814 Rees and colleagues’ conclusion that stunting (height for age) and wasting (weight for height) are independent is not true, as a non-parametric correlation coefficient demonstrates (tau= -002, p=0-02). The conclusion that the boys and girls were not significantly different in respect of stunting is also incorrect (Mann Whitney test, p = 0-036) and a lack of association between sex and wasting is doubtful (Mann Whitney test, p 0057). We are not saying that these tests should be used in place of chi-square simply because they give smaller p values: the correct test should be used whatever p value it produces. Rees et al used chi-square and Fisher’s exact tests: neither is appropriate to their study.
PROSTATE CANCER AND ANDROGENS
SIR,-Dr James (Jan 24, p 217) suggests that use of androgens as growth promoters in animals may be associated with the excess of prostatic cancer in butchers observed in England and Wales. We have done a large mortality study of over 19 000 men in a meatcutters’ union in Baltimore, Maryland. 95% of the workers had worked in only one job category throughout, and were thus classified accordingly. The job categories were: workers in abattoirs or abattoir-cum-meatpacking plants in which cattle, sheep, and pigs were slaughtered; meatpacking plants in which no slaughtering was done but cattle, sheep, and pig meat was processed; meatcutters in the meat department of supermarkets and grocery stores; workers in chicken slaughtering plants; and a control
=
Department of Clinical Epidemiology, London Hospital Medical College,
STEPHEN J. W. EVANS
London El 1BB
SIMON J. DAY
group of workers in the same union which included workers in soft drinks manufacturing, soup canning, oyster-shucking, and so on. No excess risk from prostate cancer was observed in the abattoir and meatpacking groups where exposure to cattle or beef in all
1. Rees DG. Essential statistics. London: Chapman and Hall, 1985: 122-30. 2. Siegel S. Nonparametric statistics for the behavioural sciences. Tokyo: McGraw Hill, 1956: 104-11. 3. Ingram D, Bloch R, eds. Mathematical methods in medicine. Chichester: Wiley, 1984: 41-43. 4. Moses LE, Emerson JD, Hosseini H. Analyzing data from ordered categories. N Engl J Med 1987; 311: 442-48. 5. Everitt BS. The analysis of contingency tables. London: Chapman and Hall, 1977. 6. Agresti A. Analysis of ordinal categorical data. Chichester: Wiley, 1984.
phases of production occurs (table). Thus use of hormones, whether oestrogens or androgens, in animal feed for cattle, sheep, or pigs is not
associated with excess prostate cancer in these groups.
STANDARDISED MORTALITY RATIOS* FOR PROSTATE CANCER BY JOB CATEGORY
* This letter has been shown to Mr Rees, whose reply follows.ED. L.
SiR,—Iinfer that Mr Evans and Mr Day are advising the use of Kendall’s tau for r x c contingency tables where the variables forming the tables have ordered categories and r and c are greater than 2. There are three Kendall’s tau statistics.1 Tau-a is not applicable to contingency table data. Tau-b and tau-c have the disadvantage that they have no obvious probabilistic interpretation; they are measures to be used as descriptive statistics, and questions as to their significance are relatively unimportant .2 Since tau-b may only attain the extreme values ± 1 when r c, tau-c may be used for cases when r 4= c since it can attain the values ± 1. Nevertheless I have applied tests based on tau-b and tau-c to the six tables where r and c were above 2. In five the decision to reject or not to reject the null hypothesis is unchanged. The "odd man out" is the test for independence between height for age (stunting) and weight for height (wasting). However, at this point in our paper the aim was to compare three indicators-namely, height for age (stunting), weight for height (wasting), and MUAC (wasting). For these comparisons I find: tau-b tau-c -0-08 -0-02 height for age vs weight for height
r
*Numbers in parentheses t04 deaths expected.
=
height for age vs MUAC weight for height vs MUAC
0.25 0.04
_
0.18 0.01
D. G. REES Oxford OX3 0BP 1. Kendall MG, Stuart A. The advanced theory of statistics. Vol II. London: Griffin, 1973: 582-85. 2. Everitt BS. The analysis of contingency tables. London: Chapman & Hall, 1977: 62-63. 3. Armitage P. Statistical methods in medical research. Oxford: Blackwell Scientific
Publications, 1971: 402.
number of deaths.
