- Email: [email protected]
PHENOBARBITONE AND LIVER TUMOURS
1085
Experiments
in progress with the
morphine antagonist,
naloxone, may provide additional information as to whether or not
there is
more
than
a
simple correlation between the
midbrain-thalamic cyclic-A.M.P. rise and the
onset
of
morphine analgesia.
Not only do we need good animal experimentation, and good epidemiology, we also need careful, conservative statistical analysis. National Cancer Institute, National Institutes of Health,
Bethesda, Maryland 20014,
Department of Psychiatry, U.C.L.A. Medical School, and
U.S.A.
Neurobiochemistry Laboratory, Brentwood Hospital,
GARY L. BRAMMER MICHAEL I. PAUL.
Veterans Administration, Los Angeles, California, U.S.A.
PHENOBARBITONE AND LIVER TUMOURS SiR,-My interpretation of Clemmesen’s data1 differs from yours (Sept. 14, p. 629). The liver-cancer deaths reported by Clemmesen were twice expectation. Writing about this twofold excess (5 observed compared to 2-5 expected "), Clemmesen remarked, " The slight excess for liver cancer is not statistically significant." He says this because he knows " statistically significant " does not mean " no difference ". He did not say, as your leader " no increase of liver cancer ". As for says, that there is statistical significance, the probability of having 5 or more cases when 2-5 are expected is 0-11. Usually 0-05 is used for the cut-off for statistical significance. One more case of liver cancer in that population, where underreporting is certainly a possibility, would lead to statistical significance. Other data in the Clemmesen report lead me to believe that his explanation for the less-than-expected number of cancer deaths (i.e., institutionalisation per se is followed by fewer reported cancer deaths) is more reasonable than the one offered by your leader. Both breast-cancer and cervical-cancer deaths were significantly below expectation. In an institutionalised population, likely to have less sexual intercourse than the general population, it is not surprising In a population with a to find low cervical-cancer rates. low cervical-cancer rate one usually finds high breast-cancer rates (see the studies on nuns, for example) rather than the converse. The explanation usually accepted for the lessthan-expected overall cancer deaths in an institutional population is lower risk due to being in an institution plus occasional underreporting of deaths ascribed to one cause in a population known to have another, sometimes fatal, illness. Accepting the common finding of lower cancer risk in an institutionalised population leads to asking the slightly different statistical question, " was the proportion of cancer deaths attributed to liver cancer in keeping with expectation ? (a recent paper2 in The Lancet asked this about an industrial population). The answer one gets is opposite from the conclusion in the leader-i.e., the excess in liver cancers is statistically significant. The arithmetic goes this way, after excluding the observed cases of brain tumours. Equate the total number of cases expected to the number of cases observed (this enables one to compare the proportions). One now " expects " 1-88 cases of liver cancer, rather than 2-5-a ratio of 2-7 to 1. Using the Poisson distribution for rare events, to test this difference, I find the probability of 5 or more cases, given the true " value of 188, to be 0-042, which is less than the conventional 0-05 level for " significance ". One would then call this excess " statistically significant ". This interpretation of the data from the epileptic colony is now consistent with phenobarbitone or one of the other anticonvulsants (and at least three were used) being a liver carcinogen in man. With the conclusion of statistical significance depending on how the question is asked, your conclusion that phenobarbitone does not cause cancer in man is highly questionable. "
MARVIN A. SCHNEIDERMAN.
FOOD ANTIBODIES AND MYOCARDIAL INFARCTION
SIR,-We thank Dr Galen for his letter (Oct. 5, p. 832) in which he casts doubt on the prognostic value of the detection of antibodies to cow’s milk protein and egg-white in men who have had a myocardial infarct. At’the same time, Dr Galen must realise that the prognostic value of other data which is often used in the assessment of individual patients, and sometimes used as a basis for advice, is very low indeed. Applying the method of Holland and Whitehead1 to data for similar groups of men gives the following results. Morris et al .2 published data for men followed for two to six years after a myocardial infarct. They state that " there is no general tendency for (death or re-infarction) rates to be higher in the men with higher initial cholesterol levels ". In fact, their data show that in 60 men with the highest cholesterol levels (over 324 mg. per 100 ml.) prediction of death was 22% and efficiency (considering cholesterol levels as a dichotomy above and below 324 mg. per 100 ml.) was 66%. Data of Mather et al. 3 indicate that the predictive value of a previous history of cardiovascular disease for death within 28 days is 19% and the efficiency 43%. The predictive value of smoking before infarction can be estimated from the data for men studied by a group of physicians in Newcastle upon Tyne.4 For death during the following five years this was 11% and efficiency 55%. Most of these values are in fact lower than those derived from our data for the presence of antibodies to milk and egg, namely, predictive values for death within six months of 27% and 28% and efficiencies of 58% and 70%. Holland and Whitehead are concerned with the relevance of a test to an individual patient. Clearly the misuse of their approach would condemn most experimental evidence on factors of possible relevance to aetiology. M.R.C. Epidemiology Unit
(South Wales), Road,
4 Richmond
Cardiff CF2 3AS.
