- Email: [email protected]
CANCER MORTALITY AND FLUORIDATION
150 CANCER MORTALITY AND FLUORIDATION
SIR,-On June 18, 1976, Dr Robert N. Hoover sent the National Cancer Institute data on cancer mortality and fluoridation to Dr L. J. Kinlen. In the accompanying letter he wrote: "If you are queried as to how you obtained the data, I would appreciate if you would indicate that all of the raw data are available from routine publications available to anyone." Doll and Kinlen’ did just that. They used the N.C.I.’s figures without alteration and claimed that they obtained their data from "routine publications". Unfortunately for Doll and Kinlen, the N.C.I, made a serious mathematical or tabulation error which was repeated in their paper.’1 The errors in the N.C.I./Doll-Kinlen data were discovered by Dr Dean Burk and myself and transmitted to The Lancet on Sept. 5, 1977 and to the N.C.I. shortly before the Congressional hearing in September/October investigating an alleged cover-up of the fluoridation-cancer link. Dr Hoover admitted the error and its duplication on Sept. 26, 1977, in a letter to Dr Kinlen in which he wrote: "I am sorry for this error, particularly since it seems to have been perpetuated by yourselves and the Royal Statistical Society." Dr P. D. Oldham uncovered N.C.I.’s error yet in his paper for the R.S.S. with Prof. D. J. Newell he used the erroneous data.2 More important than the mathematical errors passed on by the N.C.I. (which account for about 3500 cancer deaths a year) are the methodological errors that were also spoon-fed to Oldham and Newell and Doll and Kinlen by the N.C.I. (and which account for about 9000 cancer deaths per year). These include omission of 90-95% of the data available.and the selection as one of their data points a year (1970) during which fluoridation of the control group had already started. Using correct figures, linear regression analysis of all available cancer death-rate data (unweighted) for years in which the control cities were not fluoridated (1950 to 1968 with extrapolation to 1970), and a reference population that best reflects the cancer death-rate distribution in central cities, we have recalculated Doll and Kinlen’s table III as follows:
members of Parliament, health authorities, and water boards as evidence that fluoridation was causing many deaths from cancer. Because we wanted to be sure about the truth of the matter, and because we feared that Burk and Yiamouyiannis’s abuse of statistics might be detrimental to the future health of British children, we asked the N.C.I. to provide us with the relevant population and mortality data for the twenty cities so that we could examine the facts ourselves. In providing this information, Dr Hoover of the N.C.I. pointed out that these data were all publicly available. He also told us, in the letter from which Yiamouyiannis quotes, that "we purposely used only published data in doing the reanalysis, so we could not be accused of concealing data. Therefore, all that is necessary for anyone in this country to check our analyses is to be aware of basic principles of biostatistics and epidemiology, and have a public library card. I recognise, however, that this same ease of access to the relevant data do not exist to people in other countries. I have therefore enclosed the relevant numbers..." We consulted the sources used by the N.C.I. because we wanted to use national age, sex, and ethnic group specific mortality-rates for two years that had not been used in the American analysis (1960 and 1970). We did not, however, check all the populations and numbers of deaths in each of the cities, and in consequence perpetuated the N.C.I.’s error involving 215 deaths in non-fluoridated cities. We have already drawn attention to this error,’ when we also pointed out that it did not in any way affect our conclusions. Yiamouyiannis criticises us for choosing to make comparisons for 1970 when fluoridation had been started in two of the "non-fluoridated" cities shortly before. We preferred to do so, however, because 1970 was a census year for which accurate population figures were available. To suppose that this advantage is outweighed by the deaths that might possibly have been caused in the two "non-fluoridated cities" within months of the start of fluoridation is most implausible. This supposition can, however, be tested simply by excluding the two control cities (Atlanta and Seattle) that were fluoridated in 1969 and 1970 from the analyses. The results are unaffected: the number of cancer deaths in 1950 and 1970 were respectively 9835 and 12 229, producing observed-to-expected ratios for the non-fluoridated cities of 1-15 and 1.17, respectively-preto
"
cisely the same as in our previous analysis. the direct method, a 4-5% increase in age-adjusted cancer death-rates in the fluoridated cities over those of the non-fluoridated cities was observed.
