Diabetlgia (1992) 35:973-979 Diabetlgia 9 Springer-Verlag 1992 Trends in mrtality frm diabetes mellitus in Taiwan, 1960-1988 R. S. Lin and W. C. Lee Institute f Public Health, Cllege f Medicine, Natinal Taiwan University, Taipei, Taiwan Summary. The increasing trends f mrtality frm diabetes mellitus in Taiwan are becming a public health cncern. The age-adjusted death rate fr diabetes was 3.7 per 100,000 ppulatin in 1960, which increased t 23.2 per 100,000 in 1988, a 6.3-fld increase ver the past 30 years. The mrtality data fr diabetes in Taiwan frm 1960 t 1988 fr bth sexes are presented and analysed using an age-perid-chrt mdel in rder t gain a better understanding f the pssible determinants f the time trends f this disease. Steeply increasing trends which are particularly prminent in the elder- ly are fund in the perids studied. Females and urban dwellers demnstrated greater risks fr develping the disease. Significant chrt effects were fund which peaked in the birth:chrt 1910-1917. This declined thereafter fr females but ha s levelled since then fr males. Further study is necessary t assess the aetilgical implicatins f diet and urbanizatin n time trends f diabetes. Key wrds: Diabetes mellitus, mrtality trends, age-peridchrt analysis, urbanizatin. Over the past 30 years, Taiwan has experienced a rapid sciecnmic develpment that is reflected in changes f many aspects f life-style and dietary habits and t a large extent als in disease patterns. Chrnic diseases, e. g. cerebral and cardivascular diseases and diabetes mellitus have emerged as majr diseases threatening the lives f the inhabitants n this island. While the leading cause f death was previusly cerebrvascular disease, cancer has been respnsible fr mst deaths since 1982. Hwever, the mst remarkane increase in chrnic disease mrtality in the past 30 years has been due t diabetes. In 1960, diabetes was respnsible fr less than 0.3 T f ttal deaths, while this figure increased remarkably t 3.8 %, accunting fr mre than 3,800 deaths in 1988, Diabetes is nw the fifth leading cause f death in Taiwan next t cancer, cerebrvascular disease, accident and crnary heart disease (Table 1). In additin t being a significant cause f mrtality, diabetes is believed t be a majr cause f disability in the elderly including blindness, renal failure, and lwer limb amputatins. Cardivascular and peripheral vascular cnditins appear t be abut tw times as cmmn in diabetic patients as in nn-diabetic subjects [1]. Diabetes clearly represents an imprtant public health prblem in Taiwan as well as in ther develped cuntries. Althugh it is well-knwn that genetic factrs are invlved in the aetilgy f diabetes, the tremendus increase in diabetic mr- tality in Taiwan ver the past 20 years dentes the fact that envirnmental factrs play an imprtant rle in the causatin f Type 2 (nn-insulin-dependent) diabetes. In this study, mrtality data frm diabetes in Taiwan were analysed and presented in rder t better understand the way dietary as well as urbanizatinal factrs affect the secular trends. Subjects and methds Subjects Infrmatin n diabetes mrtality and the data f mid-year ppulatin in Taiwan frm 1960 t 1970 were taken frm the "Vital statistics" published annually by the Taiwan Prvincial Department f Table 1. Number f deaths frm diabetes mellitus in Taiwan, 1960-1988, by year Number f Prprtin f Orderinleading diabetesdeaths ttal deaths(%) causes fdeath 1960 198 0.27 29 1970 482 0.71 18 1980 1,396 1.67 13 1985 2,930 3.22 7 1988 3,883 3.84 5
