Advances in techniques of testing mycobacterial drug sensitivity, and the use of sensitivity tests in tuberculosis control programmes

Bull. Org. mond. Sante 1969, 41, 21-43 Bull. Wld Hlth Org. Advances in Techniques of Testing Mycobacterial Drug Sensitivity, and the Use of Sensitivity Tests in Tuberculosis Control Programmes G. CANETTI,1 WALLACE FOX,2 A. KHOMENKO,3 H. T. MAHLER,' N. K. MENON,5 D. A. MITCHISON,' N. RIST 1 & N. A. WMELEV 7 In a paper arising out of an informal international consultation of specialists in the bacteriology of tuberculosis held in 1961, an attempt was made to formulate criteria, and specify technical procedures, for reliable tests of sensitivity (the absolute-concentration method, the resistance-ratio method and the proportion method) to the 3 main antitubercu- losis drugs (isoniazid, streptomycin and p-aminosalicylic acid). Seven years later, a further consultation was held to review the latest developments in the field and to suggest how sensitivity tests might be put to practical use in tuberculosis control programmes. The participants reached agreement on how to define drug sensitivity and resistance, and stressed the importance of using a discrimination approach to the calibration of sensitivity tests. Their views are contained in the present paper, which also includes descriptions of the sensitivity tests used by the Medical Research Council ofGreat Britain for first- and second- line drugs (minimal inhibitory concentration and resistance-ratio methods), the two main variants ofthe proportion method developed by the Institut Pasteur, Paris, and a methodfor calibrating sensitivity tests. I. INTRODUCTION8 As the outcome of an international consultation of specialists in the bacteriology of tuberculosis, ar- ranged by WHO in 1961, Canetti et al. (1963) de- scribed criteria and techniques for reliable tests of mycobacterial resistance to tuberculostatic drugs. Since then, considerable experience has been gained in the use of the tests described and new drugs have been introduced. The technical problems involved in standardizing such methods are now much better understood. The paper by Canetti et al. (1963) had said little about how sensitivity tests might be put to practical use in the control of tuberculosis. The present article reviews the pro- gress that has been made in methods of sensitivity testing and their calibration. Further, the ways in which the results of sensitivity tests may influence the choice of regimens of chemotherapy for individ- ual patients, and their use to estimate the prevalence of drug-resistant strains in the community, are eva- luated. An attempt is made also to formulate the 1 Chef de Service, Institut Pasteur, 25 rue du Docteur Roux, Paris 15e, France. ' Director, Medical Research Council's Tuberculosis and Chest Diseases Research Unit, Brompton Hospital, Fulham Road, London, S.W.3, England. 3Formerly Medical Officer, Tuberculosis, Division of Communicable Diseases, World Health Organization, Geneva, Switzerland; present address: Professor and Head of Tuberculosis Department, Postgraduate Medical Institute, Kharkov, USSR. ' Chief Medical Officer, Tuberculosis, Division of Com- municable Diseases, World Health Organization, Geneva, Switzerland. b Director, Tuberculosis Chemotherapy Centre, Madras, India. ' Honorary Director, Medical Research Council's Unit for Research on Drug Sensitivity in Tuberculosis, Royal Post-graduate Medical School, Ducane Road, London, W.12, England. 7Director, Central Research Institute of Tuberculosis, Moscow-128, USSR. 8 Prepared by the participants in an informal consultation on the drug-resistance problem in tuberculosis, held in Geneva, Switzerland, from 23 to 26 April 1968: Dr G. Canetti, Dr Wallace Fox, Dr A. Khomenko, Dr H. T. Mahler, Dr N. K. Menon, Professor D. A. Mitchison, Dr N. Rist and Professor N. A. Smelev. 2353 -21- G. CANETTI AND OTHERS role of sensitivity tests in tuberculosis control pro- grammes at widely varying stages of development. DEFINITION OF DRUG RESISTANCE Drug resistance may be defined in bacteriological terms or in the light of the probable response of the patient to chemotherapy with the drug concerned. As it is difficult to establish a precise clinical defini- tion of resistance, a definition in purely bacteriologi- cal terms has been adopted for the purposes of this paper. The following statement (Mitchison, 1962) is considered to be adequate: "Resistance is defined as a decrease in sensitivity of sufficient degree to be reasonably certain that the strain concerned is different from a sample of wild strains of human type that have never come into contact with the drug." Such a definition assumes that the variation in sensitivity of sensitive strains to any one of the antibacterial drugs is small. (It does not apply to thioacetazone, since natural resistance to this drug occurs in a high proportion of strains obtained from certain regions of the world (Thomas et al., 1961; Rist, 1968; Mitchison & Allen, to be published).) For most of the antituberculosis drugs, there is evidence that a diminished clinical response may occur when resistance in the above-mentioned bac- teriological sense is demonstrated in the laboratory. AVAILABLE METHODS OF TESTING FOR DRUG SENSITIVITY Three widely used methods for testing sensitivity to isoniazid, streptomycin and para-aminosalicylic acid (PAS) on Lowenstein-Jensen medium-namely, the absolute-concentration method, the resistance- ratio method and the proportion method-were de- scribed by Canetti et al. (1963). Part II of the present paper describes the minimal inhibitory concentration method and the resistance-ratio method; Part III, a simplified variant and a standard variant of the proportion method; and Part IV, a method of calibrating sensitivity tests. There are, of course, other methods, including tests with agar-based semi-synthetic media and liquid media, as well as vertical diffusion methods. There remains a need for more rapid and simple but accurate techniques of sensitivity testing. Direct tests provide results more rapidly than indirect tests, which are used in most laboratories. It is therefore encouraging that progress with direct tests is being made in two directions. First, simplified direct tests, which are already available with certain methods (see Part III), are being further developed, especially for isoniazid and streptomycin. Secondly, the slide- culture method, which yields results in 7 days, is being standardized in terms of established indirect tests. THE CALIBRATION OF TESTS The necessity to standardize any sensitivity test method was emphasized by Canetti et al. (1963), but they gave no details of the ways in which this might be done. Objective methods of calibration are im- portant for several reasons. First, to clarify how the present criteria of resistance in the methods de- scribed in Parts II and III have been obtained. Secondly, for technical reasons the most efficient criterion of resistance may be slightly different from one laboratory to another. In particular, drug activ- ity may vary because of heat instability of some drugs during inspissation or storage (notably ethion- amide and cycloserine). The type of container and the type of closure for slopes containing streptomy- cin also affect the result (Krebs et al., 1962; Interna- tional Union against Tuberculosis, 1964). Thirdly, objective methods of calibration allow of compari- sons between methods of sensitivity testing in which the measures of sensitivity are expressed on different scales. In the past, a common approach was to test a sample of presumedly sensitive strains. Strains were considered to be resistant if they were slightly more resistant than the great majority of the sensitive strains. Such a procedure has the defect that strains considered to have natural or primary resistance must be removed from the population of sensitive strains, using necessarily arbitrary criteria. An alternative approach is to consider the results of the treatment of patients who have organisms of varying degrees of sensitivity at the start of treat- ment, and to try to determine at what level of sensitivity there is a change in the therapeutic re- sponse. In practice, the number of patients with initially resistant strains who are treated with the drug concerned, alone or in combination with only one other drug, is too small to allow precise criteria of resistance to be obtained. However, the best available approach is to com- pare the sensitivity of a sample of strains from untreated patients (a predominantly sensitive sam- ple) with a sample of strains from patients who have been treated with the drug for at least a few months, 22 MYCOBACTERIAL DRUG SENSITIVITY TESTS: RECENT ADVANCES AND PRACTICAL USE mostly without success (a predominantly resistant sample). The criterion of resistance that discrimi- nates between the two samples with the greatest efficiency can then be chosen. This discrimination approach is described fully in Part IV (" The calibra- tion of sensitivity tests "). For such a comparison, it is unnecessary to have either a pure population of wild strains or a pure population of resistant strains: it is essential only that there should be a substantial difference in the proportion of resistant strains with- in each sample. Unfortunately, even with this dis- crimination approach, owing to the errors inherent in any method of testing, some misclassification of sensitive strains as resistant, and vice versa, tends to occur. Since it is desirable to avoid misclassification of sensitive strains as resistant, because this may lead the physician to deprive the patient of a valuable drug, it is preferable to choose a final criterion of resistance that very rarely misclassifies a sensitive strain, although it may more often fail to detect strains with borderline degrees of resistance. As soon as possible after starting to carry out sensitivity tests, irrespective of the method of testing employed, a laboratory should make provision for the calibration of these tests. This will make it possible to establish whether the criteria of resistance proposed by the initiators of the method are entirely valid in the laboratory in question or whether some of the criteria require modification. Furthermore, it would be desirable for large laboratories to repeat this calibration procedure at intervals, especially for tests that are difficult to perform. Even though considerable work would be necessary, it would be certain that the efficiency of the tests performed in that laboratory remained high. It is desirable that samples of predominantly resistant cultures should be maintained in reference laboratories for issue to any laboratory in which it is desired to calibrate sensitivity tests. The discrimination approach should be used also in comparing methods of sensitivity testing. For this purpose, the best criterion for each method is chosen and its ability to discriminate between predominant- ly sensitive and predominantly resistant groups of cultures is expressed as a percentage, which estimates the efficiency of each method in the detection of resistance. COMPARISON OF METHODS OF SENSITIVITY TESTING Among the technical issues concerning sensitivity tests, the standardization of the size of the inoculum remains of the greatest importance. Further evi- dence is accumulating that there is a large difference from culture to culture in the number of colonies that grow from a similar inoculum specified by weight or opacity (Canetti et al., 1963; International Union against Tuberculosis, 1964; Lefford & Mit- chison, 1966; Bartmann & Galvez-Brandon, 1968). Variation in the number of culturable particles (for- merly known as the " viable count ") can be attri- buted either to differences in the actual number of organisms in the inoculum or to the degree of clumping of the bacilli. If the former interpretation is correct, the proportion method should compen- sate for variation in the total bacillary content of the inoculum. If the latter, this advantage of the pro- portion method is not gained. On the available evidence, both these factors cause variation in the number of culturable particles, but further work is recommended to establish which is the more impor- tant, and to what extent each modifies the results of tests. Of greater practical