International Journal of Epidemiology © International Epidemiological Association 1992
Vol. 21, No. 6 Printed in Greet Britain
Diabetes Diagnosed before the Age of 2 Years: Mortality in a British Cohort 8-17 Years after Onset J L BOTHA,' H PARKER," N T RAYMOND* AND P G F SWIFT*
Childhood diabetes diagnosed before the age of 24 months, while forming no more than 6%' (also M A Metcalfe—personal communication) of newly diagnosed childhood diabetes, presents very specific management problems to clinicians and parents alike. Some of the unique features which have been described in small groups of children are male predominance,2"4 a high prevalence of diabetes in the family,3 many episodes of severe hypoglycaemia,35 few episodes of ketoacidosis,3'5 lesser incidence of vascular complications before 30-34 years' duration than childhood diabetes of later onset,' possible longterm cognitive difficulties,6"10 and suggestions of early mortality.1'3 We decided these features merited investigation and the existence of the BDA Children's Register presented an opportunity to establish a large cohort. In November 1972 the BDA began a register to which notification was made of all children under 16 in the UK and Ireland with newly diagnosed diabetes mellitus. An epidemiological description of the first •Department of Epidemiology and Public Health, University of Leicester, Clinical Sciences Building, Leicester Royal Infirmary, PO Bon 65, Leicester LE2 7LX, UK. tDepartment of Paediatrics, Leicester Royal Infirmary, Leicester, UK.
1132
2 years' information on the register was published in 19752 and serum samples provided by the children have been used to investigate immunological and virological hypotheses (D R Gamble—personal communication). Further epidemiological work has not been undertaken and the register was discontinued in 1986. It seemed feasible to trace those registered children who had been diagnosed before age 24 months, either through the notifying paediatricians or via the NHSCR tracing process. In this paper we report on the process of establishing a cohort for investigation and on mortality since diagnosis. METHODS Establishing the Cohort Recruitment to the cohort from the BDA Register was restricted to notifications received during the period 1972-1981, the only period for which computerized records were available. The BDA supplied us with 427 records of children notified to the Register during that period as having developed diabetes before the age of 24 months. The information supplied was the child's surname, forenames, address (at time of notification), date of birth, sex, date of diagnosis, date of notification
Downloaded from http://ije.oxfordjournals.org/ at University of Strathclyde on April 9, 2015
Botha J L (Department of Epidemiology and Public Health, University of Leicester, Clinical Sciences Building, Leicester Royal Infirmary, PO Box 65, Leicester LE2 7LX, UK), Parker H, Raymond N T and Swift P G F. Diabetes diagnosed before the age of 2 years: mortality in a British cohort 8-17 years after onset. International Journal of Epidemiology 1992; 21: 1132-1137. Childhood diabetes diagnosed before the age of 24 months presents specific management problems. We report here on the establishment (using the British Diabetic Association [BDA] Children's Register) and mortality of a cohort of children with diabetes diagnosed before age 24 months. Children registered during the period 1972-1981 were traced by contacting consultants or by using the National Health Service Central Registers (NHSCR) of the Office of Population Censuses and Surveys (OPCS). Standardized mortality ratios (SMR) were estimated using person-years of followup for each child and age-specific death rates for the England and Wales population for the years 1972-1989. Of 339 children notified during 1972-1981, 231 were traced through consultants and 99 of the remaining 108 through the NHSCR. Twenty were found to be ineligible. The cohort available for mortality analysis comprised 310 (97%) of 319 eligible children. Their age at the time the cohort was established was 8-18 years, and their duration of diabetes 8-17 years. The male:female ratio is 1.4:1. Of 310 children studied, seven have already died: SMR 5.4 (95% Cl : 2.5-11.5). We have established a large, unique cohort of children with diabetes diagnosed before age 24 months and still living in the UK and Ireland. The natural history including mortality and occurrence of complications will be analysed prospectively in this cohort and compared to other cohorts of similar disease duration, but later age at onset.
