874
perhaps, the R.F. status of Roitt’s patients was the overriding factor, the DRw4 providing no further information, in accord with the data of Husby et al. Department of Medicine, Stanford University Medical Center, Stanford, California 94305, U.S.A.
ANDREI CALIN
numbers of patients are studied. across studies by drug regimen and posture is a statistically invalid approach to this requirement. Even if there were a linear relation between P.T.D.P. and AD.P., it may largely represent no more than regression towards the mean.2
only
prove useful if
large
However, the pooling of data
Departments of Clinical Pharmacology and Medical Statistics, Royal Postgraduate Medical School,
ANALYSIS OF EFFICACY OF ANTIHYPERTENSIVE DRUGS
London W12 OHS
GARRET FITZGERALD AVIVA PETRIE
SIR,-It has been suggested’ that pretreatment diastolic
(P.T.D.P.) and the fall in diastolic pressure (AD.P.) induced by antihypertensive therapy are linearly related. The "ideal" response line would then have a slope of 1.0 and would intercept the P.T.D.P. axis at 90 mm Hg. Dr Wilson and Professor Schumacher (March 24, p. 656) have tested this proposal, using data from forty-two published studies. They propose the coefficient of determination (r2) as an index of drug efficacy and that the width of the confidence intervals for the slope and for the P.T.D.P. intercept reflect the variability of patient re-
pressure
sponsiveness to a given regimen. Such inferences rest heavily on the assumption that a linear model accurately represents the relationship between P.T.D.P. and AD.P. Dixon and Johnson’ supported this claim with data plots whose regressions of AD.P. on P.T.D.P. appeared close to the ideal-response line. The first of these related to supine pressures and included points from four independent studies, pooled for regression analysis. The second plot was of the standing pressures from one of these studies. The correct approach to the data for supine patients would have been to analyse each study separately, in which case the approximation to the ideal-response line would have been much less impressive. The only study illustrated in the standing pressure plot corresponds poorly to the "ideal-response line" in the supine position. Dixon and Johnson then proceeded to apply their hypothesis to a study where such a pattern manifestly did not exist. Although central to the hypothesis, they present no statistical analysis on the relation of the observed to the "ideal" response lines. Because the hypothesis was based on such questionable material, it is hardly surprising that Wilson and Schumacher found that most of their "data sets" failed to fit the idealresponse line. However, instead of questioning the underlying model, they proceeded to draw inferences from it. The fallacy of this approach is shown in the table, abstracted from their data. There is no biological evidence that the antihypertensive response to propranolol is less marked and more variable in the sitting than in the standing or supine positions. The relatively high coefficient of correlation of clonidine in the sitting position may merely reflect the inappropriate application of a least-squares analysis to data from three subjects. Dixon and Johnson1 suggested that their approach might 1 Dixon, G T, Johnson, E. S. Lancet, 1976, i, 515. 2. Peto, R. Proc 9th int. Biomet. Conf. 1976, 2, 3
GUT REACTION TO ANTIGEN
SIR,-Your excellent editorial (April 7) suggests that T cells may be stimulated to assist IgA production in the gut wall by dietary antigens following oral pre-immunisation. One of the consequences of pre-immunisation is reduced transport of intact antigen, which has been observed in mice for human albumini as well as ovalbumin2 and in 100-day-old rats, but not in 40-day-old rats, for bovine IgG.3 All these experiments are concerned with the effect of a single oral immunising dose of antigen. But when the antigen is fed continuously, complete tolerance may result, in so far as the ability to produce circulating antibody, presumably of IgG and IgM, in response to parenteral immunisation is suppressed.4·5 The mechanism of this may be that IgA-antigen complexes have been shown to be
tolerogenic.6 Uptake of orally administered antigen may be reduced by oral pre-immunisation, but it is not abolished, nor is it as small as you suggest. 10 of an oral dose of bovine IgG may be demonstrated by simple Ouchterlony diffusion in the tissues of an adult rat a few hours after feeding.’ More than this may be present as large-molecular-weight breakdown products which are no longer able to react as divalent antigen, but are demonstrable by the use of tracers.8 It would seem that it is profitable for the animal to abolish its systemic immune response to these products of digestion, which flood into the circulation after every protein meal. This enables those products to reach the tissues throughout the body, where presumably the process of digestion is completed. A.R.C.
Immunology Group, Department of Zoology, Brambell Laboratories, University College of North Wales, Bangor, Gwynedd LL57 2UW
Andre, C., Lambert, R., Bazin, H., Heremans, J. F. Eur. J. Immun. 1974. 4, 701 2 Swarbrick, E. T, Stokes, C. R, Soothill, J. F Gut, 1979, 20, 121. 3 Hemmings. C., Hemmings, W. A., Patey. A. L., Wood, C. Proc R Soc B 1.
1977, 198, 439. 4. 5.
Wells, H. G., Osborne, T B. J infect. Dis. 1911, 8, 66. Bazin, H. in Protein Transmission through Living Membranes (edited by A.
between regression line and
6. André, C. Vaerman, J P. Heremans, J F in Antigen Absorption (edited by W. A. Hemmings), p. 65. Lancaster, 1978. 7. Williams, E. W. PH.D thesis, University of Wales, 1978. 8. Hemmings, W. A., Williams, E W Gut, 1978, 19, 715.
ideal-response line.
W
Hemmings, Amsterdam (in the press)).
