- Email: [email protected]
PREDICTING GENTAMICIN RESISTANCE FROM ANNUAL USAGE IN HOSPITAL
804 treated side, was observed in 5 patients on cyclosporin and in 3 controls. The treatment was well tolerated. Transient folliculitis (4 cases in each group) and occasional transient sensations of pruritus or irritation were noted, but never necessitated withdrawal of treatment. Laboratory tests (electrolytes, creatinine, uric acid), done before and after treatment, did not reveal any significant variations, and cyclosporin blood levels evaluated monthly by radioimmunoassay were below the limit of detection. The study of the peribulbar lymphocytes revealed a modification in the helper/ suppressor ratio in one patient (there being insufficient infiltrate in the other 2), with 90% T helper cells and 5% T suppressor cells before treatment and a clear inversion of the ratio after treatment in an area of regrowth of terminal hairs (helper 10%, suppressor
75%). The transient folliculitis and irritations were probably related to the oily excipient. A predominance of T helper cells in the peribulbar infiltrate has been reported in alopecia areata,2and the inversion of the helper/suppressor ratio during regrowth was an interesting finding in one patient. However, this does not necessarily directly reflect the action of cyclosporin, but may be a sign of regrowth of terminal hairs. The cyclosporin-treated group differed from the controls in that terminal hairs grew in 7 cases, all in the cyclosporin group. This regrowth was mild, delayed in onset, and never complete. However, the development of terminal hairs should encourage further trials with different concentrations of cyclosporin and different
excipients. Dermatology Unit, Höpital Necker Enfants Malades, 75015
Y. DE PROST D. TEILLAC
Paris, France
F. PAQUEZ
Dermatology Service, Höpital Henri Mondor, Creteil
L. CARRUGI
Paediatric Pathology Group, Höpital Necker Enfants Malades
H. BACHELEZ
Dermatology Service, Hopital Henri Mondor
R. TOURAINE
Relation between annual use of gentamicin and resistance of gram-negative rod isolates in subsequent year.
1. Aldridge RD, Thomson AW, Rosslyn Rankin. Inhibition of contact sensitivity reactions to DNFB by topical cyclosporin application in the guinea-pig. Clin Exp Immunol 1985; 59: 23-28. 2. Todes Taylor N, Turner R, Wood SG. T cell sub-populations in alopecia aresta. Acad Dermatol 1984; 11: 216-23.
J Am
PREDICTING GENTAMICIN RESISTANCE FROM ANNUAL USAGE IN HOSPITAL
SiR,—The bactericidal activity of aminoglycoside antibiotics against most gram-negative aerobes has led to their increased usage throughout the world. However, widespread indiscriminate use of a single aminoglycoside could be responsible for a corresponding increase in resistance by the mechanism of selective pressure. Indeed, the major mode of aminoglycoside resistance is related to the production of enzymes that inactivate the antibiotics,’ and these resistance determinants are carried on plasmids which may be transferred among different species.2 At the St Pierre Hospital, Brussels, gentamicin has been the first-choice aminoglycoside; tobramycin has been reserved for the treatment of infections due to Pseudomonas aertiginosa;3 and amikacin is used only when the causative organism was resistant to both gentamicin and
y
=
2-93 + 0’003268x (R = 09022; p <
use and the emergence of resistant organisms5 and have pointed to the value of antibiotic restriction in reducing the incidence of resistance to gentamicin6,7 and amikacin.8 However, the relation has not previously been expressed mathematically. Although we cannot be sure that the model can be applied to other hospitals or to other aminoglycosides we hope that investigators will be stimulated to this approach. A correlation does not prove cause-and-effect and it remains to be seen if an active policy of reducing gentamicin usage will lead to a reduction in resistance in hospital bacterial isolates.
tobramycin. We describe here a mathematical model for predicting gentamicin resistance of gram-negative organisms, based on the use of gentamicin in the previous year. 10 731 gram-negative organisms were isolated from 7220 inpatients between January, 1980, and June, 1986. Duplicate organisms isolated from the same patient with the same antibiogram (twelve antibiotics) were excluded. The
bacteria were: Escherichia coli 4245, Pseudomanas 1847, Proteus 1783, Klebsiella 1507, Enterobacter 605, Serratia 308, Acinetobacter 170, Citrobacter 166, and Salmonella 100. Resistance to gentamicin determined on all isolates by the method of Bauer et al.4 The yearly use of gentamicin (x, in grams) and the percentage of resistance to the drug one year later (y) were highly correlated (see figure) and the relation took the form y 2-93 + 0-003268x. Earlier studies have suggested a relation between aminoglycoside
was
=
0-005).
