International Evaluation of Cause-Specific Mortality and 1DDM
Objective: A cross-cultural study was completed to evaluate differences in mortality patterns in four population cohorts in Japan; Israel; Allegheny County, Pennsylvania; and Finland. Research Design and Methods: Cases were diagnosed between 1 January 1965 and 31 December 1979. Mortality was determined as of 1 January 1985. There were 147 deaths occurring in the 8212 insulin-dependent diabetes mellitus (IDDM) patients in the four countries. A standardized protocol for assessing causes of death (cause-specific mortality) was developed; in this article, we report the causes. Results: Major overall mortality differences by country appeared, with IDDM subjects in Japan much more likely to die than in the other countries. In Japan, the elevated mortality was the result of acute diabetesrelated complications and kidney disease. For all countries, mortality from acute diabetes-related complications accounted for a surprisingly high percentage of deaths (>25% in each country). A larger percentage of cases in Finland died as a result of suicide than for the other three countries. Conclusions: The results suggest that there are major cross-country
Diabetes Epidemiology Research International Mortality Study Group
differences in cause-specific mortality and that much of the premature mortality associated with diabetes is potentially preventable. Diabetes Care 14:55-60, 1990
nsulin-dependent diabetes mellitus (IDDM) patients have a marked increased risk of mortality from both acute and long-term diabetic complications; however, little is known concerning the geographic variations of IDDM cause-specific mortality. Studies from Steno Memorial Hospital in Denmark (1), the Joslin Clinic in the United States (2), and the Pittsburgh Children's Hospital Registry (3) have all demonstrated a decline in recent years from death due to the acute complications of diabetes. Most of the excess mortality in these IDDM populations was related to kidney disease in the first 20 yr of diabetes and, thereafter, increasingly to cardiovascular disease (1-3). If these complications of diabetes were to result mainly from the disease process itself, the IDDM mortality experience should be similar across nations with varying background mortality patterns. If, however, the development of complications and death are subject to major ethnic/ genetic or environmental influences, then the variation in cause-specific mortality should be quite large across nations having diverse background mortality patterns. To address these issues, a standardized approach to the collection of mortality data and the classification of cause of death for IDDM patients in diverse countries has been undertaken. This article presents the causespecific mortality for the four countries under study: Japan, the U.S. (Allegheny County, PA), Finland, and Israel. In an accompanying article (this issue, p. 49) the total mortality results are presented, which indicated major cross-population differences in all-cause mortal-
I
This article was prepared for the DERI Croup by Dr. Trevor J. Orchard from the Department of Epidemiology, Graduate School of Public Health, University of Pittsburgh, Pittsburgh, Pennsylvania; and the Mortality Classification Committee (T. Orchard |Chairman|, ). Tuomilehto, M. Karp, N. Tajima, E. Miki, and ). Fuller). Address correspondence and reprint requests to Trevor). Orchard, Graduate School of Public Health, Department of Epidemiology, A530 Crabtree Hall, Pittsburgh, PA 15261. Received for publication 26 February 1990 and accepted in revised form 24 August 1990. T. Kitagawa, N. Tajima, M. Aono, S. Aono, I. Hibi, Y. Ikeda, G. Isshikki, N. Kasuga, M. Kawamura, K. Kida, Y. Kohno, M. Maryuma, M. Matsushima, N. Matsuura, E. Miki, E. Mimura, A. Takeda, and T. Toyota, Japan; R.E. LaPorte, A.L. Drash, L.H. Kuller, D.|. Becker, J.S. Dorman, I.F. Gower, T.J. Orchard, S.L. Patrick, M. Rewers, and R. Vergona, Allegheny County, Pennsylvania; ). Tuomilehto, R. Lounamaa, P. Lounamaa, A. Reunanen, E. Tuomilehto-Wolf, and H. Akerblom, Finland; Z. Laron, B. Bauman, M. Karp, M. Modan, and O. Gordon, Israel; and T. Orchard (Chairman),). Tuomilehto, M. Karp, N. Tajima, E. Miki, and ). Fuller, Mortality Classification Committee.
DIABETES CARE, VOL. 14, NO. 1, JANUARY 1991
55
CAUSE-SPECIFIC MORTALITY AND IDDM
ity. This analysis of cause-specific mortality evaluates the underlying causes of death to provide insight as to why the country of residence is such a potent determinant of death among individuals with IDDM.
