Rapid Responses to:
|
|
Rapid Responses: Rapid Responses published:
-
![[Read Rapid Response]](/icons/misc/sectionDOWN.gif)
Test treatment: placebo effect, probability and more
- Alexis Descatha (1 May 2009)
-
pre-test probability
- Paul Glasziou (1 May 2009)
-
Response to Nitro has No Predictive Value BUT use it to relieve symptoms
- Malvinder S. Parmar (22 May 2009)
-
Response to GTN may be different in A&E and outpatient care
- Paul P Glasziou (3 June 2009)
|
|
|||
|
Alexis Descatha, MD PhD UVSQ-AP-HP, EMS 92 (SAMU92)-INSERM U687, Poincare teaching hospital, 92380 Garches, France
Send response to journal:
|
Sir, I read with a particular interest the paper of Paul Glasziou et al about Diagnosis using "test of treatment". However, some comments have been raised using the glyceryl trinitrate test for chest pain to illustrate our point of view. First, I think one major problem of the “test treatment” is, such as the authors mentioned too briefly, the placebo effect. In the case of chest pain, the physician should be very careful about the interpretation of a disappearance of pain after glyceryl trinitrate test: that is why, for instance, we teach our young colleagues not to induce improvement from asking “does the pain disappear?”, but to stay vague about the possible issue of the test. Second, one important way to eliminate the placebo effect is to have objective, but also specific measurements: to continue our example, to improve the accuracy of the glyceryl trinitrate test, electrocardiography (ECG) is widely used, taking into account electric modification are objective and also quite specific measures of cardiac disorders and acute coronary syndrome or ACS, whereas esophageal pain could be improve by glyceryl trinitrate but not the ECG. Third, the treatment test should only be given to a precise pretest probability. In our example, it should be done only to a patient with an intermediate pretest probability: some untrained colleague performed glyceryl trinitrate test “to be sure”, even though the probability of an ACS is very low. Sometimes, in front of false positive results, they are confused in the consequent management of the patient. Furthermore, in the particular case of high probability of ACS, we know that persistent chest pain does not decrease the probability of this diagnosis but involves probably a coronary desobstruction. In conclusion, I agree with the authors to be careful in interpretating the treatment test, such as I thought for all diagnostic testing. They suggested to multiply treatment or period, using blinding, which is possible in a scientific evaluation but difficult in routine practice. However, use treatment test correctly , with subjective and specific tests where it is only useful (appropriate prettest probability). Anyway, interpretation should be made with caution in the same way as investigating an association between an exposure (environmental, occupational, allergen or drug exposure) and a disorder/reaction. Competing interests: None declared |
|||
|
|
|||
|
Paul Glasziou, Professor OX3 7LF
Send response to journal:
|
We are pleased that Dr Descatha wants to reinforce our warning that "False positives can arise because of spontaneous remission of the condition or from placebo effects." We had then given 4 ways to improve "tests of treatment" but had not explicity mentioned the pre-test probability. In using discussing the diagnostic test sensitivity and specifity, we felt this was implicit, but we agree it may helpful to make the pre-test probability explicit. Perhaps before the test of treatment we might even record this in the notes. Competing interests: Author of article |
|||
|
|
|||
|
Malvinder S. Parmar, Associate Professor, Northern Ontario School of Medicine P4N 8P2
Send response to journal:
|
Test of Treatment or Therapeutic Challenge although imprecise is often used and is helpful in clinical practice. However, in their example of common tests of treatment, the authors gave an example of response to nitroglycerin as a test of Angina with 69-71% specificity. However, at least 2 studies [1-2] that evaluated the discriminatory value of nitroglycerin in patients presenting with chest pain, failed to show its usefulness in differentiating cardiac from non-cardiac chest pain, and response to nitroglycerin should not be used to guide the diagnosis. I think it is important to correct this myth that a chest pain relieved by nitroglycerin is angina and it incorrectly reinforces the patient and the physician. This unnecessarily results in further testing in most patients with atypical symptoms. I often tell patients that nitroglycerin is helpful in relieving symptoms by relaxing smooth muscles and discomfort causes by irritation or spasm of any smooth muscle in the body - whether in the heart (angina), oesopahgus (oesophageal spasm) or uterus (uterine contractions) would be relieved with nitroglycerin. In addition, there is always a possibility of placebo effect. It is important to advise patients to use nitroglycerin to relieve chest pain symptoms, keeping in mind that it doesn’t differentiate between the two common causes of chest pain. References: 1. Henrikson CA, Howell EE, Bush DE, Miles JS, Meininger GR, Friedlander T, Bushnell AC, Chandra-Strobos N. Chest pain relief by nitroglycerin does not predict active coronary artery disease. Ann Intern Med. 2003 Dec 16;139(12):979-86. PMID: 14678917 2. Steele R, McNaughton T, McConahy M, Lam J. Chest pain in emergency department patients: if the pain is relieved by nitroglycerin, is it more likely to be cardiac chest pain? CJEM. 2006 May;8(3):164-9. PMID: 17320010 Competing interests: None declared |
|||
|
|
|||
|
Paul P Glasziou, Professor of Evidence-Based Medicine OX3 7LF
Send response to journal:
|
We agree that the accuracy of using nitroglycerin is poor, and should not be relied on for discriminating angina from other causes of chest pain. As we showed in the Table the sensitivity and specificity are only a little better than a coin toss. The figures in the Table were derived from the 2004 systematic review by Chun & McGee [1] which had included two papers, both in outpatient settings. Chun had found pain relief within 5 minutes in 29% of non-coronary and 63% of coronary patients[2]. Wu found relief within 5 minutes for for 31% of non-coronary and 53% of coronary pain[3]. Both these papers looked at patients with recurrent chest pain in the outpatient setting, and show only a weak diagnostic accuracy of GTN. The 2004 systematic review had clearly missed the 2003 Annals paper by Henrikson[4], but that was in an emergency room setting where the accuracy appears to be even worse. So we agree that GTN should not be used as a test of treatment, particularly in the emergency room. References 1. Chun AA, McGee SR. Bedside diagnosis of coronary artery disease: a systematic review. Am J Med 2004;117:334-43. 2. Cooke RA, Smeeton N, Chambers JB. Comparative study of chest pain characteristics in patientswith normal and abnormal coronary angiograms. Heart. 1997;78:142–146. 3. Wu EB, Smeeton N, Chambers JB. A chest pain score for stratifying the risk of coronary artery disease in patients having day case cor- onary angiography. Int J Cardiol. 2001;78:257–264. 4. Henrikson CA, Howell EE, Bush DE, Miles JS, Meininger GR, Friedlander T, Bushnell AC, Chandra-Strobos N. Chest pain relief by nitroglycerin does not predict active coronary artery disease. Ann Intern Med. 2003 Dec 16;139(12):979-86. Competing interests: Author of the article |
|||