St Pieters University Hospital, B-1000 Brussels, Belgium
University of Iowa Hospitals and Clinics, Division of Clinical Epidemiology, Iowa City, Iowa, USA
HERMAN GOOSSENS G. GHYSELS Y. VAN LAETHEM S. DE WIT P. DE MOL J. LEVY N. CLUMECK J. P. BUTZLER
R. P. WENZEL
1. Shannon KP, Phillips I, King BA. Aminoglycoside resistance among Enterobacteriaceae and Acinetobacter species. J Antimicrob Chemother 1978; 4: 131-35. 2. Dickie P, Bryan LE, Pickard MA. Effect of enzymatic adenylation on dihydrostreptomycin accumulation in Escherichia coli carrying an R-factor: Model explaining aminoglycoside resistance by inactivating mechanisms. Antimicrob Agents Chemother 1978; 14: 569-74. 3. Meyers BR, Hirschman SZ. Tobramycin: in vitro antibacterial activity of a new aminoglycoside. J Clin Pharm 1972; 12: 313-17.
805 4. Bauer AW, Kirby WMM, Sherris JC, Turck M. Antibiotic susceptibility
testing by a
standardized single disk method. Am J Clin Pathol 1966; 45: 493-98. 5. Price KE, Kresel PA, Farchione LA, Siskin SB, Karpow SA. Epidemiological studies of aminoglycoside resistance in the USA. J Antimicrob Chemother 1981; 8 (suppl A): 89-105. 6. Noriega ER, Leibowitz RE, Richmond AS, et al. Nosocomial infection caused by gentamicin-resistant streptomycin sensitive Klebsiella. J Infect Dis 1975; 131
(suppl): S45-50. 7. Weinstein RA, Kabins SA. Strategies for prevention and control of multiple drug-resistant nosocomial infection. Am J Med 1981; 70: 449-54. 8. Levine JF, Maslow MJ, Leibowitz RE, et al. Amikacin resistant Gram-negative bacilli: Correlation of occurrence with amikacin use. J Infect Dis 1985; 151: 295-300.
CONSENSUS ANALYSIS
SiR,—The consensus analysis described by Professor Bull and his colleagues (Aug 16, p 377) appears to be an effective technique. However, the article does not describe the statistical procedure in sufficient detail to enable other potential users to apply it. What quantitive criteria are used for the initial rank ordering? How is the rank ordering modified by the results of patient visits? How is the consensus average calculated? 4077 Olive Hill
Drive, Claremont, California 91711, USA
SYLVAN GOLLIN
B*This letter has been shown to Professor Bull and Mr Westengard, whose reply follows.-ED.L .
SiR,-Our
paper
seems to
have confused several readers; in
particular, the terms ranking, rank order, and consensus have been frequently misinterpreted. "Ranking" as used in the title is in the commonly understoodsense of comparing one test with another. "Rank ordering" as described in Methods is used in the technical sense of taking all of the results obtained by one test method during the entire study and arranging them in order from lowest to highest. "Consensus" is used to refer to the average rank-order value of all the tests done on a patient at a single visit except for the test under scrutiny. It expressly excludes any clinical evaluation of the patient by one or more
physicians. Consider the hypothetical case of three patients on whom four (A, B, C, and D) are done every time at five clinic visits. For consensus analysis to be applicable the condition of the patients change for better or for worse, over the course of the five visits, for it is against this changing condition as reflected in the consensus that the correlation with each individual test is determined. This hypothetical study provides 15 results for test A, test B, and test C, over the 15 clinic visits and is large enough to illustrate steps 1 - 3 of the analytical procedure. Steps 4 and 5 require a data set of the size described in our Lancet paper. Step 1: To compare results that are measured in different units and on different scales the 15 results for test A are rank-ordered from low to high. Then each result is replaced for all subsequent statistical analysis by its rank-order number (here 1-15). Tests B, C, and D are treated similarly. Step 2.-The rank order of result A, B, C, and D is correlated (Spearman’s r) with the average rank order of the remaining tests done on that patient at that visit (consensus). Step 3.-After all 15 patient-visit records have been analysed all data from the test method with the poorest average correlation to the consensus are dropped and the 15 patient-visit records are reanalysed for the remaining three test methods. Step 4.