RESEARCH DESIGN AND METHODS Our accompanying article describes in depth the populations and follow-up procedures: the reader should refer to this article (this issue, p. 49). Briefly, four large population-based cohorts were examined (Table 1). These consisted of patients diagnosed as having diabetes, placed on insulin, 33% of the deaths were due to accidents, suicides, and homicides. To permit meaningful analyses by country, sex, and duration of diabetes, deaths were grouped by underlying cause into acute diabetic complications (i.e., hyperglycemia or hypoglycemia, diabetic ketoacidosis, and unspecified coma), kidney disease thought to be diabetic, accidents, suicides, or homicides, and "other." Table 5 shows these groupings of death by duration of diabetes. Kidney disease became a prominent cause of death after 10 yr duration for all countries. Nevertheless, acute complications, which accounted for more
than half the deaths in the early years, still accounted for 25% of the deaths in those dying at diabetes durations of >10 yr. The patterns of death for men and women were relatively similar, except in the area of violent deaths. Men were much more likely to die violently than women (Table 6). Annual age-adjusted death rates per 100,000 personyr of IDDM are presented in Table 7. There were very high mortality rates for acute complications in Japan. The cause-specific mortality rates across the other three countries were roughly comparable and had overlapping confidence intervals. Diabetes was determined to have contributed to the deaths in 96% of Japanese patients as opposed to only 64% in Finland. Diabetes was determined to be directly causal in 81% of Japanese patients but only 38% of Finnish patients (Table 8). Table 9 shows the important effect of the MCC procedure on the classification of death. The MCC underlying cause of death was compared to item 1C on the death certificate (or 1B or 1A if 1C or 1B was blank). Three major effects were seen. First, the MCC classification resulted in a more specific cause of death being determined in many subjects (i.e., the "other" grouping was reduced by half). Second, approximately twice as many deaths from diabetic kidney complications were identified by the MCC than by the original death certificate. Third, acute complications of diabetes were more commonly determined in this classification procedure.
TABLE 6 Deaths by cause and sex Male
Acute complications Diabetic kidney disease Accident/suicide Other Total
58
Female
Both sexes
Row %
n
Row %
n
Column %
n
47 40 86 48
21 10 30 20 81
53 60 14 52
24 15 5 22 66
31 17 24 29
45 25 35 42 147
DIABETES CARE, VOL. 14, NO. 1, JANUARY 1991
DERI GROUP
TABLE 7 Age-adjusted death rates per 100,000 person-yr of insulin-dependent diabetes mellitus by country and cause
Acute complications* Diabetic kidney disorderst Accident/suicide Other
Japan
Israel
United States
Finland
231.6(141.5-356.7) 276.7(158.3-448.2) 58.6(15.9-150.0) 113.5 (48.9-223.6)
44.65 (5.4-161.2) 0 22.0 (0.6-122.7) 64.7 (13.3-188.9)
72.1 (26.4-157.1) 22.8 (2.8-82.4) 44.6 (12.1-114.1) 90.2 (38.8-177.7)
39.4 (23.0-63.0) 15.6(6.3-32.1) 61.8 (40.4-90.8) 54.6(34.6-81.9)
The 95% confidence intervals by Poisson distribution are given in parentheses. *Rate for Japan is significantly higher than that of Finland. tRate for Japan is significantly higher than that of all other countries.
