540
aminoacid code and the codon number (eg, G551D is the replacement of glycine with aspartic acid at position 551; R553X is the introduction of a termination codon in place of
3.
4.
arginine at position 553). Putative structures for the nucleotide binding folds of CFTR and related genes (known as ABC genes for ATP-binding cassette-ie, a domain found in various proteins that binds and hydrolyses ATP) have been predicted by computer modelling.4 The deletion and substitutions observed in CFTR occur in highly conserved regions of the nucleotide binding fold. Although they are unlikely to affect ATP binding and hydrolysis directly, the aminoacid substitutions observed may affect changes in protein conformation which normally take place after nucleotide binding. Frameshift mutations in exons 13 and 19 (a two base-pair insertion and a single base deletion, respectively) introduce premature termination codons.6,7 Another series of mutations, reported in and around the first transmembrane segment,2are all associated with milder disease phenotypes. Kerem et a15 have suggested that patients with pancreatic insufficiency and a generally severe phenotype will be homozygous, or compound heterozygotes, for mutations that have a severe effect on CFTR function. Patients with a milder phenotype, including those who have no pancreatic deficiency, will carry an allele that disrupts gene function to a lesser extent. Several patients have been reported who are homozygous, or compound heterozygous, for non-sense or splice mutations.6 The generally milder phenotype in these patients suggests that loss of CFTR function is not as deleterious as a breakdown in its regulation. Although a defect in the transport of chloride ions appears to be fundamental in CF, it is unlikely that CFTR functions as a chloride channel. CFTR is homologous to a family of unidirectional active transport proteins; by contrast, ion channels permit bidirectional ion flow and do not normally require ATP hydrolysis. Moreover, all CF patients seem to have a defect in a single gene whereas different tissues appear to have chloride channels encoded by different genes. A protein as large as CFTR might be expected to transport molecules larger than ions; Ringe and Petsko8 have suggested leukotriene LTC4 and prostaglandin D2 as candidates for this ligand. Since more than 50 mutations have been found within CFTR in the past eight months, any plans to screen entire populations for CF heterozygotes for all mutations are probably over-optimistic. In some populations it is possible to identify 85% of CF mutations by analysis of only two exons (10 and 11). In the UK, population screening for heterozygotes has been thought desirable, and pilot schemes are being initiated.9 This approach contrasts with the confused picture emerging from the USA,10 where a fear of litigation seems to be slowing progress in this area of primary health care, although it has lately been reported that a National Institutes of Health working group has supported the initiation of pilot programmes.ll If the combination of mutation analysis with clinical data and structural modelling leads to a greater understanding of the function of CFTR, the gap between gene and treatment should be progressively shortened. 1. Riordan JR, Rommens JM, Kerem B-S, et al. Identification of the cystic fibrosis gene: cloning and characterization of complementary DNA. Science 1989; 245: 1066-73. 2. Dean M, Amos J, Hsu JMC, et al. Multiple mutations in highly conserved residues are found in mildly affected cystic fibrosis patients. Cell 1990; 61: 863-70.
Cutting GR, Kasch LM, Rosenstein, BJ, et al. A cluster of cystic fibrosis mutations in the first nucleotide-binding fold of the cystic fibrosis conductance regulator protein. Nature 1990; 346: 366-69. Hyde SC, Emsley P, Hartshorn MJ, et al. Structural model of ATP-binding associated with cystic fibrosis, multidrug resistance and
bacterial transport. Nature 1990; 346: 362-65. B-S, Rommens JM, Buchanan JA, et al. Identification of the cystic fibrosis gene: genetic analysis. Science 1989; 245: 1074-80. 6. Kerem B-S, Zielenski J, Markiewicz D, et al. Identification of mutations in regions corresponding to the 2 putative nucleotide (ATP)-binding folds in the cystic fibrosis gene. Proc Natl Acad Sci ( USA) (in press). 7. White MB, Amos J, Hsu JMC, Gerrard B, Finn P, Dean M. A frame-shift mutation in the cystic fibrosis gene. Nature 1990; 344: 665-67. 8. Ringe D, Petsko GA. A transport problem? Nature 1990; 346: 312-13. 9. Brock DJH. Population screening for cystic fibrosis. Am J Hum Genet 5. Kerem
1990; 47: 164-65. 10. Gilbert F. Is population screening for cystic fibrosis appropriate now? Am J Hum Genet 1990; 46: 394-95. 11. Roberts L. CF screening delayed for awhile, perhaps forever. Science 1990; 247: 1296-97.
