Migrant Asian Indians have higher prevalence ratesof diabetes, premature coronary artery disease andcardiovascular disease [CVD] mortality compared to thehost populations of these countries.1-3 Asian Indianswith diabetes also have higher mortality rates comparedto diabetic subjects of other ethnic groups.4 Accordingto recent reports, the prevalence rates of diabetes andcoronary artery disease in urban India are now similarto migrant Asian Indians.5,6 Indeed, India already hasthe largest number of people with diabetes in the worldand it is also predicted to lead in coronary artery diseasedeaths in the next fifteen years.7 A recent report publishedby the Indian Council of Medical Research (ICMR) in 2004, reported that in India, diabetes accounts for 1.09lakh deaths/ year.8 However, there are very fewpopulation-based studies on mortality rates in India andvirtually none comparing diabetic and non-diabeticsubjects. In this paper, we report on mortality data amongdiabetic and non-diabetic subjects in a population-basedstudy in urban south India.

*Rapid Publication Dr. Mohan's Diabetes Specialities Centre and Madras DiabetesResearch Foundation, Gopalapuram, Chennai, India.#Department of Epidemiology, The Tamil Nadu Dr MGRMedical University, Chennai.Received : 14.9.2005; Revised : 18.12.2005;Accepted : 12.1.2006

SUBJECTS AND METHODS

The Chennai Urban Population Study [CUPS], anongoing epidemiological study in two residentialcolonies in Chennai [formerly Madras, the largest cityin southern India] was started in 1996 and the baselinestudy was completed in 1997. The methodologicaldetails and several reports from CUPS have beenpublished.9,10 Briefly, all individuals = 20 years of agewere invited to participate in a screening programmefor diabetes. The overall response rate was 90.2%.

Anthropometric measures [weight, height, waist andhip] and blood pressure were measured using standardmethods. A resting 12-lead electrocardiogram [ECG] wascarried out on 1175 subjects [response rate - 84%].

A fasting blood sample was taken for estimation ofglucose and lipids following which an oral glucose load[75 gm] was administered to all individuals [excludingknown diabetic subjects]. All biochemical assays weredone using a Corning Express Plus Auto Analyser[Corning, Medfield, MA, USA] using kits supplied byBoehringer Mannheim [Mannheim, Germany].

Individuals were categorized based on a validatedphysical activity [PA] questionnaire as having light,moderate and heavy activity.10

From 1997, annual visits were made to everyhousehold and details of all deaths were recorded. Thelast visit was conducted between October 2003 andAugust 2004.

Causes of death: Unfortunately autopsies are usuallynot performed in India except in some medico-legalcases. We had to therefore make use of all availablesources of information to determine possible cause ofdeath. These included interview of family members andwherever available, scrutiny of hospital records or thedeceased’s medical files, police records [where relevant]and death certificates issued by the Corporation ofChennai, the official body for registration of all deathsin Chennai.

Causes of death were coded as: cardiovascular or CVD[heart failure, myocardial infarction or stroke], renal,gastrointestinal [gastrointestinal bleeding, cirrhosis orhepatitis], respiratory [tuberculosis or chronicobstructive pulmonary disease], cancer, lifestyle-related[eg alcoholism], unnatural causes [suicides, accident],others [deaths due to infection or aging] and the rest asunknown causes.

Definitions

Type 2 diabetes was diagnosed based on WHOconsulting group criteria. Hypertension was diagnosedusing the JNC-VI criteria. Coronary artery disease [CAD]was diagnosed based on a documented past history ofmyocardial infarction or drug treatment for coronaryartery disease and/or Minnesota codes 1-1-1 to 1-1-7,[Q wave changes] 4-1 to 4-2 [ST segment depression] or5-1 to 5-3 [T wave abnormalities].6

Individuals were classified as non-smokers andcurrent smokers [habitual smokers regardless of quantitysmoked]. Alcohol intake was categorized as none, social[occasional drinking] and regular [everyday intakeregardless of the quantity consumed].

STATISTICAL ANALYSIS

Mortality rates were calculated as number of deathsper 1000 person-years of follow-up. Statistical analysis was carried out using the SPSS PC Windows version10.0 [Chicago, IL]. Students “t” test was used to comparemeans and Chi Square or Fischer’s exact test to compareproportions. Cox regression analysis was used todetermine the effect of various risk factors on mortality.Survival curves were plotted by life-table analysis andthose of subjects with and without diabetes werecompared using Wilcoxon [Gehan] test.

RESULTS

Of the 1262 individuals who initially participated inCUPS, 90 shifted their residence to another colony withinChennai, 79 moved out of the state of Tamil Nadu and23 left the country. We were able to obtain follow-updata on all those who continued to reside in the samecolonies and 70 of the 90 who shifted their residence toanother colony within Chennai i.e. on a total of 1140individuals [response rate: 1140 / 1262 – 90.3%] [Fig. 1].The response rates among diabetic and non-diabeticsubjects were 94.1% and 89.8% respectively.