importance than these theo- retical considerations is the empirical comparison of the relative efficiency of the different methods in distinguishing a sample of predominantly sensitive strains from a sample of predominantly resistant strains. Largely from unpublished data, using the discrimination approach or a variant of it, the fol- lowing provisional conclusions have been drawn. First-line drugs For streptomycin, there appears to be little to choose between the minimal inhibitory concentra- tion, resistance-ratio and proportion methods, using the discrimination approach. For isoniazid, a simi- lar conclusion can be drawn about the minimal inhibitory concentration and proportion methods, but the resistance-ratio method has not been stud- ied. The relative efficiency of the three methods for testing PAS sensitivity does not appear to have been investigated. The discrimination approach has been found to be unsatisfactory for analysing sensitivity tests to thioacetazone in strains from regions where natural resistance to thioacetazone is prevalent. Second-line drugs The resistance-ratio method of measuring sensitiv- ity on Lowenstein-Jensen medium using serial 2-fold drug dilutions appears to be unsatisfactory, mainly because the critical ratio that distinguishes sensitive and resistant strains best lies between 2 and 4. Evidence on comparisons of the minimal inhibitory 23 G. CANETTI AND OTHERS concentration and proportion methods for ethion- amide and cycloserine is discrepant. Some workers find that the methods are of similar discriminative efficiency (Lefford & Mitchison, 1966); others, that the proportion method is the more efficient. For pyrazinamide-sensitivity tests, there is agreement that the proportion method is the more efficient, but, again, the extent of the advantage found differs from one group of workers to another. Since there have been few comparisons between tests done with the minimal inhibitory concentration and pro- portion methods, and no comparisons have been made for PAS, kanamycin or viomycin, further studies are needed. Only tentative suggestions can be made on methods to be adopted for testing sensitivity to two new drugs, ethambutol and rifampicin. For ethambutol, the in vitro difference between sensitive and resistant strains is very small, and the resistance-ratio method will probably prove to be unsatisfactory. The minimal inhibitory concentra- tion and proportion methods might be equivalent, but there should be only narrow intervals between the drug concentrations to be tested. Even so, there is evidence that in vitro resistance to ethambutol is often difficult to detect in cases of apparent clinical resistance. On the other hand, for rifampicin, the in vitro difference between sensitive and resistant strains is considerable, and it is likely that the resis- tance-ratio method will be as efficient as the two other methods. PROGNOSTIC SIGNIFICANCE OF DRUG RESISTANCE Prinmary drug resistance On the basis of the available evidence concerning the influence of primary isoniazid resistance on the response of patients to treatment with this drug given alone, it may be concluded that the response is often impaired by primary resistance. Nevertheless, a substantial proportion of patients with primary resistance show a bacteriological response, and some patients achieve sputum conversion. For other drugs, inadequate evidence is available to say wheth- er, in the presence of primary resistance, a measure of response can be expected with treatment with the drug concerned, when given alone. At the other extreme, there is evidence that the presence of resis- tance to a single drug has little or no effect on the outcome of treatment with the three drugs: isoniazid, streptomycin and PAS (International Union Against Tuberculosis, 1964). Furthermore, even in the pres- ence of primary resistance to 2 first-line drugs, a bacteriological response is not infrequently obtained with the 3 drugs. Acquired drug resistance When drug resistance has emerged during the course of chemotherapy, no further response is to be expected from the drug concerned, provided that the sputum is consistently positive. This is true even for isoniazid, and in all probability no further benefit is to be obtained with long-term treatment with the drug. Since the presence of persistent sputum posi- tivity implies the growth of organisms within the lesions, such strains are not only bacteriologically resistant but also clinically resistant. The degree of laboratory resistance is immaterial. Incidentally, the results of a recent investigation indicate that the addition of isoniazid to re-treatment regimens in patients whose organisms had previously acquired isoniazid resistance did not improve the result. Transitional resistance Occasionally resistance to one or more drugs of a regimen may emerge in the course of successful chemotherapy. The resistant culture, usually com- posed of only 1 or 2 colonies and seldom of more than 5 colonies, is most frequently obtained imme- diately before sputum conversion, but may also be obtained during the next few months. This finding is due to the isolation of resistant, persistent organ- isms, i.e., resistant organisms that are not multiply- ing in the lesions but have nevertheless survived (Canetti, 1965). Transitional resistance is in sharp contrast to acquired resistance in that significant multiplication of resistant organisms does not occur in the former and there is no justification for a change of treatment. It is strongly emphasized that the performance of sensitivity tests during chemotherapy in patients with decreasing bacillary populations on serial smear and culture rarely serves any useful purpose since, under these circumstances, the only resistant strains likely to be encountered are transitionally resistant strains. THE EPIDEMIOLOGY OF DRUG RESISTANCE Initial drug resistance It is important to appreciate the difference be- tween primary and initial drug resistance. Primary resistance is resistance found in patients which has not resulted from treatment of those patients with 24 MYCOBACITERIAL DRUG SENSITIVITY TESTS: RECENT ADVANCES AND PRACTICAL USE the drug concerned. It includes both natural resis- tance, which may occur in a strain that has never come into contact with the drug, and acquired resis- tance occurring as a result of exposure to the drug in another patient. Initial resistance occurs in patients who give no history of previous chemotherapy. It includes primary resistance, as well as acquired resis- tance in patients who have either concealed a history of previous chemotherapy or have received chemo- therapy-for example, isoniazid for a cough-with- out being aware of the nature of the medicament. Surveys of the prevalence of drug resistance Since it is very difficult to obtain a reliable history of previous chemotherapy from patients in the deve- loping countries, surveys of primary resistance in these countries usually cannot be carried out. Most surveys in patients disclaiming previous treatment are, in practice, surveys of initial drug resistance. On the other hand, in the technically advanced countries, surveys of true primary drug resistance can more readily be performed because the patients know if they have had previous chemotherapy and have no reason for withholding the information. Surveys of primary resistance measure the tenden- cy for resistant strains to accumulate within the community. In several of the technically advanced countries, they have been repeated at intervals of years. In general terms, the results show that there has been no increase in the prevalence of primary resistance during the past decade, and that resistance to only one drug is much more frequent than resis- tance to more than one drug. There is increasing evidence that the level of primary resistance may be a new way of assessing the amount of bacillary transmission in the community. Thus it may provide a new epidemiological parameter. Surveys of initial drug resistance measure the prevalence of resistance in patients who present themselves for treatment. They act as a guide to the formulation of chemotherapy programmes and reflect the quality and extent of chemotherapy prac- tice in the area. The available data provide no acceptable evidence of a progressive build-up of initial drug resistance to alarming levels in any developing country where the initial interrogation of patients about the history of previous chemotherapy has been carefully conducted. Indeed, as far as the trends can be followed, there is some evidence that the levels may become stable, or even decrease, as the stan- dards of the tuberculosis service and of chemo- therapeutic practice improve. In order to provide a measure of the problem that all newly registered patients present for the treat- ment services, surveys of such patients should be made. These surveys should include all patients attending a clinic for the first time for treatment, whether or not they have received treatment else- where; they would not include patients already on the clinic register who return for further treatment in the same clinic. Such surveys will be referred to as total surveys of newly registered patients. The pool of chronic drug-resistant infectious cases A number of difficulties are involved in measuring the size of the pool of chronic drug-resistant infec- tious cases in the community, a measurement that pro- vides an important indication of the total extent of drug resistance. Although it is possible to obtain information on the size and characteristics of the pool during mass radiography surveys of the com- munity, the total number of patients found to have a positive culture in such surveys is, in practice, usual- ly small. Further attempts should be made to evolve accurate methods of characterizing the pool. Drug-resistance registers In some cities, all chronic excretors of drug-resis- tant organisms are recorded in drug-resistance regis- ters. These registers are a valuable way of following one important index of the success of therapy in the community. Performance of surveys Whenever possible, a drug-resistance survey should be conducted within an entire country or region rather than in individual clinics or hospitals, so that the survey may be widely representative. Preliminary data on the notification of new cases of tuberculosis should be obtained and used to draw a statistical sample of the clinics and hospitals in the regions, so that urban and rural communities and clinics with good and with -poor treatment services are fairly represented. In developing countries, the sample should include children under the age of 15 years, since there is evidence that the pattern of drug sensitivity of their organisms may give the best measure of the prevalence of primary resistance. If the community is multiracial or contains large groups of immigrants, it is important to record the race or the country of origin of the patient, since the groups may have different prevalences of drug resis- tance. 