1230 stage 0, no goitre; stage 1-A, goitre detectable only by palpation and not visible with the neck fully extended; stage 1-B, goitre palpable but visible only with the neck fully extended (this stage also includes nodular glands); stage 2, goitre visible with the neck in the normal position. Palpation was done by two specialists. 32% of children had goitre: 25% stage 1-A, 7% stage 1-B, and 0-1% stage 2. Goitre was more prevalent in girls than in boys (36% versus 28%). 23% of boys had stage 1-A goitre and 5% goitre stage 1-B; 27% of girls had stage 1-A, 9% stage l-B,andO’3% stage 2. In 5 children in whom urinary excretion of iodine was measured, values were 146, 154, 198, 239, and 278 ug iodine/g creatinine, respectively. This study has confirmed the high prevalence of goitre among Krk children. Although the goitres were mainly grade 1-A, the prevalence of 32% indicates mild endemic goitre, despite the seemingly adequate iodine intake. The previous difference in goitre prevalence in the various parts of the island was not noted, but only a few children were investigated in some communities. Intensive population migration might also account for this change. Before the introduction of iodine prophylaxis it was assumed that other factors beside iodine deficiency, notably dietary goitrogens, might be implicated in the development of goitre on Krk. The production and consumption of kale (a known goitrogen4) in the goitrous area was at that time about twice as high as in other areas of the island. In 5 of 12 subjects with goitre the perchlorate test showed that the trapped iodine was not bound organically.s Nonetheless, the role of kale in the occurrence of goitre in Krk remained questionable since goitrogenic activity is decreased by cooking. Horvat and Maver suggested that in addition to a lack of iodine goitre might be attributable to vitamin A deficiency. A comprehensive investigation has now been initiated to reliably estimate goitre prevalence and iodine intake in Krk. Department of Nuclear Medicine and Oncology, University Hospital "Dr M. Stojanović", University of Zagreb, Zagreb, Yugoslavia; Institute of Public Health, Rijeka; and Institute of Public Health of Croatia, Zagreb Z. KUSIĆ E. MESAROŠ SIMUNČIĆ N. DAKOVIĆ A. KAIC RAK LJ LUKINAC Š. SPAVENTI 1. Matovinović J. Endemic goiter and cretinism at the dawn of the third millenium. Am Ann Rev Nutr 1983; 3: 341-412. 2. Horvat A, Maver H. The role of vitamin A in the occurrence of goitre on the island of Krk, Yugoslavia. J Nutr 1958; 66: 189-203. 3. Delange F, Bastani S, Benmiloud M, et al. Definitions of endemic goiter and cretinism: classification of goiter size and severity of endemias, and survey techniques. In: Dunn JT, Pretell EA, Daza CH, Viteri FE, eds. Towards the eradication of endemic goiter, cretinism, and iodine deficiency. Washington: PAHO/WHO (scientific publication no. 502), 1986: 373. 4. Michajlovskij N, Sedlak J, Jušić M, Buzina R. Goitrogenic substances of kale and their possible relations to the endemic goitre on the island of Krk (Yugoslavia). Endocrinol Exp 1969; 3: 65-72. 5. Buzina R, Milutinović P, Vidović V, Maver H, Horvat A. Endemic goitre of the island of Krk studied with I131. J Nutr 1959; 68: 465-71. Rapid diagnosis of testicular choriocarcinoma by urinary pregnancy tests SiR,—Urinary human chorionic gonadotropin (hCG) testing is important in the differential diagnosis of acute abdominal pain and vomiting in women.1 However, very occasionally its use in young men can prevent prolonged delay in the initiation of potentially curative treatment for metastatic choriocarcinoma. We report the use of this test in such a patient. A 21-year-old male university student was admitted to Harold Wood district general hospital from the casualty department after a transient loss of conciousness and 6 months’ history of backache and testicular swelling. He had lost about 6 kg in weight over the past 6 weeks. He had a left-sided scrotal swelling, with evidence of an underlying testicular or epididymal mass, hydrocoele, a liver enlarged to 5 fingers below the costal margin, bilateral gynaecomastia, and abnormal breath sounds. Chest radiography showed multiple rounded opacities, up to 5 cm diameter, in both lung fields. He underwent fme-needle aspiration of the scrotal fluid and liver biopsy, which showed seminoma. When