- Email: [email protected]
Thromboembolic risk and pacemakers
THE LANCET 1 2
3
4
5
Lye M. Access to advances in cardiology. Lancet 1997; 350: 1162–63. Parameshwar J, Shackell MM, Richardson A, Poole Wilson PA, Sutton GC. Prevalence of heart failure in three general practices in north west London. Br J Gen Pract 1992; 42: 287–89. Pitt B, Segal R, Martinez FA, et al. Randomised trial of losartan versus captopril in patients over 65 with heart failure (Evidence of Losartan in the Elderly Study, ELITE). Lancet 1997; 349: 747–52. The SOLVD investigators. Effect of enalapril on survival in patients with reduced left ventricular ejection fractions and congestive heart failure. N Engl J Med 1991; 325: 293–302. Cohn JN, Johnson G, Ziesche S, et al. A comparison of enalapril with hydralazineisosorbide dinitrate in the treatment of chronic congestive heart failure. N Engl J Med 1991; 325: 303–10.
HIV-1 infection associated with abnormal vaginal flora morphology and bacterial vaginosis SIR—In their response to our report 1 of an association between prevalent HIV infection and depletion of vaginal lactobacilli in Uganda, Austin Ugwumadu and colleagues (Oct 25, p 1251) 2 question whether the association between bacterial vaginosis and HIV may be confounded by concurrent cervical infections, particularly Chlamydia trachomatis. As noted in table 3 in our report, most women had no evidence of cervical infections, and no association was found between prevalent HIV and chlamydial infection, irrespective of the presence of bacterial vaginosis. Ugwumadu et al also comment on the low prevalence of C trachomatis in this population and question whether this may be due to a reduced sensitivity of the urinary ligase chain reaction (LCR) assay as a consequence of difficulties in specimen handling We have compared LCR detection of C trachomatis in urine samples with hybrid capture detection of C trachomatis from 905 self-collected vaginal swab specimens (Digene, Silver Spring, MD), and found complete agreement between the two assays and modes of sample collection. The prevalence of C trachomatis in this unselected and largely symptom-free sample of rural Ugandan women is not atypical for other unselected populations in rural East Africa (J Schachter, personal communication), and the age-specific prevalence of chlamydia shows the expected higher rates in younger women (at ages 15–19 years, the prevalence of C trachomatis was 6·6%). Furthermore, recent studies have shown
1780
stability of LCR detection of C trachomatis in urine kept at room temperature for up to 72 h before processing. We are therefore confident that the urinary LCR assay is sensitive for the detection of C trachomatis in the Rakai setting. S Ledru and colleagues (Oct 25, p 1251) 3 question our use of morphological methods for diagnosis of bacterial vaginitis and recommend the use of clinical criteria. As we noted the morphological assessment of this condition has been validated in studies of conventional clinical diagnostic criteria. Moreover, in collaboration with other investigators, we have examined the association between prevalent HIV and clinically diagnosed bacterial vaginosis among 6667 pregnant women from Blantyre, Malawi.4,5 In the Malawian study, the multivariate adjusted odds ratio of HIV infection in women with clinical bacterial vaginosis was 2·5 (95% CI 2·1–2·9),4 whereas in our Ugandan study, the odds ratio was 1·56 (1·24–1·97) with morphologically diagnosed and 2·08 (1·48–2·94) for severe infection (scores 9–10). Thus, the morphological diagnosis seems conservative. We have also analysed preliminary data on bacterial vaginosis and incident HIV infection with a nested case-control design. There were 77 incident cases of HIV who were individually matched by age (within 2·5 years) and place of residence, with 328 HIV-negative controls. The duration of observation was similar for cases and controls. These women provided a self-collected vaginal swab at the study visit before the follow-up interval during which they were at risk of seroconversion. Swabs were rolled onto microscope slides, gram-stained, and sent to the University of Pittsburgh for morphological evaluation, blinded with respect to HIV status. The findings suggest a possible increased risk of incident HIV infection associated with pre-existing bacterial infection with conditional logistic regression for matched data, the odds ratio of incident HIV was 1·50 (0·78–2·80). Although not significant, this result is consistent with the findings for HIV among women with bacterial vaginosis, 1,4,5 and suggests that depletion of vaginal lactobacilli may be a risk factor for HIV acquisition. We thank L Rabe for technical assistance.
