1134 United States, the Carribean, and Nigeria are shown in fig 2. In both men and women the mean systolic and diastolic pressures in American Negroes were slightly higher than those in West Indian or west African Negroes. Abrahams et all3 had earlier shown that blood pressures in indigenous west Africans were generally similar to those recorded in White populations in Europe and America. If mean pressures in African children in Africa are significantly higher than those in children of African descent in the United States, and mean adult pressures are practically the same in both groups, then the sharp paediatric contrast must be attributable more to an environmental than to a genetic basis. We can only speculate on the factors that contribute to the rapid rise in blood pressure with age, which eventually causes the average American Black to end up with a considerably raised arterial pressure in subsequent decades of life. In this preliminary study we have found no evidence of a meeting point in arterial pressures, when the observed established differences in adolescence are taken back to the early part of the first decade in the form of a regression, but we did not study the behaviour of blood pressure in infancy and early childhood. This forms the subject of a separate study.

BRITISH MEDICAL JOURNAL

30 APRIL 1977

References I Adams, J M, American3Journal of Medicine, 1932, 184, 342. 2 Rose, G, Journal of Chronic Diseases, 1962, 15, 373. 3

McDonough, J R, Garrison, G E, and Homes, C G, in The Epidemiology of Hypertension, ed J Stamler, R Stamler, and T N Pullman, p 167. New York, Grune and Stratton, 1962. 4 Boyle, E, jun, et al, in The Epidemiology of Hypertension, ed J Stamler, R Stamler, and T N Pullman, p 193. New York, Grune and Stratton, 1967. 5 Wilber, J A, in The Epidemiology of Hypertension, ed J Stamler, R Stamler, and T N Pullman, p 439. New York, Grune and Stratton, 1962. 6 Hutcheson, J M, jun, Hejmancik, M R, and Herrman, G R, American Heart Journal, 1953, 46, 565. 7 Graham, A W, Hines, E A, jun, and Gage, R P, American J3ournal of Diseases in Children, 1945, 69, 303. 8 Comstock, G W, American Journal of Hygiene, 1957, 65, 271. 9 Londe, S, Clinical Pediatrics, 1966, 5, 71. 0 Londe, S, Clinical Pediatrics, 1968, 7, 400. 15 Allen-Williams, G M, Archives of Disease in Childhood, 1945, 20, 125. 12 Janis, K M, Kemmerer, W T, and Hagood, C 0, of Pediatric Surgery, 1971, 6, 70. 13 Abrahams, D G, Alele, C 0, and Barnard, B G, West African Medical

jun,J7ournal

Journal, 1960, 9, 45. 14 Akinkugbe, 0 0, and Ojo, 0 A, British Medical Journal, 1969, 2, 222. 15 Miall, W E, et al, British Medical3Journal, 1962, 2, 497. (Accepted 15 February 1977)

SHORT REPORTS Water intoxication associated with carbamazepine treatment Carbamazepine has been reported to have an antidiuretic action,' though it is not known how it produces this effect. We report a patient with water intoxication caused by carbamazepine treatment in whom the antidiuretic hormone 8-arginine vasopressin (AVP) was measured by a specific and sensitive radioimmunoassay.2

Case report A 45-year-old woman was admitted as an emergency to Leeds General Infirmary on 29 August 1976 because of poorly controlled epilepsy and increasing mental confusion. She had been an epileptic for many years. On admission she was taking phenytoin 100 mg twice daily, phenobarbitone 30 mg three times a day, carbamazepine 400 mg three times a day, benzhexol 2 mg three times a day, and folic acid 5 mg three times a day. A course of erythromycin 500 mg four times a day had been started for an infected lesion in her right great toe. She was drowsy, disorientated, and dysarthric. There were no other abnormal neurological signs. In particular there was no evidence of raised intracranial pressure or head injury. The clinical picture was compatible with a postictal state or an overdose of anticonvulsant. There was no clinical evidence of fluid overload. Blood was taken for measuring urea and electrolytes. The serum sodium concentration was unexpectedly low (see table). We thought that this might be explained by inappropriate secretion of AVP caused by carbamazepine. The dose of carbamazepine was reduced to 200 mg three times a day and fluid intake restricted. The patient's clinical state improved greatly, plasma carbamazepine concentration fell, and serum electrolytes and osmolality returned to normal (see table). The change in treatment did not provoke further convulsions. AVP was determined using a recently developed specific and sensitive radioimmunoassay.2 Plasma and urinary concentrations of the hormone were very high in the presence of toxic concentrations of carbamazepine, despite considerable plasma hypotonicity. When the dose of carbamazepine was reduced the plasma sodium and AVP concentrations returned to normal within three days, by which time plasma carbamazepine concentrations were within the accepted therapeutic range.