25 G. CANETII AND OTHERS The sensitivity tests for a survey should be per- formed in a single laboratory capable of doing tests of a high standard of accuracy and, indeed, if a sufficient standard cannot be obtained the survey should not be undertaken. All mycobacteria isolat- ed should be identified as Myco. tuberculosis, Myco. bovis or some other mycobacterial species. In mak- ing a choice between the several accurate sensitivity test methods available, comparability with previous surveys in the same region is the first priority, and comparability with surveys in other countries is also of value. It is clearly desirable to reduce to a minimum the number of different sensitivity tests employed, in order to allow accurate comparisons between different countries. One or both of the following control methods are advisable. First, a detailed distribution of the sensi- tivity of a sample of wild strains (or, less satisfactori- ly, a series of tests on the standard strain, H37Rv) should be obtained; the distribution can then be compared with similar distributions obtained in any other survey using the same sensitivity test method, thus providing a check on the comparability of definitions of resistance. Secondly, a sample of pre- dominantly resistant strains, which is often automat- ically obtained in total surveys of newly registered patients (as described on page 25), may be tested. The application of the discrimination approach will then ensure that optimum definitions of resistance are used, and will facilitate comparisons with other surveys using this approach. Furthermore, in reporting the results, full technical details of the method used, including the number of culturable particles obtained on drug-free medium, should be given. Considering the prevalence of resistance to the various antituberculosis drugs, measurements of the prevalence of resistance to isoniazid have the greatest epidemiological value, since isoniazid is widely used in treatment throughout the world, the tests are comparatively easy to perform, and the results obtained by different methods of testing are likely to be closely comparable. It is also of value to estimate the prevalence of resistance to streptomy- cin, but streptomycin is more widely used in some countries than in others, the test is more difficult to perform reliably, and it is possible that the occur- rence of strains with natural resistance to the drug may obscure the epidemiological value of the data. Sensitivity tests to. PAS, thioacetazone or the second- line drugs are of less value. The ratio between the number of strains resistant to a single drug and of those resistant to more than one drug is also of epidemiological importance. THE ROLE OF SENSITIVITY TESTS IN TUBERCULOSIS CONTROL PROGRAMMES The uses of sensitivity tests Sensitivity tests have four main uses: For scientific purposes: they have contributed greatly to our understanding of the mechanisms of chemotherapy and of the reasons for success and failure. For epidemiological studies, whether at one point in time or at two different points in time in the same country, or for comparisons between countries. For planning wide-scale treatment, since, for exam- ple, the proportion of patients with fully sensitive organisms in the community can be estimated. For use in the individual patient, in order to decide on regimens that might be effective, especially in re- treatment. Too much stress has been laid on the last of these uses, compared with the other three. The wide scale on which unreliable sensitivity-test results are reported, even in the technically advanced countries, gives cause for concern. The harm that could result to the individual patient from clinical action taken on the basis of inaccurate results must be emphasized. Also, for the purposes of epidemiol- ogy, the planning of wide-scale treatment or research, inaccurate results are of very limited value. Furthermore, the use of well organized, good regimens of chemotherapy in communities with little initial drug resistance automatically limits the poten- tial value of wide-scale sensitivity testing. Studies in some developing countries have demon- strated the value of a careful interrogation of patients about their previous history of chemothera- py, including, in particular, its duration and the drugs used, to indicate whether pre-treatment drug resistance is likely to be present. This information is available at the time when decisions have to be taken concerning the regimen of chemotherapy to be given to the patient. For this reason, careful interrogation is of importance in planning treatment, whether or not a sensitivity test is performed. Sensitivity testing in countries at different stages of development In the developing countries, requirements for sen- sitivity testing necessarily vary according to the stage 26 MYCOBACTERIAL DRUG SENSITIVITY TESTS: RECENT ADVANCES AND PRACTICAL USE of technical progress and the availability of laborato- ry services. For some countries with a very small population and with limited resources there is no justification for establishing sensitivity testing in even a single laboratory. Such countries might need to obtain limited reference work for epidemiological purposes and do so in a laboratory abroad, possibly with the help of an international agency. In this connexion, efforts are currently being made by WHO to establish inter-country regional reference laboratories. At the next stage of development, there are many countries that either already have, or else require, a single laboratory able to undertake a sufficient num- ber of sensitivity tests to meet all needs for the current stage of development of the tuberculosis programme. The sensitivity tests might be used for epidemiological purposes, for the planning of wide- scale treatment programmes or, in some instances, even for individual patients. In some countries, such a laboratory might be able to meet likely require- ments for a number of years to come. Once one efficient laboratory is functioning, the need to in- crease the number of laboratories should be carefully evaluated and fully justified, not only according to the requirements of the tuberculosis programme but also in relation to competing health priorities. At the other extreme, the situation in many techni- cally advanced countries today is unsatisfactory. Only too often, there are large numbers of laborato- ries undertaking sensitivity tests, many using unsatis- factory methods or else doing tests on too small a scale to maintain an adequate control over their standards. It is encouraging to note that efforts are being made in some countries to centralize sensitivity testing in a few regional laboratories: such central- ization, where laboratories use uniform methods, offers substantial advantages. It is clearly desirable to make one laboratory a national reference centre. It should be emphasized that, whatever the stage of development of the laboratory services of a coun- try, no laboratory should embark on sensitivity test- ing until adequate staff, equipment, interest and intellectual skills are available to sustain a high standard of work. Pre-treatment sensitivity tests In technically advanced countries where it has been established that the levels of primary resistance to the first-line drugs are low and where the available resources permit the use of therapy with the three first-line drugs in all newly diagnosed cases, the role of pre-treatment sensitivity tests is limited. In such circumstances, there is little to be gained from pre- treatment sensitivity tests in patients in whom tuber- culosis has been newly diagnosed, even when sensi- tivity testing is of the highest standard. This is because failure of the standard regimen due to pre- treatment drug resistance is likely to occur only in the very small proportion of patients with strains resistant to two or three drugs. If the standards of sensitivity testing are lower, then the gains may be negligible. Indeed, a position may be reached where, paradoxically, sensitivity testing might even result in actual harm by leading to unnecessary changes of chemotherapy from effective and acceptable regimens to more toxic, less acceptable, and less effective second-line regimens. In developing countries existing knowledge does not provide adequate guidance on the best chemo- therapy policy to be adopted in the light of the widely different levels of initial and acquired drug resistance encountered in the community. One ex- treme choice of policy for an area might be to treat all patients with a standard regimen, the regimen to be changed only when it was apparent that the patient remained consistently positive bacteriologi- cally. The other extreme might be to apply several different basic regimens according to the pre-treat- ment patterns of sensitivity actually encountered in the organisms of individual patients. In a study that is now in progress in an area with a substantial level of initial drug resistance, three poli- cies of chemotherapy, either based on or ignoring the results of pre-treatment sensitivity tests, are being compared. Such studies can yield information not only on the role of sensitivity testing, and hence the laboratory services that are required, but also on the most appropriate policies of chemotherapy to apply In the past, the failure of good regimens of chemo- therapy has often been explained by the finding of resistance in pre-treatment tests for drug sensitivity This has overshadowed what is often a more impor- tant factor: the failure of the patient to co-operate in treatment, only too often because the supervisory staff do not adapt the organization and supervision of chemotherapy to the needs of the patient. Sensitivity testing during treatment Patients who are consistently culture-positive at the end of 6 months or more of treatment, especially when the cultures yield abundant growth and there has been a "fall and rise" phenomenon in the 27 G. CANETrI AND OTHERS bacterial content of the sputum, can be regarded as clear-cut " failures " of the regimen in bacteriologi- cal terms. However, these patients may have drug- sensitive organisms because they have been irregular in taking their medicament. Furthermore, patients who have had a relapse after stopping chemotherapy frequently yield sensitive organisms. In both these circumstances a rapid sensitivity test, for example, a direct test, which is available with certain methods, would be of particular value. Further, the likelihood of finding sensitive or resistant organisms varies according to the standards of treatment in the com- munity. It may well be possible, by the systematic collection of data in cases of treatment failure, as just defined, to formulate policies for further treat- ment. If drug sensitivity were common in the com- munity, treatment with the first-line drugs would normally be resumed. If resistance were common, treatment with second-line drug combinations, if available, would be given. Such an epidemiological application of sensitivity testing could be of greater general value than testing in the individual patient. Sensitivity testing and second-line regimens For patients whose treatment with the first-line drugs has failed, treatment with a second-line regimen can be safely initiated without undertaking sensitivity tests to the second-line drugs to be intro- duced. The indications for changing a second-line regimen might be the occurrence of toxicity, or the finding that the patient's organisms were still sensi- tive to more effective and acceptable first-line drugs, or evidence that the second-line regimen had itself failed. If a further course of second-line drugs is required and it is uncertain whether resistance might have been acquired to one or more of them, sensitiv- ity tests are likely to be of value. There are also problems of cross-resistance. For example, high degrees of in vitro ethionamide resis- tance have been encountered in two situations in patients who have never been treated with the drug. One is unsuccessful treatment with thioacetazone alone, or with isoniazid plus thioacetazone which has resulted in resistance to thioacetazone; the other is in the presence of very low degrees of primary isoniazid resistance (Canetti et al., 1967). Smears, cultures and sensitivity tests in control pro- grammes From the point of view of their value in a tubercu- losis control programme, smears, cultures and sensi- tivity tests may be ranked in the following order of priority: (a) smears; (b) cultures, for diagnostic purposes in patients who are bacteriologically negative on sputum smear examination; (c) sensitivity tests for epidemiological purposes; (d) sensitivity tests for individual patients. The minimum aim of a bacteriological service in a developing country should be to perform enough smear examinations to permit an accurate diagnosis on smear and also to assess the progress of therapy at intervals, the minimum frequency being at the end of 6 and 12 months of therapy. If possible, 2 smears should be performed at each assessment, because important conclusions, for example, that a patient has failed to respond adequately to a regimen of chemotherapy, should not be reached on the basis of a single positive result on smear examination. The " fall and rise " phenomenon in the bacterial content of the sputum, referred to above, is very informative. Therefore, when the results of smear examinations are read, the grade of positivity should be recorded. The aim in a developing country is to have a smear service on a large enough scale to grade the bacterial content of the sputum smears monthly in the first few months of treatment, and at 6 and 12 months. Fluorescence microscopy offers the substantial ad- vantages of allowing more specimens to be examined and the possibility of instituting adequate supervi- sion of the reading of the smears. However, its use often entails some degree of centralization of the examination of sputum specimens, and operational research is needed to see to what extent the gains might be outweighed by the disadvantages of delay in reporting results and the possible loss of interest in the outlying services. When, with the development of the service, it becomes possible to undertake cultures, their most important use is for diagnostic purposes, in order to ensure as far as possible that all patients brought under treatment are, in fact, suffering from the disease. This recommendation takes into account not only the importance of concentrating the resources of the therapeutic services on patients actually suffering from the disease but also the prior- ity of bringing all infectious cases under treatment. Sensitivity tests for epidemiological purposes (dis- cussed earlier) have a somewhat lower priority; so have sensitivity tests for the individual patient. As discussed above, the priority in the individual patient is to test cultures from those who have "failed" during or after treatment with first-line drugs. 