Using
,
National Health Federation, Delaware, Ohio 43015, U.S.A.
JOHN YIAMOUYIANNIS
further than we have above on the in the last paragraph of Yiamouyiannis’s letter, since it is not clear what he has done. Data are not available that would permit regression analyses based on data for each year, that take demographic changes adequately into account. We
cannot comment
analysis
Department of Regius Professor of Medicine, Radcliffe Infirmary, Oxford OX2 6HE
RICHARD DOLL LEO KINLEN
SIR, -Dr Yiamouyiannis alleges that
our analysis of cancermortality rates in twenty cities in the U.S. was "spoon-fed" to us by the National Cancer Institute and that it contained "methodological errors" as well as minor numerical errors. The first allegation is untrue; the second is a matter of opinion. In our view, the method we used (which was used independently by the Royal Statistical Society) was correct; the method Burk and Yiamouyiannis used-and Yiamouyiannis
continues to use in the above letter-is incorrect, because it fails to take into account adequately the demographic changes in the age, sex, and ethnic group constitution of the populations of the cities. The preparation of our paper was prompted by the concern aroused at the wide publicity that Burk and Yiamouyiannis had sought and obtained in Britain for their misleading use of crude cancer-mortality rates in fluoridated and non-fluoridated U.S. cities. Through the agency of the National Anti-Fluoridation Campaign (and perhaps others) these data were circulated 1. 2.
Doll, R., Kinlen, L. Lancet, 1977, i, 1300. Oldham, P. D., Newell, D. J. appl. Statist. 1977, 26, 125.
Six,-Dr Yiamouyiannis struggles desperately to salvage his argument that artificial fluoridation and cancer are associated. He now produces a new table whose construction is entirely obscure. He says "Using the direct method ...", which I take to imply "... of standardisation". That is to say, he implies that he now has sex and race specific death-rates in sufficiently narrow age-groups that these can be validly applied to a reference population. I hope he will publish these so that everyone interested may examine them in any way they wish. In view of Yiamouyiannis’s remarks about our paper I must make it clear that I and Professor Newell were asked by the Council of the Royal Statistical Society in July, 1976, to examine the evidence about cancer and fluoridation so that the Royal Statistical Society could give a formal opinion to the Royal College of Physicians. That evidence included population and mortality statistics provided by the National Cancer Institute and a variety of papers by Burk, Yiamouyiannis, and 1.
Kinlen, L., Doll, R. Lancet, 1977, ii, 1039.
151 others. Our report, written in January, 1977, was based on this evidence, as it stated. As Yiamouyiannis says, the vital statistical data were said by an officer of the N.C.I. to be "available from routine publications available to anyone". Through the courtesy of the Registrar General and his staff I was able, in April, 1977, after our paper was printed, to verify that this was indeed true in the U.K., but as I said in my letter to Professor Newell, mentioned by Yiamouyiannis, the exact definition of the U.S. populations used for the calculation of rates was outside my competence, and I was even unsure that what seemed to be a discrepancy of 215 deaths-out of 14 487 (Yiamouyiannis’s "serious error")-was not attributable to my unfamiliarity with U.S. demography. Yiamouyiannis should set aside his apparent conviction that there is a conspiracy to obscure the truth rather than a recognition of the fallacies in his arguments, and if he really believes his hypothesis is still tenable, let him devote his energies to gathering data capable of providing new evidence. If he has cancer deaths by age, sex, and race in the twenty cities let us see them as they are, not manipulated by "linear regression analysis" or whatever. M.R.C. Pneumoconiosis Unit,
Llandough Hospital,
P. D. OLDHAM
Penarth, Glamorgan
STABILITY OF REFERENCE THROMBOPLASTINS
SIR,-Dr Bangham and evidence2
Dr
Loeliger’
attempt
to
refute
our
reference instability of the N.I.B.S.&C. thromboplastins 67/40 and 69/223. They provide results of on