CHILDREN WHO WALK LATE For parents, few milestones in their child’s development are exciting, memorable, and easily defined as the first steps. In Britain 97% of children achieve six steps unaided by 18 months of age,’ and it has been suggested that screening for children who walk late should form part of a preschool health surveillance programme.2 What would be achieved by such a policy? In one study of 404 late walkers3 a quarter were already known to paediatricians because of cerebral palsy, mental retardation, or other neurological, developmental, or orthopaedic disorders. Of 275 new referrals from screening, only 5% had neurological signsmainly cerebral palsy, although about a fifth had other evidence of delayed development. The frequency of neurological abnormalities in late walkers rises to 56% when the perinatal period has been abnormal.4 Theoretically, screening of boys not walking at 18 months with a blood test for creatine kinase will detect half the cases of Duchenne dystrophy in the population. In practice, detection rates are unacceptably low and population screening can not be recommended. Delayed motor skills in Duchenne dystrophy are often associated with generalised developmental delay and it is this combination which necessitates creatine kinase estimation in boys. Late walking is usually a benign normal variant but differentiation of "normal" from "abnormal" late walkers is not always easy. The typical child in the normal category has a family history of late walking and bottom shuffling, and the perinatal history is uneventful. Examination reveals a hypotonic bottom shuffler with hypermobile joints, hyperextended knees, and the sitting-on-air sign when the child is suspended under the arms. Such children do not have weakness and tendon reflexes are preserved. The picture may be less clear-50% of normal late walkers have no family history and 45 % are crawlers.3 The cerebral palsy diagnosed in late walkers at 18 months is usually a diplegia;1 in 8 such children is said to be a bottom shuffler.7 However, the increased resistance to passive dorsiflexion of the foot usually found in these cases contrasts with the hypotonia of most normal late walkers. Observation of a normal gait at 18 months is recommended as the last step in the screening programme for congenital dislocation of the hip.6 Apart from this, seeking out children who walk late does not fulfil recognised criteria for screening. Most of the children detected will be normal and the benefits of early diagnosis in the remainder are uncertain. By contrast, a child surveillance programme as
541
will encourage parents and professionals to observe children carefully as they come to walk. Those skilled in surveillance will be able to assess the clinical picture in children who do walk late and offer developmental guidance or further
referral as appropriate. 1. Neligan G, Prudham D. Norms for four standard developmental milestones by sex, social class and place in family. Dev Med Child Neurol 1969; 11: 413-22. 2. Colver AF, Steiner H. Health surveillance of pre-school children. Br Med
J 1986; 293: 258-60. 3. Chaplais J de Z, MacFarlane JA. A review of four hundred and four late walkers. Arch Dis Child 1984; 59: 512-16. 4. Johnson A, Goddard O, Ashurst H. Is late walking a marker of morbidity? Arch Dis Child 1990; 65: 486-88. 5. Smith RA, Rogers M, Bradley DM, Sibert JR, Harper PS. Screening for Duchenne muscular dystrophy. Arch Dis Child 1989; 64: 1017-21. 6. Standing Medical Advisory Committee and Standing Nursing and
Midwifery Advisory Committee. Screening for detection of congenital dislocation of the hip. London: DHSS, 1986. 7. Robson P. Screening for children. Developmental Health 1978; 98: 231-37.