Comment Carbamazepine has been thought to exert its antidiuretic effect either by potentiating the action of AVP on the renal tubules,3 or by

Blood values of drugs and plasma and urinary concentrations of A VP

Phenytoin(,umol, 1) Phenobarbitone (,urmol,l) Carbamazepine ) (PlsmolaAVP Plasma (ng/l) Urine AVP (ng/mmol creatinine) Plasma sodium (mmol/l) Plasma osmolality (mmol/kg) Urine osmolality

(mmol!kg)

10 May 76

29 August 76

31(100 mg twice a day) 110 (30 mg thrice daily) 20 (200 mg thrice daily)

20

22

110

63 (400 mg

thrice

1 September 76

daily)

72

24 (200 mg

thrice daily)

1-7

0 33

17

17

120

136

261

274

375

195

Conversioni: SI to traditional tuits-Phenytoin: ,umol/l 0025 mg,'100 ml. Phenobarbitone 1 ,umol, 1 _ 0 023 mg, 100 ml. Carbamazepine 1 umol 1 0-024 mgi 100 ml. Creatinine 1 smol:l -0O0113 mg/100 ml. Sodium: 1 mmolIl= 1 mEq,l. Osmolality: 1 mmol,kg = 1 mOsm/kg.

increasing AVP secretion from the pituitary.4 Attempts to measure AVP in patients treated with carbamazepine have produced conflicting results. To some extent this may reflect the great difficulty in measuring AVP in and below the normal range. Bioassay uses a difficult preparation-the water-loaded, alcohol-anaesthetised ratand is relatively insensitive. Most radioimrnunoassays use antibodies of questionable specificity, but even a well-characterised assay,3 although providing evidence against a direct effect of carbamazepine on the pituitary, has not provided a conclusive answer. Using a specific radioimmunoassay we showed high serum and urinary AVP levels when plasma carbamazepine concentrations were in the toxic range. This suggests a derangement of the osmotic control of AVP release from the pituitary, which is seen in physiological circumstances only with considerable hypervolaemia. The administration of phenytoin and phenobarbitone in addition to the carbamazepine may have been important. Both compounds are powerful inducing agents of drug metabolism. In contrast to carbamazepine, phenytoin has been reported to inhibit AVP release.' Reduction in the dose of carbamazepine led to a rapid resolution of the low sodium state, suggesting that this drug was the major causative factor. Inappropriate production of vasopressin may have contributed to the convulsions seen in this patient before admission. At one time water intoxication induced by vasopressin was suggested as a provocative test for epilepsy. Carbamazepine has shown its value in

BRITISH MEDICAL JOURNAL

30 APRIL 1977

treating epilepsy and trigeminal neuralgia. The measurement of plasma sodium concentrations, blood concentrations of the drug, and AVP (if the assay is available), may improve the management of an individual patient and prevent the rare complication of water intoxication. Rado, J P, British Medical Journal, 1973, 3, 479. Thomas, T H, and Lee, M R, Clinical Science and Molecular Medicine, 1976, 51, 525. 3 Meinders, A E, Cjeka, V, and Robertson, G L, Clinical Science and Molecular Medicine, 1974, 47, 289. 4 Kimura, T, et al, J7ournal of Clinical Endocrinology and Metabolism, 1974, 38, 356. 6 Fichman, M P, Kleeman, C R, and Bethnine, J E, Archives of Neurology, 1970, 22, 45. I

2

(Accepted

27_January

1977)

University Departments of Medicine and Chemical Pathology, General Infirmary Leeds LS1 3EX M G ASHTON, MB, MRCP, research assistant S G BALL, MB, MRCP, MRC research training fellow T H THOMAS, BSC, PHD, research fellow M R LEE, DM, MRCP, consultant and senior lecturer in clinical pharmacology

1135 Analysis of paired samples-More correct estimates of risk and statistical significance may be obtained when using paired matched controls if the analysis maintains the pairing-that is, if each case is compared with the specific control that was paired with it. The basic difference introduced by matched-pair analysis is that pairs in which patient and control are similar with respect to the study factor are not considered, and estimates are based solely on pairs in which one member has and the other does not have the factor under study. We therefore used McNemar's marginal x2 test which takes account of the discordant pairs. Results are shown in the table.

