CORRESPONDENCE
superiority of the new agent over existing agents. If there is any theoretical advantage then this should be stated and, ideally~ tested. If there is no theoretical advantage, who, other than the drug company, benefits from the development of this drug? It m a y be argued that competition will bring down the price. In my opinion, there is already a high degree of competition in the non-ionic market. The high cost of development, animal and clinical trials makes it unlikely that the new drug will be able to significantly undercut the competition. If there is no advantage to patients, should they be asked to take part in this sort of research? There m a y be theoretical reasons for thinking that the new drug is safer than the competition. But a trial of this size would have no prospect of confirming this, given the good safety record of Iopamidol. Secondly, I take issue with the interpretation of the statistical analysis. The authors report.using a Mann-Whitney-Wilcoxon Test. This is confusing. Either a M a n n - W h i t n e y U test or a Wilcoxon rank sum test would have been appropriate to test the hypothesis that 'there is a difference between the populations treated with each agent in terms of the variables measured'. Predictably, the test(s) have failed to demonstrate any difference. This however does not 'confirm that there is no difference' in the 'average scores' (as stated) or in the populations sampled. To interpret lack of a statistically significant difference as confirmation that two populations are the same is erroneous. This m a y simply reflect too small a sample or inappropriate choice of variables.
Fig. 2
N. C H A L M E R S usual methods [1]. Shortly afterwards he developed epigastric and retrosternal pain, but a chest radiograph showed no evidence of pneumomediastinum or pleural effusions. An erect contrast study using non-ionic contrast medium revealed a large, localized 'outpouching' of the distal oesophagus (Fig. la) similar to those previously described [1]. Placing the patient supine and in the right lateral position resulted in contrast freely returning to the oesophageal lumen (Fig. 1b). Subsequent CT of the thorax following ingestion o f oral contrast medium (Fig. 2) showed localized extravasation of the contrast from the oesophagus towards the left side of the mediastinum, with consolidation in the adjacent left lower lobe. There was no evidence of free pleural or mediastinal communication. The patient was managed conservatively, by naso-gastric suction, intravenous hydration and antibiotic therapy. Following symptomatic improvement, upper GI endoscopy was performed and a 2 cm long, full thickness tear in the lateral wall of the distal oesophagus was found. Seven days after balloon dilatation the patient was discharged, and 17 days after the procedure a barium study showed complete resolution of the perforation. In at least some of our cases, the cause of the 'outpouching' of the distal oesophagus following balloon dilatation is localized perforation. However, it is still possible that in other cases, particularly those with small and very smooth 'outpouches', mucosal herniation through a partial thickness tear is the correct explanation. It is important to emphasize that, whatever the true mechanism, there is free drainage of extravasated material back into the oesophageal lumen and this group of patients will respond well to conservative management. Only when a large, free leak is demonstrated need early surgical intervention be considered. P. G. W H I T E H. A D A M S P. M. S M I T H
369
Department of Radiology Llandough Hospital Penarth South Glamorgan CF6 1XX
Reference
1 A d a m s H, Roberts GM, Smith PM. Oesophageal tears during pneumatic balloon dilatation for the treatment of achalasia. Clinical Radiology 1989;40:53 57.
15 Gladstone Terrace Edinburgh EH9 1LS
Reference
l Simmons M J, Waite DW, Galland RB, Torrie EPH. Double blind comparison of Iomepro1350 and lopamido1340 in intravenous digital subtraction angiography for peripheral vascular disease. Clinical Radiology 1992;45:338-339.
SIR T h a n k you for allowing us to reply to Dr Chalmers' comments. We are sorry that he feels complacent about the safety record of Iopamidol. Others, including Bracco themselves, whilst pleased with improved safety and tolerability of the non-ionic contrast media, still accept that there is room for improvement, not only in chemotoxicity but also in the viscosity of the higher concentrations. Hopefully, the newer contrast might be cheaper and reduce flow of papers on ionic/non-ionic drugs and high/low risk patients by enabling safer contrast media to all that require it. We think it unwise to predict the effects of new drugs, they need to be tested. Whilst it was not the remit of the paper to discuss physical properties we understand that Iomeprol has a lower osmolality and viscosity than Iopamidol. It is not 'virtually identical'. This property might therefore enable delivery of higher does of highly concentrated contrast through thinner and less traumatic catheters; and in high concentration this appears to be borne out. This might satisfy Dr Chalmers' requirement for 'benefit to the patient' (an odd prerequisite for a scientific journal). The Mann-Whitney-Wilcoxon test based on rank is appropriate for comparing the location of two populations using data from two independent samples as in this study. The test used here is the M a n n Whitney-Wilcoxon or the Wilcoxon W test which are equivalent for this situation [1]. This should not be confused with the Wilcoxon s u m of rank test which is suitable for either the one sample location problem or the comparison of matched pairs [2]. We acknowledge P. Prescot, PhD, once again for his advice on statistics. M. J. S I M M O N S D. W. W A I T E E. P. H. T O R R I E
Department of Radiology Royal Berkshire Hospital London Road Reading Berks RG1 5AN
CONTRAST MEDIA COMPARISONS SIR Simmons et al. [I] have reported a comparison of two virtually identical contrast media and have predictably failed to demonstrate any difference between them. I would like to raise two points. Firstly, who benefits from such research? Surely all medical research should offer some prospect of benefit to patients either directly or indirectly (e.g. through cost savings). The authors do not claim any
References
1 Hertmansterger TP. Statistical inJerence based on ranks. Wiley, 1984:132q35. 2 Hertmansterger TP. Statistical inference based on ranks. Wiley, 1984:31 38.