1133
MORTALITY IN A BRITISH DIABETES COHORT
Mortality Causes of death for children who had died were obtained from consultants and/or from the OPCS. We also obtained extracts from medical notes and post-mortem reports (where applicable). Using information on the vital status of all children from the BDA Register for whom it was available we calculated the SMRs for the sexes separately and together. PersonTABLE 1
years of follow-up were calculated for each child from diagnosis until the date the cohort was established (taken as 31 December 1989), or death, or last date known to be alive (from NHSCR). Age-specific death rates for the England and Wales population for the years 1972-1989 were used to calculate the expected number of deaths. RESULTS Establishing the Cohort (Table 1) We obtained complete information on 231 children as a result of our initial contact with consultants. Eight were ineligible (too old at diagnosis or notified outside the stipulated period), leaving 223 who were eligible, including five who had died. At the NHSCR 99 (92%) of the remaining 108 children were traced. Twelve were found to be ineligible, 77 were alive, two had died, one had emigrated, and for the other seven we have had no response from their FHSA/AHB or GPs. Nine children cannot be traced through the NHSCR. As a result we have identified 310 children from the BDA Register (97% of 319 eligible), comprising 302 for whom we knew their vital status at cohort establishment date, and eight for whom we knew the dates when they were last known to be alive. Cohort Characteristics Characteristics of the cohort at the time it was established are shown in Figures 1-3. Eighty-one per cent of children were diagnosed in the second year of their lives (Figure 1). There appeared to be more males than females compared with the first year of life,
Tracing procedures and results
Results of tracing New names
Names from BDA Regiiter
Sent for tracing to
Number of names
Not valid 88
Not found
Being traced
Emigrated
Died
Alive
Alive
37
Authors
427
Consultants
339
8
218
OPCS + FHSA/GPi
108
12
77
108
295"
AD
'Eight eligible for mortality analysis, last date alive known. 302 eligible for mortality analysis, vital status known.
37
Downloaded from http://ije.oxfordjournals.org/ at University of Strathclyde on April 9, 2015
and the notifying doctor's name and address. After replicated records had been excluded a potential cohort of 339 children remained. The tracing process is summarized in Table 1. From an initial feasibility study we had established that 131 of the original notifying consultants were still available for contact, and we had identified potential alternatives for the other 52. These consultants were contacted by letter and asked to provide details of each child's vital status, address, current consultant and family doctor. After one reminder letter 91.3% responded. Several of the consultants contacted indicated that they were no longer caring for the children, and suggested the names of 33 other consultants who were also contacted (26 replied, 79%). Details of 108 children left untraced after the consultant contact were sent to the NHSCR for tracing to Family Health Services Authorities (FHSAs) or Area Health Boards (AHBs) with whom children still alive were, or were last known to have been registered. Each FHSA/AHB was contacted by us, enclosing a letter to be forwarded to each child's GP, asking for details of the child's address, current consultant and GP (if sent to the incorrect one).
1134
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
although this was not statistically significant. The male: female ratio of the total cohort is 1.4:1. The median age for this cohort at time of survey was 14 years, with a range from 8 to 18 years (Figure 2). Duration of diabetes ranged from 8 to 17 years, with a median of 12 years (Figure 3). 100
30
Number of children
25
20
Number of children 15r
10
80
60
11
12
13
14
15
16
17
Duration (years) I
Sex
20
I Female 0-5
6-11
12-17
L. ; Male
FIGURE 3 Duration of DM at 31 December, 1989. Cohort with DM onset before age 2 years
18-23
Age (months) Sex I Female
5
s35
? Male
FIGURE 1 Age at diagnosis. Cohort with DM onset before age 2 years
25
Mortality The SMR for females (9.9, 95% CI : 4.1-24.2) and those for the overall cohort (5.4, 95% CI : 2.5-11.5) were significantly elevated (Table 2). TABLE 2 Standardized mortality ratios
Number of children
No.