-Surviving test methods from the above analysis are grouped into families of related tests by the behaviour of their Spearman’s r as the number of tests in the consensus decreases. Step S.-Finally, each surviving test is correlated with a new consensus limited to the survivors from steps 1-4. This consensus, to avoid the obvious hazard of circularity, excludes both the test under scrutiny and all other members of its family. It is this final consensus, tailor-made for each test, that determines how well or how poorly a test reflects the changes in patient’s disease. tests
Department of Pathology and Laboratory Medicine, School of Medicine, Loma Linda University, Loma Linda, California 92350, USA
BRIAN S. BULL J. C. WESTENGARD
TREATMENT OF ACUTE LYMPHOBLASTIC LEUKAEMIA: A TERMINOLOGICAL PLEA
SiR,—The realisation that disease in the central nervous system
(CNS) was a major obstacle to cure in children with acute lymphoblastic leukaemia (ALL) led to the systematic use, by Pinkel and colleagues at St Jude Hospital, of treatment directed specifically at the CNS.l The result was a major improvement in the proportion of disease-free survivors.2The term "CNS prophylaxis" is still frequently used to describe this phase of ALL treatment, whether with chemoradiotherapy or with chemotherapy alone.33 "Prophylaxis" means prevention, and in other medical contexts the word is used to describe treatment which, it is hoped, will prevent a disease or complication of which the patient is at risk. Craniocervical irradiation and a short course of intrathecal methotrexate might alter the cellular permeability of the bloodbrain barrier, making it more difficult for leukaemic blasts to invade the CNS, and this would constitute true prophylaxis. However, histopathological studies4,5 indicate that such treatment is eradicating disease already present within the CNS but clinically silent. Such subclinical tumour escapes detection by cytological examination of the cerebrospinal fluid and is inadequately treated by systemically delivered drugs because of poor penetration of the blood-brain barrier,but it can be eradicated by craniocervical irradiation and intrathecal methotrexate. The term "maintenance treatment" to describe chemotherapy given to patients with ALL who have achieved complete remission, via induction therapy with or without intensification, also needs reappraisal. When the phrase was introduced, some 20 years ago, the outlook for children with ALL was poor and the choice of words was perhaps influenced by the then prevalent belief that such treatment was more likely to prolong (maintain) the remisssion than eradicate residual disease and thus cure the patient. Old habits die hard but we suggest that both terms are discarded. As a substitute for "CNS prophylaxis", "CNS-directed therapy" (or "CNS therapy", the description used in the original article from St Judel) is brief and accurate. To come up with a satisfactory alternative to "maintenance therapy" is more difficult, but "continuing treatment" or "treatment in remission" are possible alternatives. J. PRITCHARD Hospital for Sick Children, London WC1N 3JH J. M. CHESSELLS 1. Aur RJA, Simone J, Hustu HO, et al. Central nervous system therapy and combination chemotherapy of childhood acute lymphocytic leukemia. Blood 1971; 37: 272-81. 2. Hustu HO, Aur RJA, Verzosa MS, et al. Prevention of central nervous system leukaemia by irradiation. Cancer 1973; 32: 585-97. 3. Green DM, Freeman AI, Sather HN, et al. Comparison of three methods of centralnervous-system prophylaxis in childhood acute lymphoblastic leukaemia. Lancet 1980; i: 1398-402. 4. Thomas LB, Chirigos MA, Humphreys SR, Goldin A. Development of meningeal leukaemia (L1210) during treatment of subcutaneously inoculated mice with methotrexate. Cancer 1964; 17: 352-60. 5. Moore EW, Thomas LB, Shaw RK, Freireich EJ. The central nervous system in acute leukemia: a post-mortem study of 117 cases with particular reference to hemorrhage, leukemic infiltrations and the syndrome of meningeal leukemia. Arch Intern Med 1960; 105: 451-68. 6. Rall DP. Experimental studies of the blood-brain barrier. Cancer Res 1962; 25: 1572-77.
