Chapter 1. Material and methods

Chapter 1. Material and methods A. R. KAGAN & K. UEMURA This chapter outlines the way in which the problems of obtaining and assessing popula- tion-related material and analysing the data were tackled. Some ofthe limits ofthe approach used, namely, the examination of nearly all deaths from several demographically defined communities, are discussed. POPULATION-RELATED MATERIAL AND DATA The problem and the solution Autopsy findings cannot be directly extrapolated to the living. But they can be related indirectly if a representative sample of all deaths in a population are autopsied. Inter-community differences found in this way would be due to differences in environment, constitution, or cause of death. If some of these factors were also assessed their role in the develop- ment of atherosclerosis could be determined. Comparisons could also be made if a represen- tative sample, or all, of a particular kind of person dying in the community or a representative sample of all persons dying of a particular disease in a community were included, provided similar kinds of people or death were studied in each community. Unfortunately neither of these conditions has been achieved in previous autopsy studies. An examination was made ofwhat at first appeared to be the simplest solution-studying all subjects dying from a particular disease in several com- munities-and the conclusion reached was that this approach would not solve the problem. For example, accidental death was thought likely to be the most accessible form of death to study in many countries. But the kinds of people who died from accidents, the kinds autopsied, and the proportion autopsied all differed between countries sufficiently to make this approach impracticable. The solution adopted was simple- studying all deaths occurring in several well-defined populations. A search was therefore made for communities in which a high proportion of all deaths already came to autopsy and a further exploration was made to see whether in these communities it would be possible to increase this proportion and to find out more about those deaths that did not come to autopsy. The study populations It proved possible to establish and maintain our requirements in 5 out of some 12 towns tested: Malm6 (Sweden), Prague, District II (Czecho- slovakia), Ryazan (Russian SSR), Yalta (Ukrainian SSR), and Tallin (Estonian SSR). Some charac- teristics of the communities studied are shown in Table 1. The total living population from all towns together was about 800 000. Deaths at age 10 and over were studied. The study started in 1963 and by 1965 there were sufficient subjects over age 79 and by 1966 sufficient over age 69 in Malmo. The autopsy rate was low for subjects aged 70 and over in Ryazan, Yalta, and Tallin. Autopsy rate Table 2 shows the number of subjects studied, by age and sex, in the five selected towns. The high autopsy rate required was maintained by continuous monitoring to find and remedy the causes of failure to include deaths in the study. The causes varied in the different towns and the remedies included ensuring the cooperation of pathology and forensic departments, providing special facilities for the autopsy of subjects dying at home, and maintaining the alertness and cooperation of doctors and others who first came into contact with the deaths. The average autopsy rate was 76.8%. for deaths between ages 10 and 69 years and was about 5% higher in the group aged 40-59 years. This in itself was not sufficient to ensure the representativeness of the material, but it was possible to identify most of the deaths that did not come into the study. Half of these had been autopsied, but not according to the study protocol; one quarter had died at home but had been ill for a long time and hospital records were available. It was only for the remaining few BULL. WORLD HEALTH ORGAN., Vol. 53, 1976- 4893464 490 A. R. KAGAN & K. UEMURA Table 1. Characteristics of communities studied Town and Study Age of No. of Stability Proportion No. of population period subjects subjects of Climate Main of deaths pathology(year) (month/year) inddied studiedpopulation Industry in hospital departments(years) stde ouain % eatet Malmo 1/63- 2/64 10 and M 3099 slow growth temperate shipbuilding, 75 3 whole above (mainly from commerce population, F 2302 immigration) port 249 000 3/64-12/65 10-79 (1965) 3/66-12/66 10-69 Prague, 3/63-12/66 10 and M 1878 stable continental commerce, 80 11 District lI,a above mixed 106 000 F 2015 industry (1961) Ryazan, 1/63-12/66 10 and M 1421 rapid growth continental mixed 60 7 whole aboveb during past industry population F 850 15 years 214 000 (1959) Yalta, 1/63- 6/67 10 and M 904 stable temperate wine 60 1 whole above production, population F 673 tourism 44 000(1959) Tallin, 1/63-12/66 10 and M 2312 stable temperate heavy 60 5 whole above industry, population F 2001 textiles, 282 000 tourism (1965) a A central part of the city, with many government offices. b The autopsy rate