974 R. S. Lin and W. C. Lee: Diabetes mellitus mrtality in Taiwan 30........../ / Data f annual per capita incme and f calrie, prtein, and fat cnsumptin in Taiwan between 1960 and 1988 were taken frm "Taiwan Statistical Data Bk" published by the Cuncil fr Ecnmic Planning and Develpment, R. O. C. [4]. S E & < 20 10 i.... i,, J, i, i,, I I,,, I i i i i I i i, 1960 1965 1970 1975 1980 1985 Fig.1. Trends f diabetes mellitus mrtality in Taiwan, 1960-1988, by sex. Male--, Female... Health, R. O. C. [2] where they were tabulated by sex and 5-year age grup. The individual health recrds f all reprted cases as well as ppulatin data fr each twnship and district in Taiwan between 1971 and 1988 were generated directly frm cmputerized gvernment data sets. The cntributing causes f death were nt available in either the annual reprt r the cmputer file. Thus, mrtality frm diabetes, in this study, was defined as cde number 250 in bth Internatinal Classificatin f Diseases (ICD)-8 and (ICD)-9 listed in the death certificates as the underlying cause f death. The degree f urbanizatin f each twnship and district in Taiwan was given an index f 1 t 8 in the study f urban-rural classificatin by Tzeng and Wu [3], in which variables f ppulatin density, age cmpsitin, mbility, ecnmic activity and family incme, educatinal level, and health and sanitatin facilities were used t estimate the urbanizatin scre. Metrplitan areas, in this study, are defined as thse with index 1-2, and cities, index 3-4, twns, index 5~6, and rural areas, index 7-8, respectively. / Statistical analysis In rder t minimize the effect f different age cmpsitins fr different perids and different areas, the secular trends f diabetes mrtality rates fr the whle ppulatin between 1960 and 1988 and the trends f diabetes mrtality rates by urbanizatin in Taiwan were age-adjusted [5] t the 1976 wrld standard ppulatin [6]. The statistical age-perid-chrt analysis (APC Mdel), which adpted the "individual recrd" methd f Rbertsn and Byle [7] t circumvent the prblem f nn-identifiability, was perfrmed using the data between 1971 and 1988. The mrtality f diabetes mellitus in a particular grup (age-perid-chrt) is assumed t fllw a Pissn distributin, with a mean g = N- r, where N is persn-years at risk in a particular grup, and r is the mrtality rate frm diabetes in that grup. The mrtality rate r is mdelled as a lg-linear functin f age, perid and chrt separately fr males and females. The baseline grups are ages 40-44, calendar year 1971-1973, and birth year 1887-1893 fr age, perid, and chrt factrs, respectively. The maximum likelihd estimatin was used t fit the mdels thrugh the statistical package Egret [8]. The significance f the factrs (age, perid, and chrt), were assessed by means f a likelihd rati test. Other methdlgical details can be fund elsewhere [9]. Results Figure i shws the trends fr age-adjusted mrtality rates f diabetes fr males and females in Taiwan between 1960 and 1988. The increasing trends in mrtality rates frm diabetes are large. Fr males, the annual age-adjusted mrtality rate increased frm 3.32 per 100,000 ppulatin in 1960 t 17.84 per 100,000 ppulatin in 1988, a 5.4-fld increase, and fr females, 4.05 per 100,000 ppulatin in 1960 t 30.01 per 100,000 in 1988, a 7.4-fld increase. The secular trends in Figure i als shw that the increases in 400 B 3 300 L 200?. 100 I----~---T-I;--T--T--T"-T-'T" I'T'~--T,,, i,, i, I, I I I I I I I 0 1960 1965 1970 1975 1980 1985 1960 1965 1970 1975 1980 1985 Fig. 2 A, B. Secular trends f age-specific mrtality frm diabetes mellitus in Taiwan, 1960-1988, by sex. A Male 70 + --, 60~59...,50-59 - - -. B Female 70 + --, 60-69...,50-59 - - -.
