May 21, 1999 - paring temporal trends in rates within and between regions, ... event, major findings in relation to trends and geographical disparities were fairly.
© International Epidemiological Association 1999
International Journal of Epidemiology 1999;28:1050–1058
Printed in Great Britain
Trends and geographical disparities in coronary heart disease in France: are results concordant when different definitions of events are used? Thierry Lang,a Pierre Ducimetière,a Dominique Arveiler,b Philippe Amouyel,c Jean Ferrières,d Jean-Bernard Ruidavets,d Michèle Montaye,c Bernadette Haasb and Annie Binghama
Objectives
To assess whether different definitions of acute coronary events yielded concordant results concerning trends and geographical disparities in coronary heart disease (CHD) mortality and morbidity in France.
Study design
Data from three French CHD registries participating in the WHO MONICA Project during the period 1985–1992.
Setting
Three areas of about one million inhabitants each in the North, South and East of France.
Subjects
About 2000 acute coronary events each year.
Main outcome measures
Mortality, annual rate of fatal and non-fatal events, incidence of first and recurrent events, case-fatality rates.
Results
For incidence and mortality, the broader the definition of the acute event, the higher the reported rates. The same tendency was not observed for case-fatality rates. Comparing between-registry rates for mortality, 28-day case-fatality and hospital case fatality yielded relatively concordant results whatever the definition of event. As a whole, the higher mortality rate in Lille and its intermediate rank in Strasbourg were related more to disparities in case-fatality rates, with only small variations in incidence rates, independently of the definition used. Comparing temporal trends in rates within and between regions, a consistent decrease in annual mortality rates and case-fatality rates was observed, whatever the definition. In contrast, the incidence of non-fatal probable myocardial infarction did not change during the period in any register.
Conclusions Although the absolute estimates of rates were variable with the definition of the event, major findings in relation to trends and geographical disparities were fairly consistent across the definitions: the North-South gradient in mortality observed in France was found to be much more pronounced for case fatality than for incidence. The proportion of milder acute myocardial infarction is currently increasing and this element should be taken into account when analysing CHD rates. Keywords
Epidemiology, coronary heart disease, acute myocardial infarction, geographical disparities, time trends
Accepted
21 May 1999
a INSERM U258, 94807 Villejuif, France. b Strasbourg MONICA-Collaborating Centre, Laboratoire d’Epidémiologie et
de Santé Publique, Faculté de Médecine, Strasbourg, France. c INSERM CJF-95–05. Lille MONICA-Collaborating Centre, Institut Pasteur
de Lille, France. d Toulouse MONICA-Collaborating Centre, Département d’Epidémiologie et
de Santé Publique, Faculté de Médecine, Toulouse, France. Reprint requests to: Dr Thierry Lang, INSERM U258, Hôpital P Brousse, 16, Avenue P Vaillant-Couturier, 94807 Villejuif, France. E-mail: lang@biomath. jussieu.fr
The evolution and geographical differences in mortality and morbidity of coronary events as observed by registries may have both research and public health implications, since they imply different strategies to improve cardiovascular health at national and regional levels. In order to produce valid comparisons between registries, between or within countries, precise definitions of coronary events are required.1 The WHO international MONICA Project, including 38 populations from 21 countries had to define such rules and algorithms for
1050
CORONARY HEART DISEASE TRENDS IN FRANCE
classification.1 According to the level of certainty desired, and whether or not the diagnosis of the clinician is used, very different estimates of incidence and case fatality might be obtained. Low case-fatality rates may result from non-specific criteria for non-fatal events but selective criteria for fatal events; high casefatality rates may result from the opposite situation.2 This phenomenon is often unavoidable in epidemiology, where strict criteria have to be defined in order to ensure standardization and comparison between centres and over time. It can, however, be puzzling for the clinicians or public health authorities, who read and use these results, to see different estimates of the occurrence of events according to definitions. In addition, betweenand within-register variability in data collection has been shown. Namely, the proportion of the definite, possible events or events with insufficient data has been reported to vary between registries and over time even in the same country.3 In addition, the percentage of unclassified deaths has been reported to be high in France.4 Due to improvement in treatment and perhaps risk factor control, a possible increase of the proportion of mild acute myocardial infarction (AMI) has been suggested. Different diagnostic thresholds in these mild cases might affect number of events and thus event rates.2 Moreover, this increasing category might account partly for the decrease of definite AMI.5 One question might thus be to explore how these results can be used and interpreted in terms of public health. For example, the debate on sex differences in case fatality has shown how the choice of definitions can affect the results.6 Failure to recognize non-fatal AMI cases has been suggested to explain partly the higher casefatality rates reported among women in some populations.6 Our objective was to analyse the geographical variations and time trends of coronary mortality within France. Eight-year (1985–1992) results from the three French MONICA registers were analysed in order to evaluate whether the interpretation and consequences of the case findings were the same whatever the definition used. Our assumption was that two main results should be consistent across definitions: geographical variations and time trends. In case of major discrepancies between results using different definitions, no clear consequence might be drawn for decision and public policy concerning coronary heart disease (CHD) in the country.