*R H Gray, M J Wawer, N Sewankambo, D Serwadda Department of Population Dynamics, School of Hygiene and Public Health, Johns Hopkins University, Baltimore, MD 21205, USA
1
2
3
4
5
Sewankambo N, Gray RH, Wawer MJ, et al. HIV-1 infection associated with abnormal vaginal flora morphology and bacterial vaginosis. Lancet 1997; 350: 546–50. Ugwumadu A, Hay P, Taylor-Robinson D. HIV-1 infection associated with abnormal vaginal flora morphology and bacterial vaginosis. Lancet 1997; 350: 1251. Ledru S, Meda N, Ledru E, Bazie AJ, Chiron JP. HIV-1 infection associated with abnormal vaginal flora morphology and bacterial vaginosis. Lancet 1997; 350: 1251–52. Taha TE, Miotti P, Dalabetta G, et al. Bacterial vaginosis during pregnancy: associations with HIV, preterm delivery and mother-to-child transmission of HIV. Presented at conference on global strategies for the prevention of HIV transmission from mothers to infants. Washington DC, Sept 3–6, 1997: 13. Gray RH, Wawer MJ, Taha TE, et al. Depleted vaginal Lactobacilli and BV increase HIV acquisition in studies in Uganda and Malawi. International Infectious Disease Society of America, 35th Annual Meeting, San Francisco, Sept 13–16, 1997: abstract.
Thromboembolic risk and pacemakers SIR—Henning Rud Andersen and coworkers (Oct 25, p 1210) 1 provide further evidence to confirm the higher incidence of atrial fibrillation, heart failure, and increased risk of thromboembolism in patients with sick-sinus syndrome who are managed with single-chamber ventricular pacing than in those managed with single-chamber atrial pacing or dualchamber pacing. We postulate that the higher adverse event rate in the ventricular-pacing group, may be the result of an abnormal haemodynamic response, in the presence of a prothrombotic or hypercoagulable state. In Andersen's study, thromboembolic events were higher in those managed with ventricular pacing, whether or not they subsequently developed atrial fibrillation. Atrial fibrillation is itself an independent risk factor for stroke and thromboembolism, with associated intra-atrial stasis, and it is well recognised that such patients develop a hypercoagulable state, which may contribute to the risk of thrombogenesis.2 These factors may, in part, explain the higher risk of thromboembolism associated with ventricular pacing in those who subsequently developed atrial fibrillation. The greater thromboembolic event rate associated with ventricular pacing in patients who did not subsequently develop atrial fibrillation may itself
Vol 350 • December 13, 1997
THE LANCET
From efficacy to costeffectiveness
reflect increased thrombogenesis; Lau and colleagues3 reported increased spontaneous systemic-platelet activation in such patients. These findings were related to the loss of atrioventricular synchrony, and to the lack of atrial pacing leading to intraatrial stasis, which confers an increased risk of thrombus formation. The presence of increased thrombogenesis or a prothrombotic state is important, since many markers of a hypercoagulable state have prognostic implications for cardiovascular events and mortality. 4 Alternatively, the high rates of heart failure and thromboembolism in these patients may be related to blood pressure load. Patients with singlechamber ventricular pacing, when compared with those with dualchamber pacing systems, lose the normal diurnal change in 24 h bloodpressure profile, with loss of the normal nocturnal blood-pressure fall (non-dippers).5 Although the role of non-dipping is uncertain, it is wellrecognised that stroke, silent cerebrovascular disease, and leftventricular hypertrophy are more common in non-dippers than in dippers, possibly because of an increased 24 h blood pressure load. Clearly, atrial pacing is preferable in patients undergoing pacemaker implantation for sick-sinus syndrome. Antithrombotic therapy should be used in all patients who are managed with single-chamber ventricular pacing who subsequently develop atrial fibrillation, and at least considered in those who do not subsequently develop atrial fibrillation.
SIR—P M Fayers and D J Hand (Oct 4, p 1025) 1 revisit the well worn pathway by outlining the problems that arise from external validity of randomised control trials, but miss the essential point. Evidence should not be derived from a hierarchy of approaches with the randomised controlled trial as the gold standard and expert opinion at the bottom of the pile, but from a spectrum of methods that matches approach to the context and perspective of the question being asked. Real life is a cognitive continuum of decision making whereby structured systems can be examined by an analytical approach, but as tasks become less organised, intuition predominates with decisions based on experience, future expectations, and complex human inter-relationships. The development of research methodology to date has been characterised by all talk and no positive action—endless philosophical debates by academics on how evidence of effectiveness and cost-effectiveness should and should not be obtained. We in the front line would value some positive, pragmatic frameworks that are useful, accessible, and acceptable to facilitate the delivery of care in an environment that is characterised by the often conflicting dictates of evidence, economics, equity, and empowerment rather than endless theoretical reiteration.
