729
vitamins and minerals they need from their food. However, if they cannot eat well-balanced meals, because of food preferences or a hurried schedule, for example, supplements that do not exceed 100% of USRDA are not known to cause any risks and may prove beneficial for selected groups. An increase in IQ seems to be one of those potential benefits, but, until our study is properly replicated, it remains tentative. Department of Sociology, California State University, Stanislas, Turlock, California 95380, USA
STEPHEN SCHOENTHALER
1. Schoenthaler SJ, Amos SP, Eysenck HJ, Peritz E, Yudkin J. Controlled trial of vitamin-mineral supplementation: effects on intelligence and performance. Personal Individ Diff 1991; 12: 351-62. 2. Schoenthaler SJ, Amos SP, Doraz WE, Kelly MA, Wakefield J. Controlled trial of vitamin-mineral supplementation on intelligence and brain function. Personal Individ Diff 1991; 12: 343-50. 3. Crombie IK, Todman J, McNeill G, Florey C du V, Menzies J, Kennedy RA. Effect of vitamin mineral supplementation on verbal and nonverbal reasoning of schoolchildren. Lancet 1990; 335: 744-47. 4. Schoenthaler SJ. Abstracts of early papers on the effects of vitamins and mineral supplementation on IQ and behavior. Personal Individ Diff 1991; 12: 335-42. 5. Schoenthaler SJ. Improve your child’s IQ and behaviour. London: BBC Books, 1991. 6. Nagheri JA Matrix analogues test short form examiner’s manual. Columbus, Ohio: Charles E Merrill Publishing, 1985. 7. Naismith DJ, Nelson M, Burley VJ, Gatenby SJ. Can children’s intelligence be increased by vitamin mineral supplements? Lancet 1988; ii: 335. 8. Benton D, Roberts G. Effects of vitamin and mineral supplementation on intelligence of a sample of schoolchildren. Lancet 1988; i: 140-43.
Lipiodol computed tomography for small hepatocellular carcinomas SIR,-We endorse the views expressed in your Feb 9 editorial with respect
to the value of ’Lipiodol’-enhanced computed tomography (CT) in the detection of small hepatocellular tumours. We have used this technique for over three years and have found it to be better then other imaging techniques, especially for the
detection of satellite nodules. We would, however, like to comment on a few points. The pharmacodynamics of lipiodol retention within tumours are not well understood, but seem to be based on more than simple haemodynamics. Iwai et all have described lipiodol accumulation within tumour extracellular space. They suggest that this might result from the absence of lymph vessels in tumour tissue, citing the use of lipiodol in lymphangiography as evidence of its lymphophilic properties. Lipiodol absorption from tissues into lymph vessels is, however, very poor,2 and there is no evidence that the lymphatic system contributes much to the clearance of lipiodol from normal hepatic tissue. Ivancev et al3,4 in in-vivo microscopy studies of the liver after lipiodol injection have shown active uptake of the lipid by both hepatocytes and Kupffer cells, and these investigators have suggested that shunting into the hepatic veins via the sinusoids is the principal route of clearance from normal liver. Electron microscopy of tumours resected in our unit has demonstrated substantial uptake of the lipid by the tumour cells themselves, and Park and colleagues5 have described globular and non-globular lipid in association with the plasma membrane of tumour cells. Lipiodol is retained by many types of tumour after intra-arterial infusion: we have lately reported its successful use in the management of locally recurrent carcinoma of the breast .6 The use of lipiodol to target chemotherapeutic agents to tumour foci has been widely reported, but so far there have been no randomised controlled trials. Two non-randomised trials have been reported;one demonstrated a modest improvement in survival whereas the other showed none. An uncontrolled pilot study on 57 patients from this unit who were treated with a lipiodol-epirubicin emulsion9 suggests some benefit in small hepatocellular carcinoma (Okuda stage I and II). Side-effects are less frequent and less severe than with systemic chemotherapy, although fulminant hepatic failure may follow administration of the lipiodol-epirubicin emulsion in patients with large tumours and poor synthetic function (Okuda stage III). Our policy is to start lipiodol-targeted
chemotherapy at initial angiography. Patients whose tumours prove inoperable are thus treated without delay, and targeting is not compromised by lipiodol retained from diagnostic administration. Initial hopes that targeted treatment of resectable tumours would improve survival after surgery have not been fulfilled. 131 I-lipiodol seems capable of delivering an ablation dose of beta radiation to small foci of hepatocellular carcinoma" and may be the treatment of choice in cirrhotic patients who are at high risk of complications after resection. Of 9 patients with hepatocellular tumours treated in this unit, 6 showed evidence of partial tumour response on computed tomography scan and by tumour markers.