the
small exercises done four years apart without definition of the calibration constant. Comment can, therefore, only be made on their normal values. Their normal control values of 67/40 have lengthened significantly, but this, in our view, supports our conclusion. Bangham and Loeliger explain this trend by changes in test conditions. Whatever the interpretation, their results are hardly evidence of stability. Contrary to their supposition, our test system remained unchanged; the same technique was used throughout by one of us. Furthermore, the greatest change in materials occurred after 1974, and during the N.I.B.S.&C. this same period, the two lyophilised preparations prepared at the National (U.K.) Reference Laboratory, which were studied in parallel, showed no evidence of change, and this indicates the reliability of our test system. Bangham and Loeliger’s statement that we chose the levels of significance to support the claim that 67/40 and 69/223 are unstable is unjustified. Any statistical test takes into account the variability of the data, and there are no grounds for saying this will lead to a falsely high level of statistical significance. There is no evidence for Bangham and Loeliger’s assertion that thromboplastins do not show the variability we have described for the N.I.B.S.&C. materials. We suggest that, before reference any organisation proposes preparations, a long-term project, similar to ours on the research and development of lyophilised thromboplastin for clinical anticoagulant control, is two
necessary. The calibration graph given by Loeliger3 is misleading. He compares the results for lyophilised B.C.T. and 67/40 obtained in his own laboratory in 1976 with the results for 69/233 and 67/40 obtained by us five years previously. As the points fit the same. line, he believes the results show that 67/40 behaves similarly in the two laboratories. It is invalid to plot 69/223 and B.c.T. on the same axis because these two reagents have always been shown to be different in their prothrombin-time results. 69/223 gave longer prothrombin-times than B.c.T. even before it deteriorated. The fact that the results with these two differ1. Bangham, D. R., Loeliger, E. A. Lancet, 1977, ii, 1280. 2. Poller, L, Thomson, J. M., Yee, K. F. ibid. p. 1019. 3. Loeliger, E. A. Ibid. p. 1279.
ent reagents fit the same line after an interval of five years could be regarded as further evidence of the deterioration of
67/40. We conclude that introduction of a new international system of anticoagulant control, in which all reagents will be based on one master preparation, requires proof of the stability of this reagent. The N.I.B.S.&C. preparations, on present evidence, are unsuitable. Reliance on B.c.T. should continue. National (U.K.) Reference Laboratory for Anticoagulant Reagents and Control,
Withington Hospital, Manchester M20 8LR
L. POLLER JEAN M. THOMSON K. F. YEE
StR,—Dr Loeliger’ wrongly implies that the British system for anticoagulant control uses prothrombin activities derived from saline dilution curves. The British system is based on prothrombin ratios obtained with the standardised thromboplastin B.C.T./M.C.R. Prothrombin activities are merely given as equivalent values. The need for the N.I.B.S.&C. reagents as reference preparations was never adequately established and now appears even more unnecessary in view of the widespread international adoption of British Comparative Thromboplastin (B.c.T.) to calibrate national reference preparations, commercial reagents, and local reagents. The need for a multiplicity of thromboplastins has not been demonstrated. There has been more characterisation of B.c.T. than of any other reagent. Its properties are widely known from its use in routine and reference work for many years. Uniformity between batches is guaranteed by the assessment provided by the British Anticoagulant Panel. The monitoring procedurez2 and production details34 have been published. Korsan-Bengsten et al.5 have demonstrated how a commercial extract (’Simplastin A’) has been successfully correlated with B.c.T. over the years. Some thromboplastins are insensitive to clotting factors reduced by oral coumarin drugs and it may, therefore, be