*Christopher R Gibbs, Andrew D Blann, Gregory Y H Lip
1
Haemostasis Thrombosis and Vascular Biology Unit, University Department of Medicine, City Hospital, Birmingham B18 7QH, UK 1
2
3
4
5
Andersen HR, Nielsen JC, Thomsen PEB, et al. Long-term follow-up of patients from a randomised trial of atrial versus ventricular pacing for sick-sinus syndrome. Lancet 1997; 350: 1210–16. Lip GYH. Does atrial fibrillation confer a hypercoagulable state? Lancet 1995; 346: 1313–14. Lau CP, Tse HF, Cheng G. Effects of atrioventricular asynchrony on platelet activation: implication of thromboembolism in paced patients. Heart 1997; 78: 358–63. Lip GYH, Lowe GDO. Fibrin D-Dimer: a useful marker of thrombogenesis? Clin Sci 1995; 89: 205–14. Lip GYH, Adeotoye AO, Zarifis J, Singh SP, Watson RDS, Beevers DG. Ambulatory blood pressure monitoring in patients with single chamber ventricular pacemakers (VVI pacing mode). Am J Hypertens 1996; 12: 1240–41.
Vol 350 • December 13, 1997
D P Kernick St Thomas Health Centre, Exeter EX4 1HJ, UK Fayers PM, Hand DJ. Generalisation from phase III clinical trials: survival, quality of life, and health economics. Lancet 1997; 350: 1025–27.
SIR—We agree with P M Fayers and D J Hand1 that generalising from the results of randomised clinical trials (RCTs) is not always easy. For healtheconomic assessment, they conclude that only basic information should be collected in most trials, because costs inside a trial do not reflect accurately costs for future patients. They therefore suggest that it may be preferable “to assess the real costs and treatment benefits as experienced in a routine, non-RCT, environment”, but do not specify how such a comparative assessment should be conducted to yield reliable answers. As has been extensively documented, non-randomised studies may provide results that are seriously biased. 2
A better approach is surely to make clinical trials more pragmatic, so that they mimic routine clinical settings and generalisation is more straightforward.3 Trials need to be designed so that the costs of the research can be separated from the costs of health care. Health-economic data are best collected on all the trial participants; it is a mistaken assumption that such evaluations can be done with sample sizes smaller than those used for assessments of clinical effectiveness. 4 In some circumstances, we may be able to predict the main components that contribute to total costs and restrict data collection to these items. In other areas, such as screening and primary prevention, costs should be viewed from a broad perspective, since complex interventions can have unexpected effects on the amount and types of resources used.5 The full variability of costs across all the individuals in the trial can then be properly taken into account when inferences are made about average costs. The purpose is to produce an unbiased estimate of the average cost difference between two treatment strategies, together with a confidence interval, not as Fayers and Hand imply simply to test a null hypothesis of no difference in costs. To aid generalisability, both use of resources and costs need to be reported so that the implications of subsequent changes in unit costs can be judged. For similar reasons, resource use should be subdivided into relevant components to provide transparency of presentation.3 Rather than abandoning RCTs as a method of health-economic assessment, we should put our efforts into making them relevant to clinical practice and therefore generalisable. The need for reliable and scientifically rigorous assessments of healtheconomic issues can only be expected to increase in the future. *Simon Thompson, Julie Barber Department of Medical Statistics and Evaluation, Imperial College School of Medicine, Hammersmith Hospital, London W12 0NN, UK 1
2 3
4
Fayers PM, Hand DJ. Generalisations from phase III clinical trials: survival, quality of life, and health economics. Lancet 1997; 350: 1025–27. Pocock SJ. Clinical trials: a practical approach. Chichester: Wiley, 1983. Drummond MF, Jefferson TO. Guidelines for authors and peer reviewers of economic submissions to the BMJ. BMJ 1996; 313: 275–83. Gray AM, Marshall M, Lockwood A, Morris J. Problems in conducting economic evaluations alongside clinical
1781
3
4
5
Lye M. Access to advances in cardiology. Lancet 1997; 350: 1162–63. Parameshwar J, Shackell MM, Richardson A, Poole Wilson PA, Sutton GC. Prevalence of heart failure in three general practices in north west London. Br J Gen Pract 1992; 42: 287–89. Pitt B, Segal R, Martinez FA, et al. Randomised trial of losartan versus captopril in patients over 65 with heart failure (Evidence of Losartan in the Elderly Study, ELITE). Lancet 1997; 349: 747–52. The SOLVD investigators. Effect of enalapril on survival in patients with reduced left ventricular ejection fractions and congestive heart failure. N Engl J Med 1991; 325: 293–302. Cohn JN, Johnson G, Ziesche S, et al. A comparison of enalapril with hydralazineisosorbide dinitrate in the treatment of chronic congestive heart failure. N Engl J Med 1991; 325: 303–10.