AWARENESS OF ONCHOCERCIASIS
SIR,—Your Sept 6 editorial on Loa loa notes that the study of this pathogenic parasite was neglected in the west until US Peace Corps workers brought the disease to Washington ten years ago. Similar studies on the clinical and pathological effects of that other filarial parasite, Onchocerca volvulus, may also be stimulated as more workers travel to affected areas. Onchocerciasis has been studied in areas of high endemicity, but perhaps more by epidemiologists and ophthalmologists than by general clinicians. Standard textbooks on tropical disease emphasise the late signs, such as widespread dry dermatitis, hanging groin, and
75%). The transient folliculitis and irritations were probably related to the oily excipient. A predominance of T helper cells in the peribulbar infiltrate has been reported in alopecia areata,2and the inversion of the helper/suppressor ratio during regrowth was an interesting finding in one patient. However, this does not necessarily directly reflect the action of cyclosporin, but may be a sign of regrowth of terminal hairs. The cyclosporin-treated group differed from the controls in that terminal hairs grew in 7 cases, all in the cyclosporin group. This regrowth was mild, delayed in onset, and never complete. However, the development of terminal hairs should encourage further trials with different concentrations of cyclosporin and different
excipients. Dermatology Unit, Höpital Necker Enfants Malades, 75015
Y. DE PROST D. TEILLAC
Paris, France
F. PAQUEZ
Dermatology Service, Höpital Henri Mondor, Creteil
L. CARRUGI
Paediatric Pathology Group, Höpital Necker Enfants Malades
H. BACHELEZ
Dermatology Service, Hopital Henri Mondor
R. TOURAINE
Relation between annual use of gentamicin and resistance of gram-negative rod isolates in subsequent year.
1. Aldridge RD, Thomson AW, Rosslyn Rankin. Inhibition of contact sensitivity reactions to DNFB by topical cyclosporin application in the guinea-pig. Clin Exp Immunol 1985; 59: 23-28. 2. Todes Taylor N, Turner R, Wood SG. T cell sub-populations in alopecia aresta. Acad Dermatol 1984; 11: 216-23.
J Am
PREDICTING GENTAMICIN RESISTANCE FROM ANNUAL USAGE IN HOSPITAL
SiR,—The bactericidal activity of aminoglycoside antibiotics against most gram-negative aerobes has led to their increased usage throughout the world. However, widespread indiscriminate use of a single aminoglycoside could be responsible for a corresponding increase in resistance by the mechanism of selective pressure. Indeed, the major mode of aminoglycoside resistance is related to the production of enzymes that inactivate the antibiotics,’ and these resistance determinants are carried on plasmids which may be transferred among different species.2 At the St Pierre Hospital, Brussels, gentamicin has been the first-choice aminoglycoside; tobramycin has been reserved for the treatment of infections due to Pseudomonas aertiginosa;3 and amikacin is used only when the causative organism was resistant to both gentamicin and
y
=
2-93 + 0’003268x (R = 09022; p <
use and the emergence of resistant organisms5 and have pointed to the value of antibiotic restriction in reducing the incidence of resistance to gentamicin6,7 and amikacin.8 However, the relation has not previously been expressed mathematically. Although we cannot be sure that the model can be applied to other hospitals or to other aminoglycosides we hope that investigators will be stimulated to this approach. A correlation does not prove cause-and-effect and it remains to be seen if an active policy of reducing gentamicin usage will lead to a reduction in resistance in hospital bacterial isolates.
tobramycin. We describe here a mathematical model for predicting gentamicin resistance of gram-negative organisms, based on the use of gentamicin in the previous year. 10 731 gram-negative organisms were isolated from 7220 inpatients between January, 1980, and June, 1986. Duplicate organisms isolated from the same patient with the same antibiogram (twelve antibiotics) were excluded. The
bacteria were: Escherichia coli 4245, Pseudomanas 1847, Proteus 1783, Klebsiella 1507, Enterobacter 605, Serratia 308, Acinetobacter 170, Citrobacter 166, and Salmonella 100. Resistance to gentamicin determined on all isolates by the method of Bauer et al.4 The yearly use of gentamicin (x, in grams) and the percentage of resistance to the drug one year later (y) were highly correlated (see figure) and the relation took the form y 2-93 + 0-003268x. Earlier studies have suggested a relation between aminoglycoside
was
=
0-005).
