216 at 78 h (normal 37+2.1 h ), and plasma clearance reduced at 1.33 ml/kg/h (normal3 3.8±0.55). Absorption halflife and apparent volume of distribution were normal. We were not able to test other drugs or any other members of the patient’s family. Previous studies in iron-deficient states have shown increased antipyrine- clearance in rats,4 but no alteration in man.5 However, there is frequently poor correlation between this and other measures of drug metabolising ability. This patient probably had grossly impaired warfarin metabolism whilst iron deficient. The plasma levels of warfarin on first administration are compatible with a half-life in excess of 200 h. It is possible therefore that while iron deficiency is not associated with impaired metabolism of the test drug antipyrine, the metabolism of other drugs may be altered. This could be due to the effect of iron deficiency on the hepatic microsomal cytochrome P-450 drug-metabolising system.
SENSITIVITY AND SPECIFICITY OF MATERNAL SERUM-A.F.P.
WEEKS OF GESTATION USING A CUT-OFF
TIMES THE NORMAL MEDIAN
We thank the late Dr G. H. H. Benham and Dr B. E. Boyes for permission to study and report this case.
University of Manchester School of Medicine, Salford M6 8HD Texas Health Science Center San Antonio,
D. P. NICHOLLS
Texas 78284, U.S.A.
Estimated as follows: 156 out of 8862 (1.8%) singleton pregnancies 16-21 weeks of gestation (the data were collected in 3-week gestational periods, so the figures for 16-20 weeks are not available), and 12 out of 64 (19%) twin pregnancies had high A.F.P. levels. These rates were combined assuming 1 pregnancy in 80 is a twin. t Estimated assuming that all twins are detected by means of ultrasound, and that 40% of singletons have negative A.F.P. results after revision of gestational age, and a further 50% have negative A.F.P. results after repeat testing. f An unknown number of women with threatened abortion were considered ineligible for amniocentesis. *
A. M. M. SHEPHERD
SCREENING FOR NEURAL-TUBE DEFECTS
SiR,—The article by Professor Ferguson-Smith and his colleagues suggests that there are inconsistencies in the results of screening for neural-tube defects derived from the Glasgow study and those that would have been expected from the U.K. Collaborative Study.2 For example, it was reported that the women in the Glasgow study who underwent amniocentesis had a four-fold higher risk of a fetal abnormality than was predicted by the collaborative study. However, the outcome of screening depends upon the serum a-fetoprotein (A.F.P.) cut-off level used and how patients are subsequently selected for amniocentesis. If this is taken into account we believe that there were no material inconsistencies between the two studies on the outcome of screening. The table shows the sensitivity and specificity of serum-A.F.p. screening for the two studies at the cut-off level used in phase n of the Glasgow study-namely, 2.8times the normal median at 16-20 weeks of gestation. (The cut-off level expressed as a multiple of the median was not given for phase I.) The data from phase ii were not included in the collaborative study as they were collected after it was completed. While the proportion of affected pregnancies with high A.F.P. levels (the screening sensitivity) in the two studies was similar the detection rate for all spina bifida (open and closed) at this cut-off level was slightly higher in the collaborative study than in the Glasgow study, and that for open spina bifida was a little lower. These small differences could be due to random error or to differences in classifying lesions as open or closed. In phase u of the Glasgow study 14/30 (47%) of spina-bifida lesions were regarded as closed, while in the collaborative study only 24/142 (17%) were closed. It would only need 4 cases of spina bifida with A.F.P. levels below 2.8 times the median regarded as closed in the Glasgow study to be regarded as open in the collaborative study-or 3 fewer cases of open spina bifida to be detected at this cut-off level in the Glasgow study-for the detection-rates to be virtually identical. 4.
