0.98] for the invited group. No differences were found in nonfatal IHD or stroke. In absolute numbers, 26 deaths were postponed in this age group.

Older females (age groups 55–57 years; mean age 56.1 years) of the intervention group did not differ from the control group in total mortality, cause-specific mortality, nonfatal IHD or stroke.

Participants versus nonparticipants Tables 3, 4a,b)

In these analyses, subjects from all invited birth cohorts were used, i.e. 30 818 men and 12 142 women. The social and demographic characteristics of the nonparticipant group differed significantly from the ones of the participant group regarding marital status, educational level, socio-economic group belonging, housing, and being foreign-born ( Table 3). Both the male and female participants differed significantly (P < 0.001) from nonparticipants in total mortality and cause-specific mortality. The risk ratio for total mortality in the nonparticipant males was 2.2 [2.1–2.3], and in the nonparticipant females 2.4 [2.1–2.8]. Participating men had more nonfatal IHD than nonparticipants (52 vs. 45 cases per 10 000 person years). Table 4a.

Table 3.  Social and demographic characteristics at baseline of participants (n = 33 346) and nonparticipants (n = 19 765) in the Malmö Prevention Programme 1974–92 a
MalesFemales
VariableParticipantsNonparticipantsParticipantsNonparticipants
  • a

    Means (SD), range.

n21 911890786763466
Age44 (32.61)43 (32.60)51 (38.57)51 (38.57)
Sweden born (%)83758477
Marital status (%)
 Cohabiting79567363
 Divorced9191521
 Never married1124710
 Widowed1156
Educational (%)
 Less than 10 years71746872
 More than 9 years29263228
Socio-economic group (SEI) (%)
 Unskilled worker21263840
 Skilled worker232066
 Lower white-collar13102824
 Middle white-collar19121312
 Higher white-collar121165
 Self-employed, entrepreneur8833
 Student or unemployed413610
 Repeated migration 3723
 Housing
‘crowded’6935
Table 4a.  Mortality, incidence of myocardial infarction and stroke in participants and nonparticipants of the Malmö Prevention Programme 1975–92
MalesParticipants
(n = 21 911) a
Nonparticipants
(n = 8907) b
  
VariablenPer 10 000
person
years
nPer 10.000
person
years
RR age-
adjusted
(95% CI)
P-value
  • a

    (367 274 person years).

  • b

    (138 520 person years).

  • c

    IHD, ischaemic heart disease (ICD 410–414); CVD, cardiovascular disease (ICD 390–448); MI, myocardial infarction; nonfatal, not dead in the same diagnosis within 28 days.

Deaths2.516681.8431322.2 (2.1,2.3)0.0001
 IHD77021515372.1 (1.8,2.3)0.0001
 Stroke10437652.3 (1.7,3.1)0.0001
 Other CVD157412892.5 (2.0,3.2)0.0001
 Total CVD1.03128719512.2 (2.0,2.4)0.0001
 Cancer81422418301.6 (1.4,1.8)0.0001
 Other deaths67118706512.0 (2.7,3.3)0.0001
Nonfatal
 MI1.92652620450.8 (0.7,0.9)0.0001
 Stroke83123346251.0 (0.9,1,2)0.703

Participating women had a lower mortality and less nonfatal strokes than nonparticipants (14 vs. 22 cases per 10 000 person years) did. Table 4b.

Discussion

We have previously reported that the identification and treatment of high-risk individuals with impaired glucose tolerance [1120], or alcohol abuse [12–16], proved to be beneficial, implying positive effects of a ‘high-risk strategy’ for such patient groups. The current analyses of the long-term outcome of the Malmö Preventive Project illustrate the effects of a high-risk approach on the outcome in the total invited population. The main findings were as follows.

Identification of subjects with cardiovascular risk factors, and the subsequent treatment of these risk factors through a structured intervention programme, had no overall significant effect on cardiovascular mortality and morbidity, above and beyond that of care available in the Swedish society at large during the period, although the difference in men in total mortality showed a trend for benefit (P = 0.089).

However, younger men and women belonging to the invited birth cohorts had a significantly lower mortality from ‘other’ causes than corresponding groups belonging to the noninvited birth cohorts. Thorough analyses of the causes of death showed that causes of death usually attributed to excess alcohol intake (intoxication, suicide, traffic accidents, liver cirrhosis and infections, especially pulmonary infections) were lower amongst men in the intervention group. In younger women the lower mortality seemed to be caused by lower cancer mortality. In these younger age groups, the intervention programme seemed to have postponed the death of 51 men and 26 women during 16 years’ mean follow-up time.

Nonparticipants of the invited birth cohorts had a higher total and cause-specific mortality than participants. Figures for incidence of nonfatal myocardial infarction and stroke gave a mixed picture. Nonparticipants were characterized by a less favourable socio-economic situation, which corroborates and extends earlier results of analyses of male nonattendees of screening programmes aimed at preventing cardiovascular disease [2425].