Hepatobiliary and Liver Transplantation Unit, Clinical Oncology Unit, Royal Free Hospital School of Medicine, London NW3 2QG, UK 1.
RICHARD NOVELL GEOFF DUSHEIKO ANDREW HILSON ROBERT DICK RICHARD BEGENT KENNETH HOBBS
Iwai K, Maeda H, Konno T. Use of oily contrast medium for selective drug targeting to tumour: enhanced therapeutic effect and X-ray image. Cancer Res 1984; 44:
2115-21. 2. Miller DL, O’Leary TJ, Girton M. Distribution of iodized oil within the liver after hepatic arterial injection. Radiology 1987; 162: 849-52. 3. Ivancev K, Lunderquist A, McCuskey R, McCuskey P, Wretlind A. Effect of intravenously injected iodinated lipid emulsions on the liver. Acta Radiol 1989; 30: 291-98. 4. Ivancev K, Lunderquist A, McCuskey R, McCuskey P, Wretlind A. Experimental investigation of a new iodinated lipid emulsion for computed tomography of the liver. Acta Radiol 1989; 30: 1-7. 5. Park C, Choi S, Kim H, et al. Distribution of lipiodol in hepatocellular carcinoma. Liver 1990; 10: 72-78. 6. Novell JR, Parbhoo SP, Dawson K, Dick R, Kelleher SM. Targeted therapy for recurrent breast carcinoma with regional ’Lipiodol’/epirubicin infusion. Lancet 1990; 336: 1383. 7. Furuta T, Kanematsu T, Matsumata T, et al. Lipiodolization prolongs survival rates in postoperative patients with a recurrent hepatocellular carcinoma. Hepatogastroenterol 1990; 37: 494-97. 8. Kalayci C, Johnson PJ, Raby N, Metivier EM, Williams R. Intraarterial adriamycin and lipiodol for inoperable hepatocellular carcinoma: a comparison with intravenous adriamycin. J Hepatol 1990; 11: 349-53. 9. Novell JR, Dusheiko G, Markham N, Reddy K, Dick R, Hobbs K. Selective regional chemotherapy of unresectable hepatic tumours using lipiodol. HPB Surg (in press). 10. Novell JR, Hilson A, Hobbs KEF. Ablation of recurrent primary liver cancer using 131I-lipiodol. Postgrad MedJ (in press).
Hypervolaemic haemodilution before major surgery SIR,-Dr Trouwborst and colleagues (Nov 24, p 1295) report the of acute hypervolaemic haemodilution to avoid blood transfusion during major surgery. Their technique was successful and they recommend it as a useful means to reduce blood loss perioperatively as well as to save time. However, their success does not mean that their method is better than conventional fluid transfusion, since no comparison was done. No evidence was shown that acute preoperative hypervolaemic haemodilution resulted in fewer red cells being lost. It is unreasonable to draw such a conclusion from the finding that the packed cell volume did not differ significantly between patients who lost less than or more than 20% of their initial blood volume. Moreover, the safety of acute hypervolaemia is questionable. Although the patients had no symptoms of pulmonary oedema, the haemodynamic data showed a high risk of pulmonary congestion due to acute overloading. Acute infusion of 1500 ml of dextran 40 and 1500 ml of Ringer’s lactate increased pulmonary wedge pressure from 5-5 to 20-8 mm Hg (a four-fold rise), in association with a slight increase in cardiac index from 2-43 to 3-10 Ifminfm2 (27-5%), suggesting that cardiac function became depressed. Since anaesthesia with fentanyl and enflurane (0-4%) is thought to have a weak cardiodepressant effect, this acute overloading itself may in part contribute to cardiac dysfunction, otherwise patients may have a compromised cardiovascular system. These haemodynamic changes in response to acute hypervolaemic haemodilution indicate to us that such a technique may have drawbacks if used routinely. This method should be used selectively and with careful use
730
RESULTS OF TWO CASE-CONTROL STUDIES
cardiovascular monitoring. We would restrict its use to patients for whom no alternative measures are available, and we also recommend the use of erythropoietin in such patients.1 Blood Transfusion Service,
Kyushu University Hospital, Fukuoka 812, Japan
S. INABA
Department of Anaesthesiology and Critical Care Medicine, Faculty of Medicine,
S. HOKA
Kyushu University
J. YOSHITAKE
1. Maeda H, Hitomi Y, Hirata R, Lancet 1989; ii: 284.
et
al.