A non-significant excess of prostate cancer among meatcutters in supermarkets who come into contact with all types of meat, including chicken, was observed for both racial groups and also in black males in chicken slaughtering plants. The fact that similar risk
Once again our conclusion is unchanged, since MUAC does not have the desired relation to the other two indicators-namely a relatively low correlation with height for age and a relatively high correlation with weight for height. Evans and Day suggest a Mann-Whitney U test for 2 x c tables where there are more than 2 columns and the column variable has ordered categories. This test is equivalent to a X2 test of independence in combination with a X2 test for trend.3 I have applied these X2 tests to the two 2 x c tables in our paper. Yet again, I find that our conclusion-that there was no significant difference between boys and girls with respect to weight for height (wasting)is unchanged. However, a comparison between boys and girls with respect to height for age (stunting) reveals a trend which is just significant at the 5% level, the proportion of boys going steadily down from 0-55 to 0’43 over the four height for age categories (table iv). This is a small effect in comparison to the previously reported age effect for height for age. Nevertheless we are grateful to Evans and Day for their assistance in uncovering it. Department of Computing and Mathematical Sciences, Oxford Polytechnic,
are
is observed in both control groups does not seem to support the hypothesis of occupational exposure as a possible cause. Meatcutters in supermarkets may well eat more meat than other workers in the meat industry as they usually receive the highest salaries. However, our data are not suitable for assessing the role of non-occupational exposures such as meat intake, so we cannot rule out this possibility completely-nor is the group of workers in chicken slaughtering plants in our study large enough to allow assessment of the role of exposure to chicken or chicken meat. Unit of Analytical Epidemiology, International Agency for Research on Cancer, 69372 Lyon, France
1.
ERIC S. JOHNSON
Johnson ES, Fischman HR, Matanoski GM, Diamond E. Cancer mortality among white males in the meat industry. J Occup Med 1986; 28: 23-32.
HALOPERIDOL METABOLISM AND ANTIPSYCHOTIC EFFECT IN SCHIZOPHRENIA
SiR,—The treatment of schizophrenic patients with neurolepdc not always successful. For some neuroleptics the
drugs is
measurement of
drug concentrations in plasma seems to be one way of improving the response to treatment.1 For haloperidol the existence of a curvilinear relation between drug levels and clinical response has proved controversialPart of the explanation could be that the drug’s sole active metabolite, reduced haloperidol, is not measured. Reduced haloperidol has a lower affinity for dopamine receptors and is less active than haloperidol. Preliminary data’ have suggested that this metabolism may influence clinical outcome. We have treated eighteen schizophrenic inpatients (mean age 35, SD 13 years) with oral haloperidol at doses of 0-2 (0-07) mg/kg (6-21 mg daily) for 4 weeks. Anticholinergic drugs and benzodiazepines were allowed only in cases of dire necessity. Before treatment and after the fourth week the brief psychiatric rating scale (BPRS)4 was completed. At the fourth week plasma haloperidol and reduced
815
MEASURED AND PREDICTED CREATININE CLEARANCE
1.73m2])
*Results
as
median
IN
(mlm7ri [per
70 WOMEN*
(first quartile, third quartile).
tMeasured/predicted.
sophisticated version of the formula, which depends on the expected excretion of creatinine per kilogram body weight at different ages, is necessary. The discrepancy was least for the lowest clearance values found in the highest age groups. Hence the prediction of clearance is a reasonable guide to the glomerular filtration rate in very elderly patients, in whom urine collection may be impossible without catheterisation. The very low clearance rates which we found in the age group 85-94 years fit in with the trend shown in the graphs of Elseviers et al. M. R. BLISS S. VELLUPILLAI P. A. C. JULIAN J. E. H. SHAW J. THOMAS
RH/H Clinical improvement and RH/H ratios.