paediatrics. J R
Soc
TAKING RISKS IN GENERAL PRACTICE It might be thought that a good general practitioner should never take risks, but Grol and colleagues have lately argued that doctors who seek to minimise uncertainty and to avoid risks in primary care are acting irrationally. These researchers compare the reported attitudes to risk-taking of general practitioners in Belgium, Holland, and the UK and suggest that Dutch general practitioners are more willing to take risks in clinical decision making because their vocational training makes them aware of the dangers of defensive medicine. These conclusions are open to the criticisms that reported attitudes do not necessarily reflect clinical practice (the evidence presented to substantiate this link is tenuous) and that the role ascribed to educational methods is speculative. Nevertheless, the paper does raise several important issues. It is potentially misleading to label a decision to "wait and see" after careful clinical assessment as "risk taking", a phrase that suggests carelessness and irresponsibility. General practitioners face the unenviable problem of needing to assess the likelihood of serious disease in a population in which such events are rare. Not only is there a problem of maintaining vigilance but also the diagnostic value of symptoms and signs in general practice is much less than in hospital medicine. A clinical sign with 90% sensitivity and 95 % specificity has a predictive value of 95 % if the underlying prevalence of disease is 50 %, but only 15 % if the prevalence is 1%. In primary care, knowledge of the proportion of patients with a certain symptom who do not have serious disease becomes more important than knowledge of the proportion who do. If general practitioners adopt the same diagnostic and investigative thresholds as do their hospital colleagues they will generate much suffering through false alarms. So, how should general practitioners achieve the correct balance between reassurance and diagnostic intervention? The issue is the appropriate management of low-level risks, a task that has intrigued epidemiologists and environmental scientists for some years and about which there have been many reports 2 The first step in risk management is simple: quantify the risk. This means assessing the significance of symptoms and signs in the community rather than working backwards from descriptions in medical textbooks which are
based on hospital case-series. In the UK introduction of on-desk computers in general practice surgeries has made this assessment possible for the first time, and it is tragic that the Government has not provided adequate support for this development in England and Wales, preferring to leave the task to commercial organisations funded by the
pharmaceutical industry. The Dutch seem to have little doubt that if one can provide better data about disease prevalence and the diagnostic significance of symptoms and signs in the community, general practitioners could be taught to make better clinical decisions. It is certainly true that doctors are ill at ease with uncertainty and Eddy has argued cogently that clinical practice would be much improved if physicians were educated to be more numerate in the sense of "learning the language" of probability.4 Pioneering work at McMaster has shown that probability theory can be used successfully in a clinical context.5 So, although there is no firm evidence that numerate doctors are better doctors at providing immediate clinical care, and the true value of numeracy may lie in planning and evaluating practice protocols for the management of disease rather than in making instant decisions in the surgery, some progress has been made already in translating theory into practice. It is also possible that knowledge-based computer systems, which can generate diagnostic hypotheses and use probabilistic data without exposing probabilistic reasoning to the doctorsmay soon revolutionise clinical practice. The most important element in medical decision making is framing the question-selecting the salient information from the morass.7 To achieve this in general practice, as in hospital medicine, the patient must be listened to and examined carefully. It is clearly an unacceptable risk not to do this. But in taking a decision to act on the information elicited (even if the action is a diagnostic test) the general practitioner must be more aware than hospital colleagues of the danger of false-positive information. The acceptable and necessary "risk" taken is the selection of a higher threshold for action. Despite Grol’s paper, it is not established that British doctors set this threshold higher, and therefore take more risk, than do their Dutch colleagues. A comparative study from general practice 10 years ago provided no evidence for this notion,8 and Payer’s overview of medicine and culture in Europe characterised the British as "doing less of everything".9 However, it is clear that the Dutch are looking ahead and teaching primary care doctors to be numerate in the belief that it will help them to deal better with uncertainty and low-level risk in the future. R, Whitfield M, De Maesener J, Mokkink H. Attitudes to risk taking in medical decision making among British, Dutch and Belgian general practitioners. Br J Gen Pract 1990; 40: 134-36. 2. Griffiths R, ed. Dealing with risk. Manchester: Manchester University Press, 1982. 3. Editorial. The selling of patients’ data. 1989; ii: 1078. 4. Eddy DM. Variations in practice: the role of uncertainty. In: Dowie J, Elstein A, eds. Professional judgement: a reader in clinical decision making. Cambridge: Cambridge University Press, 1988: 45-59. 5. Sackett D, Haynes RB, Tugwell P. Clinical epidemiology: a basic science of clinical medicine. Boston: Little, Brown, 1985. 6. Fox J. Knowledge and judgement in decision making. In: Brooke J, Rector A, Sheldon M, eds. Decision making in general practice. London: Macmillan, 1985: 107-18. 7. Schon D. From technical rationality to reflection-in-action. In: Dowie J, Elstein A, eds. Professional judgement: a reader in clinical decision making. Cambridge: Cambridge University Press, 1988: 60-77. 8. Mesker J, Mesker P. Some difficulties in comparing morbidity between countries. J R Coll Gen Pract 1979; 29: 92-96. 9. Payer L. Medicine and culture. London: Gollancz, 1989. 1. Grol
aminoacid code and the codon number (eg, G551D is the replacement of glycine with aspartic acid at position 551; R553X is the introduction of a termination codon in place of
3.