was low for age 70 years and over. that no information could be made available other than death certificate data. Thus our material included all kinds of persons and all kinds of pathological condition and our observations led us to believe that the differences established by the methods used would be unlikely Table 2. Number of subjects studied, by age and sex, in the five selected towns Age group No. studied Age group No. studied 10-19 M 160 60-69 M 2703 F 80 F 2240 20-29 M 476 70-79 M 2013 F 179 F 2216 30-39 M 763 80-89 M 629 F 290 F 1016 40-49 M 922 90-99 M 43 F 560 F 97 50-59 M 1900 Total M 9614 F 1163 F 7841 to be affected by the cases that escaped detailed study. Definitions ofanatomical andpathological terms a Aorta, abdominal. The aorta extending from a horizontal line drawn through the upper edge of the orifice of the coeliac artery to a horizontal line drawn through the inner surface of the bifurcation of the aorta. Aorta, complete. The aorta from just above the aortic valves to the bifurcation. Aorta, descending thoracic. The aorta extending from a horizontal line drawn through the first two intercostal arteries to a horizontal line drawn through the upper edge of the orifice of the coeliac artery. Aorta, average. The descending thoracic aorta and the abdominal aorta considered together. a For more detailed descriptions see HOLMAN, R. L. ET AL. Lab Invest., 7: 42 (1958), and UEMURA, K. ET AL. Bull. World Health Organ., 31: 297 (1964). MATERIAL AND METHODS 491 Atherosclerosis. A variable combination of changes in the intima of arteries, consisting of the focal accumulation of lipids, complex carbohydrates, blood and blood products, and fibrous tissue and calcium deposits, associated with medial changes. Calcification. Areas where calcium deposition is detectable either visually or by palpation, without overlying haemorrhage, ulceration, or thrombosis. Complicated lesions. Areas in which there is ulceration, haemorrhage, or thrombosis with or without calcium deposits. Coronary, circumflex. The circumflex branch of the left coronary artery from its origin, excluding auxiliary branches. (If there are two vessels appar- ently equal in size, both should be included.) Coronary, left anterior descending. The left coron- ary artery from its orifice (including the ostium) and the anterior descending branch down to the apex of the left ventricle, excluding any branches. (If there are two vessels apparently equal in size, both should be included.) Coronary, right. The right coronary artery from its origin (including the ostium) and including the flexure at the margin of the posterior interventricular septum, excluding any branches. Coronary stenosis. Narrowing of the lumen of the coronary artery by more than 50°/. Coronary, average. The circumflex coronary, left anterior descending coronary, and right coronary considered together. Fatty streak. Any intimal lesion that is stained distinctly by Sudan IV or other fat-soluble dye and does not show any other type of underlying change. Fibrous plaque. Any firm, elevated intimal lesion which in the fresh state is pale grey, glistening, and translucent. After staining it may be partially or completely covered by sudanophilic deposits. If a lesion presents any haemorrhage, thrombosis, ulceration, or calcification, that portion will be classified under those categories and not as a fibrous plaque. (For operational purposes the term " ather- oma" is included in this definition although it is realized that there is controversy about the use of that term.) Raised lesion. The total of fibrous plaque, com- plicated lesions, and calcified lesions. Basic data gathered on each subject a The basic data form completed by the pathologist for each autopsy is shown in Fig. 1. Some of the items on the form are described below. Source (entry 15). The object of this entry was to discriminate between violent death from any source and non-violent death. The latter was subclassified according to whether death occurred in hospital or elsewhere, e.g., at home. Was death sudden? (entry 17). The object of this entry was to ascertain how many deaths coming within a practical definition of " sudden " were due to coronary occlusion. Accidental death or death due to intentional trauma was excluded. The definition used was " non-violent death occur- ring unexpectedly within 6 hours in an apparently healthy subject or in a sick person whose condition was either steady or improving ". Fat (entries 34-36). The depth of subcutaneous fat, halfway between the xiphoid sternum and the umbilicus in the midline, measured to the nearest millimetre, was recorded. Weight of heart (entries 37-39). The heart was weighed (in grams) after the coronary arteries had been removed and the ventricles and auricles opened to remove blood and blood clots. The auricles and epicardial fat remaining after removal of the coronaries were included in the heart weight. Clinical hypertension (entry 40). " Benign essential hypertension" meant that a clinical diagnosis