improvement (cut-off below 20% improvement in BPRS); B = improvement (by more than 20%). A= no
concentrations
measuredand the ratio reduced haloperidol haloperidol/haloperidol (RH/H) was calculated. Patients were classed as non-responders (group A; BPRS improvement below 20%) or responders (group B; BPRS improvement more than 20%). Group A had a higher mean RH/H ratio than group B (2-87 [2’40] vs 0-94 [0’35]) (figure). The two groups did not differ in respect of age, duration of illness, BPRS before treatment, daily dosage of haloperidol, or absolute plasma concentrations of H, RH, orH+RH. Our fmdings suggest that the inconsistent fmdings in studies of plasma haloperidol and clinical response in schizophrenic patients is at least partly due to failure to assess RH/H ratios. were
A. C. ALTAMURA M. C. MAURI R. CAVALLARO
University Department of Psychiatry, Policlinico Guardia II, 20122 Milan, Italy
Department of Geriatric Medicine, Hackney Hospital, London E9 6BE 1 Cockcroft
SEROLOGICAL MARKERS FOR DELTA HEPATITIS
SIR,—Dr Aragona and colleagues (Feb 28, p 478) reported on the serological profile of hepatitis delta virus markers in 74 patients with acute delta hepatitis. Using a radioimmunoassay, they detected delta antigen in 17 patients (20 %) while IgM antibody was present in 69 (93%). They concluded that IgM anti-delta is the most important marker for delta hepatitis during acute infection. We have done a similar study in 59 patients with acute delta infection (42 co-infections of hepatitis B and delta hepatitis and 17 superinfections). At the time of the first sample, obtained within 2 weeks of the onset of symptoms delta antigen was detected in 3% and IgM antibody in 77%:
Curry SH, Montgomery S, et al. Early unwanted effects of fluphenazine esters related to plasma fluphenazine concentrations in schizophrenic patients Psychopharmacology 1985; 87: 30-33. 2. Smith RC, Schooler NR. Blood levels of haloperidol and clinical response. Psychopharmacol Bull 1985; 21: 48-72. 3 Ereshefsky L, Davis CM, Harrington CA, et al. Haloperidol and reduced haloperidol plasma levels in selected schizophrenic patients. J Clin Psychopharmacol 1984; 4: 138-42. 5. Moulin
J, Gorham D. Brief psychiatric rating scale. Psychol Rep 1962; 10: 799-812. MA, Camsonne R, Davy JP. Gas-chromatography electron impact and
chemical ionization
mass
antigen*
(with or without A. GORNI
1. Altamura AC,
4 Overall
Total
Co-infection Superinfection Delta
Mass Spectrometry Laboratory, State Health Department, Milan
DW, Gault MH. Prediction of creatinine clearance from serum creatinine.
Nephron 1976; 16: 31-41.
spectrometry of
homologue. J Chromatogr 1979; 178:
haloperidol
and its chloridrated
324-29.
MEASURED AND PREDICTED CREATININE CLEARANCE MR,—JJr blsevlers and
colleagues (-Feb 21, p 457) predicted creatinine clearance with the formula of Cockcroft and Gault! for outpatients in whom accurate urine collection may be difficult. In a study of changes of urine production with age, we have compared creatinine clearance predicted by this formula with measured clearance in 70 women without known renal disease, aged 19-100 years.
The predicted clearance underestimated the observed values by 15-20 ml/min (per 1-73 m2) between the ages of 25 to 75 (table). There was a break in the linear relation between the two sets of values under and over the age of 70, suggesting that a more
delta antibody) IgM anti-deltat
37/42 27/36
12/17 14/17
49/59 (83%) 41/53 (77%)
*’Deltassay’ (Notched, Dublin, Ireland). tMethod of Aragona et al; not tested for m 6 patients with co-infection.
The difference in delta antigen detection rate between our series and Aragona’s could be related to differences in the interval between onset of symptoms and hospital admission (not clearly stated in the report by Aragona et al) or to different sensitivities in the assays used. In our area almost all cases of delta hepatitis are in drug abusers, and such patients are difficult to follow up; thus, the diagnosis of acute delta infection should be established as soon as possible. We agree with Aragona and colleagues that IgM anti-delta is a useful diagnostic marker. However, this test is not readily available in most laboratories and repeat testing may be required. We believe that, for the present, the measurement of delta antigen by a sensitive technique is the best marker for the diagnosis of acute delta
hepatitis.’ Department of Internal Medicine, Ciudad Sanitaria Valle de Hebron, Universidad Autonoma, Barcelona, Spain 1. Buti M, Esteban R, Jardi
M. BUTI R. ESTEBAN R. JARDI J. I. ESTEBAN J. GUARDIA
R, Esteban JI, Guardia J Serological diagnosis of acute delta hepatitis. J Med Virol 1986; 1: 81-86.