4.
arginine at position 553). Putative structures for the nucleotide binding folds of CFTR and related genes (known as ABC genes for ATP-binding cassette-ie, a domain found in various proteins that binds and hydrolyses ATP) have been predicted by computer modelling.4 The deletion and substitutions observed in CFTR occur in highly conserved regions of the nucleotide binding fold. Although they are unlikely to affect ATP binding and hydrolysis directly, the aminoacid substitutions observed may affect changes in protein conformation which normally take place after nucleotide binding. Frameshift mutations in exons 13 and 19 (a two base-pair insertion and a single base deletion, respectively) introduce premature termination codons.6,7 Another series of mutations, reported in and around the first transmembrane segment,2are all associated with milder disease phenotypes. Kerem et a15 have suggested that patients with pancreatic insufficiency and a generally severe phenotype will be homozygous, or compound heterozygotes, for mutations that have a severe effect on CFTR function. Patients with a milder phenotype, including those who have no pancreatic deficiency, will carry an allele that disrupts gene function to a lesser extent. Several patients have been reported who are homozygous, or compound heterozygous, for non-sense or splice mutations.6 The generally milder phenotype in these patients suggests that loss of CFTR function is not as deleterious as a breakdown in its regulation. Although a defect in the transport of chloride ions appears to be fundamental in CF, it is unlikely that CFTR functions as a chloride channel. CFTR is homologous to a family of unidirectional active transport proteins; by contrast, ion channels permit bidirectional ion flow and do not normally require ATP hydrolysis. Moreover, all CF patients seem to have a defect in a single gene whereas different tissues appear to have chloride channels encoded by different genes. A protein as large as CFTR might be expected to transport molecules larger than ions; Ringe and Petsko8 have suggested leukotriene LTC4 and prostaglandin D2 as candidates for this ligand. Since more than 50 mutations have been found within CFTR in the past eight months, any plans to screen entire populations for CF heterozygotes for all mutations are probably over-optimistic. In some populations it is possible to identify 85% of CF mutations by analysis of only two exons (10 and 11). In the UK, population screening for heterozygotes has been thought desirable, and pilot schemes are being initiated.9 This approach contrasts with the confused picture emerging from the USA,10 where a fear of litigation seems to be slowing progress in this area of primary health care, although it has lately been reported that a National Institutes of Health working group has supported the initiation of pilot programmes.ll If the combination of mutation analysis with clinical data and structural modelling leads to a greater understanding of the function of CFTR, the gap between gene and treatment should be progressively shortened. 1. Riordan JR, Rommens JM, Kerem B-S, et al. Identification of the cystic fibrosis gene: cloning and characterization of complementary DNA. Science 1989; 245: 1066-73. 2. Dean M, Amos J, Hsu JMC, et al. Multiple mutations in highly conserved residues are found in mildly affected cystic fibrosis patients. Cell 1990; 61: 863-70.
Cutting GR, Kasch LM, Rosenstein, BJ, et al. A cluster of cystic fibrosis mutations in the first nucleotide-binding fold of the cystic fibrosis conductance regulator protein. Nature 1990; 346: 366-69. Hyde SC, Emsley P, Hartshorn MJ, et al. Structural model of ATP-binding associated with cystic fibrosis, multidrug resistance and
bacterial transport. Nature 1990; 346: 362-65. B-S, Rommens JM, Buchanan JA, et al. Identification of the cystic fibrosis gene: genetic analysis. Science 1989; 245: 1074-80. 6. Kerem B-S, Zielenski J, Markiewicz D, et al. Identification of mutations in regions corresponding to the 2 putative nucleotide (ATP)-binding folds in the cystic fibrosis gene. Proc Natl Acad Sci ( USA) (in press). 7. White MB, Amos J, Hsu JMC, Gerrard B, Finn P, Dean M. A frame-shift mutation in the cystic fibrosis gene. Nature 1990; 344: 665-67. 8. Ringe D, Petsko GA. A transport problem? Nature 1990; 346: 312-13. 9. Brock DJH. Population screening for cystic fibrosis. Am J Hum Genet 5. Kerem
1990; 47: 164-65. 10. Gilbert F. Is population screening for cystic fibrosis appropriate now? Am J Hum Genet 1990; 46: 394-95. 11. Roberts L. CF screening delayed for awhile, perhaps forever. Science 1990; 247: 1296-97.