of hypertensive cardiovzascular disease was made and that no chronic renal disease other than benign arteriolar nephrosclerosis was diagnosed either clinically or anatomically. This diagnosis was not based on the heart weight at autopsy. " Malignant hypertension" indicated a clinical diagnosis of malignant hypertension. " Other hypertension " indicated a diagnosis of hypertension secondary to another condition such as chronic renal disease, pyelonephritis, glomerulonephritis, eclampsia, or phaeochromocytoma, as established at autopsy. " Absent " indicated that clinical diagnosis of hypertension was excluded. " Unknown " was recorded when there was no clinical knowledge of the presence or absence of hypertension. In the absence of a diagnosis but the presence of blood a For more detailed descriptions see HOLMAN, R. L. ET AL. Lab. Invest., 7: 42 (1958), and UEMURA, K. ET AL. Bull. World Health Organ., 31: 297 (1964). A. R. KAGAN & K. UEMURA WORLD HEALTH ORGANIZATION 3-7 Accession number COMBINED EPIDEMIOLOGICAL AND PATHOLOGICAIL 8 Nama of hospital Code STUDIES OF ATHEROSCLEROSIS II Dote of death: 9. 12 Necropsy number F I 13 Place of residencs Urban 1 28-30 Weight r 48 Coronary occlusion Present 1 Rural 2 lk~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Ig wmitotou ittfarctton as Absent 2Rural _ .J.1.......L.. the cause of death : _ Unknown _9 31 33 Stature (length) Unknown 9 14 Co-operating pathologist ar I IlIIF Cerebro vascular accident: 34.36 Fat panniculus 49 Post- mortem Haemorrhage Ibetween xiphoid 7 recent lesion: 15 Source Accident 1 and umbilicus _j tje Inforction 2 Not aceident: Hospital 2 37.39 Heart weight after One or other 3 _ removing coronary gr (not sure which)Home 3 arteries _T 9 None found 4 Unknown 9 40 Clinical hypcrtcnsion:Brnian 1 Not exanined 9 esscntail hypertension _ o _ 16 Medical core before death Yes 1 Malignant hypertension 2 50 Post-mortem old Present 1lesion Cyst greaterNo 2 Other hypertension 3 than 0.5scm Absent 2 Not surc 9 Absent 4 Not examined 9 17 Was death sudden? Yes 1 Unknown 19 51 Disabling peripheral vascular (hor n,n arppaunr,ivr,uc,.6 No 2 41 Diabetes rdlli us: Present _1 disease (Intermittent claudication; o in .sick erson shose colidni.s _ Raynewd's syndrome, gangrene or riiristeidy*eeiagi Unknown 9 Absent 2 amputation of extremity due to 18 Was diagnosis made before Yes Unknown 9 occlusive disease of arteries) post-mortem? No 2 42 Syphilis aorta Present I (Underline appropriate categories) Unknown_ 9 (macroscopic) Absent 2 19.21 If yes, clinical diagnosis: Doubtful 3 Coronary occlusion Present nI 43 Present 1 Absent L 2 Code Absont 2 Unknown 99 ¶..ilZ Uncertoin 3 52 Other diseases due to athero- 22 23 Dote of birth Code 44 Due to thrombus sclerosis (Atherosclerotic or aortic day mo yr Not d to t 2 aneurysm; carotid, mesenteric,F NI ot due to thrombus 2 coeliac or renal artery occlusion 24 Sex Male I| 1 Not applicable 3 by atherosclerotic lesion) Femal 2 Not surc 9 (Underline appropriate categories) 25 Ethnic group 45 Recent myocardial Single 1 inforction Multiple 2 Code Non. 3 Present n I 46 Scar duo to myocardial Single 1 Absent n 2 26.27 Occupation: inforction greater than I1 or equol to 0.5 cm Multiple 2 Not sought 9 None 3 53 Pathological confirmation Yes 1 ______________ _Code 47 Small scar less thon Single 1 of clinical diagnosis: No 2 L ] 05 cm Multiple 2 No clinical diagnosi 3 F INone 3 54. 59 Pathological diagnosis Signeture WHO 5004 Fig. 1. The basic data form completed by the pathologist for each autopsy. 492 MATERIAL AND METHODS 493 pressure records, a record of pressure of 160/95 mm Hg or more in subjects under 60 years of age or of 140/90 mm Hg or more in subjects under 40 years of age was regarded as indicating the presence of essential hypertension. In the absence of a diagnosis and the presence of records showing blood pressure less than these, hypertension was regarded as absent. Other cases and subjects aged 60 years or over were included in the " Unknown " group, in the absence of a clinical diagnosis. Diabetes mellitus (entry 41). Diabetes mellitus was reported as " absent " when a reasonably adequate medical examination had been performed (urine or blood examined for sugar) and no clinical diagnosis was made. " Unknown " was recorded when no clinical history or laboratory data of diabetes were available and the presence of diabetes could not be reasonably excluded. " Present " was recorded when the information available clearly indicated the presence of diabetes. Syphilis of the aorta (entry 42). This condition was recorded as " present " if there was gross evidence of syphilitic aortitis in any portion of the aorta. The diagnosis was made on the basis of gross appearance and not on histological study of the aorta. Coronary occlusion (entries 43 and 44). The presence or otherwise of coronary occlusion was recorded for the three coronary arteries as a whole, under one of the headings " present ", " absent ", or "uncertain " (entry 43), and under one of the headings " due