CONCLUSIONS
The inadequacy of death certificate classification alone has long been recognized, both generally and in IDDM (4). The application of our standardized procedure demonstrated a major redistribution of the causes of death. These changes resulted from both the review process and the apparent inappropriate completion of the original death certificate (in many cases the last-mentioned item was clearly not itended to be the underlying cause of death). Without the MCC review, the proportion of cases due to acute complications and kidney disease would be seriously underestimated. Furthermore, the differentiation between different diabetic causes of death could not have been done without this in-depth review for each individual patient. Difficulties with comparing data between studies underscore the need for a standard mortality classification system in diabetic populations. The use of death certificates alone would have been too imprecise to determine differences in cause-specific mortality across populations. There are few data with which to compare our results because of the young age of our cohorts and the previous lack of standardized approaches for mortality classification. For example, Green and Hougaard (5) reported cause-specific mortality based on a prevalent community of 727 insulin-treated subjects
Objective: A cross-cultural study was completed to evaluate differences in mortality patterns in four population cohorts in Japan; Israel; Allegheny County, Pennsylvania; and Finland. Research Design and Methods: Cases were diagnosed between 1 January 1965 and 31 December 1979. Mortality was determined as of 1 January 1985. There were 147 deaths occurring in the 8212 insulin-dependent diabetes mellitus (IDDM) patients in the four countries. A standardized protocol for assessing causes of death (cause-specific mortality) was developed; in this article, we report the causes. Results: Major overall mortality differences by country appeared, with IDDM subjects in Japan much more likely to die than in the other countries. In Japan, the elevated mortality was the result of acute diabetesrelated complications and kidney disease. For all countries, mortality from acute diabetes-related complications accounted for a surprisingly high percentage of deaths (>25% in each country). A larger percentage of cases in Finland died as a result of suicide than for the other three countries. Conclusions: The results suggest that there are major cross-country
Diabetes Epidemiology Research International Mortality Study Group
differences in cause-specific mortality and that much of the premature mortality associated with diabetes is potentially preventable. Diabetes Care 14:55-60, 1990
nsulin-dependent diabetes mellitus (IDDM) patients have a marked increased risk of mortality from both acute and long-term diabetic complications; however, little is known concerning the geographic variations of IDDM cause-specific mortality. Studies from Steno Memorial Hospital in Denmark (1), the Joslin Clinic in the United States (2), and the Pittsburgh Children's Hospital Registry (3) have all demonstrated a decline in recent years from death due to the acute complications of diabetes. Most of the excess mortality in these IDDM populations was related to kidney disease in the first 20 yr of diabetes and, thereafter, increasingly to cardiovascular disease (1-3). If these complications of diabetes were to result mainly from the disease process itself, the IDDM mortality experience should be similar across nations with varying background mortality patterns. If, however, the development of complications and death are subject to major ethnic/ genetic or environmental influences, then the variation in cause-specific mortality should be quite large across nations having diverse background mortality patterns. To address these issues, a standardized approach to the collection of mortality data and the classification of cause of death for IDDM patients in diverse countries has been undertaken. This article presents the causespecific mortality for the four countries under study: Japan, the U.S. (Allegheny County, PA), Finland, and Israel. In an accompanying article (this issue, p. 49) the total mortality results are presented, which indicated major cross-population differences in all-cause mortal-
I
This article was prepared for the DERI Croup by Dr. Trevor J. Orchard from the Department of Epidemiology, Graduate School of Public Health, University of Pittsburgh, Pittsburgh, Pennsylvania; and the Mortality Classification Committee (T. Orchard |Chairman|, ). Tuomilehto, M. Karp, N. Tajima, E. Miki, and ). Fuller). Address correspondence and reprint requests to Trevor). Orchard, Graduate School of Public Health, Department of Epidemiology, A530 Crabtree Hall, Pittsburgh, PA 15261. Received for publication 26 February 1990 and accepted in revised form 24 August 1990. T. Kitagawa, N. Tajima, M. Aono, S. Aono, I. Hibi, Y. Ikeda, G. Isshikki, N. Kasuga, M. Kawamura, K. Kida, Y. Kohno, M. Maryuma, M. Matsushima, N. Matsuura, E. Miki, E. Mimura, A. Takeda, and T. Toyota, Japan; R.E. LaPorte, A.L. Drash, L.H. Kuller, D.|. Becker, J.S. Dorman, I.F. Gower, T.J. Orchard, S.L. Patrick, M. Rewers, and R. Vergona, Allegheny County, Pennsylvania; ). Tuomilehto, R. Lounamaa, P. Lounamaa, A. Reunanen, E. Tuomilehto-Wolf, and H. Akerblom, Finland; Z. Laron, B. Bauman, M. Karp, M. Modan, and O. Gordon, Israel; and T. Orchard (Chairman),). Tuomilehto, M. Karp, N. Tajima, E. Miki, and ). Fuller, Mortality Classification Committee.
DIABETES CARE, VOL. 14, NO. 1, JANUARY 1991
55
CAUSE-SPECIFIC MORTALITY AND IDDM
ity. This analysis of cause-specific mortality evaluates the underlying causes of death to provide insight as to why the country of residence is such a potent determinant of death among individuals with IDDM.