CHILDREN WHO WALK LATE For parents, few milestones in their child’s development are exciting, memorable, and easily defined as the first steps. In Britain 97% of children achieve six steps unaided by 18 months of age,’ and it has been suggested that screening for children who walk late should form part of a preschool health surveillance programme.2 What would be achieved by such a policy? In one study of 404 late walkers3 a quarter were already known to paediatricians because of cerebral palsy, mental retardation, or other neurological, developmental, or orthopaedic disorders. Of 275 new referrals from screening, only 5% had neurological signsmainly cerebral palsy, although about a fifth had other evidence of delayed development. The frequency of neurological abnormalities in late walkers rises to 56% when the perinatal period has been abnormal.4 Theoretically, screening of boys not walking at 18 months with a blood test for creatine kinase will detect half the cases of Duchenne dystrophy in the population. In practice, detection rates are unacceptably low and population screening can not be recommended. Delayed motor skills in Duchenne dystrophy are often associated with generalised developmental delay and it is this combination which necessitates creatine kinase estimation in boys. Late walking is usually a benign normal variant but differentiation of "normal" from "abnormal" late walkers is not always easy. The typical child in the normal category has a family history of late walking and bottom shuffling, and the perinatal history is uneventful. Examination reveals a hypotonic bottom shuffler with hypermobile joints, hyperextended knees, and the sitting-on-air sign when the child is suspended under the arms. Such children do not have weakness and tendon reflexes are preserved. The picture may be less clear-50% of normal late walkers have no family history and 45 % are crawlers.3 The cerebral palsy diagnosed in late walkers at 18 months is usually a diplegia;1 in 8 such children is said to be a bottom shuffler.7 However, the increased resistance to passive dorsiflexion of the foot usually found in these cases contrasts with the hypotonia of most normal late walkers. Observation of a normal gait at 18 months is recommended as the last step in the screening programme for congenital dislocation of the hip.6 Apart from this, seeking out children who walk late does not fulfil recognised criteria for screening. Most of the children detected will be normal and the benefits of early diagnosis in the remainder are uncertain. By contrast, a child surveillance programme as
541
will encourage parents and professionals to observe children carefully as they come to walk. Those skilled in surveillance will be able to assess the clinical picture in children who do walk late and offer developmental guidance or further
referral as appropriate. 1. Neligan G, Prudham D. Norms for four standard developmental milestones by sex, social class and place in family. Dev Med Child Neurol 1969; 11: 413-22. 2. Colver AF, Steiner H. Health surveillance of pre-school children. Br Med
J 1986; 293: 258-60. 3. Chaplais J de Z, MacFarlane JA. A review of four hundred and four late walkers. Arch Dis Child 1984; 59: 512-16. 4. Johnson A, Goddard O, Ashurst H. Is late walking a marker of morbidity? Arch Dis Child 1990; 65: 486-88. 5. Smith RA, Rogers M, Bradley DM, Sibert JR, Harper PS. Screening for Duchenne muscular dystrophy. Arch Dis Child 1989; 64: 1017-21. 6. Standing Medical Advisory Committee and Standing Nursing and
Midwifery Advisory Committee. Screening for detection of congenital dislocation of the hip. London: DHSS, 1986. 7. Robson P. Screening for children. Developmental Health 1978; 98: 231-37.