There were no significant differences in the baselinecharacteristics of the responders [n = 1140] and non-responders [n = 122] with respect to age: 43 ± 16 vs. 42 ±16 years; gender: males: 44.5% vs. 41.0%; prevalence ofdiabetes: 14.5% vs. 12.3% or prevalence of coronaryartery disease: 11.1% vs. 9.5%. However, the prevalenceof hypertension was higher among the responderscompared to non-responders [23.1% vs. 13.1%, p=0.012].

During the median follow up period of six-years, fiftydeaths [50/1140, 4.4%] were observed yielding anoverall all-cause mortality rate of 7.0 per 1,000 person-years.

The percentage of deaths was significantly higheramong diabetic subjects [17/143, 11.9%] compared tonon-diabetic subjects [33/997, 3.3%, p <0.001] [Fig. 1].This translates to an all-cause mortality of 18.9 per 1000person years among diabetic subjects compared to 5.3per 1000 person-years among non-diabetic subjects[Table 1].

Mortality due to CVD [52.9% vs. 24.2%,p=0.042] andrenal causes [23.5% vs. 6.1%,p=0.072] were higher among diabetic, compared to non-diabetic subjects.Deaths due to gastrointestinal [12.1%], respiratory [9.1%],lifestyle related [6.1%] and unnatural causes [18.2%]were observed only among non-diabetic subjects[Table 1].

Fig. 1 : Flow-chart of the study protocol.

Fig. 2 : Age wise distribution of deceased in both diabetic and non-diabetic subjects.

The mean age of death among diabetic subjects was66 years compared to 54 years among non-diabeticsubjects [Table 2]. The mean age of subjects who died ofCVD causes was 65 ±11 years and 24% were less than50 years old.

Among non-diabetic subjects who died, the proportionof smokers was significantly higher [p<0.001]. Waistcircumference [p=0.004] and triglyceride levels [p<0.05]were lower among the non-survivors compared to thesurvivors. Among diabetic non-survivors, proportion ofhypertensives were significantly higher [p<0.05] andthey also had significantly higher systolic bloodpressure values [p=0.035]. However body mass index[p=0.036], waist circumference [p=0.026] andtriglycerides were lower compared to the survivors.None of the other parameters showed any significant difference.

Mortality increased with increase in age intervals [age= 50 years: 1.8%, 51-60 years: 5.0%, >60 years: 15.6%,trend chi-square: 62.8, p<0.001]. When diabetic and non-diabetic subjects were compared, at any given ageinterval, diabetic subjects had increased death ratescompared to the non-diabetic subjects but the differencesdid not reach statistical significance [diabetic vs non-diabetic, age = 50 years : 2.0% vs. 1.7%, p=0.874, 51-60years: 7.9% vs. 4.1%,p=0.342 >60 years: 23.2% vs.12.1%,p=0.057] [Fig. 2].

Fig. 3 : Survival curves for diabetic and non-diabetic subjects

Subjects who performed light grade activity hadhigher death rates compared to subjects with heavy gradeactivity [6.5% vs. 2.7%, p=0.008]. The hazards ratio[HR] for mortality among those with light grade activitywas 2.39 [95% confidence intervals [CI]: 1.05-5.40,p=0.037] compared to heavy grade activity. However,the significance disappeared when adjusted for age [HR:1.41, 95% CI: 0. 61-3.27, p=0.418].

The number of pack-years of smoking wassignificantly higher among the non-survivors [33.3 ±14.5]compared to survivors [14.8 ± 12.8, p<0.001]. All-causemortality among smokers was 7.2% compared to 4.0%among non-smokers. The association of smoking withall-cause mortality remained significant even after ageadjustment [HR: 2.84, 95% CI: 1.44 – 5.62, p=0.003].

The survival curves were significantly different indiabetic, compared to non-diabetic, subjects [Wilcoxon[Gehan] test: p<0.001] [Fig. 3].

Hazard ratio for all-cause mortality was higher forsubjects with diabetes [HR: 3.6, 95% CI: 2.02-6.53,p<0.001] compared to non-diabetic subjects and thisremained significant even after adjusting for age [HR:1.9, 95% CI: 1.04-3.45, p=0.038]. The HR for diabetes formortality due to CVD was 7.8 [95% CI: 3.0 –20.2, p<0.001,age adjusted: HR: 3.5, 95% CI: 1.3 – 9.1, p=0.012]. Diabeticsubjects also had very high risk for renal mortality[unadjusted: HR: 14.3, 95% CI: 2.6 – 77.9, p<0.001.adjusted for age: HR: 10.7, 95% CI: 1.7 – 67.4, p=0.012].