Person- Observed Expected Approximate deaths SMR 95% CI yean deaths
Females
131
1602.77
5
0.507
9.9
4.1-24.2
Males
179
2242.46
2
0.877
2.3
0.6-9.5
3845.23
7
1.296
5.4
2.5-11.5
20
15 I
Both
310
10 I
8
I
9
10
11
12
13
14
15
16
17
18
Age (years) Sex
L
I Female
[ 2 1 Male
FIGURE 2 Age at 31 December, 1989. Cohort with DM onset before age 2 years
Five of the seven children who died were girls and two were boys (Table 3). The cause of death in three children, all of pre-school age, did not seem to be directly related to diabetes: two died of infections (viral pneumonia and acute tracheobronchitis [the child also had Down's syndrome]) and the third, who also had hypothyroidism, developed severe hepatitis (of uncertain aetiology) and died of liver failure a few months later. Of the remaining deaths, three could be directly attributable to the metabolic disturbances of diabetes. One was an infant aged 12 months with a short history
Downloaded from http://ije.oxfordjournals.org/ at University of Strathclyde on April 9, 2015
•iJ 10
MORTALITY IN A BRITISH DIABETES COHORT TABLE 3 Details of deaths of seven children Death
1 2 3 4 5 6
Diabetes duration
Age
Female Female Female Female Female Male
2 years 1 year 3 years 12 years 13 years 4 years
2 years 11 years 12 years 3 years
3 years
2 years
7 Male
< 1 year
Cause
Viral pneumonia Diabetic ketoaddosis b Liver failure, (hypothyroidism) Convulsions'5, (Down's) Suicide5 Hypoglycaemic convulsion, cerebral oedema, encephalopathyb Acute tracheobronchitis, (Down's)
* Died at time of diagnosis b Diabetes related death
of illness whose diabetic ketoacidosis was not recognized for 10 hours after hospital admission. She vomited, suffered cardiac arrest and died soon after insulin therapy was commenced. Another child, aged 4 years, developed convulsions related to poor diabetic control. He suffered a prolonged severe hypoglycaemic convulsion and died one month later from cerebral oedema and hypoxic encephalopathy, having never regained consciousness. The third child (a 12 year old with Down's syndrome) developed convulsions (possibly hypoglycaemic) and she died soon afterwards in her sleep. The seventh child's tragic suicide at 13 years was almost certainly indirectly related to the diabetes. She had previously attempted serious insulin overdoses and was said to have shown signs of depression before taking a fatal overdose of antimigraine tablets. DISCUSSION A cohort of British children with diabetes diagnosed at the unusually young age of under 2 years has been compiled from the BDA Children's Diabetic Register. This is the largest cohort of which we are aware. The successful compilation of this cohort is mainly a result of the enthusiastic response of the consultants contacted, many of whom were still caring for the children at the time. Tracing through the NHSCR at Southport has been successful (92% of 108 names submitted), although it took 12 weeks to receive replies on 75 °/o of names submitted. This delay may be reduced in future when computerized searches will be possible. It took 3
months to obtain approval from OPCS for the children to be traced, a period unlikely to be affected by their computerization. We decided not to trace all children through Southport, because we wanted to establish contact with consultants caring for the children as early as possible, and also to limit costs. Deaths since diagnosis were analysed in 310 children. Their SMR was approximately five times higher than in the general population. Given the small numbers of deaths in this cohort and the wide CI, our mortality results are similar to those reported in other studies comparing mortality in childhood diabetes to that in the general population."'12 Dorman et al.u reported SMR of 5.4 in males and 11.53 in females. Children in their USA cohort were diagnosed aged 0-17 years in the period 1950-1981 and followed for 0-30 years. Sartor et al. n reported an SMR of 2.09 (95% CI: 1.14-3.83) in a Swedish cohort of both sexes aged 0-14 at diagnosis between 1977 and 1985. Followup was relatively short (0-8.5 years), which may explain, together with reportedly good health care, the few deaths (10) which occurred in their large cohort. Five of these deaths were directly attributable to diabetes mellitus, two occurring before the age of 2 years. In a recent report on childhood diabetes13 39 deaths had occurred in a cohort of 777 children aged 3 months to 16 years at diagnosis in 1930-1985 and followed for 3-63 years, but no SMR was reported. Other reports on mortality14"17 compared early onset with later onset diabetes and they all describe increased mortality for onset before age 15. Only BorchJohnson's analysis17 compared subgroups within childhood diabetes, and they reported no differences in mortality among those subgroups diagnosed before 15 years of age. Five of the seven deaths in our cohort occurred within 3 years of diagnosis and the other two more than 10 years after diagnosis. This distribution could be due to chance, given the small numbers, but it could be that some of the 16 children not (yet) traced have already died. This theory is supported by a recent report showing that mortality among members of an occupational cohort untraced on the NHSCR was substantially greater than mortality among those found on the Register.18 Causes of death in people with diabetes are not well documented, because diabetes mellitus is not always recorded on the death certificate either as an underlying or contributory cause; for instance it was recorded on only 70