Erythropoietin and autologous blood donation. I
*Piroxicam not listed
I
pharmaceutical benefit during study 1 n study 1, tDiclofenac. diflumsal, ibuprofen, indomethacin, naproxen, and sulindac and same drugs with the addition of ketoprofen m study 2
NSAIDs and risk of upper
gastrointestinal
bleeding SIR,-We were surprised by the magnitude of the odds ratios (OR) estimated by Dr Laporte and colleagues (Jan 13, p 85) in their case-control study of non-steroidal anti-inflammatory drugs (NSAIDs) and hospital admission for upper gastrointestinal bleeding. The differences in the estimated OR for individual NSAIDs should be interpreted with caution too. We have done two case-control studies. The first covered patients aged 50 years and over admitted with upper gastrointestinal bleeding or ulcer perforation to two public hospitals in the Newcastle area of New South Wales (population 400 000) between 1985 and 1987. The second recruited consecutive admissions with
haemorrhage or perforation (irrespective of age) to all public hospitals in the area between mid-1988 and mid-1990. The aim of study 1 was to estimate the overall relative risk of ulcer complications associated with any NSAID. The aim of study 2 was to compare the OR with piroxicam to that of the other NSAIDs. By combining data from the two studies we hope to investigate the importance of other factors, such as dose-duration of therapy, co-prescription of aspirin, ulcer site, and history of peptic ulceration. Almost all cases of haemorrhage had a diagnostic endoscopy within 24 hours of admission. Cases were matched with hospital controls (1:in study 1; 1:2 in study 2) for sex and age within 5 years. Controls were new admissions with primary diagnoses not thought to be related to NSAIDs, and they were interviewed within a week of the cases. Information on use of NSAIDs by cases and controls during the week before admission was obtained by one research nurse using a structured interview, with a mixture of open followed by closed questions and prompts in the form of the Australian proprietary names of all marketed NSAIDs. This information was compared with that held in family practitioner prescribing records with simple rules to resolve conflict between these sources. In Australia drugs are little used in the community unless they are listed in the Commonwealth Government’s Schedule of Pharmaceutical Benefits. The NSAIDs listed were diclofenac, diflunisal, ibuprofen, indomethacin, naproxen, and sulindac for study 1, ketoprofen and piroxicam being added in study 2. The results are summarised in the table. OR and 95 % confidence intervals (CI) were calculated by conditional logistic regression with adjustment for social status, smoking, and alcohol consumption. The levels of use of NSAIDS (excluding aspirin) by cases and especially by controls were much higher than those recorded by Laporte et al. We have confirmation of this high use from a concurrent prevalence study of a statistically representative sample (n 2578) of the Newcastle community which established that NSAIDs had been used in the previous 4 days by 15-4% of those aged 55-64 years and by 21-4% of those aged 65 or more (unpublished). In our first study the OR for bleeding or perforation with use of any NSAID in the previous week was 2-9 (95% CI 1-8-4-8), and in the second study it was 3-1 (2-3-4-1). The comparable figures from the study of Laporte et al were 7-4 (2-4-23-0) and 8-0 (3-8-16-6) with and without a previous history of ulceration. In contrast to Laporte et al we found the estimated relative risk with piroxicam to be only slightly higher than that with
as
the other commonly used NSAIDs, and there was substantial overlap of CIs (table). The proportion of our NSAID-using cases who had taken piroxicam (21 %) was identical to that drug’s national market share during the study period (data provided by Pfizer
Australia). The OR for use of any NSAID agrees well with the pooled estimates of risk from Hawkey’s meta-analysis of four other case control studies (pooled OR 3-1, 95% CI 2-3-4-2) and our meta-analysis of eleven case control studies (34, 28-43) 2 Although our relative risk estimates for bleeding or perforation with use of any NSAID are lower than those from the Spanish study, the Australian population attributable risk (aetiological fraction) is higher (24% vs about 15%) because of the higher use of NSAIDs in the Australian community. The low use of NSAIDs by the Spanish controls results in that study having a lower power than our study, and a similar power to some previously published case-control studies despite its large size. We calculate that the minimum OR that the Spanish study could have detected with a power of 0-8 (rJ. 0-05) was 1 8.3 This calculation takes account of matching and assumes a correlation between cases and controls on exposure of 0-2. Under the same assumptions our second study would have detected an odds ratio as low as 1 ’6. The case-control studies of Somerville et all and Griffin et all could have detected OR as low as 1 and 20, respectively. We agree with Laporte et al that NSAIDs are an important cause of serious gastrointestinal bleeding. In our view it is the population risk rather than the individual risk that is most significant. Since the problem in Australia is, in the main, dependent on the prevalence of exposure to the risk factor, rather than the magnitude of the individual’s risk, prevention should in most cases be directed at reducing exposure rather than attempting to lower risk by co-prescription of protective agents. Although we do not rule out the possibility that piroxicam is associated with a significantly higher relative risk than other NSAIDs, we believe that on present evidence apparent differences between individual agents should be interpreted with caution. —
Supported by grants from the National Health and Medical Research Council of Australia and Pfizer Austrialia Pry Ltd. The full results of this study will be published later.