DISCUSSION

This is the first population-based study of mortalityrates in diabetic and non-diabetic subjects from Indiaand presents three major findings: first, mortality ratesin diabetic subjects are twice as high as in non-diabeticsubjects; second, cardiovascular and renal diseasecontributed to majority of deaths among diabetic subjectswhile lifestyle related, gastrointestinal and unnaturaldeaths were more common among non-diabetic subjects contributing to the lower age of death in the latter; finally,physical activity and smoking showed an associationwith all-cause mortality, but for the former, thesignificance disappeared when adjusted for age.

Earlier studies from India on mortality do not provideinformation on death rates in the general population.11,12In this study, we report that the overall mortality rate inthe study population is 7.0 per 1000 person-years. Themean age at death was 58 years, and 24% of CVDmortality was seen among subjects less than 50 years ofage confirming earlier reports of premature deaths dueto CVD in Indians.2,3 The lower age at death of non-diabetic subjects is perhaps fortuitous, probably due tothe small numbers and also due to gastrointestinal, life-style related and unnatural causes. This scenario is likelyto change with rapidly increasing prevalence of diabetesand control of communicable diseases in India.

Indeed, subjects with diabetes had two-fold higherrisk for mortality compared to non-diabetic subjects, evenafter age adjustment and this corroborates earlierstudies.13-16 Survival rates were markedly lower amongdiabetic subjects and this was mostly due to CVD. Thecauses for increased cardiovascular deaths amongdiabetic subjects are multiple and include elevatedglucose and blood pressure, dyslipidemia andinflammation. Multiple interventions are thereforeneeded to reduce cardiovascular risk in diabeticsubjects.17

The second commonest cause of death in diabeticsubjects was renal disease accounting for 23.5% of thedeaths, which is similar to figures reported in otherstudies.18 However, in contrast to earlier studies fromIndia,11,12,19 mortality due to infections was found to beless common. This might be due to the fact that earlierstudies were hospital-based and referral bias couldexplain excess of deaths due to infections or may be dueto misclassification bias.

In the Aerobics Center Longitudinal Study (ACLS)conducted among middle-aged men with type 2diabetes, physical inactivity was an independentpredictor of all-cause mortality.20 The National HealthInterview Survey (NHIS), which examined therelationship of walking and other physical activities toall cause mortality among US adults with diabetesreported that higher levels of physical activity wereassociated with a lower incidence of all-cause mortality.21Our finding demonstrates that physical inactivity wasassociated with all-cause mortality. However, ageappears to be an effect modifier as it abolished therelation of physical inactivity with mortality.Surprisingly, the non-survivors had lower body massindex. This may reflect inadequate control of diabetes,resulting in weight loss. The lower triglyceride levels inthe non-survivors can be explained by the use of lipidlowering drugs, as 16.7% of non-diabetic non-survivorsand 43.8% among diabetic non-survivors had coronary

artery disease at baseline and hence were on treatmentwith lipid lowering drugs. Smoking was associated withall-cause mortality and this is similar to results in otherpopulations.22

There are several limitations to this study. Theseinclude availability of soft evidence of cause of deathdue to improper filling of death certificates andinaccuracy of verbal accounts given by relatives of thedeceased. Indeed, the cause of death was unknown insome cases due to non-availability of medical records.Moreover the cause of death recorded in death certificatesmay be unreliable as diabetes is rarely recorded as acause of death.23 Finally, the number of deaths is quitesmall as the original cohort itself consisted of only 1262subjects. However, main strength of the study is that thedata collected is from the community. The other strengthsare that the response rate is very good, the data collectedhas helped to reasonably identify the cause of death inthe population and this is the first paper from India toreport on mortality due to diabetes. The results of thestudy can probably be extrapolated to most of urbanIndia, as the prevalence rates of diabetes and CVD arefairly similar across most cities in India.20 However inrural India, the prevalence of both diabetes and CVD isless than half of that in urban areas and hence mortalitydata is also likely to be quite different from that seen inurban areas.

In conclusion, this study highlights that diabeticsubjects had higher mortality rates compared to non-diabetic subjects. Cardiac disease and renal diseaseswere the commonest cause for mortality in diabeticsubjects. Smoking and physical inactivity are associatedwith all cause mortality.

REFERENCES

1.McKeigue PM, Shah B, Marmot MG. Relation of centralobesity and insulin resistance with high diabetes prevalenceand cardiovascular risk in South Asians. Lancet 1991;337:382-6.

2.Bhatnagar D, Anand IS, Durrington PN, et al. Coronary riskfactors in people from the Indian subcontinent living in westLondon and their siblings in India. Lancet 1995;345:405-9.