seen by the visiting oncologist 10 days after admission his condition was deteriorating. He had ascites and signs of inferior vena cava obstruction. Urea had risen from 13 to 26 mmol/I (normal 25-6 mmol/1), bilirubin from 30 to 179 ptnol/1 (<40 irnol!), and aspartate aminotransferase from 182 to 1300 IU/1 (<50 IUil). Computed tomography of brain and abdomen showed massive liver metastases and a tumour mass replacing the vena cava which extended into the right atrium causing bilateral lower limb oedema. The central nervous system was unaffected. Because of a hospital holiday serum testing for tumour markers was delayed by a day. A urinary pregnancy test was positive at a dilution of 1 in 1000 in saline. In view of his poor renal and hepatic function, he was treated with half doses of bleomycin, carboplatin, and etoposide. His condition further deteriorated after 48 h, followed by rapid improvement, such that at 21 days there were no ascites, palpable liver, or limb oedema. However, there had been virtually no decline in serial serum concentrations of tumour markers. Treatment was changed to bleomycin, vincristine, and cisplatin weekly for 6 weeks, then to twice weekly for 3 months. After 6 months’ treatment orchidectomy was done because of a residual mass. Histology showed only necrotic tumour. He has now been disease-free for 28 months. Non-seminomatous germ-cell tumours of the testes can present with many histological features, such as yolk sac and trophoblast as well as differentiated tissues (even bone and cartilage in some mature teratomas). A mixture of both seminomatous and non- seminomatous elements with expression of hCG is well documented and might account for our biopsy histology findings. However, the clinical features pointed towards a predominance of trophoblast (choriocarcinoma) elements. True choriocarcinomas of the testis are rare and rapidly lethal if not treated. In this case the histological pattern of seminoma was misleading. Fortunately, despite the confused clinical picture and the hospital holiday, detection and subsequent quantitative estimate of urinary intact hCG with pregnancy test kits reduced the delay in initiation of treatment. Several more days might have been saved had such a routine pregnancy test been done in the casualty department at first presentation. It is, however, possible to have false negatives with these kits if the tumour is only producing the beta-subunit.2 Therefore such tests should always be followed by quantitative assay, preferably by a supraregional assay service. Nevertheless, it is important that clinicians remember the ease and rapidity of this diagnostic procedure—especially when faced with a young adult of either sex with disseminated cancer of uncertain origin. Department of Clinical Pharmacology, St Bartholomew’s Hospital Medical College Harold Wood Hospital, London Department of Reproductive Physiology, St Bartholomew’s Hospital Medical College, London EC1A 7BE, UK Harold Wood Hospital London Hospital Medical College M. J. CAULFIELD M. G. DILKES R. K. ILES B. T. HANDEL R. T. D. OLIVER 1. Smith DB, Rustin GJS, Bagshawe KD. Don’t ignore a positive pregnancy test. Br Med J 1988 297: 1119-20. 2. Braunstein GD. hCG expression in trophoblastic and non-trophoblastic tumours In. Braunstein GD, ed. Oncodevelopmental markers: biologic, diagnostic and monitoring aspects. New York: Academic Press, 1983: 351-71. Regression dilution bias SIR,-Dr MacMahon and colleagues (March 31, p 765) discuss the importance of regression dilution bias in epidemiological studies. Their subject was blood pressure and cardiovascular disease, where the regression dilution yielded a 60% underestimate. The effect is even more striking when the variable is measured with far less precision than blood pressure. Estimates of sodium intake based on a single 24-h urine collection may result in as much as a four-fold bias in linear regression estimates.1 The difficulty increases when models more complex than a linear relation are explored. In their study in two Belgian towns, Staessen et aP fitted quadratic curves to describe the relation between sodium and blood pressure. Without correction for regression dilution, their results showed an apparently inverse sodium/blood pressure relation across the usual range of sodium intake. However, with a four-fold regression