difficult to calibrate these. The answer lies in collaborating with commercial manufacturers to improve inferior reagents. It would be a mistake to encourage the use of reagents which may result in a lack of clinical safety in anticoagulant dosage by the provision of reference materials relatively insensitive to vitamin-K-dependent clotting factors. A therapeutic range with a thromboplastin extract is obtained from clinical experience or by calibration against a proved reagent. Both approaches led to the present recommended range of ratios of 2-0-0 for B.C.T. The range of ratios of 1.8-3.0, previously recommended, is alleged by Loeliger to provide inadequate thrombosis prophylaxis. His own reference provides no basis for this statement. Loeliger’s view is, moreover, at variance with the experience of British hospitals and of our colleagues overseas. Indeed, a recent clinical trial shows that the use of the conservative range of 2.0 to 2.55 with B.c.T. gives absolute safety in prevention of postoperative thrombosis in a moderate-risk group.6 The therapeutic range with B.C.T. was revised by the British Anticoagulant Panel because of increased confidence in the lack of heemorrhagic complications with the higher ratios over the years,7-9 not because the old range was believed to be inadequate in the prevention of thrombosis. One of the main criticisms of 67/40 and 69/223 is that they have not been evaluated in clinical work. From comparative 1. Loeliger, E. A. Lancet, 1977, ii, 1279. 2. Hills, M., Ingram, G. I. C. Br. J. Hœmat. 1973, 25, 455. 3. Poller, L. Med. Lab. Tech. 1972, 29, 280. 4. Poller, L., Ingram, G. I. C. Quality Control in Hæmatology;
1975.
5.
Korsan-Bengsten, K., Johnson, M., Pehrsson, 1977, 37, 98.
p. 153. London,
N-G. Thrombos. Hœmostas.
6. Taberner, D. A., Poller, L., Burslem, R. W. B., Jones, J. B. Br. med. J. the press). 7. Barkhan, P. Thrombos. Hœmostas. 1976, 35, 751. 8. Poller, L. ibid. 1976, 36, 485. 9. Blackburn, E. K. Prescribers’ J. 1977, p. 17.
(in
SIR,-On June 18, 1976, Dr Robert N. Hoover sent the National Cancer Institute data on cancer mortality and fluoridation to Dr L. J. Kinlen. In the accompanying letter he wrote: "If you are queried as to how you obtained the data, I would appreciate if you would indicate that all of the raw data are available from routine publications available to anyone." Doll and Kinlen’ did just that. They used the N.C.I.’s figures without alteration and claimed that they obtained their data from "routine publications". Unfortunately for Doll and Kinlen, the N.C.I, made a serious mathematical or tabulation error which was repeated in their paper.’1 The errors in the N.C.I./Doll-Kinlen data were discovered by Dr Dean Burk and myself and transmitted to The Lancet on Sept. 5, 1977 and to the N.C.I. shortly before the Congressional hearing in September/October investigating an alleged cover-up of the fluoridation-cancer link. Dr Hoover admitted the error and its duplication on Sept. 26, 1977, in a letter to Dr Kinlen in which he wrote: "I am sorry for this error, particularly since it seems to have been perpetuated by yourselves and the Royal Statistical Society." Dr P. D. Oldham uncovered N.C.I.’s error yet in his paper for the R.S.S. with Prof. D. J. Newell he used the erroneous data.2 More important than the mathematical errors passed on by the N.C.I. (which account for about 3500 cancer deaths a year) are the methodological errors that were also spoon-fed to Oldham and Newell and Doll and Kinlen by the N.C.I. (and which account for about 9000 cancer deaths per year). These include omission of 90-95% of the data available.and the selection as one of their data points a year (1970) during which fluoridation of the control group had already started. Using correct figures, linear regression analysis of all available cancer death-rate data (unweighted) for years in which the control cities were not fluoridated (1950 to 1968 with extrapolation to 1970), and a reference population that best reflects the cancer death-rate distribution in central cities, we have recalculated Doll and Kinlen’s table III as follows:
members of Parliament, health authorities, and water boards as evidence that fluoridation was causing many deaths from cancer. Because we wanted to be sure about the truth of the matter, and because we feared that Burk and Yiamouyiannis’s abuse of statistics might be detrimental to the future health of British children, we asked the N.C.I. to provide us with the relevant population and mortality data for the twenty cities so that we could examine the facts ourselves. In providing this information, Dr Hoover of the N.C.I. pointed out that these data were all publicly available. He also told us, in the letter from which Yiamouyiannis quotes, that "we purposely used only published data in doing the reanalysis, so we could not be accused of concealing data. Therefore, all that is necessary for anyone in this country to check our analyses is to be aware of basic principles of biostatistics and epidemiology, and have a public library card. I recognise, however, that this same ease of access to the relevant data do not exist to people in other countries. I have therefore enclosed the relevant numbers..." We consulted the sources used by the N.C.I. because we wanted to use national age, sex, and ethnic group specific mortality-rates for two years that had not been used in the American analysis (1960 and 1970). We did not, however, check all the populations and numbers of deaths in each of the cities, and in consequence perpetuated the N.C.I.’s error involving 215 deaths in non-fluoridated cities. We have already drawn attention to this error,’ when we also pointed out that it did not in any way affect our conclusions. Yiamouyiannis criticises us for choosing to make comparisons for 1970 when fluoridation had been started in two of the "non-fluoridated" cities shortly before. We preferred to do so, however, because 1970 was a census year for which accurate population figures were available. To suppose that this advantage is outweighed by the deaths that might possibly have been caused in the two "non-fluoridated cities" within months of the start of fluoridation is most implausible. This supposition can, however, be tested simply by excluding the two control cities (Atlanta and Seattle) that were fluoridated in 1969 and 1970 from the analyses. The results are unaffected: the number of cancer deaths in 1950 and 1970 were respectively 9835 and 12 229, producing observed-to-expected ratios for the non-fluoridated cities of 1-15 and 1.17, respectively-preto
"
cisely the same as in our previous analysis. the direct method, a 4-5% increase in age-adjusted cancer death-rates in the fluoridated cities over those of the non-fluoridated cities was observed.
Using
,
National Health Federation, Delaware, Ohio 43015, U.S.A.
JOHN YIAMOUYIANNIS
further than we have above on the in the last paragraph of Yiamouyiannis’s letter, since it is not clear what he has done. Data are not available that would permit regression analyses based on data for each year, that take demographic changes adequately into account. We
cannot comment
analysis
Department of Regius Professor of Medicine, Radcliffe Infirmary, Oxford OX2 6HE
RICHARD DOLL LEO KINLEN
SIR, -Dr Yiamouyiannis alleges that
our analysis of cancermortality rates in twenty cities in the U.S. was "spoon-fed" to us by the National Cancer Institute and that it contained "methodological errors" as well as minor numerical errors. The first allegation is untrue; the second is a matter of opinion. In our view, the method we used (which was used independently by the Royal Statistical Society) was correct; the method Burk and Yiamouyiannis used-and Yiamouyiannis
continues to use in the above letter-is incorrect, because it fails to take into account adequately the demographic changes in the age, sex, and ethnic group constitution of the populations of the cities. The preparation of our paper was prompted by the concern aroused at the wide publicity that Burk and Yiamouyiannis had sought and obtained in Britain for their misleading use of crude cancer-mortality rates in fluoridated and non-fluoridated U.S. cities. Through the agency of the National Anti-Fluoridation Campaign (and perhaps others) these data were circulated 1. 2.
Doll, R., Kinlen, L. Lancet, 1977, i, 1300. Oldham, P. D., Newell, D. J. appl. Statist. 1977, 26, 125.