Methods Population Three MONICA Collaborating Centres (MCC) have participated in the WHO MONICA Project in France: MCC-Lille, MCCStrasbourg and MCC-Toulouse. The geographical areas, about one million inhabitants each, are respectively the Urban Community of Lille (LIL) and two French districts: Bas-Rhin (Strasbourg), Haute Garonne (Toulouse).1 Data collection concerns the age range 25–64 years. The information on the population living in the three geographical areas was obtained from the census data. The demographic evolution between two censuses was supposed to be linear with an adjustment based on regional migration and death rates.
Data collection Morbidity data were systematically collected by the investigators in the public and private hospitals of the area; in the emergency departments as well as in cardiologists’ private practices. General
1051
practitioners were interviewed mainly during the search for further information on causes of death. Each event diagnosed as CHD or events with symptoms compatible with CHD—such as chest pain, shock, arrhythmia, left ventricular insufficiency— were further investigated. The assessment of the prognosis 28 days after the CHD episode required additional information from the doctors in charge of the patient. In addition to these sources, death certificates were used for mortality data. Any death certificate on which a diagnosis related to CHD had been mentioned as main diagnosis, secondary diagnosis or associated cause was followed by further investigations. These ICD-9 codes include CHD (410–414); sudden death (797–799); arterial hypertensive disease (401–405); other forms of heart disease (420–429), cerebrovascular disease (430–438); disease of the arteries, arterioles and capillaries (440–447); shock (785.5); pulmonary embolism (415); heart failure (428). Other diagnoses of high cardiovascular risk, such as diabetes (250), hyperlipidaemia (272) or obesity (278) have also been investigated. Technical definitions from the MONICA manual have been published in detail.1 Diagnostic criteria included symptoms, history of previous AMI or CHD, electrocardiograms, enzyme activities and necropsy findings. Coronary events were designated as fatal and non-fatal according to the date of death, before or after midnight between the 27th and the 28th day following the first event. Record linkage was used during this 28-day period in order to avoid duplication. Episodes occurring more than 28 days after the first event were not linked. Data on previous AMI were the sole information used to distinguish first from recurrent event. Out-of-hospital deaths were defined as deaths occurring before hospitalization. External quality control was performed by the WHO Reference Centres in Dundee, Scotland, Budapest, Hungary and Helsinki, Finland.1 Coding sets of case histories and sets of test ECG were issued regularly by these Centres respectively. The Dundee centre developed computer algorithms for deriving MONICA ECG code from Minnesota codes of ECG and for deriving the MONICA diagnostic code from the diagnostic data items. The comparability of coding across the collaborative study has been shown.1 At the MONICA data centre in Helsinki, logic and consistency checks were performed. Data were judged acceptable after comparisons with routine statistics on CHD deaths.1 In addition to this international level, internal quality control was assessed by regularly exchanging French case histories and ECG between the three French registries. Regular meeting and continuous quality control by the co-ordinating centre was another quality assurance system.
Definitions of acute coronary events Several definitions of events have been used to produce rates; some of them were defined by the international WHO project, some others proposed by national teams. International MONICA definitions According to MONICA criteria, some basic categories of events were defined:1(a) definite: fatal (F1) or non-fatal (NF1) AMI; (b) possible: possible non-fatal myocardial infarction (NF2) or coronary death (F2); (c) death with insufficient data (F9). Resuscitated ischaemic cardiac arrest (NF3) was extremely rare, and the category ‘no myocardial infarction’ (NF4) was not used in our analysis. The proportion of the three basic categories
1052
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Table 1 Evolution of the different categories of events (men and women), as recorded by the registries between 1985 and 1992 Lille Definition of acute events
Strasbourg
Toulouse
All
1985
1992
1985
1992
1985
1992
1985
936
711
901
722
640
606
2477
2039
Fatal (F1) and non-fatal AMI (NF1)
33.4%
33.3%
41.6%
43.2%
41.4%
43.1%
38.5%
39.7%
Possible AMI (NF2) and coronary death (F2)
43.4%
47.5%
38.7%
42.0%
45.5%
45.0%
42.1%
45.0%
Death with insufficient data (F9)
23.3%
19.1%
19.6%
14.8%
13.6%
11.4%
19.4%
15.3%
Total
100%
100%
100%
100%
100%
100%
100%
100%
Total no. of events (F1 + NF1 + NF2 + F2 + F9)
Pa
,0.1
,0.04
,0.48
1992
,0.001
a Difference between 1992 and 1985.
