optional (and sometimes even undesirable) but with some regimens it is crucial--eg, poor compliance with immunosuppressants seems to be a leading cause of organ transplant rejection.6,7 Should the right to be non-compliant influence access to the re-transplantation queue? You noted that microprocessors in drug packages can record each opening of the package "unbeknownst to the patient". Although early researchers sought compliance data untainted by patients’ knowledge of monitoring,8,9 it is now routine for patients to be informed that the package records their dosing. Such data are probably useful as a patient management tool,lO but assessment of that is just starting and is focused on areas where compliance is crucial to avoid serious, costly problems. A determined patient could compile a false record of package use without dosing, but must do so on schedule. He would be misjudged a drug non-responder-a still too-frequent misdiagnosis due to poor compliance being undetected by less sturdy methods. Microprocessor-based compliance monitors had to await progress in microcircuitry, minicomputers, and flexible softare. The biggest share of development costs has been for software, which usually goes unnoticed. APREX Corporation, Fremont, California 94539, USA
JG, Peck CC. Chronobiology: suggestions for integrating development. Ann NY Acad Sci 1991; 618: 563-71.
2 Pullar T, Kumar S, Tindall H, Feely M. Time to stop counting the tablets? Chn Pharmacol Ther 1989; 46: 163-68. 3 Maenpaa H, Manninen V, Heinonen OP. Comparison of the digoxin marker with capsule counting and compliance questionnaire methods for measuring compliance to medication in a clinical trial Eur Heart J 1987; 8 (suppl I): 39-43. 4. Lasagna L, Hutt P. Health care, research and regulatory impact of noncompliance. In: Cramer JA, Spikler B, eds. Patient compliance in medical practice and clinical trials. New York. Raven Press, 1991: 393-403. 5. Efron B, Feldman D. Compliance as an explanatory variable in clinical trials. J Am Stat Assoc 1991; 66: 9-17. 6. Rovelli M, Palmeri D, Vossler E, Bartus S, Hull D, Schweizer R. Noncompliance m organ transplant recipients. Transpl Proc 1989; 21: 833-34. 7 Didlake RH, Dreyfus K, Kerman RH, Van Buren CT, Kahan BD. Patient noncompliance: a major cause of late graft failure in cyclosporine renal transplants. Transpl Proc 1988; 20 (suppl 3): 63-69. 8. Cheung R, Dickins J, Nicholson PW, et al. Compliance with antituberculous therapy: a field trial of a pill-box with a concealed recording device. Eur J Clin Pharmacol 1988; 35: 401-07. 9 Kass MA, Meltzer D, Gordon M, Cooper D, Goldberg J. Compliance with topical pilocarpine treatment. Am J Ophthalmol 1986; 101: 515-23. 10. Feinstein AR. On white-coat effects and the electronic monitoring of compliance. Arch Intern Med 1990; 150: 1377-78.
Mammographic dysplasia as entry criterion for breast cancer prevention trials SIR,-Several studies have shown that about 50% of women over 45 years of age have mammograms with parenchymal patterns of P2 or DY (Wolfe classification) and that they have about twice the risk of breast cancer as women not classified into these groups. This translates into an incidence rate of about 1 -3 times the population rate, which in itself is not very striking. However, if combined with another fairly common risk factor such as nulliparity it can define a group at moderately high risk of breast cancer (more than twice the population risk). About 15% of women remain nulliparous and their risk of breast cancer is about 1-5 times that of parous women, suggesting that if these factors were independent about 6% of the population would have both factors and their risk would be about three times that of women with neither factor, leading to a rate twice that of the general population. To test this hypothesis directly we re-examined data from a case-control study of subsequent breast cancer from four centres participating in the study reported by Whitehead et al.1 Breast cancer developed in 444 women, aged 45-65 at time of entry, whose parity status was known, and these were matched for age and follow-up time to two controls (only one in 16 cases). 7-4% of controls and 14-4% of cases had both risk factors, for a crude relative risk of 1-93 times that in the screened population. A matched analysis given an odds ratio of 2-2, corresponds to a relative risk of 2 05. A similar analysis was done for women with a high-risk Wolfe pattern and breast cancer in a first-degree relative, a high-risk Wolfe
pattern and a benign biopsy, or a high-risk Wolfe pattern and any of the other three factors (nulliparity, family history, biopsy). The proportion of controls in each high-risk group, the odds ratio for a matched analysis, and the relative risk compared with the total population for each group are as follows: ....... I-
For all these groups there was an odds ratio in excess of 2, but the subgroup with a benign biopsy (and the subgroup in which any one of the three factors is present) is relatively more common so the relative risk compared with the total population is just below 2. This screening population was self-selected and over-represents highrisk groups, so the size of these groups will be less in the general population. Also, only a proportion of the women with previous biopsies will have proliferative benign disease, which increases their risk of breast cancer. Brisson et al2 have shown that the Wolfe classification can be improved by quantifying the area of film showing nodular or homogeneous densities, and another way of choosing high-risk women might be to use a single factor such as 60% of the area with increased nodular or homogeneous density. However, this needs to be verified in a prospective study. Other factors that yield a risk of at least twice that in the general population and that have been considered as entry criteria for a prevention trial include premenopausal breast cancer in a first-degree relative and a personal history of benign disease with atypical hyperplasia. Both indicate increased risk and are suitable entry criteria, but they are rare and apply to at most only a few per cent of the population. Women who have P2 or DY mammograms and who are nulliparous, have a family history of breast cancer, or have had a benign lesion with proliferative potential are a large group, and if they can be approached without creating undue anxiety and wish to participate they would constitute an attractive group for prevention studies. We thank Dr T. Carlile for allowing us
Imperial Cancer Research Fund, London WC2A 3PX, UK
Department of Applied Statistics, University of Reading
JACK CUZICK DAMON BERRIDGE
J, Carlile T, Kopecky KJ, et al. The relationship between Wolfe’s classification of mammograms, accepted breast cancer risk factors, and the incidence of breast cancer. Am J Epidemiol 1985; 122: 994-1006. 2. Brisson J, Verreault R, Morrison AS, Tennina S, Meyer F. Diet, mammographic features of breast tissue, and breast cancer risk. Am J Epidemiol 1989; 130: 14-24
indicator of poor outcome SIR,-Renal itch affects about 54%1 of patients receiving long-term haemodialysis. Its importance in outlook has not been Renal itch
considered. We reviewed the current status of patients on long-term haemodialysis in whom we had documented itch 3 years earlier2 to assess whether itch was a marker of poor outcome. 10 women and 44 men, aged 25-79, on long-term haemodialysis recorded the severity of itch twice daily for a week using a visual analogue scale (V AS).2 The patients were revisited 3 years later and were classified according to whether they were dead (22), alive (20), or had received kidney transplantation (12). Non-parametric analysis, excluding those who had undergone transplantation, showed a significant association (Mann-Whitney p=0007) between itch score and 3-year outcome-high itch scores being associated with death (figure). This relation could not be accounted for by the older age of those who had died at follow-up, since no association between itch and age was demonstrated (Spearman’s rank correlation coefficient 0-01) for patients who had not