Departments of Clinical Pharmacology, Statistics, and Nursing, University of Newcastle, Newcastle, NSW 2300, Australia
DAVID HENRY ANNETTE DOBSON CATHY TURNER PAMELA HALL CAROLINE FORBES PHILIPPA PATEY
=
1.
Hawkey CJ. Non-steroidal anti-inflammatory drugs and peptic ulcers. Br Med J1990, 300: 278-84.
Henry DA. Gastrointestinal bleeding and non-steroidal anti-inflammatory drugs In Lawson DH, ed. Royal College of Physicians of Edinburgh Current medicine-3 Edinburgh: Churchill Livingstone, (in press). 3. Dupont WD, Plummer WD. Power and sample size calculations, a review and computer program. Controlled Clin Trials 1990; 11: 116-28 4. Somerville K, Faulkner G, Langman M. Non-steroidal anti-inflammatory drugs and bleeding peptic ulcer. Lancet 1986; i: 462-64. 5. Griffin MR, Ray WA, Schaffner W. Non-steroidal anti-inflammatory drug use and death from peptic ulcer in elderly persons. Ann Intern Med 1988; 109: 359-63
2.
vitamins and minerals they need from their food. However, if they cannot eat well-balanced meals, because of food preferences or a hurried schedule, for example, supplements that do not exceed 100% of USRDA are not known to cause any risks and may prove beneficial for selected groups. An increase in IQ seems to be one of those potential benefits, but, until our study is properly replicated, it remains tentative. Department of Sociology, California State University, Stanislas, Turlock, California 95380, USA
STEPHEN SCHOENTHALER
1. Schoenthaler SJ, Amos SP, Eysenck HJ, Peritz E, Yudkin J. Controlled trial of vitamin-mineral supplementation: effects on intelligence and performance. Personal Individ Diff 1991; 12: 351-62. 2. Schoenthaler SJ, Amos SP, Doraz WE, Kelly MA, Wakefield J. Controlled trial of vitamin-mineral supplementation on intelligence and brain function. Personal Individ Diff 1991; 12: 343-50. 3. Crombie IK, Todman J, McNeill G, Florey C du V, Menzies J, Kennedy RA. Effect of vitamin mineral supplementation on verbal and nonverbal reasoning of schoolchildren. Lancet 1990; 335: 744-47. 4. Schoenthaler SJ. Abstracts of early papers on the effects of vitamins and mineral supplementation on IQ and behavior. Personal Individ Diff 1991; 12: 335-42. 5. Schoenthaler SJ. Improve your child’s IQ and behaviour. London: BBC Books, 1991. 6. Nagheri JA Matrix analogues test short form examiner’s manual. Columbus, Ohio: Charles E Merrill Publishing, 1985. 7. Naismith DJ, Nelson M, Burley VJ, Gatenby SJ. Can children’s intelligence be increased by vitamin mineral supplements? Lancet 1988; ii: 335. 8. Benton D, Roberts G. Effects of vitamin and mineral supplementation on intelligence of a sample of schoolchildren. Lancet 1988; i: 140-43.