St Pieters University Hospital, B-1000 Brussels, Belgium
University of Iowa Hospitals and Clinics, Division of Clinical Epidemiology, Iowa City, Iowa, USA
HERMAN GOOSSENS G. GHYSELS Y. VAN LAETHEM S. DE WIT P. DE MOL J. LEVY N. CLUMECK J. P. BUTZLER
R. P. WENZEL
1. Shannon KP, Phillips I, King BA. Aminoglycoside resistance among Enterobacteriaceae and Acinetobacter species. J Antimicrob Chemother 1978; 4: 131-35. 2. Dickie P, Bryan LE, Pickard MA. Effect of enzymatic adenylation on dihydrostreptomycin accumulation in Escherichia coli carrying an R-factor: Model explaining aminoglycoside resistance by inactivating mechanisms. Antimicrob Agents Chemother 1978; 14: 569-74. 3. Meyers BR, Hirschman SZ. Tobramycin: in vitro antibacterial activity of a new aminoglycoside. J Clin Pharm 1972; 12: 313-17.
805 4. Bauer AW, Kirby WMM, Sherris JC, Turck M. Antibiotic susceptibility
testing by a
standardized single disk method. Am J Clin Pathol 1966; 45: 493-98. 5. Price KE, Kresel PA, Farchione LA, Siskin SB, Karpow SA. Epidemiological studies of aminoglycoside resistance in the USA. J Antimicrob Chemother 1981; 8 (suppl A): 89-105. 6. Noriega ER, Leibowitz RE, Richmond AS, et al. Nosocomial infection caused by gentamicin-resistant streptomycin sensitive Klebsiella. J Infect Dis 1975; 131
(suppl): S45-50. 7. Weinstein RA, Kabins SA. Strategies for prevention and control of multiple drug-resistant nosocomial infection. Am J Med 1981; 70: 449-54. 8. Levine JF, Maslow MJ, Leibowitz RE, et al. Amikacin resistant Gram-negative bacilli: Correlation of occurrence with amikacin use. J Infect Dis 1985; 151: 295-300.
CONSENSUS ANALYSIS
SiR,—The consensus analysis described by Professor Bull and his colleagues (Aug 16, p 377) appears to be an effective technique. However, the article does not describe the statistical procedure in sufficient detail to enable other potential users to apply it. What quantitive criteria are used for the initial rank ordering? How is the rank ordering modified by the results of patient visits? How is the consensus average calculated? 4077 Olive Hill
Drive, Claremont, California 91711, USA
SYLVAN GOLLIN
B*This letter has been shown to Professor Bull and Mr Westengard, whose reply follows.-ED.L .
SiR,-Our
paper
seems to
have confused several readers; in
particular, the terms ranking, rank order, and consensus have been frequently misinterpreted. "Ranking" as used in the title is in the commonly understoodsense of comparing one test with another. "Rank ordering" as described in Methods is used in the technical sense of taking all of the results obtained by one test method during the entire study and arranging them in order from lowest to highest. "Consensus" is used to refer to the average rank-order value of all the tests done on a patient at a single visit except for the test under scrutiny. It expressly excludes any clinical evaluation of the patient by one or more
physicians. Consider the hypothetical case of three patients on whom four (A, B, C, and D) are done every time at five clinic visits. For consensus analysis to be applicable the condition of the patients change for better or for worse, over the course of the five visits, for it is against this changing condition as reflected in the consensus that the correlation with each individual test is determined. This hypothetical study provides 15 results for test A, test B, and test C, over the 15 clinic visits and is large enough to illustrate steps 1 - 3 of the analytical procedure. Steps 4 and 5 require a data set of the size described in our Lancet paper. Step 1: To compare results that are measured in different units and on different scales the 15 results for test A are rank-ordered from low to high. Then each result is replaced for all subsequent statistical analysis by its rank-order number (here 1-15). Tests B, C, and D are treated similarly. Step 2.-The rank order of result A, B, C, and D is correlated (Spearman’s r) with the average rank order of the remaining tests done on that patient at that visit (consensus). Step 3.-After all 15 patient-visit records have been analysed all data from the test method with the poorest average correlation to the consensus are dropped and the 15 patient-visit records are reanalysed for the remaining three test methods. Step 4.-Surviving test methods from the above analysis are grouped into families of related tests by the behaviour of their Spearman’s r as the number of tests in the consensus decreases. Step S.-Finally, each surviving test is correlated with a new consensus limited to the survivors from steps 1-4. This consensus, to avoid the obvious hazard of circularity, excludes both the test under scrutiny and all other members of its family. It is this final consensus, tailor-made for each test, that determines how well or how poorly a test reflects the changes in patient’s disease. tests