The proportion of unaffected pregnancies with high A.F.P. levels (and hence screening specificity) was similar in the two studies. It was not possible to compare directly the proportions of women with unaffected pregnancies who require an amniocentesis because these data were not collected for the collaborative study. It is, however, possible to estimate this using data from the collaborative study together with those on the use of ultrasound to revise gestational age obtained from other sources. 2,3 Data from the collaborative study indicated that about 40% of the total within-week variation in A.F.P. levels per week of gestation (measured as the variance of log A.F.P.) was attributable to random assay error and within-person fluctuations in A.F.P. concentration. This means that the particular policy of repeat testing used in Glasgow would be expected to reduce the proportion of unaffected pregnancies with high A.F.P. levels by about a half. The use of ultrasound to revise gestational age will also reduce the proportion of unaffected singleton pregnancies with high A.F.P. levels, depending on the cut-off level used; at 2.0 times the median the reduction was 23%3 and at’t 2.5times the median it was 37%.4 One might expect the effects of repeat testing and correcting gestational age to act independently so that the use of both would reduce the number of women with unaffected pregnancies who actually require amniocentesis by more than a half. At a screening cut-off level of 2.8 times the median this would result in about 0.5% of such women being offered an amniocentesis. The figure from the Glasgow study was 0.3%, but this would have been higher if pregnancies with threatened abortion had been considered eligible for amniocentesis. The figures in the table which relate to the collaborative study were obtained with a single cut-off level of 2-8times the
Catz, C. S., Juchau, M. R., Yaffe, S. J. J. Pharmac. exp. Ther. 1970, 174,
197. 5. O’Malley, K., Stevenson, I.H.J. Pharm. Pharmac. 1973, 25, 339. 1. Ferguson-Smith, M. A., and others. Lancet, 1978, i, 1330. 2. U.K. Collaborative Study, ibid. 1977, i, 1323.
3. Block, D. J. H. in Prevention of Neural-Tube Defects: The role of alphafetoprotem (edited by B. F. Crandall and M. A. B. Bazier); p. 191 New 4.
York, 1978. Wald, N., Cuckle, H , Stirrat, G. Lancet, 1978, i, 495.
217 median for each week of gestation from 16 to 20 weeks. Fig. 2 in the report from Glasgow’ suggests that an average cut-off level of 2-8 times the median was used over this period so at 16 weeks it was somewhat higher (say 3-0) and at 20 weeks it was lower (say 2.7). This should not, however, make a great difference to the comparison of the results of the two studies shown in the table. The results from Glasgow’ are therefore consistent with those of the U.K. Collaborative Study.2 The expectations from any particular screening programme are dependent on the cutoff level, and this in turn will affect both the number of amniocenteses and the number of neural-tube defects detected-with a higher cut-off level fewer women would be referred for amniocentesis, but the detection-rate would also be lower. For example, less than half the number of women with unaffected pregnancies would require amniocentesis at a cut-off level of 3.0 times the median compared to 2.5 times the median at 16 to 18 weeks of gestation. At the same time, the detection-rate for all spina bifida would be reduced by 13%. In our opinion therefore it is unwise to argue that an amniocentesis-rate of less than 1% among women with unaffected pregnancies necessarily represents a better screening policy than one associated with a somewhat higher one (say, 1.5-2%). The factors which might influence the selection of cut-off levels in antenatal screening are discussed in the report of the U.K. Collaborative
screening from more than one centre to obtain better estimates of the detection-rate, and monitor the stability of the proportion of false positives. The U.K. Collaborative Study has shown that results from different centres cannot be compared directly but only by expressing them in multiples of each laboratory’s normal median; this must always be a second best, and can never tell us to what extent differences in the proportion of normal births with high serum-A.F.p. are due to assay imprecision rather than inherent differences in the screened population. A necessary condition for a national screening programme should be that the results from participating laboratories can be compared directly in terms of the same
reference preparation. Also, no such screening programme should be planned unless it is run in conjunction with a quality-control scheme to assess the precision and accuracy of the assays done in the various centres. Medical Computing and Statistics Unit, Medical School, University of Edinburgh, Edinburgh EH8 9AG