superiority of the new agent over existing agents. If there is any theoretical advantage then this should be stated and, ideally~ tested. If there is no theoretical advantage, who, other than the drug company, benefits from the development of this drug? It m a y be argued that competition will bring down the price. In my opinion, there is already a high degree of competition in the non-ionic market. The high cost of development, animal and clinical trials makes it unlikely that the new drug will be able to significantly undercut the competition. If there is no advantage to patients, should they be asked to take part in this sort of research? There m a y be theoretical reasons for thinking that the new drug is safer than the competition. But a trial of this size would have no prospect of confirming this, given the good safety record of Iopamidol. Secondly, I take issue with the interpretation of the statistical analysis. The authors report.using a Mann-Whitney-Wilcoxon Test. This is confusing. Either a M a n n - W h i t n e y U test or a Wilcoxon rank sum test would have been appropriate to test the hypothesis that 'there is a difference between the populations treated with each agent in terms of the variables measured'. Predictably, the test(s) have failed to demonstrate any difference. This however does not 'confirm that there is no difference' in the 'average scores' (as stated) or in the populations sampled. To interpret lack of a statistically significant difference as confirmation that two populations are the same is erroneous. This m a y simply reflect too small a sample or inappropriate choice of variables.
Fig. 2
N. C H A L M E R S usual methods [1]. Shortly afterwards he developed epigastric and retrosternal pain, but a chest radiograph showed no evidence of pneumomediastinum or pleural effusions. An erect contrast study using non-ionic contrast medium revealed a large, localized 'outpouching' of the distal oesophagus (Fig. la) similar to those previously described [1]. Placing the patient supine and in the right lateral position resulted in contrast freely returning to the oesophageal lumen (Fig. 1b). Subsequent CT of the thorax following ingestion o f oral contrast medium (Fig. 2) showed localized extravasation of the contrast from the oesophagus towards the left side of the mediastinum, with consolidation in the adjacent left lower lobe. There was no evidence of free pleural or mediastinal communication. The patient was managed conservatively, by naso-gastric suction, intravenous hydration and antibiotic therapy. Following symptomatic improvement, upper GI endoscopy was performed and a 2 cm long, full thickness tear in the lateral wall of the distal oesophagus was found. Seven days after balloon dilatation the patient was discharged, and 17 days after the procedure a barium study showed complete resolution of the perforation. In at least some of our cases, the cause of the 'outpouching' of the distal oesophagus following balloon dilatation is localized perforation. However, it is still possible that in other cases, particularly those with small and very smooth 'outpouches', mucosal herniation through a partial thickness tear is the correct explanation. It is important to emphasize that, whatever the true mechanism, there is free drainage of extravasated material back into the oesophageal lumen and this group of patients will respond well to conservative management. Only when a large, free leak is demonstrated need early surgical intervention be considered. P. G. W H I T E H. A D A M S P. M. S M I T H
369
Department of Radiology Llandough Hospital Penarth South Glamorgan CF6 1XX
Reference
1 A d a m s H, Roberts GM, Smith PM. Oesophageal tears during pneumatic balloon dilatation for the treatment of achalasia. Clinical Radiology 1989;40:53 57.
15 Gladstone Terrace Edinburgh EH9 1LS
Reference
l Simmons M J, Waite DW, Galland RB, Torrie EPH. Double blind comparison of Iomepro1350 and lopamido1340 in intravenous digital subtraction angiography for peripheral vascular disease. Clinical Radiology 1992;45:338-339.
SIR T h a n k you for allowing us to reply to Dr Chalmers' comments. We are sorry that he feels complacent about the safety record of Iopamidol. Others, including Bracco themselves, whilst pleased with improved safety and tolerability of the non-ionic contrast media, still accept that there is room for improvement, not only in chemotoxicity but also in the viscosity of the higher concentrations. Hopefully, the newer contrast might be cheaper and reduce flow of papers on ionic/non-ionic drugs and high/low risk patients by enabling safer contrast media to all that require it. We think it unwise to predict the effects of new drugs, they need to be tested. Whilst it was not the remit of the paper to discuss physical properties we understand that Iomeprol has a lower osmolality and viscosity than Iopamidol. It is not 'virtually identical'. This property might therefore enable delivery of higher does of highly concentrated contrast through thinner and less traumatic catheters; and in high concentration this appears to be borne out. This might satisfy Dr Chalmers' requirement for 'benefit to the patient' (an odd prerequisite for a scientific journal). The Mann-Whitney-Wilcoxon test based on rank is appropriate for comparing the location of two populations using data from two independent samples as in this study. The test used here is the M a n n Whitney-Wilcoxon or the Wilcoxon W test which are equivalent for this situation [1]. This should not be confused with the Wilcoxon s u m of rank test which is suitable for either the one sample location problem or the comparison of matched pairs [2]. We acknowledge P. Prescot, PhD, once again for his advice on statistics. M. J. S I M M O N S D. W. W A I T E E. P. H. T O R R I E
Department of Radiology Royal Berkshire Hospital London Road Reading Berks RG1 5AN
CONTRAST MEDIA COMPARISONS SIR Simmons et al. [I] have reported a comparison of two virtually identical contrast media and have predictably failed to demonstrate any difference between them. I would like to raise two points. Firstly, who benefits from such research? Surely all medical research should offer some prospect of benefit to patients either directly or indirectly (e.g. through cost savings). The authors do not claim any
References
1 Hertmansterger TP. Statistical inJerence based on ranks. Wiley, 1984:132q35. 2 Hertmansterger TP. Statistical inference based on ranks. Wiley, 1984:31 38.