to thrombus ", "not due to throm- bus ", " not applicable ", or "not sure " (entry 44). In cases where there was more than one occlusion, some due to thrombus and others not, both squares were ticked. Myocardial infarct (entries 45, 46 and 47). The cooperating pathologist indicated his opinion in the appropriate squares. One square in each of entries 45-47 was ticked in all cases. In the rare event of a large scar (> 0.5 cm) not due to myocardial infarction being present, entry 46.3 was ticked and a note was made under " pathological diagnosis (entries 54-59). Coronary occlusion without infarction as the cause of death (entry 48). " Present " was recorded in cases of sudden death in which the principal cause of death was judged by the pathologist to be myo- cardial ischaemia due to coronary stenosis or occlusion (which in turn could be due to thrombosis, haemorrhage into a plaque, or growth of a plaque) but in which no myocardial necrosis was grossly detectable. This entity could, furthermore, be considered " present " in cases with myocardial scars but no congestive failure as a result of the previous infarction. In such cases it may be presumed that a second occlusive episode developed that resulted in sudden death. Cases in which the oc- clusion was clearly due to embolism (e.g., bacterial endocarditis) were excluded from this category. " Absent " was ticked for these cases and the diag- nosis listed in the final diagnosis. Cerebrovascular accident (entries 49 and 50). The cooperating pathologist indicated his macroscopic post-mortem findings in relation to a recent lesion (entry 49) and an old lesion (entry 50). " Cyst " was recorded as present only if it was greater than 0.5 cm in diameter (entry 50). Disabling peripheral vascular disease (entry 51). "Present " was recorded in any case in which there was a history of intermittent claudication, Raynaud's syndrome, gangrene, or amputation of an extremity due to occlusive disease of the arteries of the ex- tremities, or if the condition was known to be due to trauma or embolism not related to atherosclerosis. Other diseases due to atherosclerosis (entry 52). "Present " was recorded for atherosclerotic aneur- ysm with or without rupture or occlusion of carotid, mesenteric, coeliac, renal, or other arteries by atherosclerotic lesions. Calcified aortic stenosis, calcified mitral ring, medial necrosis of the aorta with or without dissecting aneurysm, Monckeberg's sclerosis, or congenital aneurysm of the circle of Willis were not included in this category. Pathological confirmation of clinical diagnosis (entry 53). The object of this entry was to confirm (" yes ") or refute (" no ") a pre-autopsy diagnosis, if made (entries 19-21). If there was no pre-autopsy diagnosis, " no clinical diagnosis " was recorded. Pathological diagnosis (entries 54-59). Ample space was provided to enter all the pathological findings. Later a special coding form with 300 rubrics was designed and used in Malmo and Prague.a a Copies of this form are available from Cardiovascular Diseases, World Health Organization, 1211 Geneva 27, Switzerland. 494 A. R. KAGAN & K. UEMURA Missing data It was not possible to obtain all data on some of the subjects, and some of the data obtained were unsuitable for assessment. Efforts were made to monitor such losses before and during the study, to reduce them where possible, and where this was not possible to take special measures to establish how the losses should be taken account of in the analysis so as to avoid arriving at false conclusions. Aorta and coronary arteries. These vessels were not always suitable for assessment when they arrived at the central processing laboratory. The wastage varied from town to town and for anatomical part, but it was reasonably slight-7Y to less than 1 % according to vessel-and did not differ much according to pathological diagnosis. The wastage was negligibly small for age group 40-69, which was analysed most intensively. Basic data items. Information on sex was always reported, but there were a few cases for which the date of birth was unknown; age in years had to be estimated for them. Items frequently missing were body weight in Ryazan (almost all), thickness of fat in Malmo (22%Y, i.e., nearly all forensic cases), hypertension and diabetes mellitus in all towns (about 20%Y), and peripheral vascular disease in Ryazan (37°/). Information on hypertension (8-31 %) and diabetes mellitus (5-38%4) was often missing because it was not possible to make special arrangements for observations to be made before death. Because of the importance of studying these two factors in relation to atherosclerosis, special tests were made on the reliability of assuming that when information was absent it could be assumed that this condition was absent. Personal environment factors. The last occupation, as described in commonly used terms in each location, was entered on the basic data sheet (Fig. 1, entries 26-27). Because many of the older subjects were described as " pensioner " and many of the women were described as " housewife ", and because there were relatively few deaths below 40 years, the analysis of occupation and atheroscler- osis was confined to males aged 40-59 years. In Chapter 18 it is explained how occupation in this age and sex group was assessed in terms of physical activity and the steps taken to ensure reasonable reliability. Alcohol consumption and cigarette smoking. These factors were assessed in Yalta only, fromclin ical records and from an interview of relatives. COLLECTION AND PREPARATION OF MATERIAL AND DATA a The basic data form (Fig. 1) was completed in duplicate at the pathology department during the course of each autopsy. One copy was sent to WHO, Geneva, and one was retained by the pathology department. The coronary arteries and aorta were removed, prepared, and fixed in a standardized way, placed in plastic bags with formalin, and sent to the Central Laboratory, Malmo, for further pro- cessing. From this stage onwards the vessels were identified only by the tag-number enclosed in the bag, which corresponded to the accession number on the basic data form and which was supplied in random order by the central statistical service in Geneva. WHO was responsible for checking the completion of the basic data form and the Central Laboratory was responsible for checking the quality of the prepared specimens, processing them by staining with Sudan IV according to a standardized pro- cedure, replacing them in numbered plastic bags with the same number and storing them for further use. A continuous quality control was exercised by the Central Laboratory on the state of material and data, and defects could readily be identified and their cause often corrected. METHODS OF ASSESSMENT, TESTS OF PRECISION AND BIAS, AND QUALITY CONTROL In order to achieve the objective of comparing atherosclerosis of the type defined earlier in this chapter according to community, age, sex, cause of death etc., methods had to be devised that were sufficiently precise. It was necessary to know how imprecise and how biased the methods were before and during the study and if necessary to reduce these errors.3 The term " precision " is used in connexion with this study to mean the closeness to one another of repeated measurements of the same thing. Errors of precision, or imprecision, can be in the direction a For more detailed descriptions see HOLMAN, R. L. ET AL. Lab Invest., 7: 42 (1958), and UEMURA, K. ET AL. Bull. World Health Organ., 31: 297 (1964). b For a further description of these and other errors and of some of the tests used, see the following authors. KAGAN, A. R. In: SOMMERS, S. C., ed. Pathology annual, New York, Appleton-Century-Crofts, 1969, p. 1. KAGAN, A. R. In: SHERRY, S. ET AL. ed., Thrombosis, Washington, DC, National Academy of Sciences, 1969, p. 236. UEMURA, K. ET AL. Bull. World Health Organ., 31: 297 (1964). MATERIAL AND METHODS 495 of too high or too low; too much or too little, present or absent. If many observations are made on the same thing or group of things, the errors tend to cancel each other out. Errors of bias differ from imprecision in that they tend to be in one direction only, too high, too low, too large, too small, absent or present. Bias can be considered as the direction and amount of the tendency to read away from the true value or from another observer (i.e., inter- observer bias). Such errors do not cancel each other out. Assessment of atherosclerosis in the aorta and coronary arteries Each of the four types of atherosclerosis and also the total area of atherosclerosis were graded visually as the percentage of the surface of the defined artery. Coronary stenosis was assessed as present or absent. The record forms used are shown in Fig. 2. From time to time a batch of specimens identified only by number was assigned to pathologists at random, except that each examiner received an equal proportion from each community. Prior to the commencement of the grading session the pathologists met together, reviewed the definitions and rules, and independently assessed a standard set of specimens. Reasons for differences in their assessment were discussed. This preliminary session enabled the pathologists each time to clarify the definitions of atherosclerosis and the way in which they applied them. During the grading session it was arranged that each pathologist graded some specimens twice (some he had graded already at that session, some that each of his colleagues had graded, and some that had been graded at a previous session). Precision. Through monitoring and practice, precision was greater during this study period than during the trials.a The range of average intra- observer variation over 16 grading sessions was 1.0-6.3% for different atherosclerotic lesions of the aorta and 0.3-6.5% for lesions of the coronary arteries. Inter-observer bias was kept fairly low. For comparison of groups of vessels from different communities this bias was completely neutralized by the method of