The Malmö Preventive Project has a few characteristics worth noting. First, the project was successful in keeping up a high participation rate from the start to the end of the screening programme. The overall participation rate was 71%, somewhat higher in the beginning when middle-aged subjects were invited, and down to just over 60% at the end when younger age groups were screened. The high participation rate should mean that the participants are representative of that part of the population willing to take part in preventive efforts including screening for case finding and subsequent treatment [7].

Secondly, the Malmö Preventive Project is somewhat unique in utilizing a mixed bag of interventions directed toward several diseases such as cardiovascular disease, alcohol abuse, diabetes as well as prediabetic states, and breast cancer through the affiliated mammography screening programme. Thus, a large proportion of those attending the screening examination, 30–35%, was taken care of in the various intervention programmes. This is illustrated in Fig. 1, which depicts the totality of the preventive measures taken within the Malmö Preventive Project.

Figure 1.

Schematic description of the baseline examination and intervention with percentage of participants referred to the various treatment clinics. By permission of Preventive Medicine[9].

Intervention against cardiovascular diseases was the main objective of the project. Why were the preventive efforts unsuccessful? The Malmö project shares this with several other contemporary population-based intervention programmes aimed at the prevention of cardiovascular disease. The North Karelia Project [3], the Minnesota Heart Health Programme [4], the Multiple Risk Factor Intervention Programme [5] and the Primary Preventive Trial in Gothenburg, Sweden [6] all failed to show a significant preventive effect by intervention on cardiovascular disease. An extended follow-up of the original intervention and usual care groups of the Multiple Risk Factor Intervention Programme [26] did, though, show a reduction in mortality from myocardial infarction. In at least one study, the Helsinki Businessmen Study, a detrimental effect was shown after 15 years of follow-up [27]. Although these projects used somewhat different approaches to achieve their goal, in all the changes in the usual care of cardiovascular diseases in society, as reflected in the control areas or control populations, equalled the changes brought about by the special intervention programmes set up in the intervention areas. The main results in our study are in line with these earlier findings, and corroborate the somewhat dismal conclusions of the recent review of the matter by the Swedish Council on Technology Assessment in Health Care [8]. A total of 30–35% of all participants was subject to intervention programmes. As the overall attendance rate was 71%, this means that only about 25% of all invited subjects participated in the programmes. This may be a too small proportion for effects to be shown. The intervention toward cardiovascular risk factors, although pursued for many years in most patients, was relying on a conventional medical treatment model with rather limited attempts to change behaviour. However , the prevention of cardiovascular disease seems to be slowly on its way through development in the affluent societies and their health services, as judged by the recent decline in mortality from these diseases [8]. Such secular trends, which were evident in Malmö during the study period, might also diminish the power to detect preventive effects of the intervention.

Another possible reason for the lack of effect by the intervention programme on cardiovascular mortality and morbidity, might be the use of inefficient methods, including contemporary drugs, to achieve the desired treatment goals in those with detected risk factors for cardiovascular disease. The above cited studies [3–6] have shown the large problems involved in normalizing risk factors in a majority of those detected to have such factors. New measures for pharmacological and nonpharmacological therapy, are, however, constantly emerging and are being scientifically tested. The fact that preventive cardiovascular population-based projects so far have been unsuccessful, should not discourage us to start new preventive projects, if and when new effective treatment modalities have emerged. The combination of drugs and lifestyle modifications should be expected to be more effective than either intervention alone.

The lower mortality in alcohol-related diseases amongst younger men and in cancer amongst younger women, must be interpreted with caution as it constitutes a posthoc analysis without a preformed hypothesis. However, our interpretation that intervention, at least in part, might contribute to the reduction in mortality, is supported by earlier analyses within our project on the effect of intervention in subjects with high alcohol intake [12–16], and on the reduction of mortality from breast cancer in groups of women offered repeated mammography screening examinations [23]. Further analyses on these aspects are underway.

Conclusions

We conclude that screening for risk factors for major diseases, such as cardiovascular disease, alcohol abuse, and breast cancer, and the subsequent treatment of the detected risk factors/diseases – The Malmö Preventive Project – did not reduce total or cause-specific mortality in the intervention group as a whole compared to the controls, even if benefits have been registered in specific targeted high-risk groups. However, in subjects under 40 years of age at the screening examination, total mortality was lower in the intervention group than in the control group. In men this seemed to be due to a reduction of alcohol-related deaths, whilst in women death from cancer was reduced. Nonparticipants had a higher total and cause-specific mortality than participants, and were characterized by a less favourable socio-economic situation.

Acknowledgements

This study was supported by the City of Malmö, the Swedish Medical Research Council, and the Swedish Heart and Lung Foundation.

Received 21 December 1998; accepted 15 April 1999.

Ancillary