of fatal cases was not similar across registries or stable over time (Table 1). Some combinations of these events were used to produce international MONICA definitions.1 The MONICA definition 2 (F1 + NF1 + F2) is a strict one using only definite AMI and coronary death. MONICA definition 1 (F1 + NF1 + F2 + F9) includes in addition death with insufficient data. The MONICA definition 3 (F1 + NF1 + F2 + NF2 + F9), which includes in addition possible AMI (NF2), was not used in this paper. MONICA—France definition In addition to these MONICA criteria, the clinical diagnosis was recorded by the registries for each event. The trends and regional disparities using diagnosis of AMI as stated by the clinician were also assessed as such (ICD-9: 410). In addition to the international MONICA definitions, a French definition was used. In addition to MONICA definition 2, MONICA possible AMI (NF2) was added whenever a clinician’s diagnosis of AMI was recorded: (F1 + NF1 + F2 + (NF2 and ICD-9: 410). ‘Mild’ myocardial infarction According to a Finnish proposal, a category ‘non-fatal probable myocardial infarction’ was defined as a subset of non-fatal AMI (NF2), which in addition to typical symptoms had ECG or enzyme changes suggesting cardiac ischaemia, but were not severe enough to fulfil the criteria for non-fatal definite myocardial infarction.5
Statistical methods Event rates were adjusted using the world standard population.1 Incidence was calculated as the number of first events divided by the total number of person-years exposed during this period. Case-fatality rate was computed as (fatal events/fatal + non-fatal events). The analysis was performed in each registry independently and after cumulating data. A multiple logistic regression analysis, approximating a Poisson regression analysis, was performed, using age and region as independent variables and the incidence and case-fatality rates as dependent variables. To compare one region with another, odds ratios (OR) were used and Toulouse was chosen as the reference group. Adding period as an independent variable, the annual trends of rates were calculated from this logistic regression analysis, assuming a linear trend during the 8-year period. Confidence intervals were calculated using the normal approximation of the binomial distribution. Statistical analysis was performed with the SAS statistical package.7
Results Distribution of the categories of events in the registries and over time Deaths with insufficient data accounted for 19.4% of events in the three registries in 1985. This proportion decreased over time and was 15.3% in 1992. At any period, variations between centres were observed, the proportion decreasing from Lille to Strasbourg and Toulouse. The percentages were 19.1% in Lille, 14.8% in Strasbourg and 11.4% in Toulouse, in 1992.
Comparison of absolute rates according to various definitions within a region As expected, for incidence and mortality, the broader the definition of the acute event, the higher the reported rates (Tables 2 and 3). The increase in rates with a broader definition was more pronounced for mortality than for annual event rates. For example, in 1992, the mortality rates were 64 and 111/ 100 000 with MONICA definitions 2 and 1 respectively (+73%); the corresponding rates were 194 and 241/100 000 for annual event rates (+24%). The same tendency was not observed for case-fatality rates since events added by the broader definition were found both in the numerator and denominator (Tables 4 and 5). The lowest case-fatality rate was observed with clinician’s diagnosis (Table 4). The same was true for withinhospital restricted rate (Table 5).
Comparison of rates between registries Fatal and non-fatal events The annual rates of acute coronary events were higher in Strasbourg than in Toulouse whatever the definition used (OR 1.15–1.25). It did not differ between Lille and Toulouse when clinician’s diagnosis was used. In contrast, using MONICA definitions, rates were slightly higher in Lille than in Toulouse with definition 2 (OR = 1.07) and more importantly when deaths with unsufficient data were included (OR = 1.25) (Table 2). Mortality and case-fatality The results on mortality, 28-day case-fatality rates and hospital case fatality rates were relatively concordant whatever the definition and the period used (Tables 3, 4 and 5). These three rates were higher in Lille than in Toulouse. OR ranged from 1.4 to 1.8, 1.7 to 2.1 and 1.6 to 2.3 for mortality, 28-day casefatality rates and hospital case-fatality rates respectively. Strasbourg was in an intermediate position, between Lille and
CORONARY HEART DISEASE TRENDS IN FRANCE
1053
Table 2 Age-standardized annual event rates of acute coronary event according to the definitions of incidence/100 000 (95% CI). Men (25–64 years) Odds ratioa Total three registers
Lille
Strasbourg
Toulouse
Lille/ Toulouse
Strasbourg/ Toulouse
1985
315 (299–331)
333 (305–361)
352 (322–382)
261 (235–288)
1.25
1.25
1992
241 (227–255)
268 (243–294)
250 (226–274)
204 (182–227)
–3.1% P , 0.001 –2.6% P , 0.0001 –4.6% P , 0.0001
–1.7% P = 0.03
P = 0.0001
P = 0.0001
1.07
1.17
Definition of the acute event MONICA definition 1 (F1 + NF1 + F2 + F9)b
slopec MONICA definition 2 (F1 + NF1 + F2)b 1985
240 (226–255)
235 (212–259)
266 (240–292)
222 (198–246)
1992
194 (181–206)
203 (181–225)
201 (180–223)
175 (155–196)
–2.6% P , 0.001
–2% P , 0.001
–4.1% P , 0.001
–1.7% P = 0.05
P = 0.008
P , 0.001
1.04
1.15
P = 0.18
P , 0.001
0.99
1.17
P = 0.66
P , 0.001
slopec
MONICA France (F1 + NF1 + F2 + [NF2 and ICD-9: 410])b 1985
264 (250–279)
262 (237–286)
282 (255–309)
252 (226–278)
1992
208 (195–221)
211 (189–234)
221 (199–244)
190 (169–211)
slopec
–3.0% P , 0.001 –2.6% P = 0.0001 –3.8% P = 0.0001 –2.5% P = 0.002
Clinician’s diagnosis (ICD 410) 1985
246 (232–260)
241 (218–265)
260 (234–286)
239 (214–269)
1992
190 (177–202)
174 (154–195)
227 (204–250)
165 (145–185)
–3.6% P , 0.001
–4.4% P , 0.001
slopec
–2.9% P , 0.001 –3.6% P , 0.001
a Difference between registers in 1992. b F1: fatal AMI; NF1: non-fatal AMI; F2: coronary death; NF2: possible non-fatal AMI; F9: death with insufficient data. c Logistic regression analysis. Adjustment for age. Linear effect for one calendar year.