R. S. Lin and W. C. Lee: Diabetes mellitus mrtality in Taiwan 975 1000- A 0 0 100 /... 10 1 1 J I I I [ I I I I 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85+ A g e (years) J~ s sss~sssslsls/ssss s/" i ~'~,.4"'.",,, 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85+ A g e (years) Fig. 3 A, B. Age-specific mrtality rate frm diabetes mellitus in Taiwan, fr the perids f 1971- i976, 1977-1982 and 1983 1988, by sex. A Male 1983-1988 - - -, 1977-1982--, 1971-1976.... B Female 1983-1988 - - -, 1977-1982--, 1971-1976.... Mrtality rate is expressed n a lgarithmic scale mrtality frm diabetes were particulary remarkable in the last 10 years fr bth sexes. Further analysis f the age-specific mrtality rates f three different age grups were pltted in Figure 2 fr males and females. It can be seen that the ldest age grup f 70 and ver in bth sexes shwed the mst dramatic secular trends, especially females (15.8-fld increase). Depicted in Figure 3 are the crss-sectinal age distributins f mrtality frm diabetes fr three cnsecutive perids f 1971-1976, 1977-1982 and 1983-1988 fr bth sexes. It can be seen that the mrtality rates increased prgressively with age fr males but shwed a dwnward curvature after age 75 years fr females. The separate effects due t age, chrt and perid indices, respectively (Fig. 4-6) were estimated frm the APC mdel [7] and were all statistically significant (p < 0.01) as judged by the likelihd rati test (Table 2) The ldest age grup, (80-84 years) had 46.8 and 60.1 times the risk f dying frm diabetes as cmpared t the yungest age grup f 40-44 years fr males and females, respectively. The dwnward curvature after ages 75 years fr females bserved in the crss-sectinal age distributins (Fig. 3) disappeared in this birth-chrt analysis (Fig. 4). As regards chrt effects, females brn between 1910-1917 achieved the highest risk amng the birth chrts studied (the relative risk is 2.04 as cmpared t the 1887-1893 birth chrt) and the effect declined in the recent chrts. Fr males, the birth-chrt effect was less cnspicuus. It was increasing befre the year 1909, reached a peak in 1910-1917 and remained at abut the same level f risk thereafter (1.5 times the risk as cmpared t the 1887-1893 birth chrt). We als analysed the chrt effects using finer gruping in the age, perid, chrt indices and the results were essentially the same. The age-adjusted diabetes mrtality rates by the degree f urbanizatin between 1971-1988 fr males and re-
976 R. S. Lin and W. C. Lee: Diabetes mellitus mrtality in Taiwan 80 7O 45~ i'"'"'"" 9 ~..'" 30 20 T.'" 10.,'" 0,L,,, 40-44 45-49 50-54 50-59 60-64 65-69 70-74 75-79 80-84 A g e (years) Fig.4. Age effect (with 95 % cnfidence interval) in diabetes mellitus mrtality in Taiwan, 1971-1988, by sex. Male.--., Female 9..., re T 0, I I i I I I I 1887 1894 1902 1910 1918 1926 1934 1942 i 1 I / I I I I 1893 1901 1909 1917 1925 1933 1941 1948 Fig. 5. Chrt effect (with 95 % cnfidence interval) in diabetes mellitus mrtality amng birth chrts f 1887-1948 in Taiwan, by sex. M ae9 1 F ema le 9... 9. T T that residents in rural areas will reach diabetes mrtality rates as high as the metrplitan figures in abut 6 years. Figure 8 shws the trends f annual per capita incme and f calrie, prtein, and fat cnsumptin in Taiwan between 1960 and 1988. The increasing trends f fat cnsumptin and calrie intake seem very similar t the trends f diabetes mrtality (Pearsn crrelatin cefficients, 0.97 and 0.82, respectively). Discussin In the absence f cmprehensive incidence and prevalence data abut diabetes, studies based upn vital statistics cmpiled frm death certificates seemed a reasnable first step in estimating trends f diabetic risk. Hwever, several studies have demnstrated that diabetes was reprted as the underlying cause in nly abut 7 % and as a cntributing r "ther" cause in abut ne-third f all death certificates [10-12]. Underestimatin f mrtality rates frm diabetes was, therefre, inevitable, if vital statistic data were used. It shuld be kept in mind that even in cmplete ascertainment frm death certificates, frm bth