EFFECT OF POSTURE ON RESPONSE TO ANTIHYPERTENSIVE DRUGS
*No significant difference I P
perhaps, the R.F. status of Roitt’s patients was the overriding factor, the DRw4 providing no further information, in accord with the data of Husby et al. Department of Medicine, Stanford University Medical Center, Stanford, California 94305, U.S.A.
ANDREI CALIN
numbers of patients are studied. across studies by drug regimen and posture is a statistically invalid approach to this requirement. Even if there were a linear relation between P.T.D.P. and AD.P., it may largely represent no more than regression towards the mean.2
only
prove useful if
large
However, the pooling of data
Departments of Clinical Pharmacology and Medical Statistics, Royal Postgraduate Medical School,
ANALYSIS OF EFFICACY OF ANTIHYPERTENSIVE DRUGS
London W12 OHS
GARRET FITZGERALD AVIVA PETRIE
SIR,-It has been suggested’ that pretreatment diastolic
(P.T.D.P.) and the fall in diastolic pressure (AD.P.) induced by antihypertensive therapy are linearly related. The "ideal" response line would then have a slope of 1.0 and would intercept the P.T.D.P. axis at 90 mm Hg. Dr Wilson and Professor Schumacher (March 24, p. 656) have tested this proposal, using data from forty-two published studies. They propose the coefficient of determination (r2) as an index of drug efficacy and that the width of the confidence intervals for the slope and for the P.T.D.P. intercept reflect the variability of patient re-
pressure
sponsiveness to a given regimen. Such inferences rest heavily on the assumption that a linear model accurately represents the relationship between P.T.D.P. and AD.P. Dixon and Johnson’ supported this claim with data plots whose regressions of AD.P. on P.T.D.P. appeared close to the ideal-response line. The first of these related to supine pressures and included points from four independent studies, pooled for regression analysis. The second plot was of the standing pressures from one of these studies. The correct approach to the data for supine patients would have been to analyse each study separately, in which case the approximation to the ideal-response line would have been much less impressive. The only study illustrated in the standing pressure plot corresponds poorly to the "ideal-response line" in the supine position. Dixon and Johnson then proceeded to apply their hypothesis to a study where such a pattern manifestly did not exist. Although central to the hypothesis, they present no statistical analysis on the relation of the observed to the "ideal" response lines. Because the hypothesis was based on such questionable material, it is hardly surprising that Wilson and Schumacher found that most of their "data sets" failed to fit the idealresponse line. However, instead of questioning the underlying model, they proceeded to draw inferences from it. The fallacy of this approach is shown in the table, abstracted from their data. There is no biological evidence that the antihypertensive response to propranolol is less marked and more variable in the sitting than in the standing or supine positions. The relatively high coefficient of correlation of clonidine in the sitting position may merely reflect the inappropriate application of a least-squares analysis to data from three subjects. Dixon and Johnson1 suggested that their approach might 1 Dixon, G T, Johnson, E. S. Lancet, 1976, i, 515. 2. Peto, R. Proc 9th int. Biomet. Conf. 1976, 2, 3
GUT REACTION TO ANTIGEN
SIR,-Your excellent editorial (April 7) suggests that T cells may be stimulated to assist IgA production in the gut wall by dietary antigens following oral pre-immunisation. One of the consequences of pre-immunisation is reduced transport of intact antigen, which has been observed in mice for human albumini as well as ovalbumin2 and in 100-day-old rats, but not in 40-day-old rats, for bovine IgG.3 All these experiments are concerned with the effect of a single oral immunising dose of antigen. But when the antigen is fed continuously, complete tolerance may result, in so far as the ability to produce circulating antibody, presumably of IgG and IgM, in response to parenteral immunisation is suppressed.4·5 The mechanism of this may be that IgA-antigen complexes have been shown to be
tolerogenic.6 Uptake of orally administered antigen may be reduced by oral pre-immunisation, but it is not abolished, nor is it as small as you suggest. 10 of an oral dose of bovine IgG may be demonstrated by simple Ouchterlony diffusion in the tissues of an adult rat a few hours after feeding.’ More than this may be present as large-molecular-weight breakdown products which are no longer able to react as divalent antigen, but are demonstrable by the use of tracers.8 It would seem that it is profitable for the animal to abolish its systemic immune response to these products of digestion, which flood into the circulation after every protein meal. This enables those products to reach the tissues throughout the body, where presumably the process of digestion is completed. A.R.C.
Immunology Group, Department of Zoology, Brambell Laboratories, University College of North Wales, Bangor, Gwynedd LL57 2UW
Andre, C., Lambert, R., Bazin, H., Heremans, J. F. Eur. J. Immun. 1974. 4, 701 2 Swarbrick, E. T, Stokes, C. R, Soothill, J. F Gut, 1979, 20, 121. 3 Hemmings. C., Hemmings, W. A., Patey. A. L., Wood, C. Proc R Soc B 1.
1977, 198, 439. 4. 5.
Wells, H. G., Osborne, T B. J infect. Dis. 1911, 8, 66. Bazin, H. in Protein Transmission through Living Membranes (edited by A.
between regression line and
6. André, C. Vaerman, J P. Heremans, J F in Antigen Absorption (edited by W. A. Hemmings), p. 65. Lancaster, 1978. 7. Williams, E. W. PH.D thesis, University of Wales, 1978. 8. Hemmings, W. A., Williams, E W Gut, 1978, 19, 715.
ideal-response line.
W
Hemmings, Amsterdam (in the press)).
EFFECT OF POSTURE ON RESPONSE TO ANTIHYPERTENSIVE DRUGS
*No significant difference I P