Table 3 Age-standardized mortality rates in the French MONICA Project, according to the definitions of acute coronary events. Annual mortality rate/100 000 (95% CI). Men (aged 25–64) Odds ratioa Total three registers
Lille
Strasbourg
Toulouse
Lille/ Toulouse
Strasbourg/ Toulouse
1985
167 (155–179)
201 (179–222)
181 (160–203)
118 (101–136)
1.8
1.5
1992
111 (102–121)
153 (134–172)
109 (94–125)
72 (59–85)
–4.2% P = 0.0001
–2.0% P = 0.02
–6.5% P = 0.0001
–4.9% P = 0.0001
P , 0.0001
P , 0.0001
1.6
1.4
Definition of the acute event MONICA definition 1 (F1 + F2 + F9)b
slopec
MONICA definition 2 and MONICA France (F1 + F2)b 1985
93 (84–101)
103 (88–118)
95 (80–111)
80 (66–94)
1992
64 (57–71)
87 (73–102)
62 (50–74)
43 (33–53)
–3.9% P = 0.0001
0%
–6.4% P = 0.0001
–6.7% P = 0.0001
P , 0.0001
P , 0.0001
1985
82 (74–90)
92 (77–106)
84 (69–98)
71 (57–84)
1.4
1.4
1992
49 (43–55)
56 (44–68)
59 (47–70)
32 (24–41)
–6.7% P , 0.0001 –5.5% P , 0.001
–5.9% P , 0.001
–9.5% P , 0.001
P , 0.0001
P , 0.0001
slopec Clinician’s diagnosis (ICD-9: 410)
slopec
a Difference between registers in 1992. b F1: fatal AMI; NF1: non-fatal AMI; F2: coronary death; NF2: possible non-fatal AMI; F9: death with insufficient data. c Logistic regression analysis. Linear effect for one calendar year, adjusted for age.
Toulouse, whatever the definition used, strict or broad. The only exception was found with the clinical diagnosis (ICD-9: 410), for which Lille and Strasbourg did not differ with respect to mortality (Table 3). Out-of-hospital death rates Out-of-hospital death rates were significantly higher in Strasbourg and Lille as compared with Toulouse. This was true with a strict definition of coronary death (MONICA definition 2):
OR = 1.4, P , 0.0001 and 1.3, P , 0.0002) respectively, as well as with a broad definition (MONICA definition 1): OR = 1.5, P , 0.0001 and 1.8, P , 0.0001) respectively. First and recurrent events The incidence of first events did not differ between Lille and Toulouse with definitions 2 and MONICA-France; it was slightly higher with definition 1: OR = 1.1, P , 0.01. In contrast, the incidence of first events was higher in Strasbourg than in Toulouse,
1054
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Table 4 Age-standardized 28-day case-fatality rates according to the definitions of acute events. Rate/100 (95% CI). Men (25–64 years) Odds ratioa Definition of the acute event
Total three registers
Lille
Strasbourg
Toulouse
Lille/ Strasbourg/ Toulouse Toulouse
MONICA definition 1 (F1 + F2 + F9)/(F1 + NF1 + F2 + F9)b 1985
54 (51–56)
61 (57–65)
52 (48–57)
47 (42–52)
1992
47 (44–49)
59 (54–63)
44 (40–49)
37 (31–42)
–2.3% P = 0.005
+1.5% P = 0.28
–3.7% P = 0.007
–5.6% P = 0.0004
slopec
2.1
1.4
P , 0.0001 P , 0.0001
MONICA definition 2 (F1 + F2)/(F1 + NF1 + F2)b 1985
40 (37–43)
45 (40–50)
37 (32–42)
38 (33–43)
1992
34 (31–38)
45 (41–50)
32 (27–37)
26 (21–32)
–2.1% P = 0.04
+3.6% P = 0.03
–3.8% P = 0.02
–7.3% P = 0.0001
slopec
1.8
1.3
P , 0.0001 P , 0.0001
MONICA France (F1 + F2)/(F1 + NF1 + F2 + [NF2 and ICD-9: 410])b 1985
36 (34–39)
41 (36–45)
35 (30–39)
33 (29–38)
1992
32 (30–35)
44 (38–49)
29 (24–34)
24 (20–29)
slopec
–1.4% P = 0.14
+4.6% P = 0.005
–4.0% P = 0.01
–6.0% P = 0.001
1985
34 (32–37)
39 (35–44)
33 (28–37)
31 (26–36)
1992
27 (24–30)
34 (28–40)
27 (22–32)
21 (16–26)
–4.3% P = 0.0001
–1.5% P = 0.39
1.8
1.3
P , 0.0001 P , 0.0001
Clinician’s diagnosis (ICD-9: 410)
slopec
–4.0% P = 0.02 –8.4% P , 0.0001
1.7
1.3
P , 0.0001 P , 0.0001
a Difference between registers in 1992. b F1: fatal AMI; NF1: non-fatal AMI; F2: coronary death; NF2: possible non-fatal AMI; F9: death with insufficient data. c Logistic regression analysis. Adjustment for age.