28 MYCOBACTERIAL DRUG SENSITIVITY TESTS: RECENT ADVANCES AND PRACTICAL USE 29 II. SENSITIVITY TESTS USED BY THE MEDICAL RESEARCH COUNCIL OF GREAT BRITAIN FOR FIRST- AND SECOND-LINE DRUGS 1 MEDIUM Sensitivity tests are done on Lowenstein-Jensen medium without potato starch (Jensen, 1955). The drugs are added before inspissation, in the concentrations noted below. The medium is dispensed in about 5-ml amounts in 1-fl oz (28-ml) screw-capped bottles and is inspissated once for 50 minutes at 85°C. TESTS FOR ISONIAZID, STREPTOMYCIN AND PAS Inoculum The sensitivity tests are set up with an inoculum, referred to below as the standard inoculum, prepared from the growth on primary diagnostic Lowenstein-Jensen medium slopes. Tests are set up not later than 2 weeks after the slopes become positive (usually within 3 days), and older growths are subcultured if necessary to ensure that the inoculum is composed of young viable organisms. With a 22 SWG 2 (wire diameter 0.7 mm) Nichrome loop, a representative sweep from the growth is taken on the loop, with a volume, judged by eye, of 2 mm3 (approximately 2 mg moist weight of bacilli). The growth taken on the loop is then discharged into 0.4 ml of sterile distilled water contained in 0.25-fl oz (7-mI) screw-capped bottles together with 6 glass beads 3 mm in diameter. Standardization of the size of the inoculum is important and depends on estimating the amount of growth on the loop as 2 mm3: it is advisable that the worker should initially weigh a number of such loopfuls of growth to ensure that there are actually 2 mg of growth on his loop. A suspension is prepared by shaking the 7-ml bottle for 1 minute on a mechanical shaker, and then, with a 3-mm external diameter 27 SWG (wire diameter 0.4 mm) Nichrome loop, a loopful of the suspension is spread on the surface of each slope of the sensitivity test. As a control, a drug-free slope is set up for each strain tested. The standard sensitive strain, H37Rv, is tested in each set of tests and again within each set if the batch of medium is changed. I Prepared by Professor D. A. Mitchison. 'British Standard Wire Gauge. Incubation and reading of tests The slopes are incubated at 37°C. A reading may be made at 2 weeks to give a preliminary indication of the presence of resistant strains, but the definitive reading is made at 4 weeks, and a report that a strain is sensitive should not be given earlier. For all tests, " growth " is d2fined as the presence of 20 or more colonies. The resistance ratio (RR) is the minimal concentration inhibiting growth (as defined above) of the test strain divided by the minimal concentration inhibiting growth (as defined above) of the standard sensitive strain, H37Rv, in the same set of tests. Short test After the test has been in use for a period, the range of drug concentrations may be shortened with little loss of information. The concentrations that could be omitted in tests at the Royal Postgraduate Medical School, London, are indicated with an asterisk below. However, it is possible that slightly different concentrations might be needed in other laboratories. The range required for the test strain is determined by the variation in the minimal inhibitory concentration (MIC) of H37Rv, and by the need to determine a resistance ratio of 2 or less for sensitive strains and a resistance ratio of 8 or more for resistant strains. Thus, if H37Rv is inhibited by either 8 pg/ml or 4 ,ug/ml of streptomycin in a series of batches of tests, then it would be necessary to have 8 ,ug/ml as the lowest streptomycin concentration in the test strain range (test strain MIC = 8 ,tg/ml; H37Rv MIC= 4,g/ml; RR= 2 or less) and 32 jug/ml as the highest concentration (test strain grows on 32 jug/ml; H37Rv MIC= 8 jug/ml; RR= 8 or more). Storage Stock solutions of isoniazid, streptomycin and PAS are usually prepared at monthly intervals and stored at 4°C. Drug-containing media as well are stored at 4°C and can be used for at least 2 months after preparation. Procedure with different drugs Isoniazid. Stock solutions are prepared in distilled water and sterilized with a membrane, candle or sintered-glass G. CANETTI AND OTHERS filter, or by very brief autoclaving (5 minutes at 10 lbf/in2 (0.7 kgf/cm2)). Seitz filtration should not be practised as it removes some of the isoniazid. Concentrations in test medium. Test strain: 0.2, 1, 5* and 50* Htg/ml isoniazid H37Rv: 0.025, 0.05, 0.1, 0.2 and 1 sg/ml iso- niazid * Not necessary in the short test. However, the inclusion of 50 gug/ml is helpful in the identification of atypical mycobacteria, since tubercle bacilli growing on this concentration are catalase- negative, whereas atypical mycobacteria are often catalase-positive. Definition of resistance. Sensitive: no growth (less than 20 colonies) on 0.2 fig/ml isoniazid. Resistant: growth on 1 cg/ml isoniazid or growth on 0.2 ,Ag/ml followed by growth on 0.2 ug/ml in a further test done from the control slope or on a culture from the same patient. Streptomycin. Stock aqueous solutions are prepared with aseptic precautions from sterile ampoules of streptomycin sulfate. Concentrations (before inspissation) in test medium. Test strain: 8, 16, 32, 64 * and 1024 * tg/ml strepto- mycin H37Rv: 2, 4, 8 and 16 * ,tg/ml streptomycin * Not necessary in short test. Definition of resistance. Sensitive: a resistance ratio of 2 or less Resistant: a resistance ratio of 8 or more or a resistance ratio of 4 followed by a resistance ratio of 4 or more in a further test done from the control slope or on another culture from the same patient. PAS. Stock solutions are prepared as for isoniazid or from sterile ampoules of sodium PAS dihydrate. Concentration in test medium. Test strain: 1*, 2, 4, 8 and 16* ,tg/ml sodium PAS dihydrate H37Rv: 0.25*, 0.5, 1 and 2 ,ug/ml sodium PAS dihydrate * Not necessary in short test. Definition of resistance. As for streptomycin. TESTS FOR ETHIONAMIDE, CYCLOSERINE, KANAMYCIN AND VIOMYCIN Inoculum The standard inoculum is prepared and used in the same way as in tests for isoniazid, streptomycin and PAS. Stock solutions and storage ofmedium Cycloserine and especially ethionamide are both heat-labile. For these two drugs, stock solutions should be made up fresh for each batch of medium. Sterile water and containers are used but the final solution is not sterilized. Ethionamide is made up and diluted in 0.1 N HCI. Drug-containing medium should be transferred to a refrigerator as soon as possible after inspissation and, at all stages in the test, it should be kept at room temperature for the minimum time possible. Ethionamide-containing medium should be used within 2 weeks of preparation, but cycloserine-containing medium is more stable. Kanamycin and viomycin are both heat-stable. Preliminary titration A preliminary titration of culture from 20-25 patients is done in several batches of medium containing drug concentrations above and below the usual end-point for sensitive strains. Thus, for ethionamide, the concentrations should be 0, 10, 20, 40 and 80 ,ug/ml ethionamide. The geometric mean of the usual 20-colony minimal inhibitory concentra- tions obtained in these titrations is calculated. The mean is then multiplied by 1.07, 1.5 and 2.1 (a 1.4- fold series), to give the three drug concentrations to be used in the routine test. Thus, if the geometric mean minimal inhibitory concentration for ethiona- mide is 27 ,tg/ml, the concentrations would be 28 ,tg/ml, 40 ,tg/ml and 56 jug/ml. These concen- trations are referred to below as the lower, the middle and the upper concentrations, respectively. Procedure Test strains are set up on a control, drug-free slope and on one slope containing the lower, the middle and the upper concentrations of the drug concerned. In each batch of tests, 2 pre-treatment control strains are also included, set up on slopes containing the range of concentrations used in the preliminary titration. The results with these control strains are collected at intervals of 6 months to 1 year, and the geometric means calculated as a 30 MYCOBACTERIAL DRUG SENSITIVITY TESTS: RECENT ADVANCES AND PRACTICAL USE check that the lower, middle and upper concentra- tions are still correct. In any batch of tests, the results on these controls also serve as a check that the batch of medium used has not deteriorated (particularly important for ethionamide tests). Incubation and reading of tests All tests are incubated at 37°C and read at 4 weeks. As for the first-line drugs, preliminary readings at 2 weeks may indicate the presence of a resistant strain. Definition of resistance. Sensitive: no growth on the lower concentration (e.g., in the example, 28 ug/ml ethion- amide). Resistant: growth on the middle concentration or the upper concentrations or growth on the lower concentration followed by a similar growth in a further test done on the control slope or on another culture from the same patient. TEST FOR PYRAZINAMIDE Inoculum, incubation and reading of tests Pyrazinamide sensitivity tests are set up with a 1:10 dilution of the standard inoculum suspension. The medium must be acid and colonies take longer to grow. Hence, readings should be made at 6 weeks after inoculation. Medium Acid medium is prepared by adding N HCl very gradually and with continuous shaking to the medium before inspissation (approximately 50 ml N HCl is required for 1.0 litre medium) until the medium is at pH 4.85 0.05, as measured with a glass electrode. Once acidified, the medium should be distributed without delay in 7-ml amounts in 28- ml screw-capped bottles and inspissated, as usual, for 50 minutes at 85°C. The medium may be stored for I month. Procedure Test strains and strain H37Rv are inoculated on to slopes containing 0, 25, 50 and 100 ,ug/ml pyrazinamide. The test may be simplified by using only one drug concentration (100 ,tg/ml pyrazin- amide) but a duplicate pair of drug-free and drug- containing slopes should then be set up from each strain. When inoculating the slopes, care must be taken not to dig the loop into the soft surface of the medium and not to inoculate the margins of the slope near the bottle walls, or in the water of condensation. Readings should be made with a lens, taking into account the most minute colonies. Definition of resistance. Sensitive: growth of less than 10 colonies on 100 ,ug/ml pyrazinamide. (Growth is usually inhibited by 25 jug/ml, but occasional sensitive strains may grow on slopes containing 25 ,sg/ml or even 50 ug/ml pyrazinamide). Resistant: growth of 10 colonies or more on 100 jig/ml pyrazinamide. It is probable that strains with acquired resistance to pyrazinamide always have very high degrees of resistance. Medium at a pH more alkaline than 4.85 is a little less efficient in distinguishing between sensitive and resistant strains. Medium at pH 4.85 has been found to support the growth of the great majority of strains from British, South Indian and Chinese (Hong Kong) patients. However, a slightly more alkaline pH medium (4.95) was necessary to get satisfactory growth of many strains from patients from Uganda. Thus, regional variation in the characteristics of Myco. tuberculosis renders it necessary to determine the optimum pH for strains from different areas of the world. TESTS FOR THIOACETAZONE Tests for thioacetazone sensitivity are of little prognostic value in areas of the world where a proportion of wild strains are naturally resistant to the drug since it is difficult to distinguish between strains with natural and acquired resistance. This is because there is no strong association between natural thioacetazone sensitivity of pre-treatment strains and the response of patients to treatment with a regimen of isoniazid and thioacetazone (Mitchison, 1968). In these areas, it is therefore unprofitable to do sensitivity tests to the drug. Fewer studies using the Medical Research Council sensitivity test methods have been done in areas where natural resistance is rare. However, the following method can be used. Inoculum, incubation and reading of tests Slopes are inoculated with the standard inoculum and read after 4 weeks' incubation. 