661
pre-eclampsia in our maternal population. Kilpatrick et al speculate that the association they observed may not be with HLA DR4 per se, but with a gene influencing susceptibility to pre-eclampsia that is in linkage disequilibrium with HLA DR4. This could be a characteristic of British people in contrast to the southern US maternal population. However, we have some doubts about this possibility since most of the white and a portion of the black gene pool in our cohort is derived from a population of British origin. In view of the small sample size in Kilpatrick and colleagues’ family study, we believe it is premature to suggest a causal association between HLA DR4 or a closely linked susceptibility increased risk for
gene and risk for
CAUSES OF DEATH POSSIBLY RELATED TO DIABETES 1981-88
pre-eclampsia in pregnant women.
We thank the National Institute of Child Health and Human Development (grant HD 21144) for financial support. Partial support was also provided by the US Department of Energy under contract DE-AC06-76RLO-1830. Battelle Pacific Northwest Laboratories, Richland, WA 99352, USA
CHARLES HOFF RICHARD G. STEVENS
College of Medicine, University of South Alabama, USA
HIRAM MENDENHALL RAYMOND D. A. PETERSON
Department of Obstetrics and Gynaecology, University of Louisville, USA
JOSEPH A. SPINNATO
Organ Procurement Foundation. SOEPF reference manual. 2nd ed. Richmond, Virginia: SOEPF, 1985: 18-8. 2. Pritcnard JA, MacDonald PC, Gant NF. Williams obstetrics. 17th ed. Norwalk, Connecticut: Appleton-Century-Crofts, 1985. 1. South-Eastern
Hypoglycaemia as cause of death insulin
in human
era
SIR,-The contention that treatment with human insulin carries an increased risk of severe hypoglycaemia’ has caused much concern. 75% or more of diabetic patients on insulin in Britain now inject human insulin. Since only 6% of insulin sold in 1985-86 was of human type, one might expect any potentially fatal adverse effect associated with human insulin to be reflected by an increase in deaths from hypoglycaemia. The issue is complicated by changes in managment other than the switch to human insulin. More intensive regimens aimed at tighter blood glucose control came in at the same time, and intensive treatment itself has been identified as a possible cause of increased hypoglycaemia unawareness.2 The Office of Population Censuses and Surveys (OPCS) confirms that evidence for or against such an increase is not easily obtained by examining routine statistics on cause of death. In the UK only about 60% of known diabetic patients have diabetes recorded on their death certificates;3 of these, only about 34% have diabetes recorded as the underlying cause of death,4 from which the routine OPCS statistics are derived. However, the likelihood of diabetes being assigned the underlying cause of death is much higher if death occurs before 45 years.3 Although deaths due to diabetes (ICD code 250) are subclassified according to information on the death certificate about specific complications such as "diabetes with ketoacidosis" (250.1) or "diabetes with coma" (250.2), there is no specific provision for diabetes with hypoglycaemia. Diabetes with coma includes ketoacidotic, hyperosmolar, and unspecified comas, which may be the result of unrecognised hypoglycaemia, but it does not specifically include diabetes with hypoglycaemic coma.s Hypoglycaemia as an underlying cause of death is found under ICD 251, which includes "insulin coma" but not diabetes. Assigning an underlying cause of death in such cases depends on the exact form of words used on the death certificate and what the coder accepts as a sequence of events leading to death. Thus the coding of a death described in part I of the death certificate as "diabetic hypoglycaemia" may vary—eg, one permissible interpretation is to assume coma and thus code the death under 250.2, or the death might be coded as "diabetes without mention of complications" (250.0), and in either interpretation hypoglycaemia will not be reflected in the underlying-cause statistics. The table summarises annual deaths, from 1981 to 1988, under relevant cause codes. The rise after 1983 in deaths due to diabetes
No data available for code 962 3 m 1985 *Provisional data for 1988.
(250.0) is probably an artifact, reflecting a change to the coding rules introduced in 1984, which had the effect of assigning the underlying cause of death to a major disease in part II of the death certificate, such as diabetes, in preference to a single condition considered to be a terminal event in part I (eg, bronchopneumonia). However, deaths in patients under 45 years show no such increase, which is especially relevant because much of the controversy has centred around deaths in young patients. Furthermore, there was no clear trend within this code over the years 1985 to 1988, the period which saw most
of the transfer from animal
to
human insulin. No
increasing trend is discernible for any of the other causes of death. We do not know which, or how many, of the deaths were caused by hypoglycaemia in diabetic patients, nor whether the patients were taking insulin, human or otherwise, at the time, but we might expect a genuine and substantial rise in deaths from hypoglycaemia in the diabetic population to be identifiable among the generally small numbers. Whilst the data we have presented cannot provide definitive
reassurance
about any increase in deaths due
to
hypoglycaemia since the introduction of human insulin, it does not substantiate such
a
claim.