Appendicectomy state of male and female diabetics and controls Diabetics

Controls

Appendicectomy Appendicectomy No appendicec-

No appendicectomy

Total pairs

8 (r)

48 (s)

56

tomy

34 (t)

161 (u)

195

Total

42

209

251

(s

-

t)2 (14)2

X2 =---s+t

82

= 2-4.

DF= 1; P>0 05.

Comment

Diabetes and appendicectomy: testing a hypothesis So-called "pressure diseases" such as varicose veins, haemorrhoids, diverticular disease, appendicitis, and hiatus hernia may be associated with a fibre-deficient diet.' 2Some, however, are sceptical about these sweeping conclusions.:' Burkitt4 states that there is abundant epidemiological evidence relating appendicitis to both the removal of fibre from diet and the addition of sugar. Fibre deficiency may cause raised intraluminal pressure in the colon, particularly in the appendix when blocked with faecoliths from constipation. This may devitalise the mucosa, while excess sugar alters the faecal flora, probably causing the inflammatory process. The association between conditions attributed to low-residue diet and those, such as diabetes, attributed to excess consumption of refined carbohydrates, may be readily explained.4 The removal of unabsorbed fibre from carbohydrate food may cause increased consumption of the refined product to satisfy appetite. Excessive consumption of refined carbohydrate and fibre depletion are therefore two sides of the same coin. Since an increased incidence of diabetes among patients with diverticulitis has been shown; it was suggested that the prevalence of appendicectomy among patients with maturity-onset diabetes should be greater than among a control group without maturity-onset diabetes. This hypothesis prompted us to test the statement2 that the causal relation of two or more diseases to a single common environmental factor may predispose people exposed to this factor to develop several of the diseases, and that these diseases would therefore be more frequently associated with one another in single patients than would otherwise be expected. We wanted to see if there was any association between appendicectomy and diabetes with reference to the fibre depletion and refined carbohydrate theory. Patients and methods We studied 104 men and 147 women with maturity-onset diabetes, aged 40 years and upwards, who had not been diagnosed as suffering from diabetes mellitus before the age of 40, and who had been consecutively admitted to the outpatient diabetic clinic in St Finbarr's Hospital, Cork. Controls were 104 men and 147 women matched with the diabetics for sex and age (±2 years) who had attended the accident and emergency department, St Finbarr's Hospital, or who had been admitted to the Orthopaedic Hospital, Cork, and in whom diabetes had been excluded. Subject to these criteria the controls were also taken consecutively. Patients and controls were matched for age and sex as these are two of the most important factors associated with the prevalence of appendicectomy and maturity-onset diabetes. Appendicectomy is commonest in young people, especially women, and maturity-onset diabetes is found more commonly in middle-aged and

elderly women.

The mean age at appendicectomy in male diabetics was 315 years and 34 9 in the controls; in women it was 26-5 and 30 3 years, respectively. In both sexes more controls than diabetics had had an appendicectomy but the differences between the diabetics and controls were not significant, and the hypothesis that the prevalence of appendicectomy among patients with maturity-onset diabetes would be greater than among non-diabetic controls matched for age and sex was not sustained. Thus any difference that exists may be due to chance. I

Cleave, T L, Campbell, G D, and Painter, N S, Diabetes, Coronary Thrombosis and the Saccharine Disease, 2nd edn, Bristol, John Wright, 1969. 2 Burkitt, D P, and Trowell, H C, Refined Carbohydrate Foods and Disease. London, Academic Press, 1975. 3 Mendeloff, A I, American Journal of Digestive Diseases, 1976, 21, 109. 4 Burkitt, D P, Cancer, 1971, 28, 3. 5 Schowengerdt, C G, et al, Archives of Surgery, 1969, 98, 500.

(Accepted 24 January 1977) Department of Social Medicine, University College, Cork, and Department of Medicine, St Finbarr's Hospital and University College, Cork J P CORRIDAN, MD, professor of social medicine J P O'REGAN, MB, MRCP, registrar in medicine (present appointment: assistant professor of medicine, St Louis University Medical Center, Missouri, USA) D J O'SULLIVAN, MD, FRCP, professor of medicine

Miliary Crohn's disease The diagnosis of Crohn's disease is difficult, particularly preoperatively.1 Even the classic form of Crohn's disease requires differentiation from ileocaecal tuberculosis.2 Previous reports have described five cases of Crohn's disease in which the dominant or sole pathological feature observed at laparotomy was peritoneal nodules resembling tuberculosis.3-5 Heaton et al3 named this form "miliary Crohn's disease". We present here the sixth case of this type. Case report A 23-year-old woman gave a two-year history of vague abdominal pain and intermittent diarrhoea without blood. She was admitted to hospital in