underlying and cntributing causes, using mrtality data t study trends in diabetes had the drawback f separating the effects f imprvement in medical care accessibility and the reductin in case fatality rate due t better treatment frm true incidence change. Thus, studies based upn vital statistics give nly an imprecise picture f trends in diabetes. The dramatic increase in mrtality frm diabetes in Taiwan between 1960 and 1988 as revealed in ur study, did, hwever, prvide sme clues abut the risk factrs fr the disease. It was unlikely that imprved case ascertainment r better diagnsis frm the imprvement in medical care facilities accunt fr all f the increases, When these trends did ccur, they ften started at a yunger age, and shifted t the elderly. Hwever, ur bservatins shw that the mrtality rates in the yung are stable, while thse in the elderly increase dramatically. The results f the prevalence survey f diabetes cnducted in Taipei City (the largest city in Taiwan) which als..,r... 0 I I I I I I 1971-73 74-76 77-79 80-82 83-85 86-88 Fig. 6. Perid effect (with 95 % cnfidence interval) in diabetes mellitus mrtality in Taiwan, 1971-1988, by sex. Male.--., Female males are presented in Figure 7. Generally, metrplitan residents had a 1.5-fld higher diabetes mrtality rate than thse in rural areas. The cmparisns f time trends fr diabetes mrtality rates by urbanizatin als revealed.~..' Table 2. Variable selectin in age-perid-chrt mdels f diabetes mellitus mrtality in Taiwan, 1971-1988 Mdel df Deviance Likelihd rati statistic (dr) ~ Male Age 85 1713.7 14764.6 (8) u Age, perid 80 104.9 1608.8 (5) b Age, perid, chrt 73 84.6 20.3 (7) b Female Age 85 2401.9 25047.8 (8) b Age, perid 80 238.7 2163.0 (5) b Age, perid, chrt 73 114.7 123.9 (7) u a Each likelihd rati statistic refers t the last term entered int the mdel and has an apprximate chi-square distributin with the apprpriate degree f freedm (dr). bp < 0.01
R. S. Lin and W. C. Lee: Diabetes mellitus mrtality in Taiwan 25 A 977 s 2 15."... >" ~ / 9 //;7 s".f.....:::2_ j /...'?'i..j,/.,cf f,.: //,-.../i," /...'" /:/.~ s f /... "_2... "." 9 9 jsz /~---i. m g 0 I I. I I / I I I I 1971-72 73-74 75-76 77-78 79-80 81-82 83-84 85-86 87-88 I I I I I L I I I 1971-72 73-74 75-76 77-78 79-80 81-82 83-84 85-86 87-88 Fig.7A, B. Secular trends f diabetes mellitus mrtality in Taiwan, 1971-1988 by urbanizatin and sex. A Male Metrplitan --, City..., Twn - - -, Rural.... B Female Metrplitan--, City..., Twn - - -, Rural... gm r % 160 140 120 100 80 60 40 //-... Calrie Y ] 3,100 ~- 13,000 /... 2,900 2,800 " 2,700 2,600 1 2,500 20... [... i' 2,400 \ / L./ L i l i i 1 I-g. i i I i i I i I I i I i ] i i i I I I i i 1960 1965 1970 1975 1980 1985 2,300 T O Fig. 8. Secular trends f per capita incme, and f calrie, prtein, and fat cnsumptin in Taiwan, 1960-1988. Fat (g) --, Prtein (g) ---, Calrie..., Incme.... Index year fr per capita incme is 1981 as 100 (US$ 2, 443) supprt ur cntentin, shwed that the age-adjusted prevalence rates fr thse aged 40 years r ver in Taipei City were 5.05 %, 7.10 %, 8.17 % fr 1970, 1979, and 1986, respectively [13]. As well as this, ur analysis, which by means f the age-perid-chrt mdel revealed that, in additin t age and perid effects, birth-chrt effect als played a significant rle in the trends, which might accunt fr part f this phenmenn because the elderly represent an earlier generatin when the risk f dying frm diabetes increased in successive birth-chrts, whereas the yung live in a generatin when rising trends have levelled ff. Alternatively, the elderly wh experienced r accumulated sufficient amunts f diabetic risk factrs in their lifetime, and received additinal triggers frm the current affluent envirnment, wuld be mst likely t develp the disease, and therefre, shw the mst cnspicuus rising mrtality trends9 Increasing fat cnsumptin and calrie intake as well as less physical activity assciated with urbanizatin might accunt fr the dramatic increase in diabetes mrtality in Taiwan which deserves further study. The significant birth-chrt effects fr diabetes fund in this study imply that sme imprtant determinants f the disease might ccur early in life, but nt shw their effects until sme time later. Fr example, vcatinal preferences, dietary habits and the nutritinal status f a birth chrt may be determined in early life. But we must als cnsider pssible generatin-related reprductive experiences r age-specific expsures such as war, epidemics r lifestyle changes, which might enhance r be de-