Lipiodol computed tomography for small hepatocellular carcinomas SIR,-We endorse the views expressed in your Feb 9 editorial with respect
to the value of ’Lipiodol’-enhanced computed tomography (CT) in the detection of small hepatocellular tumours. We have used this technique for over three years and have found it to be better then other imaging techniques, especially for the
detection of satellite nodules. We would, however, like to comment on a few points. The pharmacodynamics of lipiodol retention within tumours are not well understood, but seem to be based on more than simple haemodynamics. Iwai et all have described lipiodol accumulation within tumour extracellular space. They suggest that this might result from the absence of lymph vessels in tumour tissue, citing the use of lipiodol in lymphangiography as evidence of its lymphophilic properties. Lipiodol absorption from tissues into lymph vessels is, however, very poor,2 and there is no evidence that the lymphatic system contributes much to the clearance of lipiodol from normal hepatic tissue. Ivancev et al3,4 in in-vivo microscopy studies of the liver after lipiodol injection have shown active uptake of the lipid by both hepatocytes and Kupffer cells, and these investigators have suggested that shunting into the hepatic veins via the sinusoids is the principal route of clearance from normal liver. Electron microscopy of tumours resected in our unit has demonstrated substantial uptake of the lipid by the tumour cells themselves, and Park and colleagues5 have described globular and non-globular lipid in association with the plasma membrane of tumour cells. Lipiodol is retained by many types of tumour after intra-arterial infusion: we have lately reported its successful use in the management of locally recurrent carcinoma of the breast .6 The use of lipiodol to target chemotherapeutic agents to tumour foci has been widely reported, but so far there have been no randomised controlled trials. Two non-randomised trials have been reported;one demonstrated a modest improvement in survival whereas the other showed none. An uncontrolled pilot study on 57 patients from this unit who were treated with a lipiodol-epirubicin emulsion9 suggests some benefit in small hepatocellular carcinoma (Okuda stage I and II). Side-effects are less frequent and less severe than with systemic chemotherapy, although fulminant hepatic failure may follow administration of the lipiodol-epirubicin emulsion in patients with large tumours and poor synthetic function (Okuda stage III). Our policy is to start lipiodol-targeted
chemotherapy at initial angiography. Patients whose tumours prove inoperable are thus treated without delay, and targeting is not compromised by lipiodol retained from diagnostic administration. Initial hopes that targeted treatment of resectable tumours would improve survival after surgery have not been fulfilled. 131 I-lipiodol seems capable of delivering an ablation dose of beta radiation to small foci of hepatocellular carcinoma" and may be the treatment of choice in cirrhotic patients who are at high risk of complications after resection. Of 9 patients with hepatocellular tumours treated in this unit, 6 showed evidence of partial tumour response on computed tomography scan and by tumour markers.
Hepatobiliary and Liver Transplantation Unit, Clinical Oncology Unit, Royal Free Hospital School of Medicine, London NW3 2QG, UK 1.
RICHARD NOVELL GEOFF DUSHEIKO ANDREW HILSON ROBERT DICK RICHARD BEGENT KENNETH HOBBS
Iwai K, Maeda H, Konno T. Use of oily contrast medium for selective drug targeting to tumour: enhanced therapeutic effect and X-ray image. Cancer Res 1984; 44:
2115-21. 2. Miller DL, O’Leary TJ, Girton M. Distribution of iodized oil within the liver after hepatic arterial injection. Radiology 1987; 162: 849-52. 3. Ivancev K, Lunderquist A, McCuskey R, McCuskey P, Wretlind A. Effect of intravenously injected iodinated lipid emulsions on the liver. Acta Radiol 1989; 30: 291-98. 4. Ivancev K, Lunderquist A, McCuskey R, McCuskey P, Wretlind A. Experimental investigation of a new iodinated lipid emulsion for computed tomography of the liver. Acta Radiol 1989; 30: 1-7. 5. Park C, Choi S, Kim H, et al. Distribution of lipiodol in hepatocellular carcinoma. Liver 1990; 10: 72-78. 6. Novell JR, Parbhoo SP, Dawson K, Dick R, Kelleher SM. Targeted therapy for recurrent breast carcinoma with regional ’Lipiodol’/epirubicin infusion. Lancet 1990; 336: 1383. 7. Furuta T, Kanematsu T, Matsumata T, et al. Lipiodolization prolongs survival rates in postoperative patients with a recurrent hepatocellular carcinoma. Hepatogastroenterol 1990; 37: 494-97. 8. Kalayci C, Johnson PJ, Raby N, Metivier EM, Williams R. Intraarterial adriamycin and lipiodol for inoperable hepatocellular carcinoma: a comparison with intravenous adriamycin. J Hepatol 1990; 11: 349-53. 9. Novell JR, Dusheiko G, Markham N, Reddy K, Dick R, Hobbs K. Selective regional chemotherapy of unresectable hepatic tumours using lipiodol. HPB Surg (in press). 10. Novell JR, Hilson A, Hobbs KEF. Ablation of recurrent primary liver cancer using 131I-lipiodol. Postgrad MedJ (in press).
