257
European drug regulation SIR,—Your June 29 editorial on drug regulation in Europe voices fears that the establishment of a European market for medicines might result in a lowering in the level of consumer protection which we expect for medicinal products. You refer to the British system as though it has evolved independently and will be eroded when a European system is developed. However, although the UK was not part of the European Community when the Medicines Act was passed in 1968, the British system was modelled on an EC directive on the registration of pharmaceutical products drawn up in 1965. Subsequent EC directives and guidelines have been drawn up with full UK involvement and their provisions are reflected in our national legislation and regulatory requirements. The British system is already an integral part of an EC system in which member states have agreed on what testing is required and on what standards of assessment will be applied, and in which pharmacovigilance is
already seen as a joint responsibility. There is little evidence that differences in assessment are such that public safety is at risk but the EC system proposed for 1993 has already built in the safeguard which you suggest. Although member states are expected to allow products on to their market on the basis of an assessment carried out in another country they can refuse to do so if they have reason to believe that the product will present a serious risk to the health of their population. They will be able to evaluate products for that risk since they will continue to receive full dossiers and assessment reports from the country which first assessed the product. Products may not be marketed until that evaluation has been done. If there are major differences between member states they will be resolved by a European body, whose decision will be binding. I agree with you on the need for a measure of transparency in the system. The European Commission proposes that the "summary of product characteristics" issued when a product is approved for marketing should become a public document. Those issued centrally will be published in the Official Journal of the European Communities. The Proprietary Association of Great Britain represents the manufacturers of non-prescription medicines. Our contribution towards EC harmonisation is to help develop European summaries of product characteristics for key ingredients, taking into account all the national summaries issued for products after the reviews of medicines carried out in the member states. This project, undertaken with our partners in the European Proprietary Medicine Manufacturers Association (AESGP), is a major industry initiative. The first four summaries, for aspirin, paracetamol, ibuprofen, and bisacodyl, were published at the beginning of June by AESGP and have been presented to the European Commission. They will be sent to the health authorities of the member states and will be discussed in the Committee for Proprietary Medicinal Products which coordinates the regulatory work of the EC. This approach is similar to the monograph system developed in the United States, which is a public document, and if it is adopted by the EC will ensure that everyone can see that all medicines containing these ingredients will be evaluated to the same standard of safety and efficacy and will have consumer information which is the same throughout the EC. Proprietary Association
of Great Britain,
Vernon House, Sicilian Avenue, London WC1 N 2QH, UK
SHEILA KELLY, Executive director
Decision analysis in immunohistochemistry SIR,-Immunohistochemistry is being used increasingly as an aid to histological diagnosis. In tumour diagnosis the specific expression of certain antigens on particular cell types is sought. However, such techniques raise a practical problem: how much weight should be attached to the immunohistochemical findings in reaching a final tissue diagnosis? The assumption seems to be that the use of such antibodies provides an objectivity that standard histological methods lack but the histopathologist may find it
difficult to interpret those cases where the antibody staining seems conflict with his prior diagnosis by light microscopy. The answer lies in decision analysis based on Bayes’ rule. The prior probability attached to a diagnosis is revised in the light of a further diagnostic to
test.