Vol. 21, No. 6 Printed in Greet Britain
Diabetes Diagnosed before the Age of 2 Years: Mortality in a British Cohort 8-17 Years after Onset J L BOTHA,' H PARKER," N T RAYMOND* AND P G F SWIFT*
Childhood diabetes diagnosed before the age of 24 months, while forming no more than 6%' (also M A Metcalfe—personal communication) of newly diagnosed childhood diabetes, presents very specific management problems to clinicians and parents alike. Some of the unique features which have been described in small groups of children are male predominance,2"4 a high prevalence of diabetes in the family,3 many episodes of severe hypoglycaemia,35 few episodes of ketoacidosis,3'5 lesser incidence of vascular complications before 30-34 years' duration than childhood diabetes of later onset,' possible longterm cognitive difficulties,6"10 and suggestions of early mortality.1'3 We decided these features merited investigation and the existence of the BDA Children's Register presented an opportunity to establish a large cohort. In November 1972 the BDA began a register to which notification was made of all children under 16 in the UK and Ireland with newly diagnosed diabetes mellitus. An epidemiological description of the first •Department of Epidemiology and Public Health, University of Leicester, Clinical Sciences Building, Leicester Royal Infirmary, PO Bon 65, Leicester LE2 7LX, UK. tDepartment of Paediatrics, Leicester Royal Infirmary, Leicester, UK.
1132
2 years' information on the register was published in 19752 and serum samples provided by the children have been used to investigate immunological and virological hypotheses (D R Gamble—personal communication). Further epidemiological work has not been undertaken and the register was discontinued in 1986. It seemed feasible to trace those registered children who had been diagnosed before age 24 months, either through the notifying paediatricians or via the NHSCR tracing process. In this paper we report on the process of establishing a cohort for investigation and on mortality since diagnosis. METHODS Establishing the Cohort Recruitment to the cohort from the BDA Register was restricted to notifications received during the period 1972-1981, the only period for which computerized records were available. The BDA supplied us with 427 records of children notified to the Register during that period as having developed diabetes before the age of 24 months. The information supplied was the child's surname, forenames, address (at time of notification), date of birth, sex, date of diagnosis, date of notification
Downloaded from http://ije.oxfordjournals.org/ at University of Strathclyde on April 9, 2015
Botha J L (Department of Epidemiology and Public Health, University of Leicester, Clinical Sciences Building, Leicester Royal Infirmary, PO Box 65, Leicester LE2 7LX, UK), Parker H, Raymond N T and Swift P G F. Diabetes diagnosed before the age of 2 years: mortality in a British cohort 8-17 years after onset. International Journal of Epidemiology 1992; 21: 1132-1137. Childhood diabetes diagnosed before the age of 24 months presents specific management problems. We report here on the establishment (using the British Diabetic Association [BDA] Children's Register) and mortality of a cohort of children with diabetes diagnosed before age 24 months. Children registered during the period 1972-1981 were traced by contacting consultants or by using the National Health Service Central Registers (NHSCR) of the Office of Population Censuses and Surveys (OPCS). Standardized mortality ratios (SMR) were estimated using person-years of followup for each child and age-specific death rates for the England and Wales population for the years 1972-1989. Of 339 children notified during 1972-1981, 231 were traced through consultants and 99 of the remaining 108 through the NHSCR. Twenty were found to be ineligible. The cohort available for mortality analysis comprised 310 (97%) of 319 eligible children. Their age at the time the cohort was established was 8-18 years, and their duration of diabetes 8-17 years. The male:female ratio is 1.4:1. Of 310 children studied, seven have already died: SMR 5.4 (95% Cl : 2.5-11.5). We have established a large, unique cohort of children with diabetes diagnosed before age 24 months and still living in the UK and Ireland. The natural history including mortality and occurrence of complications will be analysed prospectively in this cohort and compared to other cohorts of similar disease duration, but later age at onset.