assigning equal proportions of vessels from each community to each grader. Since specimens were randomly allocated, inter-observer a See UEMURA, K. ET AL. Bull. World Health Organ., 31: 297 (1964). bias was also neutralized with respect to age, sex, and cause of death. This in effect transformed the bias into imprecision-an advantage, because imprecision is decreased by increasing the number of specimens in each group to be compared, whereas the effect of bias is not. The range of average inter- observer bias over 13 sessions was 1.3-8.7%y for different atherosclerotic lesions of the aorta and 0.3-8.4%o for lesions of the coronary arteries. Bias arising from knowledge of the source of the specimen. It is quite probable, because of the sub- jective method of grading atherosclerosis, that a grader would tend to read higher if he knew the specimen related to a condition with which he associated high degrees of atherosclerosis and lower if he knew it related to a condition with which he associated low degrees of atherosclerosis. This bias was completely obviated by withholding from the graders all information on the source. Bag bias could arise because the two parts of an individual's aorta were always in one bag and therefore each part was in the grader's mind when he assessed the other. Similarly, the three coronaries of a subject were in the same bag. But the aorta and coronary arteries of the same subject were in different bags and what was seen in one would not affect the grading of the other, even by the same observer.b Inter-session variations. In spite of all precautions there was a tendency to grade lesions differently in the different sessions, especially lesions in the aorta, and to give a lower assessment for the total amount, fatty streaks, and fibrous plaques as time went on. This effect was neutralized in so far as specimens from different sources were equally allocated in each session. However, the results of the first two sessions were found to differ so widely from those of the later sessions that they were discarded and all the specimens were regraded at later sessions. It was concluded that the precision of observation was sufficient to discriminate with respect to the extent of atherosclerosis and coronary stenosis between groups of subjects of different age, sex, and geographical origin. Bias was kept within reasonably known limits and could be allowed for when necessary. b For details of how this effect was assessed, how it overestimated correlations of lesion between the two parts of the aorta and between the three coronary branches, and how the effect was allowed for, see STERNBY, N. H. ET AL. Bull. World Health Organ., 33: 741 (1965). A. R. KAGAN & K. UEMURA WORLD HEALTH ORGANIZATION Combined Epidemiological and Pathological Studies of Atherosclerosis Visual grading - Aorta 1-2. Study number: 30 8. Grading session: 3-7. Accession number. 9. Observer: M5HO/I/53&.63 WORLD HEALTH ORGANIZATION Combined Epidemiological and Pathological Studies of Atherosclerosis 1-2. Study number: 30 30. Grading session: Visual grading - Coronaries 3-7. Accession number: 31. Observer: Diffuse Coronary Total amount redness Coronary stenosis atherosclerosis Faity streak Fibrous plaque Complicated Calcified (tick if 50% ~~~~~~~~~~~~~~~~~~~~~~~~present) 1 None 32 3334 35-3G; 37-38 39-40 41-42 Right 0 3 >1 Area 4 Unknown 143 46-47 48-49 52-53 Left 0I None anterior 2 1Area descending 3 >1 Area 4 Unknown I None 54 57- - 1-2 63-64 Left 2 1 Area circumflex 0 3 >I Area 4 Unknown Tick appropriate square MH0//N3b.63 Fig. 2. The record forms used in the coronary arteries. study for assessing atherosclerosis in the aorta and 496 Diffuse Total amount rednessAorta atherosclerosis Fatty streak Fibrous plaque Complicated Calcified (tick if present) 10-11 1 2-13 14-15 16-17 18-19 Descending thoracic 2021 22-23 2256-7 28-29 Abdominal MATERIAL AND METHODS 497 In order to compare the findings of the present study with those of the International Athero- sclerosis Project," 10 sets of aortas and 10 sets of coronary arteries from each study were exchanged and their grading in the two studies compared. It was found that precision (intra-observer variation) was similar in the two groups and that the Inter- national Atherosclerosis Project group tended to assess the total amount of atherosclerosis and the prevalence of fatty streaks higher and the prevalence of coronary stenosis less frequently. Diagnosis of myocardial infarction, cerebral haemor- rhage, and cerebral infarct The collecting pathologist's diagnosis of myo- cardial infarction, cerebral haemorrhage, and cere- bral infarct was used exclusively. During the course of the study, a test was carried out on the reliability of diagnosis in Malmo, Prague, and Yalta. Diagnosis of fresh myocardial infarction, large myocardial scar, recent cerebral haemorrhage, and recent cerebral infarct was found to be completely reliable but assessment of small myocardial scars and old cerebrovascular lesions was not. These two factors were