Department of Pathology and Laboratory Medicine, School of Medicine, Loma Linda University, Loma Linda, California 92350, USA
BRIAN S. BULL J. C. WESTENGARD
TREATMENT OF ACUTE LYMPHOBLASTIC LEUKAEMIA: A TERMINOLOGICAL PLEA
SiR,—The realisation that disease in the central nervous system
(CNS) was a major obstacle to cure in children with acute lymphoblastic leukaemia (ALL) led to the systematic use, by Pinkel and colleagues at St Jude Hospital, of treatment directed specifically at the CNS.l The result was a major improvement in the proportion of disease-free survivors.2The term "CNS prophylaxis" is still frequently used to describe this phase of ALL treatment, whether with chemoradiotherapy or with chemotherapy alone.33 "Prophylaxis" means prevention, and in other medical contexts the word is used to describe treatment which, it is hoped, will prevent a disease or complication of which the patient is at risk. Craniocervical irradiation and a short course of intrathecal methotrexate might alter the cellular permeability of the bloodbrain barrier, making it more difficult for leukaemic blasts to invade the CNS, and this would constitute true prophylaxis. However, histopathological studies4,5 indicate that such treatment is eradicating disease already present within the CNS but clinically silent. Such subclinical tumour escapes detection by cytological examination of the cerebrospinal fluid and is inadequately treated by systemically delivered drugs because of poor penetration of the blood-brain barrier,but it can be eradicated by craniocervical irradiation and intrathecal methotrexate. The term "maintenance treatment" to describe chemotherapy given to patients with ALL who have achieved complete remission, via induction therapy with or without intensification, also needs reappraisal. When the phrase was introduced, some 20 years ago, the outlook for children with ALL was poor and the choice of words was perhaps influenced by the then prevalent belief that such treatment was more likely to prolong (maintain) the remisssion than eradicate residual disease and thus cure the patient. Old habits die hard but we suggest that both terms are discarded. As a substitute for "CNS prophylaxis", "CNS-directed therapy" (or "CNS therapy", the description used in the original article from St Judel) is brief and accurate. To come up with a satisfactory alternative to "maintenance therapy" is more difficult, but "continuing treatment" or "treatment in remission" are possible alternatives. J. PRITCHARD Hospital for Sick Children, London WC1N 3JH J. M. CHESSELLS 1. Aur RJA, Simone J, Hustu HO, et al. Central nervous system therapy and combination chemotherapy of childhood acute lymphocytic leukemia. Blood 1971; 37: 272-81. 2. Hustu HO, Aur RJA, Verzosa MS, et al. Prevention of central nervous system leukaemia by irradiation. Cancer 1973; 32: 585-97. 3. Green DM, Freeman AI, Sather HN, et al. Comparison of three methods of centralnervous-system prophylaxis in childhood acute lymphoblastic leukaemia. Lancet 1980; i: 1398-402. 4. Thomas LB, Chirigos MA, Humphreys SR, Goldin A. Development of meningeal leukaemia (L1210) during treatment of subcutaneously inoculated mice with methotrexate. Cancer 1964; 17: 352-60. 5. Moore EW, Thomas LB, Shaw RK, Freireich EJ. The central nervous system in acute leukemia: a post-mortem study of 117 cases with particular reference to hemorrhage, leukemic infiltrations and the syndrome of meningeal leukemia. Arch Intern Med 1960; 105: 451-68. 6. Rall DP. Experimental studies of the blood-brain barrier. Cancer Res 1962; 25: 1572-77.
AWARENESS OF ONCHOCERCIASIS
SIR,—Your Sept 6 editorial on Loa loa notes that the study of this pathogenic parasite was neglected in the west until US Peace Corps workers brought the disease to Washington ten years ago. Similar studies on the clinical and pathological effects of that other filarial parasite, Onchocerca volvulus, may also be stimulated as more workers travel to affected areas. Onchocerciasis has been studied in areas of high endemicity, but perhaps more by epidemiologists and ophthalmologists than by general clinicians. Standard textbooks on tropical disease emphasise the late signs, such as widespread dry dermatitis, hanging groin, and