Department of Community Medicine, University College London, London WC1E 6BT, UK
JUDITH STEPHENSON JOHN FULLER
A, Berger WG. Hypoglycaemia unawareness in diabetics transferred from beef/porcine insulin to human insulin. Lancet 1987; ii: 382-85. 2. Gale EAM. Hypoglycaemia and human insulin. Lancet 1989; i: 1264-66. 3. Fuller JH. Causes of death in diabetes mellitus. Horm Metab Res 1985; 15 (suppl). 4. Mortality statistics: cause 1986 (series DH2 no 13). London: OPCS. 5. Manual of the international statistical classification of diseases, injuries and causes of death: vol I. Geneva: World Health Organisation, 1977. 1. Teuscher
in epidemiology of diabetes mellitus in Africans
Suitability of reflectometry
SrR,-Your correspondents have once again raised the controversy about the epidemiology of diabetes mellitus in Mricans.1-3 One of the points made was whether reflectometry is suitable for the epidemiological study of hyperglycaemia in Africa. We have investigated the reliability of this method under tropical climatic conditions in Mali (West Africa). The temperature varied between 25°C and 32°C, and humidity between 26% and 60%. Test-solutions (’Preciset’) and blood samples from patients at Point G National Hospital in Bamako were tested with three reflectometers (’Reflolux’, Boehringer Mannheim) and a reference method (’God-Pap’ with deproteinisation). To test the precision of the reflectometers glucose was assayed ten times daily for 20 successive working days, in three test-solutions of different concentrations. The variation coefficient was 6% or less for a glucose concentration of 4-5 mmol/1, and less than 3% for concentrations of 90 and 150 mmol/1. Variation between apparatuses was negligible. These results were compared with those
pre-eclampsia in our maternal population. Kilpatrick et al speculate that the association they observed may not be with HLA DR4 per se, but with a gene influencing susceptibility to pre-eclampsia that is in linkage disequilibrium with HLA DR4. This could be a characteristic of British people in contrast to the southern US maternal population. However, we have some doubts about this possibility since most of the white and a portion of the black gene pool in our cohort is derived from a population of British origin. In view of the small sample size in Kilpatrick and colleagues’ family study, we believe it is premature to suggest a causal association between HLA DR4 or a closely linked susceptibility increased risk for
gene and risk for
CAUSES OF DEATH POSSIBLY RELATED TO DIABETES 1981-88
pre-eclampsia in pregnant women.
We thank the National Institute of Child Health and Human Development (grant HD 21144) for financial support. Partial support was also provided by the US Department of Energy under contract DE-AC06-76RLO-1830. Battelle Pacific Northwest Laboratories, Richland, WA 99352, USA
CHARLES HOFF RICHARD G. STEVENS
College of Medicine, University of South Alabama, USA
HIRAM MENDENHALL RAYMOND D. A. PETERSON
Department of Obstetrics and Gynaecology, University of Louisville, USA
JOSEPH A. SPINNATO
Organ Procurement Foundation. SOEPF reference manual. 2nd ed. Richmond, Virginia: SOEPF, 1985: 18-8. 2. Pritcnard JA, MacDonald PC, Gant NF. Williams obstetrics. 17th ed. Norwalk, Connecticut: Appleton-Century-Crofts, 1985. 1. South-Eastern
Hypoglycaemia as cause of death insulin
in human
era
SIR,-The contention that treatment with human insulin carries an increased risk of severe hypoglycaemia’ has caused much concern. 75% or more of diabetic patients on insulin in Britain now inject human insulin. Since only 6% of insulin sold in 1985-86 was of human type, one might expect any potentially fatal adverse effect associated with human insulin to be reflected by an increase in deaths from hypoglycaemia. The issue is complicated by changes in managment other than the switch to human insulin. More intensive regimens aimed at tighter blood glucose control came in at the same time, and intensive treatment itself has been identified as a possible cause of increased hypoglycaemia unawareness.2 The Office of Population Censuses and Surveys (OPCS) confirms that evidence for or against such an increase is not easily obtained by examining routine statistics on cause of death. In the UK only about 60% of known diabetic patients have diabetes recorded on their death certificates;3 of these, only about 34% have diabetes recorded as the underlying cause of death,4 from which the routine OPCS statistics are derived. However, the likelihood