RESEARCH DESIGN AND METHODS Our accompanying article describes in depth the populations and follow-up procedures: the reader should refer to this article (this issue, p. 49). Briefly, four large population-based cohorts were examined (Table 1). These consisted of patients diagnosed as having diabetes, placed on insulin, 33% of the deaths were due to accidents, suicides, and homicides. To permit meaningful analyses by country, sex, and duration of diabetes, deaths were grouped by underlying cause into acute diabetic complications (i.e., hyperglycemia or hypoglycemia, diabetic ketoacidosis, and unspecified coma), kidney disease thought to be diabetic, accidents, suicides, or homicides, and "other." Table 5 shows these groupings of death by duration of diabetes. Kidney disease became a prominent cause of death after 10 yr duration for all countries. Nevertheless, acute complications, which accounted for more
than half the deaths in the early years, still accounted for 25% of the deaths in those dying at diabetes durations of >10 yr. The patterns of death for men and women were relatively similar, except in the area of violent deaths. Men were much more likely to die violently than women (Table 6). Annual age-adjusted death rates per 100,000 personyr of IDDM are presented in Table 7. There were very high mortality rates for acute complications in Japan. The cause-specific mortality rates across the other three countries were roughly comparable and had overlapping confidence intervals. Diabetes was determined to have contributed to the deaths in 96% of Japanese patients as opposed to only 64% in Finland. Diabetes was determined to be directly causal in 81% of Japanese patients but only 38% of Finnish patients (Table 8). Table 9 shows the important effect of the MCC procedure on the classification of death. The MCC underlying cause of death was compared to item 1C on the death certificate (or 1B or 1A if 1C or 1B was blank). Three major effects were seen. First, the MCC classification resulted in a more specific cause of death being determined in many subjects (i.e., the "other" grouping was reduced by half). Second, approximately twice as many deaths from diabetic kidney complications were identified by the MCC than by the original death certificate. Third, acute complications of diabetes were more commonly determined in this classification procedure.
TABLE 6 Deaths by cause and sex Male
Acute complications Diabetic kidney disease Accident/suicide Other Total
58
Female
Both sexes
Row %
n
Row %
n
Column %
n
47 40 86 48
21 10 30 20 81
53 60 14 52
24 15 5 22 66
31 17 24 29
45 25 35 42 147
DIABETES CARE, VOL. 14, NO. 1, JANUARY 1991
DERI GROUP
TABLE 7 Age-adjusted death rates per 100,000 person-yr of insulin-dependent diabetes mellitus by country and cause
Acute complications* Diabetic kidney disorderst Accident/suicide Other
Japan
Israel
United States
Finland
231.6(141.5-356.7) 276.7(158.3-448.2) 58.6(15.9-150.0) 113.5 (48.9-223.6)
44.65 (5.4-161.2) 0 22.0 (0.6-122.7) 64.7 (13.3-188.9)
72.1 (26.4-157.1) 22.8 (2.8-82.4) 44.6 (12.1-114.1) 90.2 (38.8-177.7)
39.4 (23.0-63.0) 15.6(6.3-32.1) 61.8 (40.4-90.8) 54.6(34.6-81.9)
The 95% confidence intervals by Poisson distribution are given in parentheses. *Rate for Japan is significantly higher than that of Finland. tRate for Japan is significantly higher than that of all other countries.
CONCLUSIONS
The inadequacy of death certificate classification alone has long been recognized, both generally and in IDDM (4). The application of our standardized procedure demonstrated a major redistribution of the causes of death. These changes resulted from both the review process and the apparent inappropriate completion of the original death certificate (in many cases the last-mentioned item was clearly not itended to be the underlying cause of death). Without the MCC review, the proportion of cases due to acute complications and kidney disease would be seriously underestimated. Furthermore, the differentiation between different diabetic causes of death could not have been done without this in-depth review for each individual patient. Difficulties with comparing data between studies underscore the need for a standard mortality classification system in diabetic populations. The use of death certificates alone would have been too imprecise to determine differences in cause-specific mortality across populations. There are few data with which to compare our results because of the young age of our cohorts and the previous lack of standardized approaches for mortality classification. For example, Green and Hougaard (5) reported cause-specific mortality based on a prevalent community of 727 insulin-treated subjects