31 32 G. CANETTI AND OTHERS Stock solutions and storage ofmedium A stock solution of 6400 ,tg/ml thioacetazone is prepared by dissolving 64 mg thioacetazone in 10 ml triethylene glycol (Trigol, British Drug Houses), and dilutions of this solution are also made in triethylene glycol. The triethylene glycol should be warmed at 37°C to reduce its viscosity, and fresh pipettes should be used for each dilution step. The stock solutions are self-sterilizing, and are added to L6wenstein-Jensen medium to give a final concentra- tion of 0.5% triethylene glycol in the medium. All stock solutions and medium should be stored in the dark, but are heat-stable. Procedure Test strains are inoculated on to slopes containing 0, 0.5, 1, 2, 4 and 8 ,g/ml thioacetazone. Strain H37Rv may also be inoculated on to slopes containing 0.12 ,uglml and 0.25 ,tg/ml, but since this strain is not fully sensitive to the drug, its use as a control strain is not recommended. Definition of resistance. Sensitive: no growth (of 20 colonies or more) on 2 yg/ml thioacetazone. Resistant: growth on 2 /ugfml thioacetazone or more. III. THE PROPORTION METHOD (THE SIMPLIFIED AND THE STANDARD VARIANTS) 1 PRINCIPLE OF THE METHOD All strains of tuberculosis contain some bacilli that are resistant to antibacillary drugs. However, in resistant strains, the proportion of such bacilli is considerably higher than in sensitive strains. The proportion method consists in calculating the proportion of resistant bacilli present in a strain. Two appropriate bacillary dilutions, a high and a low one, are inoculated on drug-containing and drug-free medium, in order to provide numerable colonies on both media. The ratio of the number of colonies obtained on the drug-containing medium to the number of colonies obtained on the drug-free medium indicates the proportion of resistant bacilli present in the strain. Below a certain proportion, the strain is classified as sensitive; above, as resistant. There exist two variants of the proportion method: the simplified variant and the standard variant.2 The simplified variant is suitable for current practice: with the criteria adopted, its use under adequate technical conditions causes an almost negligible proportion of sensitive strains to be misclassified as resistant and a very small proportion of resistant strains to be misclassified as sensitive. The standard variant, which requires a wider range of drug-containing media and more work, should be used only in laboratories that are highly specialized I Prepared by Dr G. Canetti and Dr N. Rist. 2 Both variants of the proportion method were worked out in collaboration with Dr J. Grosset. in sensitivity testing; it has relatively little advan- tage over the simplified variant, except for research purposes. THE SIMPLIFIED VARIANT 3 Culture medium; incorporation of drugs into the medium; concentrations of drugs employed L6wenstein-Jensen medium with or without potato starch is used for all the resistance tests. The slopes used are 17 mm x 170 mm and contain 7 ml of medium. The drugs, dissolved in distilled water,4 are incorporated in the medium before coagulation. The control medium without drugs is prepared at the same time as the drug-containing media. After the drug has been added to the medium and after the medium has been distributed in the slopes, the medium is coagulated at 85°C for 50 minutes. The slopes are left at room temperature for 24 hours with cotton-wool plugs and are then covered with rubber caps and stored at 4°C. The period of validity of the media stored at 4°C is 2 months; for media with ethionamide, 1 month. Concentrations of drugs employed Only one concentration is employed per drug. The concentrations are: ' Canetti et al. (1964). 'For ethionamide, thioacetazone and rifampicin, whose solubility in water is very low, 1 % solutions in pure ethylene glycol are first prepared. The subsequent dilutions are made in distilled water. MYCOBACTERIAL DRUG SENSITIVITY TESTS: RECENT ADVANCES AND PRACTICAL USE 33 Isoniazid 0.2 ,g/ml Streptomycin1 4 ,g/ml PAS 2 0.5 ug/ml Thioacetazone 2 pg/ml Ethionamide 20 ug/ml Kanamycin 20 ,ug/ml Cycloserine 30 ug/ml Viomycin 30 jug/ml Capreomycin 20 tg/ml Pyrazinamide 3 1oo jLg/ml Ethambutol 2 pg/ml Rifampicin 40 pg/ml Inoculum With a spatula, a representative sample of 5 mg-10 mg is taken from the primary culture and placed in a spherical, flat-bottomed flask containing 30 glass beads 3 mm in diameter. The flask is shaken for 20-30 seconds; 5 ml of distilled water are added slowly under continuous shaking. The opacity of the bacterial suspension is then adjusted by the addition of distilled water to that of a standard suspension containing 1 mg/ml of tubercle bacilli (or BCG). For further processing and for inoculation, either a calibrated loop or pipettes may be used. Loop. The loop should be of platinum (wire diam- eter 0.7 mm) and have an internal diameter of 3 mm; such a loop delivers 0.01 ml. The 2 bacterial dilutions requiredfor inoculation with the loop are 10-2 mg/ml and 10-4 mg/ml; the 2 inocula are respectively 10-4 mg and 10-6 mg of bacilli for each slope. The dilution 10-2 mg/ml is produced by discharging 2 loopfuls of the bacterial suspension standardized at 1 mg/ml into a small tube containing 2 ml of distilled water and shaking. Similarly, the dilution 10-4 mg/ml is produced by discharging 2 loopfuls of the dilution 10-2 mg/ml into a small tube containing 2 ml of distilled water and shaking. Two slopes of medium without drug and 2 slopes of medium with drug are inoculated with a loopful of each dilution. Pipettes. If pipettes are used, the inoculum for each slope is 0.1 ml. Accordingly, the 2 bacterial dilutions required for inoculation with the pipette are 10-3 mg/ml and 10-5 mg/ml; again, the 2 inocula are 10-4 mg and 10-6 mg of bacilli, respectively, for each slope. The dilutions are prepared by 10-fold dilution 1 Dihydrostreptomycin sulfate, at a concentration corre- sponding to 4 zg/ml base. 'If sodium PAS is used to prepare the solution, 1.38 g of the salt is employed for 1 g of p-aminosalicylic acid. ' Lowenstein-Jensen medium at pH 4.9 is used (see under " Test of sensitivity to pyrazinamide ", p. 36). steps (0.5 ml of the bacterial suspension 1 mg/ml, discharged into 4.5 ml of distilled water, produces the dilution 10-1 mg/ml; 0.5 ml of the bacterial sus- pension 10-1 mg/ml, discharged into 4.5 ml of distilled water, produces the dilution 10-2 mg/ml, etc., down to 10-5 mg/ml). The pipette is changed for each dilu- tion. Two slopes of medium without drug and 2 slopes of medium with drug are inoculated with 0.1 ml of the 2 chosen dilutions. The inoculated slopes are plugged with cotton- wool; no rubber caps are placed on them. The slopes are put in a stand at a very slight angle from the horizontal and placed in the incubator at 37°C. The liquid should cover as much surface of the medium as possible, without touching the cotton- wool plug. When the liquid part of the inoculum has evaporated (24-48 hours) the slopes are cov- ered with rubber caps and left in the incubator at 370C. Reading of tests The results are read for the first time on the 28th day. The colonies are counted only on the slopes seeded with the lowest inoculum that has produced growth. This inoculum either may be the same for the control slopes and the drug-containing slopes; or it may be the low inoculum (10-6 mg of bacilli) for the control slopes and the high inoculum (10-4 mg of bacilli) for the drug-containing slopes. The average number of colonies obtained for the 2 control slopes indicates the number of culturable particles contained in the inoculum. The average number of colonies obtained for the drug-containing slopes indicates the number of resistant bacilli contained in the inoculum. The ratio between the second figure and the first indicates the proportion of resistant bacilli existing in the strain. Below a certain proportion (the critical proportion) the strain is classified as sensitive; above, as resistant. The proportions are reported in terms of percentages. If, according to the criteria indicated below, the result of the reading made on the 28th day is " resistant ", no further reading of the test for that drug is required: the strain is classified as resistant. If the result at the 28th day is " sensitive ", a second reading is made on the 42nd day: this provides the definitive result. Criteria of resistance Any strain showing a proportion of resistant bacilli equal or superior to that indicated below is classified as resistant to the corresponding drug: 34 G. CANETTI AND OTHERS Drug concentration Critical proportion (jig/ml) for resistance (%) Isoniazid 0.2 1 Streptomycin 4 10 PAS 0.5 1 Thioacetazone 2 10 Ethionamide 20 10 Kanamycin 20 10 Cycloserine 30 10 Viomycin 30 10 Capreomycin 20 10 Pyrazinamide 100 10 Ethambutol 1 2 10 Rifampicin 40 1 For the calculation of the proportion of resistant bacilli, the highest count obtained on the drug-free and on the drug-containing medium should be taken, regardless of whether this count is obtained on the 28th day, on the 42nd day, or on the 28th day with one medium and on the 42nd with the other. THE STANDARD VARIANT The preparation of the culture medium, the preparation of the inoculum, the bacillary dilutions employed for inoculation and the numbers of control slopes inoculated with each dilution are exactly the same as for the simplified variant (see pp. 32-33). The only difference consists in the greater number of drug concentrations tested. Concentrations of drugs employed The 3 (or sometimes 2) concentrations of drugs employed are shown in the middle column of Table 1. The concentrations within boxes, which are those used in the simplified variant, are the main concentrations; the others are subsidiary. With each of the 2 bacillary dilutions used as inoculum (see under 'Inoculum ", p. 33) 1 slope of the main drug concentration and 1 of each subsidiary drug concentration are inoculated. Criteria of resistance The critical proportions indicating resistance are shown in Table 1.2 Reading of the results The results are read for the first time on the 28th day. The procedure of counting the number of I The criterion indicated is tentative, owing to limited experience. 