Hypervolaemic haemodilution before major surgery SIR,-Dr Trouwborst and colleagues (Nov 24, p 1295) report the of acute hypervolaemic haemodilution to avoid blood transfusion during major surgery. Their technique was successful and they recommend it as a useful means to reduce blood loss perioperatively as well as to save time. However, their success does not mean that their method is better than conventional fluid transfusion, since no comparison was done. No evidence was shown that acute preoperative hypervolaemic haemodilution resulted in fewer red cells being lost. It is unreasonable to draw such a conclusion from the finding that the packed cell volume did not differ significantly between patients who lost less than or more than 20% of their initial blood volume. Moreover, the safety of acute hypervolaemia is questionable. Although the patients had no symptoms of pulmonary oedema, the haemodynamic data showed a high risk of pulmonary congestion due to acute overloading. Acute infusion of 1500 ml of dextran 40 and 1500 ml of Ringer’s lactate increased pulmonary wedge pressure from 5-5 to 20-8 mm Hg (a four-fold rise), in association with a slight increase in cardiac index from 2-43 to 3-10 Ifminfm2 (27-5%), suggesting that cardiac function became depressed. Since anaesthesia with fentanyl and enflurane (0-4%) is thought to have a weak cardiodepressant effect, this acute overloading itself may in part contribute to cardiac dysfunction, otherwise patients may have a compromised cardiovascular system. These haemodynamic changes in response to acute hypervolaemic haemodilution indicate to us that such a technique may have drawbacks if used routinely. This method should be used selectively and with careful use
730
RESULTS OF TWO CASE-CONTROL STUDIES
cardiovascular monitoring. We would restrict its use to patients for whom no alternative measures are available, and we also recommend the use of erythropoietin in such patients.1 Blood Transfusion Service,
Kyushu University Hospital, Fukuoka 812, Japan
S. INABA
Department of Anaesthesiology and Critical Care Medicine, Faculty of Medicine,
S. HOKA
Kyushu University
J. YOSHITAKE
1. Maeda H, Hitomi Y, Hirata R, Lancet 1989; ii: 284.
et
al.
Erythropoietin and autologous blood donation. I
*Piroxicam not listed
I
pharmaceutical benefit during study 1 n study 1, tDiclofenac. diflumsal, ibuprofen, indomethacin, naproxen, and sulindac and same drugs with the addition of ketoprofen m study 2
NSAIDs and risk of upper
gastrointestinal
bleeding SIR,-We were surprised by the magnitude of the odds ratios (OR) estimated by Dr Laporte and colleagues (Jan 13, p 85) in their case-control study of non-steroidal anti-inflammatory drugs (NSAIDs) and hospital admission for upper gastrointestinal bleeding. The differences in the estimated OR for individual NSAIDs should be interpreted with caution too. We have done two case-control studies. The first covered patients aged 50 years and over admitted with upper gastrointestinal bleeding or ulcer perforation to two public hospitals in the Newcastle area of New South Wales (population 400 000) between 1985 and 1987. The second recruited consecutive admissions with
haemorrhage or perforation (irrespective of age) to all public hospitals in the area between mid-1988 and mid-1990. The aim of study 1 was to estimate the overall relative risk of ulcer complications associated with any NSAID. The aim of study 2 was to compare the OR with piroxicam to that of the other NSAIDs. By combining data from the two studies we hope to investigate the importance of other factors, such as dose-duration of therapy, co-prescription of aspirin, ulcer site, and history of peptic ulceration. Almost all cases of haemorrhage had a diagnostic endoscopy within 24 hours of admission. Cases were matched with hospital controls (1:in study 1; 1:2 in study 2) for sex and age within 5 years. Controls were new admissions with primary diagnoses not thought to be related to NSAIDs, and they were interviewed within a week of the cases. Information on use of NSAIDs by cases and controls during the week before admission was obtained by one research nurse using a structured interview, with a mixture of open followed by closed questions and prompts in the form of the Australian proprietary names of all marketed NSAIDs. This information was compared with that held in family practitioner prescribing records with simple rules to resolve conflict between these sources. In Australia drugs are little used in the community unless they are listed in the Commonwealth Government’s