The analysis presented here is an adaptation of that suggested for histopathology in general by Schwartz et al.2 Faced, for example, with a specimen of metastatic adenocarcinoma which may arise from a primary in the breast, the histopathologist estimates, on the basis of its appearance under the microscope and his knowledge of
the range of appearances of breast and other tumours, that there is a 10% probability that this metastatic adenocarcinoma arises from breast and a 90% probability that it is from another primary site. Suppose that an antibody is available that stains positively in 90% of breast carcinomas and in 30% of a panel of adenocarcinomas from other sites. If the stain is positive, the posterior probability-ie, the probability of each diagnosis after antibody staining-is then calculated as follows:
(A) = Possible diagnoses from light microscopy. (B) = Estimate of probability of (A). (C) Probability of positive antibody staining for each diagnosis in (A). (D)=(B)x(C). (E) = Revised estimate of probability of diagnosis. =
Thus a new set of diagnostic probabilities is obtained after the application of immunohistochemistry. The method can be expanded for more than two diagnoses in column A, as long as each can be given a relative probability on the basis of conventional histology, and for any number of immunohistochemical tests, provided that, for each antibody, as much information as possible is available about staining with a representative panel of tumours. This tabular approach exposes many of the underlying mechanisms of analysis which are often subjective. It shows that if an antibody is to be used to solve a particular diagnostic problem the likelihood of a positive (or negative) result for each of the diagnostic possibilities must be known; if they are not the test may be being used inappropriately. A further advantage is that the histopathologist can decide, before the test is done, how much difference a positive (or negative) result will make to the probabilty of the diagnosis arrived at by light microscopy. This can be assessed readily with a nomogram3 and should allow immunohistochemical techniques to be used in the most cost-effective way. It is perhaps not always appreciated that a histopathologist’s assessment of a specimen creates a set of diagnostic probabilities which can be altered by a further test, such as immunohistochemistry, but not in some way "over-ruled" because of the apparent objectivity of that test. As for clinical tests, the predictive value depends on the prior probability of the disease. If that probability is only 2%, a test which is 99% sensitive and 99% specific has a predictive value of only 67%. If an antibody has 100% specificity true objectivity may be achieved but in most cases the histopathologist has to do a calculation such as the above to assess the importance of immunohistochemical stains. He also has to select antibodies for the most cost-effective route to a diagnosis with a high degree of probability in the light of probable diagnoses suggested by light microscopy, and this too can be helped by decision analysis.
Department of Histopathology, St Mary’s Hospital Medical School, London W2 1PG, UK
H. T. COOK
1. Pauker SG, Kassirer JP. Decision analysis. N Engl J Med 1987; 316: 250-58. 2. Schwartz WB, Wolfe HJ, Pauker SG. Pathology and probabilities. A new approach t interpreting and reporting biopsies. N Engl J Med 1981; 305: 917-23. 3.
Fagan TJ. Nomogram for Bayes’s theorem. N Engl J Med 1975; 293:
257.
European drug regulation SIR,—Your June 29 editorial on drug regulation in Europe voices fears that the establishment of a European market for medicines might result in a lowering in the level of consumer protection which we expect for medicinal products. You refer to the British system as though it has evolved independently and will be eroded when a European system is developed. However, although the UK was not part of the European Community when the Medicines Act was passed in 1968, the British system was modelled on an EC directive on the registration of pharmaceutical products drawn up in 1965. Subsequent EC directives and guidelines have been drawn up with full UK involvement and their provisions are reflected in our national legislation and regulatory requirements. The British system is already an integral part of an EC system in which member states have agreed on what testing is required and on what standards of assessment will be applied, and in which pharmacovigilance is
already seen as a joint responsibility. There is little evidence that differences in assessment are such that public safety is at risk but the EC system proposed for 1993 has already built in the safeguard which you suggest. Although member states are expected to allow products on to their market on the basis of an assessment carried out in another country they can refuse to do so if they have reason to believe that the product will present a serious risk to the health of their population. They will be able to evaluate products for that risk since they will continue to receive full dossiers and assessment reports from the country which first assessed the product. Products may not be marketed until that evaluation has been done. If there are major differences between member states they will be resolved by a European body, whose decision will be binding. I agree with you on the need for a measure of transparency in the system. The European Commission proposes that the "summary of product characteristics" issued when a product is approved for marketing should become a public document. Those issued centrally will be published in the Official Journal of the European Communities. The Proprietary Association of Great Britain represents the manufacturers of non-prescription medicines. Our contribution towards EC harmonisation is to help develop European summaries of product characteristics for key ingredients, taking into account all the national summaries issued for products after the reviews of medicines carried out in the member states. This project, undertaken with our partners in the European Proprietary Medicine Manufacturers Association (AESGP), is a major industry initiative. The first four summaries, for aspirin, paracetamol, ibuprofen, and bisacodyl, were published at the beginning of June by AESGP and have been presented to the European Commission. They will be sent to the health authorities of the member states and will be discussed in the Committee for Proprietary Medicinal Products which coordinates the regulatory work of the EC. This approach is similar to the monograph system developed in the United States, which is a public document, and if it is adopted by the EC will ensure that everyone can see that all medicines containing these ingredients will be evaluated to the same standard of safety and efficacy and will have consumer information which is the same throughout the EC. Proprietary Association
of Great Britain,
Vernon House, Sicilian Avenue, London WC1 N 2QH, UK
SHEILA KELLY, Executive director
Decision analysis in immunohistochemistry SIR,-Immunohistochemistry is being used increasingly as an aid to histological diagnosis. In tumour diagnosis the specific expression of certain antigens on particular cell types is sought. However, such techniques raise a practical problem: how much weight should be attached to the immunohistochemical findings in reaching a final tissue diagnosis? The assumption seems to be that the use of such antibodies provides an objectivity that standard histological methods lack but the histopathologist may find it
difficult to interpret those cases where the antibody staining seems conflict with his prior diagnosis by light microscopy. The answer lies in decision analysis based on Bayes’ rule. The prior probability attached to a diagnosis is revised in the light of a further diagnostic to
test.