P. C. ELWOOD D. F. DAVIES.
"
"
"
"
PROTEIN AND HEART-DISEASE
SIR,—Medical thought is turning towards a missing link or " third force " in the incomplete picture we have of coronary atherogenesis. Laboratory and clinical work is being undertaken beyond the fat/sugar platform to explore the effects in heart-disease of animal-protein foods which are known to have immunological importance. A clinical study has previously been conducted into the results of the withdrawal of specific protein foods from the diet of angina sufferers, with striking benefit to the patient, and it is proposed to extend this work internationally to obviate local factors and establish further evidence. I am seeking the participation of any physician who has care of patients with angina pectoris, no matter the number. 1. 2. 3.
1.
Clemmesen, J., Frederikson, V. F., Plum, C. M. Lancet, 1974, i,
705. 2. Monson, R.
p. 397.
R., Peters, J. M., Johnson, M. N. ibid. Aug. 17, 1974, 4.
Holland, W. W., Whitehead, T. P. Lancet, Aug. 17, 1974, p. 391. Report of a Research Committee to the Medical Research Council. ibid. 1968, ii, 693. Mather, H. G., Pearson, N. G., Read, K. L. Q., Shaw, D. B., Steed, G. R., Thorne, M. G., Jones, S., Guerrier, C. J., Erault, C. D., McHugh, P. M., Choudhury, N. R., Jafary, M. H., Wallace, T. J. Br. med. J. 1971, iii, 334. ibid. 1971, iv, 767.
Experiments
in progress with the
morphine antagonist,
naloxone, may provide additional information as to whether or not
there is
more
than
a
simple correlation between the
midbrain-thalamic cyclic-A.M.P. rise and the
onset
of
morphine analgesia.
Not only do we need good animal experimentation, and good epidemiology, we also need careful, conservative statistical analysis. National Cancer Institute, National Institutes of Health,
Bethesda, Maryland 20014,
Department of Psychiatry, U.C.L.A. Medical School, and
U.S.A.
Neurobiochemistry Laboratory, Brentwood Hospital,
GARY L. BRAMMER MICHAEL I. PAUL.
Veterans Administration, Los Angeles, California, U.S.A.
PHENOBARBITONE AND LIVER TUMOURS SiR,-My interpretation of Clemmesen’s data1 differs from yours (Sept. 14, p. 629). The liver-cancer deaths reported by Clemmesen were twice expectation. Writing about this twofold excess (5 observed compared to 2-5 expected "), Clemmesen remarked, " The slight excess for liver cancer is not statistically significant." He says this because he knows " statistically significant " does not mean " no difference ". He did not say, as your leader " no increase of liver cancer ". As for says, that there is statistical significance, the probability of having 5 or more cases when 2-5 are expected is 0-11. Usually 0-05 is used for the cut-off for statistical significance. One more case of liver cancer in that population, where underreporting is certainly a possibility, would lead to statistical significance. Other data in the Clemmesen report lead me to believe that his explanation for the less-than-expected number of cancer deaths (i.e., institutionalisation per se is followed by fewer reported cancer deaths) is more reasonable than the one offered by your leader. Both breast-cancer and cervical-cancer deaths were significantly below expectation. In an institutionalised population, likely to have less sexual intercourse than the general population, it is not surprising In a population with a to find low cervical-cancer rates. low cervical-cancer rate one usually finds high breast-cancer rates (see the studies on nuns, for example) rather than the converse. The explanation usually accepted for the lessthan-expected overall cancer deaths in an institutional population is lower risk due to being in an institution plus occasional underreporting of deaths ascribed to one cause in a population known to have another, sometimes fatal, illness. Accepting the common finding of lower cancer risk in an institutionalised population leads to asking the slightly different statistical question, " was the proportion of cancer deaths attributed to liver cancer in keeping with expectation ? (a recent paper2 in The Lancet asked this about an industrial population). The answer one gets is opposite from the conclusion in the leader-i.e., the excess in liver cancers is statistically significant. The arithmetic goes this way, after excluding the observed cases of brain tumours. Equate the total number of cases expected to the number of cases observed (this enables one to compare the proportions). One now " expects " 1-88 cases of liver cancer, rather than 2-5-a ratio of 2-7 to 1. Using the Poisson distribution for rare events, to test this difference, I find the probability of 5 or more cases, given the true " value of 188, to be 0-042, which is less than the conventional 0-05 level for " significance ". One would then call this excess " statistically significant ". This interpretation of the data from the epileptic colony is now consistent with phenobarbitone or one of the other anticonvulsants (and at least three were used) being a liver carcinogen in man. With the conclusion of statistical significance depending on how the question is asked, your conclusion that phenobarbitone does not cause cancer in man is highly questionable. "
MARVIN A. SCHNEIDERMAN.