Table 5 Age-standardized within-hospital case-fatality rates according to four definitions of case fatality. Rate/100 (95% CI). Men (aged 25–64) Odds ratioa Definition of the acute event
Total three registers
Lille
Strasbourg
Toulouse
Lille/ Strasbourg/ Toulouse Toulouse
MONICA definition 1 (F1 + F2 + F9/F1 + NF1 + F2 + F9)b 1985
21 (19–24)
23 (19–28)
23 (18–27)
18 (14–23)
1992
16 (14–19)
23 (17–28)
14 (10–18)
12 (8–16)
–5.0% P = 0.0001
–1.3% P = 0.02
slopec
–5.1% P = 0.02–10.9% P = 0.0001
2.3
1.4
P , 0.0001 P , 0.0001
MONICA definition 2 (F1 + F2/F1 + NF1 + F2)b 1985
24 (21–27)
27 (22–32)
24 (19–28)
21 (16–27)
1992
18 (15–21)
24 (18–30)
16 (11–20)
13 (9–18)
–3.8% P = 0.005
+1% P = 0.57
slopec
–5.1% P = 0.02 –8.9% P = 0.0003
2.2
1.5
P , 0.0001 P , 0.0001
MONICA France (F1 + F2/F1 + NF1 + F2 + [NF2 and ICD-9: 410])b 1985
21 (19–23)
22 (18–26)
22 (18–26)
17 (13–21)
1992
15 (13–17)
21 (17–25)
13 (9–17)
11 (7–15)
–3.3% P = 0.02
+2.2% P = 0.32
1985
15 (13–17)
15 (11–20)
15 (11–20)
14 (10–18)
1992
9 (7–11)
10 (5–14)
9 (5–12)
9 (5–13)
–6.8% P , 0.001
–3.6% P = 0.21
–7.7% P , 0.01
–9.7% P , 0.01
slopec
–5.6% P = 0.01 –8.7% P = 0.002
2.2
1.5
P , 0.0001 P , 0.0001
Clinician’s diagnosis (ICD-9: 410)
slopec
1.6
1.3
P , 0.0001
P , 0.01
a Difference between registers in 1992. b F1: fatal AMI; NF1: non-fatal AMI; F2: coronary death; NF2: possible non-fatal AMI; F9: death with insufficient data. c Logistic regression analysis. Adjustment for age.
either with MONICA definition 2, MONICA-France definition, or clinician’s diagnosis (+10%, P , 0.001 each), or MONICA definition 1 (+17%, P , 0.001). Recurrent events occurred more frequently in Lille than in Toulouse for MONICA 2 and MONICA-France (OR = 1.2, P , 0.001 both) or MONICA 1 (OR
= 1.4, P = 0.0001). A similar tendency, although not significant, was observed with clinician’s diagnosis (OR = 1.1, P = 0.12). The annual rates of recurrent events were higher in Strasbourg than in Toulouse with definitions MONICA 1, 2 and France (OR = 1.4, P , 0.001 each) or clinician’s diagnosis (OR = 1.5, P , 0.001).
CORONARY HEART DISEASE TRENDS IN FRANCE
1055
Table 6 Trends in first and recurrent acute coronary events according to the definitions. 1985–1992. Men Definition of the acute event
Total three registersb
Lilleb
Strasbourgb
Toulouseb
MONICA definition 1 (F1 + NF1 + F2 + F9)a First event
–2.0% P = 0.0001
–1.4% P = 0.12
–3.3% P = 0.0001
–1.2% P = 0.19
Recurrent event
–5.9% P = 0.0001
–4.8% P = 0.004
–8.3% P = 0.0001
–3.7% P = 0.05
MONICA definition 2 (F1 + NF1 + F2)a First event
–1.4% P = 0.0004
–1.1% P = 0.26
–2.4% P = 0.007
–0.7% P = 0.45
Recurrent event (same as def.1)
–5.9% P = 0.0001
–4.8% P = 0.004
–8.3% P = 0.0001
–3.7% P = 0.05
MONICA France (F1 + NF1 + F2 + [NF2 and ICD-9: 410])a First event
–2.0% P = 0.0001
–1.9% P = 0.03
–2.4% P = 0.001
–1.8% P = 0.05
Recurrent event
–5.8% P = 0.0001
–4.8% P = 0.004
–8.0% P = 0.001
–3.8% P = 0.50
Clinician’s diagnosis (ICD-9: 410) First event Recurrent event
–2.3% P = 0.0001
–2.7% P , 0.01
–1.6% P = 0.06
–2.8% P , 0.01
–7%b P , 0.001
–6.2% P , 0.01
–8.1% P , 0.001
–5.3% P = 0.02
a F1: fatal AMI; NF1: non-fatal AMI; F2: coronary death; NF2: possible non-fatal AMI; F9: death with insufficient data. b Logistic regression analysis. Linear effect for one calendar year, adjusted for age.
Non-fatal probable myocardial infarction Neither the rate of first or recurrent myocardial infarction changed between 1985 and 1992. The incidence of first nonfatal probable AMI was the highest in Strasbourg (38 [95% CI : 29–48]/100 000) as compared with Toulouse (30 [21–38]/ 100 000) (OR = 1.3 [95% CI : 1.2–1.5], P , 0.001) and did not differ between Lille (26 [18–33]/100 000) and Toulouse. The incidence of recurrent non-fatal probable AMI was higher in Strasbourg (20 [13–27]/100 000) (OR = 1.7 [95% CI : 1.4–2.1], P , 0.001) and Lille (16 [9–22]/100 000) (OR = 1.3 [95% CI : 1.1–1.6], P , 0.05) as compared with Toulouse (13 [8–19]/ 100 000). As a whole, independently of the definition used, the higher mortality rate in Lille and its intermediate rank in Strasbourg were mainly related to the disparities in case-fatality rates, with smaller variations in incidence rates.