The analysis presented here is an adaptation of that suggested for histopathology in general by Schwartz et al.2 Faced, for example, with a specimen of metastatic adenocarcinoma which may arise from a primary in the breast, the histopathologist estimates, on the basis of its appearance under the microscope and his knowledge of
the range of appearances of breast and other tumours, that there is a 10% probability that this metastatic adenocarcinoma arises from breast and a 90% probability that it is from another primary site. Suppose that an antibody is available that stains positively in 90% of breast carcinomas and in 30% of a panel of adenocarcinomas from other sites. If the stain is positive, the posterior probability-ie, the probability of each diagnosis after antibody staining-is then calculated as follows:
(A) = Possible diagnoses from light microscopy. (B) = Estimate of probability of (A). (C) Probability of positive antibody staining for each diagnosis in (A). (D)=(B)x(C). (E) = Revised estimate of probability of diagnosis. =
Thus a new set of diagnostic probabilities is obtained after the application of immunohistochemistry. The method can be expanded for more than two diagnoses in column A, as long as each can be given a relative probability on the basis of conventional histology, and for any number of immunohistochemical tests, provided that, for each antibody, as much information as possible is available about staining with a representative panel of tumours. This tabular approach exposes many of the underlying mechanisms of analysis which are often subjective. It shows that if an antibody is to be used to solve a particular diagnostic problem the likelihood of a positive (or negative) result for each of the diagnostic possibilities must be known; if they are not the test may be being used inappropriately. A further advantage is that the histopathologist can decide, before the test is done, how much difference a positive (or negative) result will make to the probabilty of the diagnosis arrived at by light microscopy. This can be assessed readily with a nomogram3 and should allow immunohistochemical techniques to be used in the most cost-effective way. It is perhaps not always appreciated that a histopathologist’s assessment of a specimen creates a set of diagnostic probabilities which can be altered by a further test, such as immunohistochemistry, but not in some way "over-ruled" because of the apparent objectivity of that test. As for clinical tests, the predictive value depends on the prior probability of the disease. If that probability is only 2%, a test which is 99% sensitive and 99% specific has a predictive value of only 67%. If an antibody has 100% specificity true objectivity may be achieved but in most cases the histopathologist has to do a calculation such as the above to assess the importance of immunohistochemical stains. He also has to select antibodies for the most cost-effective route to a diagnosis with a high degree of probability in the light of probable diagnoses suggested by light microscopy, and this too can be helped by decision analysis.
Department of Histopathology, St Mary’s Hospital Medical School, London W2 1PG, UK
H. T. COOK
1. Pauker SG, Kassirer JP. Decision analysis. N Engl J Med 1987; 316: 250-58. 2. Schwartz WB, Wolfe HJ, Pauker SG. Pathology and probabilities. A new approach t interpreting and reporting biopsies. N Engl J Med 1981; 305: 917-23. 3.
Fagan TJ. Nomogram for Bayes’s theorem. N Engl J Med 1975; 293:
257.