of diabetes being assigned the underlying cause of death is much higher if death occurs before 45 years.3 Although deaths due to diabetes (ICD code 250) are subclassified according to information on the death certificate about specific complications such as "diabetes with ketoacidosis" (250.1) or "diabetes with coma" (250.2), there is no specific provision for diabetes with hypoglycaemia. Diabetes with coma includes ketoacidotic, hyperosmolar, and unspecified comas, which may be the result of unrecognised hypoglycaemia, but it does not specifically include diabetes with hypoglycaemic coma.s Hypoglycaemia as an underlying cause of death is found under ICD 251, which includes "insulin coma" but not diabetes. Assigning an underlying cause of death in such cases depends on the exact form of words used on the death certificate and what the coder accepts as a sequence of events leading to death. Thus the coding of a death described in part I of the death certificate as "diabetic hypoglycaemia" may vary—eg, one permissible interpretation is to assume coma and thus code the death under 250.2, or the death might be coded as "diabetes without mention of complications" (250.0), and in either interpretation hypoglycaemia will not be reflected in the underlying-cause statistics. The table summarises annual deaths, from 1981 to 1988, under relevant cause codes. The rise after 1983 in deaths due to diabetes
No data available for code 962 3 m 1985 *Provisional data for 1988.
(250.0) is probably an artifact, reflecting a change to the coding rules introduced in 1984, which had the effect of assigning the underlying cause of death to a major disease in part II of the death certificate, such as diabetes, in preference to a single condition considered to be a terminal event in part I (eg, bronchopneumonia). However, deaths in patients under 45 years show no such increase, which is especially relevant because much of the controversy has centred around deaths in young patients. Furthermore, there was no clear trend within this code over the years 1985 to 1988, the period which saw most
of the transfer from animal
to
human insulin. No
increasing trend is discernible for any of the other causes of death. We do not know which, or how many, of the deaths were caused by hypoglycaemia in diabetic patients, nor whether the patients were taking insulin, human or otherwise, at the time, but we might expect a genuine and substantial rise in deaths from hypoglycaemia in the diabetic population to be identifiable among the generally small numbers. Whilst the data we have presented cannot provide definitive
reassurance
about any increase in deaths due
to
hypoglycaemia since the introduction of human insulin, it does not substantiate such
a
claim.
Department of Community Medicine, University College London, London WC1E 6BT, UK
JUDITH STEPHENSON JOHN FULLER
A, Berger WG. Hypoglycaemia unawareness in diabetics transferred from beef/porcine insulin to human insulin. Lancet 1987; ii: 382-85. 2. Gale EAM. Hypoglycaemia and human insulin. Lancet 1989; i: 1264-66. 3. Fuller JH. Causes of death in diabetes mellitus. Horm Metab Res 1985; 15 (suppl). 4. Mortality statistics: cause 1986 (series DH2 no 13). London: OPCS. 5. Manual of the international statistical classification of diseases, injuries and causes of death: vol I. Geneva: World Health Organisation, 1977. 1. Teuscher
in epidemiology of diabetes mellitus in Africans
Suitability of reflectometry
SrR,-Your correspondents have once again raised the controversy about the epidemiology of diabetes mellitus in Mricans.1-3 One of the points made was whether reflectometry is suitable for the epidemiological study of hyperglycaemia in Africa. We have investigated the reliability of this method under tropical climatic conditions in Mali (West Africa). The temperature varied between 25°C and 32°C, and humidity between 26% and 60%. Test-solutions (’Preciset’) and blood samples from patients at Point G National Hospital in Bamako were tested with three reflectometers (’Reflolux’, Boehringer Mannheim) and a reference method (’God-Pap’ with deproteinisation). To test the precision of the reflectometers glucose was assayed ten times daily for 20 successive working days, in three test-solutions of different concentrations. The variation coefficient was 6% or less for a glucose concentration of 4-5 mmol/1, and less than 3% for concentrations of 90 and 150 mmol/1. Variation between apparatuses was negligible. These results were compared with those