HIV-1 infection associated with abnormal vaginal flora morphology and bacterial vaginosis SIR—In their response to our report 1 of an association between prevalent HIV infection and depletion of vaginal lactobacilli in Uganda, Austin Ugwumadu and colleagues (Oct 25, p 1251) 2 question whether the association between bacterial vaginosis and HIV may be confounded by concurrent cervical infections, particularly Chlamydia trachomatis. As noted in table 3 in our report, most women had no evidence of cervical infections, and no association was found between prevalent HIV and chlamydial infection, irrespective of the presence of bacterial vaginosis. Ugwumadu et al also comment on the low prevalence of C trachomatis in this population and question whether this may be due to a reduced sensitivity of the urinary ligase chain reaction (LCR) assay as a consequence of difficulties in specimen handling We have compared LCR detection of C trachomatis in urine samples with hybrid capture detection of C trachomatis from 905 self-collected vaginal swab specimens (Digene, Silver Spring, MD), and found complete agreement between the two assays and modes of sample collection. The prevalence of C trachomatis in this unselected and largely symptom-free sample of rural Ugandan women is not atypical for other unselected populations in rural East Africa (J Schachter, personal communication), and the age-specific prevalence of chlamydia shows the expected higher rates in younger women (at ages 15–19 years, the prevalence of C trachomatis was 6·6%). Furthermore, recent studies have shown
1780
stability of LCR detection of C trachomatis in urine kept at room temperature for up to 72 h before processing. We are therefore confident that the urinary LCR assay is sensitive for the detection of C trachomatis in the Rakai setting. S Ledru and colleagues (Oct 25, p 1251) 3 question our use of morphological methods for diagnosis of bacterial vaginitis and recommend the use of clinical criteria. As we noted the morphological assessment of this condition has been validated in studies of conventional clinical diagnostic criteria. Moreover, in collaboration with other investigators, we have examined the association between prevalent HIV and clinically diagnosed bacterial vaginosis among 6667 pregnant women from Blantyre, Malawi.4,5 In the Malawian study, the multivariate adjusted odds ratio of HIV infection in women with clinical bacterial vaginosis was 2·5 (95% CI 2·1–2·9),4 whereas in our Ugandan study, the odds ratio was 1·56 (1·24–1·97) with morphologically diagnosed and 2·08 (1·48–2·94) for severe infection (scores 9–10). Thus, the morphological diagnosis seems conservative. We have also analysed preliminary data on bacterial vaginosis and incident HIV infection with a nested case-control design. There were 77 incident cases of HIV who were individually matched by age (within 2·5 years) and place of residence, with 328 HIV-negative controls. The duration of observation was similar for cases and controls. These women provided a self-collected vaginal swab at the study visit before the follow-up interval during which they were at risk of seroconversion. Swabs were rolled onto microscope slides, gram-stained, and sent to the University of Pittsburgh for morphological evaluation, blinded with respect to HIV status. The findings suggest a possible increased risk of incident HIV infection associated with pre-existing bacterial infection with conditional logistic regression for matched data, the odds ratio of incident HIV was 1·50 (0·78–2·80). Although not significant, this result is consistent with the findings for HIV among women with bacterial vaginosis, 1,4,5 and suggests that depletion of vaginal lactobacilli may be a risk factor for HIV acquisition. We thank L Rabe for technical assistance.