paediatrics. J R
Soc
TAKING RISKS IN GENERAL PRACTICE It might be thought that a good general practitioner should never take risks, but Grol and colleagues have lately argued that doctors who seek to minimise uncertainty and to avoid risks in primary care are acting irrationally. These researchers compare the reported attitudes to risk-taking of general practitioners in Belgium, Holland, and the UK and suggest that Dutch general practitioners are more willing to take risks in clinical decision making because their vocational training makes them aware of the dangers of defensive medicine. These conclusions are open to the criticisms that reported attitudes do not necessarily reflect clinical practice (the evidence presented to substantiate this link is tenuous) and that the role ascribed to educational methods is speculative. Nevertheless, the paper does raise several important issues. It is potentially misleading to label a decision to "wait and see" after careful clinical assessment as "risk taking", a phrase that suggests carelessness and irresponsibility. General practitioners face the unenviable problem of needing to assess the likelihood of serious disease in a population in which such events are rare. Not only is there a problem of maintaining vigilance but also the diagnostic value of symptoms and signs in general practice is much less than in hospital medicine. A clinical sign with 90% sensitivity and 95 % specificity has a predictive value of 95 % if the underlying prevalence of disease is 50 %, but only 15 % if the prevalence is 1%. In primary care, knowledge of the proportion of patients with a certain symptom who do not have serious disease becomes more important than knowledge of the proportion who do. If general practitioners adopt the same diagnostic and investigative thresholds as do their hospital colleagues they will generate much suffering through false alarms. So, how should general practitioners achieve the correct balance between reassurance and diagnostic intervention? The issue is the appropriate management of low-level risks, a task that has intrigued epidemiologists and environmental scientists for some years and about which there have been many reports 2 The first step in risk management is simple: quantify the risk. This means assessing the significance of symptoms and signs in the community rather than working backwards from descriptions in medical textbooks which are
based on hospital case-series. In the UK introduction of on-desk computers in general practice surgeries has made this assessment possible for the first time, and it is tragic that the Government has not provided adequate support for this development in England and Wales, preferring to leave the task to commercial organisations funded by the
pharmaceutical industry. The Dutch seem to have little doubt that if one can provide better data about disease prevalence and the diagnostic significance of symptoms and signs in the community, general practitioners could be taught to make better clinical decisions. It is certainly true that doctors are ill at ease with uncertainty and Eddy has argued cogently that clinical practice would be much improved if physicians were educated to be more numerate in the sense of "learning the language" of probability.4 Pioneering work at McMaster has shown that probability theory can be used successfully in a clinical context.5 So, although there is no firm evidence that numerate doctors are better doctors at providing immediate clinical care, and the true value of numeracy may lie in planning and evaluating practice protocols for the management of disease rather than in making instant decisions in the surgery, some progress has been made already in translating theory into practice. It is also possible that knowledge-based computer systems, which can generate diagnostic hypotheses and use probabilistic data without exposing probabilistic reasoning to the doctorsmay soon revolutionise clinical practice. The most important element in medical decision making is framing the question-selecting the salient information from the morass.7 To achieve this in general practice, as in hospital medicine, the patient must be listened to and examined carefully. It is clearly an unacceptable risk not to do this. But in taking a decision to act on the information elicited (even if the action is a diagnostic test) the general practitioner must be more aware than hospital colleagues of the danger of false-positive information. The acceptable and necessary "risk" taken is the selection of a higher threshold for action. Despite Grol’s paper, it is not established that British doctors set this threshold higher, and therefore take more risk, than do their Dutch colleagues. A comparative study from general practice 10 years ago provided no evidence for this notion,8 and Payer’s overview of medicine and culture in Europe characterised the British as "doing less of everything".9 However, it is clear that the Dutch are looking ahead and teaching primary care doctors to be numerate in the belief that it will help them to deal better with uncertainty and low-level risk in the future. R, Whitfield M, De Maesener J, Mokkink H. Attitudes to risk taking in medical decision making among British, Dutch and Belgian general practitioners. Br J Gen Pract 1990; 40: 134-36. 2. Griffiths R, ed. Dealing with risk. Manchester: Manchester University Press, 1982. 3. Editorial. The selling of patients’ data. 1989; ii: 1078. 4. Eddy DM. Variations in practice: the role of uncertainty. In: Dowie J, Elstein A, eds. Professional judgement: a reader in clinical decision making. Cambridge: Cambridge University Press, 1988: 45-59. 5. Sackett D, Haynes RB, Tugwell P. Clinical epidemiology: a basic science of clinical medicine. Boston: Little, Brown, 1985. 6. Fox J. Knowledge and judgement in decision making. In: Brooke J, Rector A, Sheldon M, eds. Decision making in general practice. London: Macmillan, 1985: 107-18. 7. Schon D. From technical rationality to reflection-in-action. In: Dowie J, Elstein A, eds. Professional judgement: a reader in clinical decision making. Cambridge: Cambridge University Press, 1988: 60-77. 8. Mesker J, Mesker P. Some difficulties in comparing morbidity between countries. J R Coll Gen Pract 1979; 29: 92-96. 9. Payer L. Medicine and culture. London: Gollancz, 1989. 1. Grol