Six,-Dr Yiamouyiannis struggles desperately to salvage his argument that artificial fluoridation and cancer are associated. He now produces a new table whose construction is entirely obscure. He says "Using the direct method ...", which I take to imply "... of standardisation". That is to say, he implies that he now has sex and race specific death-rates in sufficiently narrow age-groups that these can be validly applied to a reference population. I hope he will publish these so that everyone interested may examine them in any way they wish. In view of Yiamouyiannis’s remarks about our paper I must make it clear that I and Professor Newell were asked by the Council of the Royal Statistical Society in July, 1976, to examine the evidence about cancer and fluoridation so that the Royal Statistical Society could give a formal opinion to the Royal College of Physicians. That evidence included population and mortality statistics provided by the National Cancer Institute and a variety of papers by Burk, Yiamouyiannis, and 1.
Kinlen, L., Doll, R. Lancet, 1977, ii, 1039.
151 others. Our report, written in January, 1977, was based on this evidence, as it stated. As Yiamouyiannis says, the vital statistical data were said by an officer of the N.C.I. to be "available from routine publications available to anyone". Through the courtesy of the Registrar General and his staff I was able, in April, 1977, after our paper was printed, to verify that this was indeed true in the U.K., but as I said in my letter to Professor Newell, mentioned by Yiamouyiannis, the exact definition of the U.S. populations used for the calculation of rates was outside my competence, and I was even unsure that what seemed to be a discrepancy of 215 deaths-out of 14 487 (Yiamouyiannis’s "serious error")-was not attributable to my unfamiliarity with U.S. demography. Yiamouyiannis should set aside his apparent conviction that there is a conspiracy to obscure the truth rather than a recognition of the fallacies in his arguments, and if he really believes his hypothesis is still tenable, let him devote his energies to gathering data capable of providing new evidence. If he has cancer deaths by age, sex, and race in the twenty cities let us see them as they are, not manipulated by "linear regression analysis" or whatever. M.R.C. Pneumoconiosis Unit,
Llandough Hospital,
P. D. OLDHAM
Penarth, Glamorgan
STABILITY OF REFERENCE THROMBOPLASTINS
SIR,-Dr Bangham and evidence2
Dr
Loeliger’
attempt
to
refute
our
reference instability of the N.I.B.S.&C. thromboplastins 67/40 and 69/223. They provide results of on
the
small exercises done four years apart without definition of the calibration constant. Comment can, therefore, only be made on their normal values. Their normal control values of 67/40 have lengthened significantly, but this, in our view, supports our conclusion. Bangham and Loeliger explain this trend by changes in test conditions. Whatever the interpretation, their results are hardly evidence of stability. Contrary to their supposition, our test system remained unchanged; the same technique was used throughout by one of us. Furthermore, the greatest change in materials occurred after 1974, and during the N.I.B.S.&C. this same period, the two lyophilised preparations prepared at the National (U.K.) Reference Laboratory, which were studied in parallel, showed no evidence of change, and this indicates the reliability of our test system. Bangham and Loeliger’s statement that we chose the levels of significance to support the claim that 67/40 and 69/223 are unstable is unjustified. Any statistical test takes into account the variability of the data, and there are no grounds for saying this will lead to a falsely high level of statistical significance. There is no evidence for Bangham and Loeliger’s assertion that thromboplastins do not show the variability we have described for the N.I.B.S.&C. materials. We suggest that, before reference any organisation proposes preparations, a long-term project, similar to ours on the research and development of lyophilised thromboplastin for clinical anticoagulant control, is two
necessary. The calibration graph given by Loeliger3 is misleading. He compares the results for lyophilised B.C.T. and 67/40 obtained in his own laboratory in 1976 with the results for 69/233 and 67/40 obtained by us five years previously. As the points fit the same. line, he believes the results show that 67/40 behaves similarly in the two laboratories. It is invalid to plot 69/223 and B.c.T. on the same axis because these two reagents have always been shown to be different in their prothrombin-time results. 69/223 gave longer prothrombin-times than B.c.T. even before it deteriorated. The fact that the results with these two differ1. Bangham, D. R., Loeliger, E. A. Lancet, 1977, ii, 1280. 2. Poller, L, Thomson, J. M., Yee, K. F. ibid. p. 1019. 3. Loeliger, E. A. Ibid. p. 1279.