2 The rationale for these criteria is discussed by Canetti et al. (1963). TABLE 1 CRITERIA OF RESISTANCE APPLICABLE IN THE STANDARD VARIANT OF THE PROPORTION METHOD a Concentration Critical proportionDrug (ig/ml) for resistance I_ Isoniazid 0.1 1 0.2 1 1 0.21 b Streptomycin 4 10 8 0.1b PAS 0.25 10 0.51 1 1 0.1b Thioacetazone 1 50 2 1 0 4 1 Ethionamide 10 50 20- 10 40 1 Kanamycin 10 50 20 1 0 30 1 Cycloserine 20 50 30 10 40 1 Viomycin 20 50 30 1 0 40 1 Capreomycin 20 10 40 1 Pyrazinamide 50 50 1100 10 400 1 Ethambutol c 1 50 2 10 3 1 Rifampicin 20 10 40 1 a The main drug concentrations and the criteria for them are boxed. b This criterion cannot be employed for the diagnosis of sensitivity (<0.1 % bacilli resistant to the corresponding drug concentration) if the average number of colonies on the control- slopes seeded with the low inoculum is less than 20. c The criterion indicated is tentative, owing to limited ex- perience. MYCOBACTERIAL DRUG SENSITIVITY TESTS: RECENT ADVANCES AND PRACTICAL USE 35 colonies and calculating the proportion of bacilli resistant to a given drug is the same as with the simplified variant. However, for the slopes with drugs, the counting on the 28th day may be restricted to those containing the main drug concentration. If the result is " resistant ", the strain is classified as such: no further reading of the test is required for that drug. If the result is " sensitive ", the test is read again on the 42nd day. On the 42nd day, the colonies are counted first on the slopes without drug and on the slopes containing the main drug concentration. As stated above, for the calculation of the proportion of resistant bacilli the highest count obtained on the drug-free and on the drug-containing medium should be taken, regardless of the date at which this count is obtained. If the result is " resistant ", no count- ing of the colonies on the other drug concentra- tions is needed: the strain is classified as resistant. If the result for the main drug concentration is " sensitive ", the colonies are also counted on the subsidiary drug concentrations, and the correspond- ing criteria for resistance are applied for definitive classification of the strain, as indicated below. Application of criteria of resistance for subsidiary drug concentrations For mathematical and biological reasons, criteria of resistance for the subsidiary drug concentrations may be slightly less reliable than those for the main drug concentrations. Consequently, if the criteria applied under the conditions specified above indicate "resistance" for both subsidiary drug concentra- tions, the strain is definitively classified as resistant. If they indicate " resistance " for one subsidiary drug concentration, the strain is retested. If, at the second test, the criteria again indicate " resistance " for one subsidiary drug concentration (or for both, or for the main drug concentration), the strain is definitively classified as resistant. If they indicate " resistance " for no subsidiary drug concentration (nor for the main drug concentration) the strain is definitively classified as sensitive. However, for isoniazid and streptomycin, strains showing resistance at one subsidiary drug concentra- tion are classified as resistant without further retesting. THE DIRECT TEST Any sputum specimen in which microscopic examination demonstrates a sufficient number of bacilli may be submitted to a direct test by the proportion method. The saving of time, in comparison to the indirect test, is 3-4 weeks. A smear is prepared from the purulent part of the sputum. If the smear shows at least 1 bacillus per oil-immersion-objective field, a direct test may be made. Two millilitres of sputum are placed in a sterile mortar. An equal quantity of 4 % NaOH and 3 drops of tincture of litmus are added. The mixture is triturated with the pestle for 1-2 minutes, wrapped in sterile paper, incubated at 37°C forN 50 minutes and neutralized with a few drops of 15% sulfuric acid. The product thus obtained constitutes dilution 1.1 Further dilutions are prepared by 10-fold steps," using pipettes. If the smear has shown 1-10 bacilli per field, the 2 dilutions chosen for inoculation are 1 and 10-2 (10-3 may also be inoculated on 2 tubes without drugs). If the smear has shown more than 10 bacilli per field, the 2 dilutions chosen for inoculation are 10-1 and 10-3. The number of tubes inoculated with each dilution, the reading of the results and the application of the criteria for resistance 2 are exactly the same as in the previously described tests. The direct test may be used as the simplified variant (inoculation of one concentration per drug) or as the standard variant (inoculation of more than one concentration per drug) in exactly the same way as the indirect tests described previously. However, since the frequency of failures of the direct test owing to a dearth of culturable particles in the inoculum is not negligible, the simplified variant should be generally preferred, for the sake of economy. ADDITIONAL RECOMMENDATIONS Inoculum The inocula indicated above (p. 33) usually provide satisfactory numbers of culturable particles for the test: more than 10 and less than 100 colonies per control slope seeded with the smaller inoculum. However, in some laboratories, the yield may prove smaller, owing to different growth conditions. In I If a method involving centrifugation of the treated sputum specimen is used, dilution 1 will be provided by the centrifugation-sediment resuspended in 4 ml of distilled water. 2The criteria for ethambutol are not valid in direct tests. 36 G. CANETrI AND OTHERS such cases, a different set of inocula may be adopted for current use, for instance: 5x10-4 mg and 5x10-6 mg or 10-3 mg and 10-5 mg For tests on single colonies, the set of inocula employed should always be 10-3 mg and 10-5 mg owing to the smaller yield of culturable particles from single colonies. Any indirect or direct test showing an average of less than S colonies in the control slopes seeded with the smaller inoculum should be repeated (from the controls). Dysgonic growth A certain amount of dysgony on the drug- containing media, with normal growth on the controls, is a frequent occurrence in resistance tests and does not impair the reading. However, some strains grow very dysgonically in the drug- containing and in the drug-free slopes. Since the negative influence of dysgony on the number of colonies that develop is usually more pronounced in the slopes seeded with the smaller inoculum than in those seeded with the larger inoculum (and may vary considerably between slopes seeded with the same inoculum) the proportions calculated from one inoculum are sometimes highly inconsistent with those calculated from the other, and no reliable figures can be given. In such cases, the definitive reading of the test should be postponed for 2 weeks. If the situation has not changed, the test should be repeated (from the control slope). Test of sensitivity to streptomycin Growth on dihydrostreptomycin-containing Lowenstein-Jensen medium is highly dependent on the oxygen supply available in the container: the less oxygen, the more growth.' The amount of oxygen available in different medium containers (slopes, bottles, etc.) depends on several factors (amount of medium, surface area of the medium, size of the containers, and principally tightness of the cap). The criteria for streptomycin resistance indicated above apply to a type of container, an amount of medium and a type of slope closure that have been specified (p. 33). The use of the same criteria under different culture conditions without previous assess- ment may be highly misleading. 1 With streptomycin, the oxygen effect is even more pronounced. Test ofsensitivity to pyrazinamide For tests of sensitivity to pyrazinamide, acid L6wenstein-Jensen medium at pH 4.9 is used. With many strains, growth on such a medium is less abundant than on standard L6wenstein-Jensen medium (pH 6.8). For tests of sensitivity to pyrazinamide, control slopes with acid medium should be inoculated, in addition to the normal controls. If the number of bacilli growing on the acid controls amounts to less than 10% of the number of bacilli growing on the standard controls, the test should be repeated. The 2 inocula used for the test are 10 times stronger than the usual ones: 10-3 mg of bacilli and 10- mg of bacilli. The number of culturable particles on which the calculation of the proportion of resistant bacilli is based is the number counted on the acid controls. The results of tests of resistance to pyrazinamide are less reliable than those of tests of sensitivity to other drugs. The use of direct tests often leads to the misclassification of resistant strains as sensitive. CONTROL TESTS Routine check of drug-containing media Each new batch of drug-containing medium should be checked for drug activity, in order to disclose errors in the amount of drug dispensed or in heating. For the control test, the same wild strain -for instance, H37Rv-should be used throughout. The amount of growth produced by a known inoculum on the new drug-containing medium should fall within a range of values established previously. The control test is done only once for each batch (a repetition each time the same medium is employed is useless). The type of control test used at the Institut Pasteur, Paris, and its results are illustrated in Table 2. A 6-8-day-old culture of H37Rv in Dubos liquid medium is standardized opacimetrically to 1 mg/ml (= dilution 1): under the growth conditions prevailing, dilution 1 contains 5 to 100 x 10-6 culturable particles per ml (approximately 90% of the tests). Growth from a Lowenstein-Jensen culture standardized to 1 mg/ml may be used instead of a Dubos culture, but the results may be slightly different. Appropriate dilutions are inoculated on 2 slopes (per dilution) of the new batch of drug- containing medium and on 2 control slopes: the dilutions most suitable for each drug and for the MYCOBACTERIAL DRUG SENSITIVITY TESTS: RECENT ADVANCES AND PRACTICAL USE TABLE 2 NORMAL RANGE OF RESULTS a OBTAINED AT THE INSTITUT PASTEUR WITH CONTROL TEST FOR NEW BATCHES OF DRUG-CONTAINING LWWENSTEIN-JENSEN-MEDIUM, USING STRAIN H37Rv AND THE MAIN DRUG CONCENTRATIONS SHOWN IN TABLE 1 Dilution b Controls soniazid Strepto-i PAS Ethionamide 1 Cyclo- Viomycin Kanamycininoculatedb no (0.2 .tLgl/mI) I m cn serine(4 Ag/mI) (0.5 Ag/mI) (20 pg/mI) (30 Ag/mI) i(30 ug/ml) (20 Ag/ml) I I~~~~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 5->100 < 1-20 2-50 0-5 1-20 0-2 ++ 0-10 <100 ++ 0 0-10 <100 <100 0-10 0 ++ >100 ++ <100 >100 0-10 <100 oc 0 <100 0-10 <50 0-5 a Results expressed as growth (number of colonies) on the 28th day; +++ + = confluent eugonic growth; ++ = innumer- able distinct colonies, sometimes minute. b Dilution I = 6-8-day-old Dubos culture standardized at I mg/mi. Inoculum per slope: 0.2 ml. c Reading at 42 days: 0-50 colonies. controls are those against which data have been given in Table 2. The inoculum per slope is 0.2 ml (instead of the routine 0.1 ml), in order to obtain more culturable particles: it produces < 1-20 colonies on a control slope inoculated with a 10-6 dilution. At the 18th to 21st day, a preliminary reading of the test is made. If growth on the drug-containing slopes is equal to growth on the controls, an error is sure to have occurred, and the batch is discarded. The definitive reading of the test is made on the 28th day. Table 2 indicates for each drug the normal range of results obtained at the Institut Pasteur: they are based on 100-120 batches ofmedium for isoniazid, streptomycin and PAS, and 40-60 batches ofmedium for the other drugs (main drug concentrations). On the 28th day, the colonies on the drug- containing media are often considerably smaller than those on the controls: they may be of borderline visibility. The range of findings for isoniazid, streptomycin, cycloserine, viomycin and kanamycin is narrow. For PAS and ethionamide, the range is wider: for each, 3 subcolumns in Table 2 summarize the different types of findings, none of which is incompatible with the expected drug concentration in the medium; these differences are due mainly to fluctuations of the test strain. Insufficient drug concentrations (through errors in the quantity of drug added, or excessive heating) are disclosed by considerably more growth on the drug- containing slopes. In addition, the colonies are usually ofgreater size. Excessive drug concentrations are more difficult to detect, since for most drugs the lower limit of growth of the strain H37Rv compatible with the expected drug concentration is close to 0 with the chosen inoculum (see Table 2). In laboratories using more than one concentration per drug (standard variant of the proportion method) an excessive amount of drug added to all the corresponding media may be revealed by insufficient growth on the medium with the lowest concentration used, since the lower limit of growth of strain H37Rv compatible with the expected drug concentration is higher on that concentration. In laboratories using only one drug concentration per drug (simplified variant of the proportion method), an excessive drug concentration in the new batch may be disclosed either by increasing several times the inoculum of