Comparison of temporal trends in rates within and between regions Mortality rates Whatever the definition, a decline was observed in Strasbourg and Toulouse. In Lille, this declining trend was observed only if the category F9 (fatal cases with insufficient data) was included or clinician’s diagnosis used (Table 2). It should be noted that in the same period, a decrease in the category F9 (death with insufficient data) and an increase in the category NF2 and F2 (possible AMI and coronary death) were observed in the three registries. This suggests that, over the years, a better collection of data might have taken place, increasing the number of fatal cases classified as coronary deaths. As a result, this category of events decreased more slowly than the total number of fatal cases. Case-fatality rates The decrease in the 28-day case-fatality rates was observed whatever the definition used in Strasbourg and Toulouse. In Lille, 28-day case fatality rates increased significantly if strict definitions were used at the numerator. Within hospital casefatality rates did not change over time, whatever the definition used in Lille; the decrease was steeper in Toulouse than in Strasbourg (Tables 4 and 5).
Out-of-hospital death rates The out-of-hospital mortality rates decreased in Strasbourg (–5.8%, P , 0.0001 and –5.8%, P = 0.0001) and Toulouse (–3.2%, P = 0.03 and –3.2%, P = 0.02) with definitions 2 and 1 respectively. In contrast, the rates did not change in Lille (P = 0.35 and P = 0.13). Fatal and non-fatal events The decrease in annual event rates was concordant between clinician’s diagnosis, definitions 2 and 1 of AMI, in Lille and Strasbourg (Table 2). In Toulouse, the annual decrease was less marked, in particular using definitions 1 and 2. First and recurrent events Results according to definitions were concordant. A steeper decrease was observed for recurrent events than for first events. In Toulouse and Lille, MONICA definitions showed a relative stability of incidence of first events. In contrast, decrease trends were observed in Strasbourg whatever the definition (Table 6). Non-fatal probable myocardial infarction The annual rates of these events did not change over the years in any of the registries.
Discussion Some geographical and temporal trends in CHD appear to have been concordant whatever the definition of the acute events. Although the absolute levels in mortality, case-fatality and incidence rates have varied considerably—from one to two— according to the definitions, the disparities between regions were stable. Mortality rates were higher in Lille than in Toulouse, and Strasbourg was in an intermediate position whatever the definition, with the exception of clinician’s diagnosis. Moreover, with each definition, differences in case fatality, much more than incidence, has accounted for these disparities. Rate of first and recurrent events was the highest in Strasbourg. Lille did not differ from Toulouse with regard to incidence of first events, although the rate of recurrent events was slightly higher in Lille. These differences were however modest when compared with the difference found for case-fatality rates. At most, incidence was
1056
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
20% higher in Strasbourg than in Toulouse. In contrast, 28-day case-fatality rate was between 30% and 40% higher in Strasbourg compared with Toulouse and from 70% to 110% higher in Lille than in Toulouse. With any of the definitions, the analysis of trends is mostly concordant. Mortality rates have decreased in the 8-year period and this decrease appears to have been less pronounced in Lille. The 28-day case-fatality rate and the hospital case-fatality rate have decreased in Strasbourg, even more in Toulouse and have been stable in Lille. Conversely, the incidence has decreased in Lille and Strasbourg and has been fairly stable in Toulouse. The recurrent events have decreased less in Toulouse than in the other two registries. Several sources of error may affect the interpretation of rates produced by registries. Modification over the years in the collection of events and/or the relevant information to categorize them may affect trends. Differences between registries might produce artefactual geographical disparities. Every effort has been made to reduce these sources of error, although some of them have been described in the first years of the registry functioning.4 Some differences can be seen in 1992, namely the proportion of the different categories of events was not similar across centres. For example, some differences were observed according to the definition used. In Lille, strictly defined events were found to exhibit a lower decline than broader definitions, given the transfer, over the years, of cases from insufficient data to possible or definite AMI. Similarly, using the clinician’s diagnosis in Lille has introduced an underestimation of the differences found between that area and other registries. One can hypothesize that physicians’ habits, e.g. using less often the ICD-9 code 410 in this region, have accounted for this result. Another hypothesis is that for a lot of fatal cases, information on the circumstances of death was lacking. The relative trends according to the definition may provide in itself some valuable information. In our data, the incidence of non-fatal probable AMI was found to be stable between 1985 and 1992, in contrast with other categories of coronary events which decreased. This figure is likely to be related to the increasing number of cases with CHD or possible AMI diagnosed with angiography. This result underlines the importance of taking into an account mild infarction cases which may represent an increasing part of AMI cases, due to the development of diagnostic procedures and prevention.5 Due to earlier recognition and medical intervention, the categories of stable and unstable angina have increased from 15% to 37% of coronary artery disease discharges in US hospitals between 1980 and 1989.8 Despite these misclassifications, stable results across definitions were observed which can be used for research hypotheses or public health decisions. These results support the usefulness of monitoring trends and geographical disparities using registries, provided that procedures of data quality assessment are organized continuously and rigorously as in the experience reported in France from 1985 to 1992. The main result was that, unlike incidence of first events, case-fatality rates were found to play a major role in the disparities in mortality observed in France and in the North-South mortality gradient within France. The main effect of case-fatality rates on mortality contrasting with nearly stable incidence rates has been reported in the US8–10 and Finland.11 The reverse has been observed in Canada,12 Iceland13 and in a large US