Study. I.C.R.F. Cancer Epidemiology and Clinical Trials Department of the Regius Professor of Medicine, Radcliffe Infirmary, Oxford OX2 6HE
Department of Human Genetics, Western General Hospital, Edinburgh EH4 2XU
NICHOLAS J. WALD HOWARD S. CUCKLE DAVID J. H. BROCK
Paediatric Research Unit,
Guy’s Hospital, London SE1 9RT
ENDOGENOUS ACTIVATOR OF
PAUL E. POLANI
SIR,-Professor Ferguson-Smith and his colleagues describe how their
screening of 11 500 pregnancies for raised serdetected 93% of those births with neural-tube defects, with a false-positive rate of only 1-4%. However, if their results are to be used to predict the consequences of a national screening programme, we need to know to what extent an equivalent intervention level might produce similar results in other centres, and in particular we must make an assessment of the possible biases and the standard errors of these estimates of percentages. Their detection-rate of 93% is based on the detection of 27 anencephalics out of 27, and 13 spina bifidas out of 16. These percentages should be examined separately rather than together, as there were more cases of anencephaly than spina bifida in this study although the incidence of the two defects is known to be similar; this difference in incidence-rates must itself cast some doubt on the adequacy of the follow-up in the Glasgow study. The true detection-rate for spina bifida should have been quoted as 12 out of 16, because 1 case was missed at screening and only retrospectively recognised as above the intervention limit. Thus the sensibilities of the screening test for the two defects should be given as 100% (95% confidence limits 87-100%) for anencephaly and 75% (95% confidence limit 47-92%) for spina bifida. The width of the confidence intervals is a result of the small numbers on which these percentages are based. In contrast the false-positive rate of 1.6% is based on large numbers and thus well determined for
the Glasgow population. However, this rate is likely to vary both over time, and at different centres. The intervention level of 2.8 times the median used in this study was supposed to estimate the 97th percentile for normal singleton pregnancies and hence should have produced a false-positive rate of at least 3%. These points emphasise the need to combine the results of
SiR,—The apparent concentration of renin in human plasma can be increased by dialysis to pH 3.0-3.3 followed by dialysis to pH 7.4.1-3 This increase is due to activation of an inactive form of the enzyme. It has been argued that the activating process cannot be important physiologically since plasma is not exposed to a pH of 3.0in vivo. However, inactive renin is also activated by trypsin at pH 7.4,4,5 and an endogenous protease can activate the inactive renin of amniotic fluid.5 The experiments described below show that acid-activation of inactive renin in human plasma is probably mediated by an endogenous protease. In the course of this work I learned that Dr M. A. D. H. Schalekamp and his group had obtained similar results (see accompanying letter). Plasma from normal males was pooled. Samples were assayed for renin before and after the following procedures: (a) Dialysis to pH 3.0 and back to pH 7.4 (b) Dialysis to pH 3.0 in the presence of protease inhibitors, followed by dialysis to pH 7.4. ’Trasylol’ (aprotinin, Bayer) and soybean trypsin inhibitor (S.B.T.I.) in pH 7.4 phosphate-saline buffer were added directly to 10 x the volume of plasma and to the pH 3.0 glycine/HCl dialysis buffer at the concentrations stated in the table, and the mixture was incubated for 3 min at 37°C before acid dialysis. N-ethylmaleimide and o-phenanthroline were included in the pH 3.0 glycine/HCl buffer.3 (c) Plasma was also treated as in (a) and then assayed for renin in the presence of trasylol and S.B.T.I.
The kallikrein inhibitor trasylol and S.B.T.I. partially inhibited the acid-induced increase in renin concentration (table). Increasing concentrations of inhibitor increased its effect. N-ethyimaleimide had some inhibitory effect but o-phenanthroline had no effect. Neither trasylol nor S.B.T.I. affected the reaction between renin and renin substrate. These results suggest that acid-activation of inactive renin in plasma is mediated by a serine protease. The increase in plasma-renin 1. 2.
Skinner, S. L., Lumbers, E. R., Symonds, E. M. Clin. Sci. 1972, 42, 479. Leckie, B. J., McConnell, A.J. Endocr. 1975, 65, 7p. 3. Leckie, B. J., McConnell, A., Grant, J., Morton, J. J., Tree, M., Brown, J. J. Circulation Res. 1977, 40, suppl. 1, p. 46. 4. Cooper, R. M., Osmond, D. H., Scaiff, K. D., Ross, L. J. Fedn Proc. 1974,
33, 584. 5. Leckie, B. J., McConnell, A., Jordan, J. Adv. exp. Med. Biol. 1977, 95, 249. Morris, B. J., Lumbers, E. R., Biochim biophys. Acta. 1972, 289, 385. Millar, J. A., Leckie, B., J., Semple, P. F., Morton, J. J., Sonkodi, S., Robertson, J. I. S. Circulation Res. (in the press). 8. Osmond, D. H., Loh, A. Y. Lancet, 1978, i, 102.