ent reagents fit the same line after an interval of five years could be regarded as further evidence of the deterioration of
67/40. We conclude that introduction of a new international system of anticoagulant control, in which all reagents will be based on one master preparation, requires proof of the stability of this reagent. The N.I.B.S.&C. preparations, on present evidence, are unsuitable. Reliance on B.c.T. should continue. National (U.K.) Reference Laboratory for Anticoagulant Reagents and Control,
Withington Hospital, Manchester M20 8LR
L. POLLER JEAN M. THOMSON K. F. YEE
StR,—Dr Loeliger’ wrongly implies that the British system for anticoagulant control uses prothrombin activities derived from saline dilution curves. The British system is based on prothrombin ratios obtained with the standardised thromboplastin B.C.T./M.C.R. Prothrombin activities are merely given as equivalent values. The need for the N.I.B.S.&C. reagents as reference preparations was never adequately established and now appears even more unnecessary in view of the widespread international adoption of British Comparative Thromboplastin (B.c.T.) to calibrate national reference preparations, commercial reagents, and local reagents. The need for a multiplicity of thromboplastins has not been demonstrated. There has been more characterisation of B.c.T. than of any other reagent. Its properties are widely known from its use in routine and reference work for many years. Uniformity between batches is guaranteed by the assessment provided by the British Anticoagulant Panel. The monitoring procedurez2 and production details34 have been published. Korsan-Bengsten et al.5 have demonstrated how a commercial extract (’Simplastin A’) has been successfully correlated with B.c.T. over the years. Some thromboplastins are insensitive to clotting factors reduced by oral coumarin drugs and it may, therefore, be difficult to calibrate these. The answer lies in collaborating with commercial manufacturers to improve inferior reagents. It would be a mistake to encourage the use of reagents which may result in a lack of clinical safety in anticoagulant dosage by the provision of reference materials relatively insensitive to vitamin-K-dependent clotting factors. A therapeutic range with a thromboplastin extract is obtained from clinical experience or by calibration against a proved reagent. Both approaches led to the present recommended range of ratios of 2-0-0 for B.C.T. The range of ratios of 1.8-3.0, previously recommended, is alleged by Loeliger to provide inadequate thrombosis prophylaxis. His own reference provides no basis for this statement. Loeliger’s view is, moreover, at variance with the experience of British hospitals and of our colleagues overseas. Indeed, a recent clinical trial shows that the use of the conservative range of 2.0 to 2.55 with B.c.T. gives absolute safety in prevention of postoperative thrombosis in a moderate-risk group.6 The therapeutic range with B.C.T. was revised by the British Anticoagulant Panel because of increased confidence in the lack of heemorrhagic complications with the higher ratios over the years,7-9 not because the old range was believed to be inadequate in the prevention of thrombosis. One of the main criticisms of 67/40 and 69/223 is that they have not been evaluated in clinical work. From comparative 1. Loeliger, E. A. Lancet, 1977, ii, 1279. 2. Hills, M., Ingram, G. I. C. Br. J. Hœmat. 1973, 25, 455. 3. Poller, L. Med. Lab. Tech. 1972, 29, 280. 4. Poller, L., Ingram, G. I. C. Quality Control in Hæmatology;
1975.
5.
Korsan-Bengsten, K., Johnson, M., Pehrsson, 1977, 37, 98.
p. 153. London,
N-G. Thrombos. Hœmostas.
6. Taberner, D. A., Poller, L., Burslem, R. W. B., Jones, J. B. Br. med. J. the press). 7. Barkhan, P. Thrombos. Hœmostas. 1976, 35, 751. 8. Poller, L. ibid. 1976, 36, 485. 9. Blackburn, E. K. Prescribers’ J. 1977, p. 17.
(in