*R H Gray, M J Wawer, N Sewankambo, D Serwadda Department of Population Dynamics, School of Hygiene and Public Health, Johns Hopkins University, Baltimore, MD 21205, USA
1
2
3
4
5
Sewankambo N, Gray RH, Wawer MJ, et al. HIV-1 infection associated with abnormal vaginal flora morphology and bacterial vaginosis. Lancet 1997; 350: 546–50. Ugwumadu A, Hay P, Taylor-Robinson D. HIV-1 infection associated with abnormal vaginal flora morphology and bacterial vaginosis. Lancet 1997; 350: 1251. Ledru S, Meda N, Ledru E, Bazie AJ, Chiron JP. HIV-1 infection associated with abnormal vaginal flora morphology and bacterial vaginosis. Lancet 1997; 350: 1251–52. Taha TE, Miotti P, Dalabetta G, et al. Bacterial vaginosis during pregnancy: associations with HIV, preterm delivery and mother-to-child transmission of HIV. Presented at conference on global strategies for the prevention of HIV transmission from mothers to infants. Washington DC, Sept 3–6, 1997: 13. Gray RH, Wawer MJ, Taha TE, et al. Depleted vaginal Lactobacilli and BV increase HIV acquisition in studies in Uganda and Malawi. International Infectious Disease Society of America, 35th Annual Meeting, San Francisco, Sept 13–16, 1997: abstract.
Thromboembolic risk and pacemakers SIR—Henning Rud Andersen and coworkers (Oct 25, p 1210) 1 provide further evidence to confirm the higher incidence of atrial fibrillation, heart failure, and increased risk of thromboembolism in patients with sick-sinus syndrome who are managed with single-chamber ventricular pacing than in those managed with single-chamber atrial pacing or dualchamber pacing. We postulate that the higher adverse event rate in the ventricular-pacing group, may be the result of an abnormal haemodynamic response, in the presence of a prothrombotic or hypercoagulable state. In Andersen's study, thromboembolic events were higher in those managed with ventricular pacing, whether or not they subsequently developed atrial fibrillation. Atrial fibrillation is itself an independent risk factor for stroke and thromboembolism, with associated intra-atrial stasis, and it is well recognised that such patients develop a hypercoagulable state, which may contribute to the risk of thrombogenesis.2 These factors may, in part, explain the higher risk of thromboembolism associated with ventricular pacing in those who subsequently developed atrial fibrillation. The greater thromboembolic event rate associated with ventricular pacing in patients who did not subsequently develop atrial fibrillation may itself
Vol 350 • December 13, 1997
THE LANCET
From efficacy to costeffectiveness
reflect increased thrombogenesis; Lau and colleagues3 reported increased spontaneous systemic-platelet activation in such patients. These findings were related to the loss of atrioventricular synchrony, and to the lack of atrial pacing leading to intraatrial stasis, which confers an increased risk of thrombus formation. The presence of increased thrombogenesis or a prothrombotic state is important, since many markers of a hypercoagulable state have prognostic implications for cardiovascular events and mortality. 4 Alternatively, the high rates of heart failure and thromboembolism in these patients may be related to blood pressure load. Patients with singlechamber ventricular pacing, when compared with those with dualchamber pacing systems, lose the normal diurnal change in 24 h bloodpressure profile, with loss of the normal nocturnal blood-pressure fall (non-dippers).5 Although the role of non-dipping is uncertain, it is wellrecognised that stroke, silent cerebrovascular disease, and leftventricular hypertrophy are more common in non-dippers than in dippers, possibly because of an increased 24 h blood pressure load. Clearly, atrial pacing is preferable in patients undergoing pacemaker implantation for sick-sinus syndrome. Antithrombotic therapy should be used in all patients who are managed with single-chamber ventricular pacing who subsequently develop atrial fibrillation, and at least considered in those who do not subsequently develop atrial fibrillation.
SIR—P M Fayers and D J Hand (Oct 4, p 1025) 1 revisit the well worn pathway by outlining the problems that arise from external validity of randomised control trials, but miss the essential point. Evidence should not be derived from a hierarchy of approaches with the randomised controlled trial as the gold standard and expert opinion at the bottom of the pile, but from a spectrum of methods that matches approach to the context and perspective of the question being asked. Real life is a cognitive continuum of decision making whereby structured systems can be examined by an analytical approach, but as tasks become less organised, intuition predominates with decisions based on experience, future expectations, and complex human inter-relationships. The development of research methodology to date has been characterised by all talk and no positive action—endless philosophical debates by academics on how evidence of effectiveness and cost-effectiveness should and should not be obtained. We in the front line would value some positive, pragmatic frameworks that are useful, accessible, and acceptable to facilitate the delivery of care in an environment that is characterised by the often conflicting dictates of evidence, economics, equity, and empowerment rather than endless theoretical reiteration.