1133
MORTALITY IN A BRITISH DIABETES COHORT
Mortality Causes of death for children who had died were obtained from consultants and/or from the OPCS. We also obtained extracts from medical notes and post-mortem reports (where applicable). Using information on the vital status of all children from the BDA Register for whom it was available we calculated the SMRs for the sexes separately and together. PersonTABLE 1
years of follow-up were calculated for each child from diagnosis until the date the cohort was established (taken as 31 December 1989), or death, or last date known to be alive (from NHSCR). Age-specific death rates for the England and Wales population for the years 1972-1989 were used to calculate the expected number of deaths. RESULTS Establishing the Cohort (Table 1) We obtained complete information on 231 children as a result of our initial contact with consultants. Eight were ineligible (too old at diagnosis or notified outside the stipulated period), leaving 223 who were eligible, including five who had died. At the NHSCR 99 (92%) of the remaining 108 children were traced. Twelve were found to be ineligible, 77 were alive, two had died, one had emigrated, and for the other seven we have had no response from their FHSA/AHB or GPs. Nine children cannot be traced through the NHSCR. As a result we have identified 310 children from the BDA Register (97% of 319 eligible), comprising 302 for whom we knew their vital status at cohort establishment date, and eight for whom we knew the dates when they were last known to be alive. Cohort Characteristics Characteristics of the cohort at the time it was established are shown in Figures 1-3. Eighty-one per cent of children were diagnosed in the second year of their lives (Figure 1). There appeared to be more males than females compared with the first year of life,
Tracing procedures and results
Results of tracing New names
Names from BDA Regiiter
Sent for tracing to
Number of names
Not valid 88
Not found
Being traced
Emigrated
Died
Alive
Alive
37
Authors
427
Consultants
339
8
218
OPCS + FHSA/GPi
108
12
77
108
295"
AD
'Eight eligible for mortality analysis, last date alive known. 302 eligible for mortality analysis, vital status known.
37
Downloaded from http://ije.oxfordjournals.org/ at University of Strathclyde on April 9, 2015
and the notifying doctor's name and address. After replicated records had been excluded a potential cohort of 339 children remained. The tracing process is summarized in Table 1. From an initial feasibility study we had established that 131 of the original notifying consultants were still available for contact, and we had identified potential alternatives for the other 52. These consultants were contacted by letter and asked to provide details of each child's vital status, address, current consultant and family doctor. After one reminder letter 91.3% responded. Several of the consultants contacted indicated that they were no longer caring for the children, and suggested the names of 33 other consultants who were also contacted (26 replied, 79%). Details of 108 children left untraced after the consultant contact were sent to the NHSCR for tracing to Family Health Services Authorities (FHSAs) or Area Health Boards (AHBs) with whom children still alive were, or were last known to have been registered. Each FHSA/AHB was contacted by us, enclosing a letter to be forwarded to each child's GP, asking for details of the child's address, current consultant and GP (if sent to the incorrect one).
1134
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
although this was not statistically significant. The male: female ratio of the total cohort is 1.4:1. The median age for this cohort at time of survey was 14 years, with a range from 8 to 18 years (Figure 2). Duration of diabetes ranged from 8 to 17 years, with a median of 12 years (Figure 3). 100
30
Number of children
25
20
Number of children 15r
10
80
60
11
12
13
14
15
16
17
Duration (years) I
Sex
20
I Female 0-5
6-11
12-17
L. ; Male
FIGURE 3 Duration of DM at 31 December, 1989. Cohort with DM onset before age 2 years
18-23
Age (months) Sex I Female
5
s35
? Male
FIGURE 1 Age at diagnosis. Cohort with DM onset before age 2 years
25
Mortality The SMR for females (9.9, 95% CI : 4.1-24.2) and those for the overall cohort (5.4, 95% CI : 2.5-11.5) were significantly elevated (Table 2). TABLE 2 Standardized mortality ratios
Number of children
No.