Schedule of Pharmaceutical Benefits. The NSAIDs listed were diclofenac, diflunisal, ibuprofen, indomethacin, naproxen, and sulindac for study 1, ketoprofen and piroxicam being added in study 2. The results are summarised in the table. OR and 95 % confidence intervals (CI) were calculated by conditional logistic regression with adjustment for social status, smoking, and alcohol consumption. The levels of use of NSAIDS (excluding aspirin) by cases and especially by controls were much higher than those recorded by Laporte et al. We have confirmation of this high use from a concurrent prevalence study of a statistically representative sample (n 2578) of the Newcastle community which established that NSAIDs had been used in the previous 4 days by 15-4% of those aged 55-64 years and by 21-4% of those aged 65 or more (unpublished). In our first study the OR for bleeding or perforation with use of any NSAID in the previous week was 2-9 (95% CI 1-8-4-8), and in the second study it was 3-1 (2-3-4-1). The comparable figures from the study of Laporte et al were 7-4 (2-4-23-0) and 8-0 (3-8-16-6) with and without a previous history of ulceration. In contrast to Laporte et al we found the estimated relative risk with piroxicam to be only slightly higher than that with
as
the other commonly used NSAIDs, and there was substantial overlap of CIs (table). The proportion of our NSAID-using cases who had taken piroxicam (21 %) was identical to that drug’s national market share during the study period (data provided by Pfizer
Australia). The OR for use of any NSAID agrees well with the pooled estimates of risk from Hawkey’s meta-analysis of four other case control studies (pooled OR 3-1, 95% CI 2-3-4-2) and our meta-analysis of eleven case control studies (34, 28-43) 2 Although our relative risk estimates for bleeding or perforation with use of any NSAID are lower than those from the Spanish study, the Australian population attributable risk (aetiological fraction) is higher (24% vs about 15%) because of the higher use of NSAIDs in the Australian community. The low use of NSAIDs by the Spanish controls results in that study having a lower power than our study, and a similar power to some previously published case-control studies despite its large size. We calculate that the minimum OR that the Spanish study could have detected with a power of 0-8 (rJ. 0-05) was 1 8.3 This calculation takes account of matching and assumes a correlation between cases and controls on exposure of 0-2. Under the same assumptions our second study would have detected an odds ratio as low as 1 ’6. The case-control studies of Somerville et all and Griffin et all could have detected OR as low as 1 and 20, respectively. We agree with Laporte et al that NSAIDs are an important cause of serious gastrointestinal bleeding. In our view it is the population risk rather than the individual risk that is most significant. Since the problem in Australia is, in the main, dependent on the prevalence of exposure to the risk factor, rather than the magnitude of the individual’s risk, prevention should in most cases be directed at reducing exposure rather than attempting to lower risk by co-prescription of protective agents. Although we do not rule out the possibility that piroxicam is associated with a significantly higher relative risk than other NSAIDs, we believe that on present evidence apparent differences between individual agents should be interpreted with caution. —
Supported by grants from the National Health and Medical Research Council of Australia and Pfizer Austrialia Pry Ltd. The full results of this study will be published later.
Departments of Clinical Pharmacology, Statistics, and Nursing, University of Newcastle, Newcastle, NSW 2300, Australia
DAVID HENRY ANNETTE DOBSON CATHY TURNER PAMELA HALL CAROLINE FORBES PHILIPPA PATEY
=
1.
Hawkey CJ. Non-steroidal anti-inflammatory drugs and peptic ulcers. Br Med J1990, 300: 278-84.
Henry DA. Gastrointestinal bleeding and non-steroidal anti-inflammatory drugs In Lawson DH, ed. Royal College of Physicians of Edinburgh Current medicine-3 Edinburgh: Churchill Livingstone, (in press). 3. Dupont WD, Plummer WD. Power and sample size calculations, a review and computer program. Controlled Clin Trials 1990; 11: 116-28 4. Somerville K, Faulkner G, Langman M. Non-steroidal anti-inflammatory drugs and bleeding peptic ulcer. Lancet 1986; i: 462-64. 5. Griffin MR, Ray WA, Schaffner W. Non-steroidal anti-inflammatory drug use and death from peptic ulcer in elderly persons. Ann Intern Med 1988; 109: 359-63
2.