dilution 1 in the control test described above, or by running a parallel series of control tests with a second wild strain, slightly less sensitive than H37Rvto most drugs. The new batch of medium is obviously used before the results of the control test are available. If the test shows that an error has occurred, the strains for which the batch has already been employed should be retested on a new batch. The findings indicated in Table 2 apply to the culture conditions (duration of inspissation of the 10-1 10-2 10-3 10-5 106 37 38 G. CANETTI AND OTHERS medium, size of slopes, amount of medium, slope closure, etc.) indicated above (pp. 32-33): the findings may be different under different culture conditions. Even if the culture conditions are apparently the same, the findings may vary to some extent from one laboratory to another. Consequent- ly, each laboratory producing drug-containing media has to build up its own experience with regard to results of control tests for drug-containing media. Table 2 is given merely for preliminary orientation. The results obtained with H37Rv are not average results for wild strains, at least with regard to certain drugs. H37Rv (Pasteur variant) is more sensitive to streptomycin and less sensitive to PAS, ethionamide and thioacetazone than most other wild strains. H37Rv is nevertheless used for checking the drug- containing med,ia because of its relative stability. Modification of the criteria of resistance The resistance criteria indicated above have been arrived at by comparing large numbers of wild strains and of strains from patients treated with the corresponding drug during increasing lengths of time. They apply to specific culture conditions which have been described previously (pp. 32-33): their validity under these conditions has been confirmed in many laboratories. Under different culture conditions they may require modification. Attention has already been drawn to the importance of this factor in the case of streptomycin (see under " Test of sensitivity to streptomycin ", p. 36). In the case of ethionamide and thioacetazone also, culture conditions tend to change, although to a lesser degree. The risk is minimal with isoniazid. The best way of assessing the validity of resistance criteria to be used in a given laboratory is to apply the calibration method described in Part IV below. However, this approach, which should be applied eventually in all laboratories, requires familiarity with the method of testing; i.e., it cannot be used at the very beginning of testing practice. A preliminary estimation of the validity of the proposed criteria under the existing culture conditions may be made in two ways. Growth of the H37Rv strain A control test (as described on p. 36) should be performed with H37Rv on 3 different, carefully prepared, batches of drug-containing Lowenstein- Jensen medium. If the results are within the limits indicated in Table 2, the criteria of resistance, as given in Table 2, are probably valid for the laboratory concerned. Testing of sensitive and resistant strains receivedfrom a highly qualified reference laboratory using the resistance criteria indicated in Table I If the proportions of strains classified as sensitive and as resistant are significantly different from those indicated by the reference laboratory, and if technical shortcomings can be excluded, the corresponding criterion of resistance most likely requires a modification in the laboratory concerned. The new criterion should be elaborated with the calibration method described in Part IV, and expressed in terms of proportions. (For instance, the new criterion of resistance to ethionamide might be 5% on 20 ug/ml, instead of 10%.) The fact that, under certain culture conditions, the criterion of resistance needs to be changed does not interfere with the principles of the proportion method, but merely with the validity of certain figures given in Table 1. IV. THE CALIBRATION OF SENSITIVITY TESTS1 The method for obtaining definitions of drug resistance to be described compares a population of strains that are predominantly sensitive with a population of strains that are predominantly resistant, and obtains the definitions of resistance 1 Prepared by Professor D. A. Mitchison. Some of the statistical procedures were suggested by Dr S. Radhakrishna of the Tuberculosis Chemotherapy Centre, Madras, and Miss Ruth Tall of the Medical Research Council's Tuber- culosis and Chest Diseases Research Unit, London. that best distinguish between two populations. The method has the following advantages: (1) Pure populations of sensitive and resistant strains are not required, so that no preconceptions about definitions of resistance are necessary. (2) No correlation with the clinical progress of the patient under treatment with the drug concerned is necessary, though such information may provide confirmation of the definition obtained. MYCOBACTERIAL DRUG SENSITIVITY TESTS: RECENT ADVANCES AND PRACTICAL USE (3) The efficiency of radically different methods of sensitivity testing may be compared. It has the following disadvantages: (1) Large numbers of sensitivity tests must be done. (2) A large sample of strains that may have acquired resistance must be available. POPULATIONS OF STRAINS The following groups of strains are required. Probably sensitive strains These strains are conveniently obtained from any patient who has not been treated with the drug concerned previously. At least 100 (and preferably more) strains are required. It is advisable, though not essential, for each strain to be obtained from a different patient. The presence of a few strains with primary or acquired resistance to the drug concerned is immaterial. Probably resistant strains At least 50 strains, and preferably as many as the probably sensitive strains, should be obtained from patients who have remained sputum-positive during treatment with the drug concerned. In selecting these strains it is possible to use more than 1 strain from the same patient provided that they are obtained at different intervals of time following the start of chemotherapy. The aim should be to include not only strains that are very likely to be resistant, but also some in which resistance is emerging, and a few that may well still be sensitive. The efficiency of the procedure is improved by the inclusion of numerous strains of doubtful resistance in the probably resistant population. SENSITIVITY TESTS If the method is to be used for the calibration of one sensitivity test, several sets of tests are set up, each on a separate batch of medium. Each batch should contain a similar ratio of probably sensitive to probably resistant strains. When several methods of testing sensitivity (to the same drug) are being compared, the tests should be set up simultaneously on the same populations of probably sensitive and probably resistant strains, or, when feasible, the different methods can be made part of one larger test. Thus, a single " population study " test, including a range of inoculum dilutions and drug concentrations, can be read by the absolute-concentration and proportion methods, allowing a comparison of these two methods. In making comparisons between laboratories (for instance, between a central reference laboratory and an outlying laboratory) it is unnecessary to take special precautions to guard against instability of resistance in the probably resistant strains. EVALUATION OF RESULTS The results of the sensitivity tests are set out as cumulative distributions of the results with the probably sensitive and the probably resistant populations. An example for streptomycin sensitivity tests, read by the absolute-concentration and the proportion methods, is set out in Table 3 (R. W. Riddell, personal communication). For each level of sensitivity in these distributions, the cumulative percentage for the probably sensitive strains is subtracted from the corresponding cumulative percentage for the probably resistant strains, to give the " difference between percentages ". The latter percentage is a measure of the efficiency of each level of sensitivity in distinguishing between sensitive and resistant strains. As the level of sensitivity in the distribution increases, the difference between percentages usually increases, to reach a peak, and then decreases again. The peak value indicates the most efficient definition of resistance. However, it is often advisable to choose a definition at a level of sensitivity very slightly higher than this optimum value so as to diminish the chance of misclassifying sensitive strains as resistant. Alternatively, if tests are available on at least 2 strains from a patient, the definition of resistance giving the peak difference between percentages may be used as an indication of resistance provided that it shows all the strains from the same patient to be resistant. Considering first the minimal inhibitory concen- tration obtained in the streptomycin sensitivity tests of the example (Table 3), the peak difference between percentages (52.4%) was obtained by using a definition of resistance of a minimal inhibitory concentration of 16 ,tg/ml dihydrostreptomycin or more. Correspondingly, with proportion tests, the peak difference between percentages (55.4 %) was obtained with proportions on 4 ,ug/ml dihydrostrep- tomycin of 0.5% or more and of 1 % or more. In the application of this procedure to different sensitivity-test methods, the effect of grouping of the data should be recognized. In some methods of 39 G. CANETI1 AND OTHERS TABLE 3 CUMULATIVE DISTRIBUTIONS OF SENSITIVITY TO DIHYDROSTREPTOMYCIN Measure of sensitivity Minimal inhibitory concen- tration (Mg/ml dihydro- streptomycin) 8 or more 16 or more 32 or more Proportion on 4 AgIml dihydrostreptomycin 0.01 % or more 0.1 % or more 0.5 % or more 1 % or more 5 % or more 50 % or more Total no. of strains tested Number and cumulative percentage of strains Probably sensitive(PS) No. % 39 20.3 9 4.7 5 2.6 81 29 14 12 9 4 42.2 15.1 7.3 6.2 4.7 2.1 192 Wild No. % 32 4 73 23 9 7 4 0 18.8 2.4 0.6 42.9 13.5 5.3 4.1 2.4 0.0 170 Probably resistant (PR) No. % 118 66.7 101 57.1 84 47.5 149 84.2 117 66.1 111 62.7 109 61.6 98 55.4 48 27.1 177 reading tests, the measure of sensitivity can be taken anywhere on a continuous scale of sensitivity. For instance, in the proportion method, it is possible for routine use to choose any value of the proportion (such as, say, 1.23 %) to examine as a potential definition of resistance, though it is desirable in practice to choose a round number that can be remembered easily. On the other hand, many methods of reading sensitivity are discontinuous. For instance, in the absolute-concentration method, readings of the minimal inhibitory concentration cannot be obtained between the drug concentrations used in the test. Bearing this in mind, it is evident from Table 3 that a definition of resistance of a minimal inhibitory concentration between 8 ttg/ml dihydrostreptomycin or more and 16 ,ug/ml dihy- drostreptomycin or more might have been more efficient in detecting resistance than a minimal inhibitory concentration of 16 jug/ml or more. However, no such readings are available since slopes with concentrations between 4 ,ug/ml and 8 jug/ml dihydrostreptomycin were not used in the tests. Cleaily, intermediate concentrations between 4 ,ug/ml and 8 tg/ml and, possibly, between 8 ,ug/ml and 16 ,ug/ml should have been included if the absolute-concentration method for streptomycin were to have been calibrated with greater efficiency. A further feature of the grouping effect is that it tends to make accurate comparisons difficult between methods with continuous distributions of sensitivity and those with discontinuous distribu- tions. One way of allowing for this difficulty is as follows. The distributions of the probably sensitive strains for the two types of test are examined in the range of optimum discrimination, to find some measure of sensitivity which yields the same number of probably sensitive strains consid'ered as " resis- tant" by both methods. Thus, in Table 3, 9 probably sensitive strains would be considered as " resistant " if they had a minimal inhibitory concentration of 16 ,ug/ml dihydrostreptomycin or more or if they had a proportion on 4 ,ug/ml dihydrostreptomycin of 5% or more. If the two methods are of fairly