Person- Observed Expected Approximate deaths SMR 95% CI yean deaths
Females
131
1602.77
5
0.507
9.9
4.1-24.2
Males
179
2242.46
2
0.877
2.3
0.6-9.5
3845.23
7
1.296
5.4
2.5-11.5
20
15 I
Both
310
10 I
8
I
9
10
11
12
13
14
15
16
17
18
Age (years) Sex
L
I Female
[ 2 1 Male
FIGURE 2 Age at 31 December, 1989. Cohort with DM onset before age 2 years
Five of the seven children who died were girls and two were boys (Table 3). The cause of death in three children, all of pre-school age, did not seem to be directly related to diabetes: two died of infections (viral pneumonia and acute tracheobronchitis [the child also had Down's syndrome]) and the third, who also had hypothyroidism, developed severe hepatitis (of uncertain aetiology) and died of liver failure a few months later. Of the remaining deaths, three could be directly attributable to the metabolic disturbances of diabetes. One was an infant aged 12 months with a short history
Downloaded from http://ije.oxfordjournals.org/ at University of Strathclyde on April 9, 2015
•iJ 10
MORTALITY IN A BRITISH DIABETES COHORT TABLE 3 Details of deaths of seven children Death
1 2 3 4 5 6
Diabetes duration
Age
Female Female Female Female Female Male
2 years 1 year 3 years 12 years 13 years 4 years
2 years 11 years 12 years 3 years
3 years
2 years
7 Male
< 1 year
Cause
Viral pneumonia Diabetic ketoaddosis b Liver failure, (hypothyroidism) Convulsions'5, (Down's) Suicide5 Hypoglycaemic convulsion, cerebral oedema, encephalopathyb Acute tracheobronchitis, (Down's)
* Died at time of diagnosis b Diabetes related death
of illness whose diabetic ketoacidosis was not recognized for 10 hours after hospital admission. She vomited, suffered cardiac arrest and died soon after insulin therapy was commenced. Another child, aged 4 years, developed convulsions related to poor diabetic control. He suffered a prolonged severe hypoglycaemic convulsion and died one month later from cerebral oedema and hypoxic encephalopathy, having never regained consciousness. The third child (a 12 year old with Down's syndrome) developed convulsions (possibly hypoglycaemic) and she died soon afterwards in her sleep. The seventh child's tragic suicide at 13 years was almost certainly indirectly related to the diabetes. She had previously attempted serious insulin overdoses and was said to have shown signs of depression before taking a fatal overdose of antimigraine tablets. DISCUSSION A cohort of British children with diabetes diagnosed at the unusually young age of under 2 years has been compiled from the BDA Children's Diabetic Register. This is the largest cohort of which we are aware. The successful compilation of this cohort is mainly a result of the enthusiastic response of the consultants contacted, many of whom were still caring for the children at the time. Tracing through the NHSCR at Southport has been successful (92% of 108 names submitted), although it took 12 weeks to receive replies on 75 °/o of names submitted. This delay may be reduced in future when computerized searches will be possible. It took 3
months to obtain approval from OPCS for the children to be traced, a period unlikely to be affected by their computerization. We decided not to trace all children through Southport, because we wanted to establish contact with consultants caring for the children as early as possible, and also to limit costs. Deaths since diagnosis were analysed in 310 children. Their SMR was approximately five times higher than in the general population. Given the small numbers of deaths in this cohort and the wide CI, our mortality results are similar to those reported in other studies comparing mortality in childhood diabetes to that in the general population."'12 Dorman et al.u reported SMR of 5.4 in males and 11.53 in females. Children in their USA cohort were diagnosed aged 0-17 years in the period 1950-1981 and followed for 0-30 years. Sartor et al. n reported an SMR of 2.09 (95% CI: 1.14-3.83) in a Swedish cohort of both sexes aged 0-14 at diagnosis between 1977 and 1985. Followup was relatively short (0-8.5 years), which may explain, together with reportedly good health care, the few deaths (10) which occurred in their large cohort. Five of these deaths were directly attributable to diabetes mellitus, two occurring before the age of 2 years. In a recent report on childhood diabetes13 39 deaths had occurred in a cohort of 777 children aged 3 months to 16 years at diagnosis in 1930-1985 and followed for 3-63 years, but no SMR was reported. Other reports on mortality14"17 compared early onset with later onset diabetes and they all describe increased mortality for onset before age 15. Only BorchJohnson's analysis17 compared subgroups within childhood diabetes, and they reported no differences in mortality among those subgroups diagnosed before 15 years of age. Five of the seven deaths in our cohort occurred within 3 years of diagnosis and the other two more than 10 years after diagnosis. This distribution could be due to chance, given the small numbers, but it could be that some of the 16 children not (yet) traced have already died. This theory is supported by a recent report showing that mortality among members of an occupational cohort untraced on the NHSCR was substantially greater than mortality among those found on the Register.18 Causes of death in people with diabetes are not well documented, because diabetes mellitus is not always recorded on the death certificate either as an underlying or contributory cause; for instance it was recorded on only 70
