regression coefficients are slightly less for the within-county analysis than for the between-county analysis. Random errors in the measurement of natural radiation can bias regression coefficients the
and affect standard errors; the influence of the is likely to be greater for the analysis based on districts than that based on counties, since fewer measurements are available within districts. If we correct for the bias in an analysis between districts based on ungrouped measures of natural radiationregression coefficients increase by about 40% for indoor gamma and by just over 60% for radon, with similar increases in the associated standard errors. Thus, although random errors in the statistical measurement of natural radiation do increase uncertainties there is still evidence of discrepancies between the results from the between-county and within-county analyses. The differences in the trend with radon and indoor gamma between the analysis based on counties and that based on the smaller districts within them suggest that the between-county analysis is affected by geographical confounding factors. Correlation studies based on large areas are often confounded in this way, and it is better to look at small areas or to do case-control or cohort studies. The Imperial Cancer Research Fund, NRPB, and the Office of Population Censuses and Surveys are collaborating in a cohort study of cancer and radon in Devon and Cornwall. The results presented here do not support the claims of Henshaw et al. Furthermore, in contrast to the analysis by Knox et al,8 based on NRPB data on outdoor gamma dose rates but with non-standard statistical methods, there is no evidence that the rate of childhood leukaemia and NHL increases significantly with gamma dose rate.
Protection Board, Chilton, Didcot, Oxon OX11 0RQ, UK
C. R. MUIRHEAD B. K. BUTLAND B. M. R. GREEN
Childhood Cancer Research University of Oxford
G. J. DRAPER
1. Henshaw DL, Eatough JP, Richardson RB. Radon as a causative factor in induction of myeloid leukaemia and other cancers. Lancet 1990; 335: 1008-12. 2. Butland BK, Muirhead CR, Draper GJ. Radon and leukaemia. Lancet 1990; 335: 1338-39. 3. Green BMR, Lomas, PR, Bradley EJ, Wrixon AD. Gamma-radiation levels outdoors in Great Britain (NRPB-R191). London: HM Stationery Office, 1989 4. Wrixon AD, Green BMR, Lomas PR, et al. Natural radiation exposure in UK dwellings (NRPB-R190). London: HM Stationery Office, 1988 5. GLIM (Generalised Linear Interactive Modelling) system. Release 3.77 London: Royal Statistical Society, 1985. 6. Alexander FE, McKinney PA, Cartwright RA. Radon and leukaemia. Lancet 1990; 335: 1336-37. 7. Stefanski LA. The effects of measurement error on parameter estimation Biometrika 1985; 72: 583-92. 8. Knox EG, Stewart AM, Gilman EA, Kneale GW. Background radiation and childhood cancers. J Radiol Prot 1988, 8: 9-18.
injection sclerotherapy of a bleeding peptic ulcer
SIR,-Several studies have shown that injection sclerotherapy decreases both rebleeding rates and mortality in patients with bleeding peptic ulcers.1,2 We report a case in which this procedure may have resulted in a fatal complication. A 34-year-old man presented in 1986 with dyspnoea, clubbing, and cyanosis and proved to have cryptogenic cirrhosis complicated
by microscopic intrapulmonary arteriovenous shunts (found at open lung biopsy), which resulted in permanent hypoxoaemia and arterial desaturation (P02 8 kPa, Sa02 88%). In October, 1990, he was admitted shocked with haematemesis, and had a Pa02 of 38 kPa. He was given four units of blood, and at endoscopy a large anterior duodenal ulcer was seen with an adherent clot. The ulcer was injected with a total of 5 ml of 1 in 10 000 adrenaline and 6 ml of 3% sodium tetradecylsulphate (STD) into four quadrants around the immediate area of the ulcer. Oral ranitidine was started. On day 11 of admission he became shocked, was found to have disseminated intravascular coagulation, and died despite surgery. Necropsy revealed circumferential full-thickness necrosis and severe confluent ulceration on the gastric antrum, lesser curve, and duodenum. Microscopic examination and culture showed
invasion by clostridial species and fungi but no evidence of antemortem thrombosis within gastric blood vessels was found. Such extensive ulceration and necrosis after injection of a duodenal ulcer has not been reported before and necrosis of the lesser curve and gastric antrum is not a recognised complication of such ulceration. We believe that injection of adrenaline and sclerosant into the duodenal wall of a patient with profound pre-existing hypoxia led to ischaemic necrosis of the gastric mucosa and duodenum, perhaps because of intravascular injection into the gastroduodenal artery. Arterial spasm secondary to intravascular adrenaline may have led to necrosis without arterial thrombosis and subsequent bacterial and fungal colonisation. We recommend that injection sclerotherapy for bleeding peptic ulcers should be carried out with caution in patients with severe hypoxaemia.
JEREMY LEVY Department of Gastroenterology, Whittington Hospital, London N19 5NF, UK 1
SALIM KHAK00 ROGER BARTON ROBIN VICARY
Chung SCS, Leung JWC, Steele RJC, Crofts TJ, Li AKC. Endoscopic injection of adrenaline for actively bleeding ulcers: a randomised trial. Br Med J 1988; 296:
1631-33. 2 Panés J, Viver
J, Forné M, Garcia-Olivares E, Marco C, Garau J. Controlled trial of endoscopic sclerosis in bleeding peptic ulcers Lancet 1987; ii : 1292-94
Neural tube defects in 18th century SIR,-Dr Duff and colleagues’ observation (Jan 12, p 120) of an increased incidence of neural tube defects (NTD) after a natural catastrophe, a hurricane, has a curious 18th century parallel. Midwife Catherina Schrader, who practised in Hallum and Dokkum, Friesland, Holland, between 1693 and 1745, carefully recorded the 3100 deliveries with which she assisted.1 She noted 6 cases of NTD or 1-9 per 1000 deliveries, a rate much higher than the 0-5 per 1000 seen in the Netherlands today. These cases occurred in two clusters during the years 1722/23 and 1732/33, both years following unusually poor crops in the region. The children with NTD came from urban families. She charged 6 guilders for a delivery on average and her fees ranged between none and 66 guilders. In the NTD families she charged only 0-5 guilders. These families were among the poorest, as paternal occupations indicate, and probably did not have access to the food, especially folate sources, that rural families would have. Although later in her career her reputation ensured that she assisted at a greater proportion of abnormal deliveries, her record of a 46% maternal mortality and 5-4 per 1000 infant mortality are similar to those in England at the time. These unusually accurate records suggest that there was a nutritional component, including folate deficiency, in the incidence of NTD in 18th century Holland. ICRF Tumour
Immunology Group, University College and Middlesex Schools of Medicine, London W1 P 8BT, UK
C. A. MICHIE
1. Marland H (translator) Mother and child were saved the memoirs of the Frisian midwife Catherina Schrader. Amsterdam. Rodopi, 1987. 2. Hibbard ED, Smithells RW. Folic acid and human embryopathy Lancet 1965, 1 1254
hypervolaemlc haemodilution to avord blood transfusion this article by Dr Trouwborst and colleagues (Nov 24, p 1295) the type of dextran 40 solution given to patients was the iso-oncotic form-ie, 50 g/1.
during major surgery.-In
Direct diagnosis of carriers of Duchenne and Becker muscular dystrophy by amplification of lymphocyte RNA.-Dr R. G. Roberts and his colleagues draw attention to two primer sequences given incorrectly in their paper (Dec 22/29, p 1523). The correct sequences (changes in bold for primers DMD7a and DMD7c were GCAACGCCTGTGGAAACGCTG and CAGGAAGCTCTCTCCCAGC (6431-6440 (6404-6424) respectively.