978 trimental t sme develpmental stages f diabetes. It is f interest t nte after the 1910-1917 birth chrt a cntinuusly declining risk in the recent birth chrts fr females. It raises a pssible suggestin that envirnmental factrs, regardless f what they were, had diminished the intensity in later birth chrts fr females but nt fr males. The declining parity f wmen is thught t be assciated with this birth chrt effect fr females as the general fertility rate in Taiwan has shwn a declining trend after 1952. The present findings can als be cmpared t ur previus study f pancreatic cancer [9] where mrtality was shwn t have increasing chrt effects up t the 1917-1926 birth-chrt and declined thereafter fr bth sexes. It is wrthwhile, t further explre what, besides the reprductive factrs, differentiates the patterns f chrt effects fr diabetes between males and females and thse between diabetes and pancreatic cancer in Taiwan. Excess risk fr diabetes in urban areas, repeatedly demnstrated in studies cnducted in almst all parts f the wrld [14-16], a previus study in Taiwan [17], and again in this study seem cnsistent with the hypthesis f an urbanizatin factr in the aetilgy f diabetes. Obesity [13, 18-20], physical activity [21-24], diet [25-27], psychscial stress due t mdernizatin [28, 29], and ther variables [30-32] have been prpsed as risk factrs fr diabetes which are assciated with urbanizatin. Hwever, ther pssible factrs in an urban envirnment perating independently f the abve remain t be explred. While diabetic incidence studies demnstrate n clear evidence that gender influences the risk fr the develpment f diabetes [31], high mrtality rates in females have been repeatedly shwn in mrtality statistics frm varius cuntries [33, 34]. This apparent paradx may be attributable t the higher fatality rate fr female diabetic patients [33]. Using prevalence data, which may be cnfunded by the duratin f disease, further cmplicates the aetilgical inference. Therefre, a cmprehensive mdel [35] with which t describe prevalence, mrtality, and incidence f diabetes in the general ppulatin is clearly warranted. Acknwledgements. The authrs are indebted t Ms. H-WLin and Mr. S-C Hung fr their excellent secretarial assistance. References 1. Natinal diabetes data grup (1983) Harris M, Hammn R (eds) Diabetes in America. NIH Pub, N. 85-1468. Gvernment Printing Office, Washingtn 2.Taiwan Prvincial Department f Health (1971-1988) Vital statistics. Chung-Hsin New Village, Prvincial Department f Health, Taiwan, ROC 3.Tzeng GH, Wu TY (1986) Characteristics f urbanizatinal levels in Taiwan districts. Gegraph Res 12:287-323 4. Cuncil fr ecnmic planning and develpment, ROC (1991) Taiwan statistical data bk. Taipei, Taiwan. ROC 5.Breslw NE, Day NE (1987) Statistical methds in cancer research, vl. II.Internatinal Agency fr Research n Cancer Scientific Publicatins, Lyn R. S. Lin and W. C. Lee: Diabetes mellitus mrtality in Taiwan 6.Waterhuse RJ, Muir C, Crrea R Pwell J (1976) Cancer incidence in five cntinents, vl III. Internatinal Agency fr Research n Cancer Scientific Publicatins, Lyn 7. Rbertsn C, Byle P (1986) Age, perid, and chrt mdels: the use f individual recrds. Stat Med 5:527-538 8.Statistics and epidemilgy research crpratin (1988) Egret user's manual, cpyright 1985-1988. Washingtn 9. Lee WC, Lin RS (1990) Age-perid-chrt analysis f pancreatic cancer mrtality in Taiwan, 1971-1986. Int J Epidemil 19: 839-847 10.Palumb PJ, Elveback LR, Chu CR Cnnlly DC, Kurland LT (1976) Diabetes mellitus: incidence, prevalence, survivrship, and causes f death in Rchester, Minnesta 1945-1970. Diabetes 25:566-569 11. Tkuhata GH, Miller W, Dign E, Hartnab T (1975) Diabetes mellitus: an underestimated public health prblem. J Chrn Dis 28:23-35 12.Fuller JH, Elfrd J, Gldblatt R Adelstein AM (1983) Diabetes mrtality: new light n an underestimated public health prblem. Diabetlgia 24:336-341 13.Tai TY, Yang CL, Chang CJ et ai. (1987) Epidemilgy f diabetes mellitus amng adults in Taiwan, ROC. J Med Ass Thailand 70 [SuppI 2]: 42-48 14.Zimmet P, Taylr R, Ram Pet al. (1983) Prevalence f diabetes and impaired glucse tlerance in the biracial (Melanesian and Indian) ppulatin f Fiji: a rural-urban cmparisn. Am J Epidemil 118:673-688 15. King H, Zimmet R Raper LR, Baldau B (1984) Risk factrs fr diabetes in three Pacific ppulatins. Am J Epidemil 119: 396-409 16. Papz L, Khalifa B, Eschwege E, Ayed HB (1988) Diabetes mellitus in Tunisia: descriptin in urban and rural ppulatins. Int J Epidemi] 17:419-422 17.Tai TY, Yang CL, Chang CJ et al. (1987) Epidemilgy f diabetes mellitus in Taiwan, ROC: cmparisn between urban and rural areas. J Med Ass Thailand 70 [Suppl 2]: 49-53 18. Knwler WC, Pettitt D J, Savage P J, Bennett PH (1981) Diabetes incidence in Pima Indians: cntributins f besity and parental diabetes. Am J Epidemi1113:144-156 19. Ohlsn LO, Larssn B, Svardsudd K et al. (1985) The influence f bdy fat distributin n the incidence f diabetes mellitus: 13.5 years f fllw up f the participants in the study f men brn in 1913. Diabetes 34:1055-1058 20.Mdan M, Karasik A, Halkin H et al. (1986) Effect f past and cncurrent bdy mass index n prevalence f glucse intlerance and type 2 (nn-insulin-dependent) diabetes and n insulin respnse: the Israel study f glucse intlerance, besity and hypertensin. Diabetlgia 29:82-89 21. Lipman RL, Raskin R Lve T et al. (1972) Glucse intlerance during decreased physical activity in man. Diabetes 21:101-107 22. Bgardus C, Ravussin E, Rbbins DC, Wlfe RR, Hrtn ES, Sims EAH (1984) Effects f physical training and diet therapy n carbhydrate metablism in patients with glucse intlerance and nn-insulin-dependent diabetes mellitus. Diabetes 33: 311-318 23. Taylr R, Ram R Zimmet R Raper LR, Ringrse H (1984) Physical activity and prevalence f diabetes in Melanesian and Indian men in Fiji. Diabetlgia 27:578-582 24. Frisch RE, Wyshak G, Albright TE, Albright NL, Schiff I (1986) Lwer prevalence f diabetes in female frmer cllege athletes cmpared with nnathletes. Diabetes 35:1101-1105 25.Trwell H (1975) Dietary-fiber hypthesis f the etilgy f diabetes mellitus. Diabetes 24:762-765 26. West KM (1978) Epidemilgy f diabetes and its vascular cmplicatins. Elsevier, New Yrk 27. Kawate R, Yamakid M, Nishimt Y. (1979) Diabetes mellitus and its cmplicatins in Japanese migrants n the island f Hawaii. Diabetes Care 2:161-170 28.Prte D Jr, Berthud HR, Jeanrenaud B, Wds SC (1981) Diabetes and the nervus system. Diabetlgia 20 [Suppl]: 305 313
R. S. Lin and W. C. Lee: Diabetes mellitus mrtality in Taiwan 29. Taylr R, Zimmet P (1983) Migrant studies in diabetes epidemilgy. In: Mann JI, Pyrala K, Teuscher A (eds): Diabetes in epidemilgical perspective. Churchill Livingstne 30. Zimmet P (1982) Type 2 (nn-insulin-dependent) diabetes -an epidemilgical verview. Diabetlgia 22:399411 31.Nelsn RG, Everhart JE, Knwler WC, Bennett PH (1988) Incidence, prevalence and risk factrs fr nn-insulin-dependent diabetes mellitus. Primary Care 15:227-250 32. Taylr R (1989) Aetilgy f nn-insulin dependent diabetes. Br Med Bull 45:73-91 33.Hamman RF (1983) Diabetes in affluent scieties. In: Mann JI, Pyrala K, Teuscher A (eds) Diabetes in epidemilgicat perspective. Churchill Livingstne 34. Entmacher PS, Krall LP, Kranczer SN (1985) Diabetes mrtality frm vital statistics. In: Marble A, Krall LP, Bradley RE Christlieb AR, Seldner JS (eds) Jslin's diabetes mellitus, 12th ed. Lea & Febiger, Philadelphia 979 35.Herman WH, Sinnck R Brenner E (1984) An epidemilgic mdel fr diabetes mellitus: incidence, prevalence, and mrtality. Diabetes Care 7:367-371 Received: 10 April 1992 and in revised frm: 10 June 1992 Dr.R.S.Lin Institute f Public Health Cllege f Medicine Natinal Taiwan University 1 Jen-Ai Rd. 1st Sec. Taipei Taiwan ROC