3.Heng DM, Lee J, Chew SK, Tan BY, Hughes K, Chia KS.Incidence of ischaemic heart disease and stroke in Chinese,Malays and Indians in Singapore: Singapore CardiovascularCohort Study. Ann Acad Med Singapore 2000;29:231-6.

4.Ma S, Cutter J, Tan CE, Chew SK, Tai ES. Associations ofdiabetes mellitus and ethnicity with mortality in a multiethnicAsian population: data from the 1992 Singapore NationalHealth Survey. Am J Epidemiol 2003; 158:543–52.

5.Mohan V, Shanthirani CS, Deepa R. Glucose intolerance(diabetes and IGT) in a selected south Indian populationwith special reference to family history, obesity and life stylefactors – The Chennai Urban Population Study (CUPS 14).J Assoc Physicians India 2003;51;771-7.

6.Mohan V, Deepa R, Rani SS, Premalatha G. Prevalence ofcoronary artery disease and its relationship to lipids in aselected population in south India. The Chennai Urbanpopulation Study (CUPS No. 5). J Am Coll Cardiol 2001;38;682-7.

7.Wild S, Roglic G, Green A, Sicree R, King H. Global prevalenceof diabetes, estimates for the year 2000 and projections for2030. Diabetes Care 2004;27:1047–53.

8.Assessment of burden of non-communicable diseases. Finalreport of project WR/SE IND RPC 001 RB 02 – SE / 02 /419575. Indian Council of Medical Research, New Delhi. 2004,pp 8-17 and 23-28.

9.Shanthi Rani CS, Rema M, Deepa R, et al. The Chennai UrbanPopulation Study (CUPS)–Methodological details (CUPSPaperNo.1). Int J Diab Dev Countries 1999;19:149-57.

10.Mohan V, Gokulakrishnan K, Deepa R, Shanthirani CS, DattaM. Association of physical inactivity with components ofmetabolic syndrome and coronary artery disease—theChennai Urban Population Study (CUPS no. 15). Diabet Med2005;22:1206-11.

11.Sridhar CB, Ahuja MM. Pattern of mortality amongstdiabetics in North India. Indian J Pathol Bacteriol 1966;9:222-7.

12.Zargar AH, Wani AI, Masoodi SR, Laway BA, Bashir MI.Mortality in diabetes mellitus—data from a developing regionof the world. Diabetes Res Clin Pract 1999;43:67-74.

13.Morgan CL, Currie CJ, Peters JR. Relationship betweendiabetes and mortality: a population study using recordlinkage. Diabetes Care 2000;23:1103-7.

14.Haffner SM, Lehto S, Ronnemaa T, Pyorala K, Laakso M.Mortality from coronary heart disease in subjects with type2 diabetes and in nondiabetic subjects with and withoutprior myocardial infarction. N Engl J Med 1998;330:229-34.

15.Sicree RA, Ram P, Zimmet P, Cabealawa S, King H. Mortalityand health service utilization amongst Melanesian and Indiandiabetics in Fiji. Diabetes Res Clin Pract 1985;1:227-34.

16.Gu K, Cowie CC, Harris MI. Mortality in adults with andwithout diabetes in a national cohort of the U.S. population,1971-1993. Diabetes Care 1998;21:1138-45.

17.Gaede P, Vedel P, Larsen N, Jensen GV, Parving HH, PedersenO. Multifactorial intervention and cardiovascular disease inpatients with type 2 diabetes. N Engl J Med 2003;348:383–93.

18.Simmons D, Schaumkel J, Cecil A, Scott DJ, Kenealy T. Highimpact of nephropathy on five-year mortality rates amongpatients with Type 2 diabetes mellitus from a multi-ethnicpopulation in New Zealand. Diabet Med 1999;6:926-31.

19.Das S, Mishra RK, Jena BB, Mishra BK, Misra KC, Sarangi B.Mortality events amongst non insulin dependent diabetesmellitus patients in Orissa. J Assoc Physicians India1991;39:519–20.

20.Wei M, Gibbons LW, Kampert JB, Nichaman MZ, Blair SN.Low cardiorespiratory fitness and physical inactivity aspredictors of mortality in men with type 2 diabetes. AnnIntern Med 2000;132:605-11.

21.Gregg EW, Gerzoff RB, Caspersen CJ, Williamson DF, NarayanKM. Relationship of walking to mortality among US adultswith diabetes. Arch Intern Med. 2003;163:1440–7.

22.LaCroix AZ, Lang J, Scherr P, et al. Smoking and mortalityamong older men and women in three communities N Engl JMed 1991;324:1619-25.

23.Andresen EM, Lee JA, Pecoraro RE, Koepsell TD, HallstromAP, Siscovick DS. Under reporting of diabetes on deathcertificates, King County, Washington, Am J Public Health1993;83:1021–4.