company14 where incidence had a prominent role in explaining the decreasing CHD mortality. Divergent contributions of incidence and survival to the decreasing mortality rates have been observed between regions within the same country in Finland15 and between occupational categories in France.16 The 10-year results of the WHO MONICA Project were heterogeneous in this respect. From the results of the pooled 37 populations, coronary event rates contributed two-thirds and case fatality one-third.4 In some countries, most of the decline in mortality rates has occurred in out-of-hospital sudden deaths, predominantly without previous history of CHD. Thus, this decline cannot be attributed to secondary prevention.17 In our data, out-ofhospital sudden deaths did not change in two registries and even increased in Lille during the study period, suggesting poor results from primary prevention in the country as a whole and more particularly in the latter area. The incidence of events with previous history of AMI is the highest in Strasbourg and the lowest in Toulouse, Lille being in an intermediate position. The lower incidence of events with previous history of AMI in Toulouse might be related to a lower incidence of first events, which is not the case. They might be related to a higher case-fatality rate of the events leading to selection of the less severe cases, less likely to suffer a new event. These two conditions were not met in Toulouse; therefore, the most likely explanation is a lower rate of recurrent events after a first event. A better secondary prevention and/or an earlier diagnosis of CHD might thus be an explanation for these findings. Conversely, the phenomenon of selection of the more severe cases might explain the lower incidence of events with previous history of AMI in Lille compared with Strasbourg, since the case-fatality rate is higher and the incidence of recurrent events lower in Lille than in Strasbourg. What is behind these differences is far from simple, since case-fatality rates do not reflect medical care only. They also reflect severity of the events and pre-existing conditions of the patients. Lack of decrease in out-of-hospital deaths, stability of incidence of first events suggest poor effectiveness of primary prevention. In addition, there is little evidence to think that the distribution of ‘classical’ cardiovascular risk factors such as cholesterol, tobacco consumption or arterial hypertension account for the differences observed between regions.18 Although medical management of AMI has improved during the study period, differences in medical care procedures and treatment between registries have been reported to be unrelated to the disparities in mortality between registries in France.19 This may suggest that factors which are non-specific to CHD mortality may play an important role. The regional differences in total mortality observed for CHD, very similar to disparities reported for other diseases, is another argument for this hypothesis.20 For example, social factors, such as social support, have been shown to be related to mortality through case fatality more than through incidence of the disease21 and to be related to a better survival not only after myocardial infarction,22,23 but after cancer as well.24 Ecological approaches suggest that income distribution and social capital within geographical areas or countries are related to mortality from multiples causes including cardiovascular diseases.25–27 The North-South distribution of diseases in France is in keeping with a higher proportion of low social groups, higher unemployment rate, lower educational level,
CORONARY HEART DISEASE TRENDS IN FRANCE
higher frequency of poor housing in the North.28 There is thus room and need for explaining variations in CHD besides the classical risk factors and coronary care.29 From a public health point of view, an approach based on one disease independently of the broader picture, is incomplete. In 27 countries, from 1950 to 1987, trends in CHD, cancer, stroke were found to be similar. As suggested by cohort studies, adverse health risk factors accumulate for an individual over decades and may affect multiple outcomes.30 Some common causes amenable to the same preventive measures might thus exist.31
Conclusion Although the absolute estimates of the occurrence of coronary events were found to be variable depending on the definition of the event used in the calculations, the major findings in relation to trends and geographical disparities were consistent across the definitions, despite a relatively high percentage of unclassifiable deaths.4 The North to South European gradient in mortality observed in France was found to be much more pronounced for case-fatality rates than for incidence. Differences in case-fatality rates and recurrence of events suggest that higher severity of the disease in Lille and earlier detection of CHD in Toulouse might occur. The explanations for this North-South gradient in CHD case-fatality rate should not point to medical care only since socioeconomic factors follow this gradient and survival rates do not reflect treatment alone.4,19
Acknowledgement This work was supported by a grant from the Institut National de la Santé et de la Recherche Médicale (INSERM) and from the Direction Générale de la Santé (DGS). MCC Lille is supported by INSERM, DGS, Institut Pasteur de Lille and the Centre Hospitalier et Universitaire de Lille. We would like to thank the investigators of the three French centres for their invaluable contribution in the careful collection and validation of the data, the physicians and cardiologists who helped in this process.
References 1 The WHO MONICA Project (prepared by Tunstall-Pedoe H,
Kuulasmaa K, Amouyel P, Arveiler D, Rajakangas AM, Pajak A). Myocardial infarction and coronary deaths in the World Health Organization MONICA Project. Registration procedures, event rates, and case-fatality rates in 38 populations from 21 countries in four continents. Circulation 1994;90:583–612. 2 Tunstall-Pedoe H. Perspectives on trends in mortality and case fatality
from coronary heart attacks : the need for a better definition of acute myocardial infarction. Heart 1998;80:112–13. 3 Richard J, Arveiler D, Cambou J et al. Between-register and within
register variability in event-finding and rates in the MONICA-France Project. Rev Epidemiol Santé Publique 1990;38:403–10. 4 Tunstall-Pedoe H, Kuulasmaa K, Mähönen M, Tolonen H, Ruokokoski
E, Amouyel P for the WHO-MONICA Project. Contribution of trends in survival and coronary heart disease mortality: 10-year results from 37 WHO MONICA Project populations. Lancet 1999;353:1547–57. 5 The WHO MONICA Project (prepared by Salomaa V, Miettinen H,
Rajakangas AM, Kuulasmaa K). Mild myocardial infarction. A classification problem in epidemiologic studies. J Clin Epidemiol 1997;50:3–13.