therefore excluded from the analysis. Coronary occlusion due to thrombosis The diagnosis of coronary occlusion due to thrombus or otherwise as made by the collecting pathologists was checked in a special study of 10 000 specimens made by a panel of pathologists. Table 3 shows that the collecting pathologists seldom missed coronary occlusion with or without thrombus when coronary heart disease or myo- cardial infarction was diagnosed clinically or at autopsy but missed about half the cases when these diseases were absent. For this reason an " alertness " factor was calculated to adjust the prevalence of coronary occlusion with or without thrombosis when populations or disease conditions were being compared. Hypertension and diabetes mellitus In 8-31% of the autopsied cases (according to population) the diagnosis of hypertension was recorded as " unknown ", and in 5-38%/ the diag- nosis of diabetes mellitus was " unknown ". It was important to know whether the diagnosis could be relied on if made or excluded and to what extent we could assume that " unknown " meant absent. a See MCGILL, H. C. JR, ed. Lab. Invest., 18(5): 463(1968) (reprinted, Baltimore, MD, Williams & Watkins). Table 3. Routine assessment of thrombosis in extra- myocardial coronary arteries, according to clinical or autopsy diagnosis, compared with special inquiry when the clinical and autopsy diagnoses were not known (men and women aged 40-59 years in Malmo, Prague (District 11), Yalta, Ryazan, and Tallin) No. in which Alert- No. of thrombus found nessDiagnosis deaths __________factor, routine/ Routine Special special Clinically, coronary heart disease 200 32 35 0.91 Clinically, not coronary heart disease 2300 18 34 0.53 Autopsy, 'myocardial infarct' 478 44 56 0.79 Autopsy, not 'myo- cardial infarct" 2022 6 13 0.46 Source: Kagan, A. R. In: Sherry, S. et al., ed. Thrombosis. Washington, DC, National Academy of Sciences, 1969, p. 236. A random sample of clinical case histories (without the basic data records) were reviewed by clinicians of the relevant department and their con- clusion was requested on the assessment of several factors including hypertension and diabetes mellitus. This enquiry showed that a basic data recording of " present " or " absent " was reliable for hyper- tension or diabetes mellitus, that " unknown " for diabetes mellitus could be regarded as diabetes mellitus absent, and that " unknown " for hyper- tension could not be regarded as hypertension absent. Thus in one population 91% of the unknowns remained unknown, while in the others 12% and 17% of the unknowns were regarded as hypertensive. Thus, in analyses that seek to show the effect of diabetes mellitus or hypertension comparison is made with the group known to be non-hypertensive or non-diabetic. But the large intermediate group of hypertension or diabetes " unknown " usually behaves in the same way as the hypertension or diabetes " absent " group. METHODS OF ANALYSIS Data entered on the basic data form, the form for detailed study of pathological findings, and the form for visual grading were transferred to punched cards, matched for each subject, and stored on a computer tape. The detailed pathology form was A. R. KAGAN & K. UEMURA used in Malmo and Prague. For the other three towns the details given in the basic data form (items 54-59) were coded according to the detailed pathol- ogy form. In this way the principal diseases could be similarly coded in all the centres. However, concurrent diseases were not recorded uniformly between the two groups of towns, except those recorded specifically on the basic data form (hyper- tension, diabetes, fresh myocardial infarction, myocardial scars, occlusion, thrombus, cerebro- vascular accidents, diseases of peripheral arteries, and other atherosclerotic lesions). The results of the grading of atherosclerotic lesions were verified by a computer programme for their internal consistency and whenever dis- crepancies were detected between the total number of lesions and the sum total of fatty streak, fibrous plaques, complicated lesions, and calcified lesions, the specimen was included with the normal spe- cimens at the subsequent session for regrading. The degree and the extent of atherosclerotic lesions were assessed in terms of the prevalence and mean (see equations (1) and (2) below). Binomial formulae were used in the computation of the standard error and statistical testing for the prevalence rate. The formulae for the two-tailed t-distribution were used in computing the standard error and statistical testing with regard to means. In some instances inter-town differences were studied by summarizing the results of the above- mentioned tests in an index giving the number of times the atherosclerosis measurement was higher for a particular town in comparisons with the other towns. The main purpose of using this index was to have a quick overview of the general tendency. The statistical significance of the index was determined by computing the combinational