*Christopher R Gibbs, Andrew D Blann, Gregory Y H Lip
1
Haemostasis Thrombosis and Vascular Biology Unit, University Department of Medicine, City Hospital, Birmingham B18 7QH, UK 1
2
3
4
5
Andersen HR, Nielsen JC, Thomsen PEB, et al. Long-term follow-up of patients from a randomised trial of atrial versus ventricular pacing for sick-sinus syndrome. Lancet 1997; 350: 1210–16. Lip GYH. Does atrial fibrillation confer a hypercoagulable state? Lancet 1995; 346: 1313–14. Lau CP, Tse HF, Cheng G. Effects of atrioventricular asynchrony on platelet activation: implication of thromboembolism in paced patients. Heart 1997; 78: 358–63. Lip GYH, Lowe GDO. Fibrin D-Dimer: a useful marker of thrombogenesis? Clin Sci 1995; 89: 205–14. Lip GYH, Adeotoye AO, Zarifis J, Singh SP, Watson RDS, Beevers DG. Ambulatory blood pressure monitoring in patients with single chamber ventricular pacemakers (VVI pacing mode). Am J Hypertens 1996; 12: 1240–41.
Vol 350 • December 13, 1997
D P Kernick St Thomas Health Centre, Exeter EX4 1HJ, UK Fayers PM, Hand DJ. Generalisation from phase III clinical trials: survival, quality of life, and health economics. Lancet 1997; 350: 1025–27.
SIR—We agree with P M Fayers and D J Hand1 that generalising from the results of randomised clinical trials (RCTs) is not always easy. For healtheconomic assessment, they conclude that only basic information should be collected in most trials, because costs inside a trial do not reflect accurately costs for future patients. They therefore suggest that it may be preferable “to assess the real costs and treatment benefits as experienced in a routine, non-RCT, environment”, but do not specify how such a comparative assessment should be conducted to yield reliable answers. As has been extensively documented, non-randomised studies may provide results that are seriously biased. 2
A better approach is surely to make clinical trials more pragmatic, so that they mimic routine clinical settings and generalisation is more straightforward.3 Trials need to be designed so that the costs of the research can be separated from the costs of health care. Health-economic data are best collected on all the trial participants; it is a mistaken assumption that such evaluations can be done with sample sizes smaller than those used for assessments of clinical effectiveness. 4 In some circumstances, we may be able to predict the main components that contribute to total costs and restrict data collection to these items. In other areas, such as screening and primary prevention, costs should be viewed from a broad perspective, since complex interventions can have unexpected effects on the amount and types of resources used.5 The full variability of costs across all the individuals in the trial can then be properly taken into account when inferences are made about average costs. The purpose is to produce an unbiased estimate of the average cost difference between two treatment strategies, together with a confidence interval, not as Fayers and Hand imply simply to test a null hypothesis of no difference in costs. To aid generalisability, both use of resources and costs need to be reported so that the implications of subsequent changes in unit costs can be judged. For similar reasons, resource use should be subdivided into relevant components to provide transparency of presentation.3 Rather than abandoning RCTs as a method of health-economic assessment, we should put our efforts into making them relevant to clinical practice and therefore generalisable. The need for reliable and scientifically rigorous assessments of healtheconomic issues can only be expected to increase in the future. *Simon Thompson, Julie Barber Department of Medical Statistics and Evaluation, Imperial College School of Medicine, Hammersmith Hospital, London W12 0NN, UK 1
2 3
4
Fayers PM, Hand DJ. Generalisations from phase III clinical trials: survival, quality of life, and health economics. Lancet 1997; 350: 1025–27. Pocock SJ. Clinical trials: a practical approach. Chichester: Wiley, 1983. Drummond MF, Jefferson TO. Guidelines for authors and peer reviewers of economic submissions to the BMJ. BMJ 1996; 313: 275–83. Gray AM, Marshall M, Lockwood A, Morris J. Problems in conducting economic evaluations alongside clinical
1781