1097 Symmetrical cutaneous lymphoma SiR,—Professor Goudie and colleagues (Feb 10, p 316) hypothesise that the peripheral distribution of the cutaneous lesions of B-cell and T-cell lymphoma might be due to site-specific migration from the circulation and/or to preferential proliferation of neoplastic lymphocytes at defined anatomical sites. They do not suggest why this might happen. The distribution of the lesions illustrated is strikingly similar to that seen in lepromatous leprosy and it coincides with areas of low body surface temperature. Mycobacterium leprae multiplies preferentially at 3G-33OC. In advanced lepromatous disease not only is the skin involved but also the underlying subcutaneous tissue and sometimes muscle and bone. Another factor that may contribute to the proliferation of M leprae in the cooler areas of the skin is that immune surveillance, as mediated by macrophages and T-cells, is inefficient at low temperatures. I wonder whether lymphoma cells in the patients described multiply as well or better at 30-33°C and whether there is any evidence that at these sites they are escaping immunological attack. Hospital for Tropical Diseases, London NW1 0PE, UK ANTHONY BRYCESON Risk of cancer among laboratory workers S1R,-The observation of rare cancers (sarcomas, lymphomas) among workers under the age of 50 and doing biomedical research at the Institut Pasteur, Paris, prompted the establishment of a committee* to find out if work in biology laboratories, especially that involving biotechnology, might carry a risk of cancer. Investigations of occupational histories and working conditions included an epidemiological study. We report here on the first results of a mortality study in Paris among 3765 people who had worked for at least 6 months in the Institut Pasteur between 1971 and 1986. This cohort was followed up until the end of 1987, with 42 384 person-years of observation: 66 people (0-2%) could not be traced. 145 deaths occurred, and for 123 it was possible to obtain information on cause of death from death certificates (courtesy of Dr F. Hatton and Dr E. Michel). Expected deaths were calculated, with national rates as a reference, by sex, five-year age groups, and calendar year from 1971 until 1987, and observed and expected deaths were compared by a one-sided Poisson test. Selected results, for both sexes combined, are shown in the table.t t Total mortality was less than expected. This is not surprising in view of the socioeconomic background of the population studied. Deaths from cancer and cardiovascular disease were less than expected while accidental deaths were as expected, as were deaths from cancer of the breast and ovary. For cancers at sites observed in excess among laboratory workers or chemists in previous reports1 (haemopoietic, bone, brain, pancreas) mortality from leukaemia or lymphoma was close to expected. For bone cancer there was an excess among males (p=006); pancreatic cancer (p=0’02) and brain tumours were in excess among females. These cancers were in people who had worked mostly ill bacteriology laboratories, and their work practices will be studied more closely. Among other cancers there was a deficit of lung cancer and cancer of the larynx, probably reflecting low tobacco consumption. This mortality study shows that, despite an overall deficit in cancer deaths compared with national rates, this group of workers may carry an excess risk of death from bone, brain, and pancreatic cancer. It should be noted that pancreas and brain cancers are difficult to diagnose, and a proportion of bone neoplasms on death certificates are in reality secondary. (Considerations of confidentiality bar access to the clinical records of those who died.) * Prof Jean Bernard (chairman), Ms F. Conso, S. Cordier, G. Flandrin, R.Latarjet, G. Lenoir, C. S. Muir, A. Picot, P. Potier, A. J. Sasco, M. Tubiana, and J. C. Zerbib. t A more detailed table, including data for males and females separately and information on less usual cancers and on other causes of death, may be obtained from The Lancet. OBSERVED (0) AND EXPECTED (E) DEATHS *p<O 01; tp<O 10 We are now studying detailed occupational histories and chemical, radiation, and biological exposures and searching for incident cases of cancer (especially leukaemia and lymphoma) since several reported incident cases occurring after 1986 were not included in the mortality study. I thank the following: from the Institut Pasteur Dr M. L. Mousel, N. Namiech, 0. Soumah and the medical staff; P. Des Pres de la Morlais, P. Pollet, R. Bariisso and the administrative staff; and all the researchers who gave time to this work, especially G. Gachelin, G. Gareau, and G. Michaud; from the National Institute for Health and Medical Research (INSERM) SC8 F. Hatton and E. Michel; and from INSERM U170 B. Bonnet, C. Le Goaster, N. Le Moual, F. Pietri, and G. Vasseur. INSERM U170, 94807 Villejuif, France SYLVAINE CORDIER 1. Sasco AJ. Risques pour la santé dans les laboratoires de recherche biologique et médicale. Med Sci 1989; 5: 489-98. Trends in schizophrenia SiR,—To the question "Is schizophrenia disappearing?" the answer "We don’t know" is more accurate than the affirmative reply offered by Mr Der and colleagues (March 3, p 513). It is not usually accurate to generalise information on treated incidence to population incidence. This is akin to saying that the number of new cases of peptic ulcer seen by physicians is equal to the number of new cases in a population, an inference that is patently false. Treated incidence is a subset of population incidence. The evidence that Der et al marshal in support of the notion that treated incidence is decreasing needs to be critically assessed. We would offer the following thoughts. Better to understand first admission and readmission rates we need to consider system size and length of stay. For example, patients admitted in the past to State mental hospitals in the US had long periods of stay in each episode; a significant number remained for life. Those hospitals had a year-end census of about 559 000 patients in 1955, compared with 107 000 in 1987. In a large system with long stays, the first-admission rate will be high and the readmission rate low. By contrast, a smaller inpatient care system with short stays will have a higher readmission rate (more ex-patients to be readmitted) while first admissions will be fewer because of efforts to keep new admissions out of the system. Thus, treated incidence can decrease independently of any hypothesised decrease in population incidence. A case register for a single community care system (Aberdeen) and a small sample from another (Nottingham) are not adequate for system-wide inferences for England and Wales. Community mental health care systems have become progressively more complex. They now encompass specialty providers, health care providers, social service and housing agencies, family groups, community shelters, and so on. Research is needed to clarify the role of all these organisations in the care of people with schizophrenia. It is unclear what components of care Der et al considered when drawing their inferences. The age distribution of the base population is important too. In the US the "baby boom" cohort (those born between 1946 and 1964) has now largely passed the prime age for onset of schizophrenia. Hence, Kramer’ and others expect increases in population and treated prevalence, even if age-specific population