1057
6 Chambless L, Keil U, Dobson A, Mähönen M, Kuulasmaa K,
Rajakangas AM, Löwel H, Tunstall-Pedoe H for the WHO-MONICA Project. Population versus clinical view of case fatality from acute cornary heart disease. Results from the WHO-MONICA Project 1985–90. Circulation 1997;96:349–59. 7 SAS/STAT User’s Guide, Version 6, Fourth Edn. Cary, NC: SAS Institute Inc.,
1989. 8 Whitney EJ, Shear CL, Mantell G, Lydick E, Wolf R. The case for
unstable angina pectoris as a primary end point in primary prevention studies. Am J Cardiol 1992;70:738–43. 9 Sytkowski P, Kannel W, D’Agostino RB. Changes in risk factors and
the decline in mortality from cardiovascular disease. The Framingham Heart Study. N Engl J Med 1990;322:1635–41. 10 DeStefano F, Merritt RK, Anda RF, Casper ML, Eaker ED. Trends in
nonfatal coronary heart disease in the United States, 1980 through 1989. Arch Intern Med 1993;153:2489–94. 11 Salomaa V, Arstila M, Kaarsalo E et al. Trends in the incidence of and
mortality from coronary heart disease in Finland, 1983–1988. Am J Epidemiol 1992;136:1303–14. 12 Bata I, Eastwood B, Gregor R et al. Decreasing mortality from acute
myocardial infarction. J Clin Epidemiol 1997;50:787–91. 13 Sigfusson N, Sigvaldason H, Steingrimsdottir L et al. Decline in
ischaemic heart disease in Iceland and change in risk factor levels. Br Med J 1991;302:1371–75. 14 Pell S, Fayerweather WE. Trends in the incidence of myocardial
infarction and in associated mortality and morbidity in a large employed population, 1957–83. N Engl J Med 1985;312:1005–11. 15 Salomaa V, Miettinen H, Kuulasmaa K et al. Decline of coronary heart
disease mortality in Finland during 1983 to 1992: roles of incidence, recurrence, and case-fatality. The FINMONICA MI Register Study. Circulation 1996;94:3130–37. 16 Lang T, Ducimetière P, Arveiler D et al. Occupational categories and
cardiovascular risk factors: distinct relationships with incidence and case fatality from acute coronary heart disease. Results from the French WHO-MONICA project, 1985–89, in men aged 30–59. Int J Epidemiol 1997;26:47–57. 17 Beaglehole R. International trends in coronary heart disease
mortality, morbidity, and risk factors. Epidemiol Rev 1990;12:1–15. 18 Registres français des cardiopathies ischémiques. Facteurs de Risque et
Comportements de Prévention dans la Population des Trois Registres MONICA-France. Enquête de Population 1994–97. Paris: Fédération Française de Cardiologie Edn, 1998. 19 Amouyel P, Arveiler D, Cambou JP et al. Myocardial infarction case-
fatality gradient in three French regions: the influence of coronary care. Int J Epidemiol 1994;23:700–09. 20 Haut Comité de la Santé Publique. La Santé des Français. Paris: Docu-
mentation Française Edn, 1995. 21 Berkman L, Syme S. Social networks, host resistance and mortality:
a nine-year follow-up study of Alameda county residents. Am J Epidemiol 1979;109:186–204. 22 Ruberman W, Weinblatt E, Goldberg JD, Chaudhary BS. Psychosocial
influences on mortality after myocardial infarction. N Engl J Med 1984;311:552–59. 23 Williams RB, Barefoot JC, Califf RM et al. Prognostic importance of
social and economic resources among medically treated patients with angiographically documented coronary artery disease. JAMA 1992; 267:520–24. 24 Maunsell E, Brisson J, Deschenes L. Social support and survival among
women with breast cancer. Cancer 1995;76:631–37. 25 Kawachi I, Kennedy BP, Lochner K, Prothrow-Stith D. Social
capital, income inequality, and mortality. Am J Public Health 1997;87: 1491–98.
1058
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
26 Kennedy BP, Kawachi I, Prothrow-Stith D. Income distribution and
29 Watson P. Explaining rising mortality among men in Eastern Europe.
mortality: cross sectional ecological study of the Robin Hood index in the United States. Br Med J 1996;312:1004–06.
30 Power C, Matthews S. Origins of health inequalities in a national
27 Wilkinson RG. Income distribution and life expectancy. Br Med J
1992;304:165–68. 28 Haut Comité de la Santé Publique. Allocation Régionale des Ressource et
Réduction des Inégalités de Santé. Rennes: ENSP Edn, 1999.
Soc Sci Med 1995;41:923–34. population sample. Lancet 1997;350:1584–89. 31 Thom TJ, Epstein FH. Heart disease, cancer, and stroke mortality
trends and their interrelations: an international perspective. Circulation 1994;90:574–82.