FOOD ANTIBODIES AND MYOCARDIAL INFARCTION
SIR,-We thank Dr Galen for his letter (Oct. 5, p. 832) in which he casts doubt on the prognostic value of the detection of antibodies to cow’s milk protein and egg-white in men who have had a myocardial infarct. At’the same time, Dr Galen must realise that the prognostic value of other data which is often used in the assessment of individual patients, and sometimes used as a basis for advice, is very low indeed. Applying the method of Holland and Whitehead1 to data for similar groups of men gives the following results. Morris et al .2 published data for men followed for two to six years after a myocardial infarct. They state that " there is no general tendency for (death or re-infarction) rates to be higher in the men with higher initial cholesterol levels ". In fact, their data show that in 60 men with the highest cholesterol levels (over 324 mg. per 100 ml.) prediction of death was 22% and efficiency (considering cholesterol levels as a dichotomy above and below 324 mg. per 100 ml.) was 66%. Data of Mather et al. 3 indicate that the predictive value of a previous history of cardiovascular disease for death within 28 days is 19% and the efficiency 43%. The predictive value of smoking before infarction can be estimated from the data for men studied by a group of physicians in Newcastle upon Tyne.4 For death during the following five years this was 11% and efficiency 55%. Most of these values are in fact lower than those derived from our data for the presence of antibodies to milk and egg, namely, predictive values for death within six months of 27% and 28% and efficiencies of 58% and 70%. Holland and Whitehead are concerned with the relevance of a test to an individual patient. Clearly the misuse of their approach would condemn most experimental evidence on factors of possible relevance to aetiology. M.R.C. Epidemiology Unit
(South Wales), Road,
4 Richmond
Cardiff CF2 3AS.
P. C. ELWOOD D. F. DAVIES.
"
"
"
"
PROTEIN AND HEART-DISEASE
SIR,—Medical thought is turning towards a missing link or " third force " in the incomplete picture we have of coronary atherogenesis. Laboratory and clinical work is being undertaken beyond the fat/sugar platform to explore the effects in heart-disease of animal-protein foods which are known to have immunological importance. A clinical study has previously been conducted into the results of the withdrawal of specific protein foods from the diet of angina sufferers, with striking benefit to the patient, and it is proposed to extend this work internationally to obviate local factors and establish further evidence. I am seeking the participation of any physician who has care of patients with angina pectoris, no matter the number. 1. 2. 3.
1.
Clemmesen, J., Frederikson, V. F., Plum, C. M. Lancet, 1974, i,
705. 2. Monson, R.
p. 397.
R., Peters, J. M., Johnson, M. N. ibid. Aug. 17, 1974, 4.
Holland, W. W., Whitehead, T. P. Lancet, Aug. 17, 1974, p. 391. Report of a Research Committee to the Medical Research Council. ibid. 1968, ii, 693. Mather, H. G., Pearson, N. G., Read, K. L. Q., Shaw, D. B., Steed, G. R., Thorne, M. G., Jones, S., Guerrier, C. J., Erault, C. D., McHugh, P. M., Choudhury, N. R., Jafary, M. H., Wallace, T. J. Br. med. J. 1971, iii, 334. ibid. 1971, iv, 767.