similar efficiency, the difference Difference between percentages (PR-PS) 46.4 52.4 44.9 42.0 51.0 55.4 55.4 50.7 25.0 40 MYCOBACTERIAL DRUG SENSITIVITY TESTS: RECENT ADVANCES AND PRACTICAL USE between percentages at these cuts (52.4% for the minimal inhibitory concentration definition and 50.7% for the proportion definition) may be considered to be approximately equivalent. An alternative statistical procedure is to attempt to fit curves to the cumulative difference between percent- ages for each method of sensitivity testing, but it has been found difficult so far to get a sufficiently good fit to all of the curves. Finally, many bacteriologists will wish to know what proportion of wild strains will be classified as resistant by any of the calculated definitions of resistance, since they will wish to know how the definition obtained by this discrimination procedure compares with previous experience. For this purpose it is useful to " purify " the probably sensitive population by removal of any strain found resistant by all of the definitions of resistance chosen for each method of measuring sensitivity. This procedure removes strains which are clearly resistant from the probably sensitive population, leaving a group of strains assumed to be wild. In the example with streptomycin (Table 3) it is evident that definitions of resistance of either a minimal inhibitory concentration of 16 ,g/ml dihydrostrepto- mycin or more, or a proportion on 4 jug/ml dihydrostreptomycin of 5% or more, would each classify 4 (2.4y%) of the wild strains as resistant. In interpreting this result, one should recall that the strains from the 4 patients concerned may have been resistant by one measure but not by the other. In any case, the number of wild strains found resistant is satisfactorily low. APPLICATIONS When a laboratory is initiating sensitivity tests to a drug, it is envisaged that it will first carry out tests for a trial period by whichever method it favours, keeping records of whether the strains have been obtained before treatment, and therefore belong to the probably sensitive population, or have been obtained after treatment for several months with the drug concerned, and therefore belong to the probably resistant population. Having accumulated a sufficient number of results, an analysis would be done by the procedure indicated above to see whether the provisional definition of resistance adopted for these preliminary tests was of maximum efficiency. If not, the definition of resistance, and possibly the detailed design of the sensitivity tests (for instance, the drug concentrations used), would be changed. Once the tests are established, succes- sive checks on the efficiency of the definitions in use should be done at intervals of, say, a few years. COMPARISONS BETWEEN LABORATORIES The most efficient method of comparing the work of different laboratories, e.g., a peripheral laboratory and a central laboratory, is to carry out the calibration procedure described above in both laboratories. In practice, the central laboratory would maintain batches of probably resistant strains for each drug and would have established what was the degree of discrimination (i.e., the value of the " difference between percentages ") obtainable when comparing these strains with a probably sensitive population in its own tests. The second laboratory would then test the same probably resistant population, and the efficiency of its tests would be measured by the value of the " difference between percentages" obtained. An alternative method is for the central laboratory to send out a number of strains with varying degrees of sensitivity which it has tested, for testing at the peripheral laboratory. Such a procedure has the advantage that it requires less work to be done by the peripheral laboratory. However, it has the following disadvantages: (1) Great care must be taken to see that no reversion towards sensitivity occurs in the strains during storage in the central laboratory, during transport to the peripheral laboratory, or during the period before the sensitivity test is set up in the peripheral laboratory. For an accurate comparison, the sensitivity tests, and all previous subculturing of the exchange strains, should be done under identical conditions in both laboratories-conditions which are difficult to fulfil. (2) The tests, especially those in the peripheral laboratory, are likely to be done with greater care than in the performance of routine tests. In consequence, the results of the comparison may not be representative of those obtained in routine tests. 4 41 G. CANETI AND OTHERS RI-SUMI PROGRES DES TECHNIQUES DESTINE-ES A EVALUER LA SENSIBILITt DES MYCOBACtRIES AUX MEDICAMENTS; EMPLOI DES EPREUVES DE SENSIBILITE AU COURS DES PROGRAMMES DE LUlTE ANTITUBERCULEUSE Un groupe de consultants de l'OMS s'est reuni a Geneve du 23 au 26 avril 1968. Ces specialistes se pro- posaient de faire le point des progres recents intervenus en matiere de mesure de la pharmacoresistance des mycobacteries tuberculeuses et de suggerer des moda- lites pratiques de recours aux epreuves de sensibilit6 aux differents stades des operations de lutte antituber- culeuse. Le present document expose leurs conclusions. II est preferable de definir la resistance en se fondant sur des criteres purement bacteriologiques. I1 est en effet demontr6 que pour la plupart des medicaments anti- tuberculeux la resistance au sens bacteriologique a souvent comme corollaire clinique une diminution de l'efficacite therapeutique. L'etude comparative des pro- cedes de recherche de la sensibilite fait apparaitre que la methode de la concentration inhibitrice minimale, la methode des proportions et la methode des rapports de r6sistance (resistance ratio) permettent avec un egal succes d'evaluer la resistance d'une souche a la strepto- mycine. On peut egalement appliquer les deux premi&res de ces methodes a la mesure de la resistance a l'isoniazide; la methode des rapports de resistance n'a jamais et utilis6e a cet effet. Quant a la valeur comparative des trois procedds pour l'evaluation de la sensibilite au PAS, elle n'a apparemment fait l'objet d'aucune recherche. La m6thode des rapports de r6sistance ne semble pas avantageuse pour l'etude des medicaments de (' deuxieme ligne *. On peut en revanche recourir aux deux autres methodes, bien que les avis different au sujet de leur valeur comparative et que de nouvelles recherches soient ndcessaires 'a cet egard. II est particulierement important, dans la pratique des tests de resistance, d'opdrer avec des semences dont la teneur en unit6s viables varie dans des limites relative- ment etroites et soit au moins approximativement connue. Des semences de poids et d'opacite identiques peuvent donner des nombres de colonies tres differents, ce qui tient non seulement aL des differences dans le nombre d'unites viables pr6sent, mais egalement a des differences dans la proportion des bacilles isoles et celle des bacilles groupes en amas. Lorsqu'un laboratoire adopte telle ou telle methode de mesure de la sensibilite, il importe qu'il verifie au bout d'un certain temps que les criteres de resistance proposes par les auteurs de la methode sont valables dans les conditions locales: une methode d'eta- lonnage des tests permettant d'y parvenir est decrite dans le rapport. Le pronostic qui s'attache a la resistance est different selon qu'il s'agit de resistance primaire ou de resistance acquise. II a ete demontre que l'incidence de la resistance primaire sur les resultats therapeutiques est pratiquement nulle s'il s'agit de resistance a un seul des trois medica- ments de premiere ligne, ai condition que le malade soit trait6 par ces trois medicaments ai la fois. En cas de resistance acquise - c'est-a-dire apparue en cours de traitement - il est vain d'attendre du medicament homologue la moindre action th6rapeutique. L'epidemiologie de la resistance se pr6sente differem- ment dans les pays d6veloppes et dans les pays en voie de developpement. Dans les pays d6veloppes, les enquetes sur la resistance chez les malades non encore traites renseignent reellement sur la r6sistance primaire. Celle-ci n'atteint nulle part des taux alarmants, et comme il s'agit dans la grande majorite des cas de resistance a une seule drogue et que tous les malades peuvent etre traites par les trois drogues de premiere ligne a la fois, ces enquetes n'ont guere d'utilite therapeutique. Elles fournissent cependant une indication sur l'intensite des transmis- sions bacillaires qui se produisent dans la collectivite etudiee. Dans les pays en voie de d6veloppement, les enquetes correspondantes apportent une donnee plus complexe, la resistance initiale, ofu se melent deux 616- ments: d'une part, la r6sistance primaire, d'autre part, la resistance acquise lors de traitements non avoues ou ignor6s du malade. La resistance initiale reflate jus- qu'"a un certain point la diffusion et la qualitd de la chimiotherapie pratiqu6e dans le territoire correspondant. Sa connaissance est tres utile pour la formulation de programmes de chimiotherapie efficace dans ce territoire. La connaissance du nombre total de malades infec- tieux 'a bacilles resistants qui existent dans une commu- naute serait fort utile: mais on ne la possede que pour quelques rares villes, oiu ces malades figurent sur un registre des cas r6sistants. La mise en ceuvre de tests de resistance chez des malades en cours de traitement est surtout utile dans deux circons- tances: chez les malades qui gardent une expectoration positive apres 6 mois de traitement, et chez les malades dont l'expectoration, apres s'etre negativee, redevient positive. L'institution d'un traitement par des drogues de seconde ligne peut etre faite sans test de resistance a ces drogues, si le malade ne les a encore jamais reques. I1 est tout a fait inutile de faire des tests de resistance en cours de traitement lorsque les examens directs et les cultures montrent une diminution de la population bacil- laire. On risque meme alors, sur la foi de resultats errones ou mal interpr6tes, de nuire au malade, en rem- placant un regime efficace par un regime plus toxique ou moins facilement accepte. Les tests de resistance ne devraient etre pratiques que dans des laboratoires bien equipes, assures d'une activite suffisante pour pouvoir maintenir un niveau technique 42 MYCOBACTERIAL DRUG SENSITIVITY TESTS: RECENT ADVANCES AND PRACTICAL USE 43 eleve, et soucieux de controler constamment ce niveau. La centralisation de la pratique des tests de resistance devrait etre fortement encouragee. Dans la perspective de leur utilit6 globale dans un programme de lutte contre la tuberculose, les divers examens de laboratoire peuvent etre classes dans l'ordre suivant: 1) les examens directs; 2) les cultures pratiquees dans un but diagnos- tique chez les malades dont 1'expectoration est negative a 1'examen direct; 3) les dpreuves de sensibilite pratiqu6es dans un but 6pid6riiologique; 4) les epreuves de sensi- bilite pratiqu6es pour des cas individuels. Dans un pays en voie de developpement, il est particulierement impor- tant de respecter cet ordre de priorit6s lorsqu'on procede a l'implantation des organes d'ex6cution n6cessaires. REFERENCES Bartmann, K. & Galvez-Brandon, J. (1968) Scand. J. resp. Dis., 49, 141-152 Canetti, G. (1965) Amer. Rev. resp. Dis., 92, 687-703 Canetti, G., Froman, S., Grosset, J., Hauduroy, P., Langerovi, M., Mahler, H. T., Meissner, G., Mitchi- son, D. A. & gula, L. (1963) Bull. Wld Hlth Org., 29, 565-578 Canetti, G., Grosset, J. & Cetrangolo, J. (1964) Arch. Inst. Pasteur Alger., 42, 14 Canetti, G., Kreis, B., Thibier, R., Gay, P. & Le Lirzin, M. (1967) Rev. Tuberc. (Paris), 31, 433-474 Canetti, G., Rist, N. & Grosset, J. (1963) Rev. Tuberc. (Paris), 27, 217-272 International Union against Tuberculosis (1964) Bull. in!. Un. Tuberc., 34, 83 Jensen, K. A. (1955) Bull. int. Un. Tuberc., 25, 89 Krebs, A., Kappler, W. & Ferreira, 0. T. (1962) Beitr. Klin. Tuberk., 126, 84-97 Lefford, M. J. & Mitchison, D. A. (1966) Tubercle (Edinb.), 47, 250-262 Mitchison, D. A. (1962) Bull. int. Un. Tuberc., 32, 81-99 Mitchison, D. A. (1968) Tubercle (Edinb.), 49, Suppl., p. 38 Rist, N. (1968) Tubercle (Edinb.), 49, Suppl., pp. 36-46 Thomas, K. L., Joseph, S., Subbaiah, T. V. & Selkon, J. B. (1961) Bull. Wld Hith Org., 25, 747-758