probabilities. Age- and sex-standardized rates The degree and the extent of atherosclerotic lesions vary considerably according to the age and sex of the subjects. For instance, fatty streaks first increase in relatively young ages followed by a decline, while all the other types of lesion generally increase with age. In order to summarize the com- parison of groups of different age and sex composi- tion, standardized rates and means were computed for the age range 40-69 years by using 5-year age groups (see equations (3) and (4) below). The standard error (SE) of the standardized rates was estimated using equation (5) below. The standard rates (or means) were thus the overall rate or mean referring to a hypothetical group composed of an equal number of subjects in all age (or age-sex) groups. Generally the stan- dardized rates (or means) were computed only when at least 10 subjects were available per subgroup. Similar formulae were used to compute rates (or means) standardized for age, sex, and town. Correlation coefficients were computed between lesions in different vessel parts. In Chapter 17 the correlation of atherosclerotic lesions is studied together with age, sex, town, diseases conditions, and body measurements. Some of these factors were expressed as " yes " or " no " and in such cases the value 1 was assigned to "yes " and the value 0 to " no ". In a more detailed analysis of these multivariate relations, stepwise regression analysis and analysis of variance were also performed No. of subjects with that type of lesion Prevalence of a specific type of lesion =. .x 100%° No. of subjects examined Total of the extent of specific lesions over oll subjects Mean of a specific type of lesion - T _ _ __:_ Age-standardized rate (or mean) No. ot subjects examinecd Total of age-specific rates (or means) over all age groups No. of age groups Total of age- and sex-specific rates (or mean) over all age-sex groups (4)Age- and sex-standardized rate (or mean) No. of age-sex groups SE (age-standardized rate) V' E [ SE (age-specific rate) ]No. of age groups where I stands for summation over all the age groups. 498 MATERIAL AND METHODS to show the proportion of the variance of each lesion attributable to particular factors or groups of factors. Reference groups In evaluating the degree and extent of athero- sclerosis in subjects suffering from a specific disease, it was thought convenient to compare them with reference groups. Three reference groups were used in the data analysis, as follows: 1. A low atherosclerosis group composed of subjects satisfying the following two conditions: (a) heart weight under 400 g in men and under 350 g in women, and (b) absence of certain specified diseases (diabetes mellitus, syphilis, occlusion of the coronary arteries, fresh or old myocardial infarction, multiple diffuse myocardial scar, fresh or old cerebrovascular lesion, disease of peripheral vessels, other diseases related to atherosclerosis, rheumatic heart disease, cor pulmonale, malignant tumour). 2. A standardized average atherosclerosis group constructed artificially by including equal proportions of 5 groups of causes of death-coronary heart disease, atherosclerotic disease as defined in group 3 below, cancer, violent causes, and other diseases (mostly infectious diseases). 3. A high atherosclerosis group composed of cases where death was due to coronary heart disease, aneurysm of the aorta, amputation for gangrene, thrombus of any artery, cerebral haemor- rhage, encephalomalacia, or a pseudocyst greater than 0.5 cm in diameter, recorded as principal disease, but where neither hypertension nor diabetes mellitus was mentioned. The observed degree and extent of atherosclerosis in subjects suffering from a specific disease were compared with the expected degree in the three reference groups by standardizing them indirectly for age and sex and for each town, i.e., the prevalence rate and the mean were computed for each reference group having the same age (5-year groups), sex, and town composition as the specific disease group and used as reference values. The prevalence rate and the mean for the disease group were then measured on the three different scales, for each of the three reference groups, by dividing the observed rate or mean for the disease group by the expected rate or mean for the reference group having the same age and sex composition as the disease group. This method has advantages over direct stan- dardization in that the ratio can be computed for a disease group even when it contains some empty age/sex groups, and that it can be used to standardize for age, sex, and town. UME CHAPITRE 1. MATiRIEL ET METHODES Ce chapitre decrit la fagon dont ont ete abordes les problemes poses par l'obtention et l'evaluation d'un materiel demographique, ainsi que par I'analyse des donn6es. Certaines des insuffisances de la methode utilisee, qui a consiste a examiner la quasi-totalite des deces survenus dans plusieurs collectivites demographiquement definies, sont exposees. 499