Rapid Responses to:
|
|
Rapid Responses: Rapid Responses published:
-
![[Read Rapid Response]](/icons/misc/sectionDOWN.gif)
Summarizing the evidence on corticosteroids for the acute respiratory distress syndrome (ARDS) – some methodological annotations
- Matthias Briel, Francois Lamontagne (Intensive Care Fellow) (4 May 2008)
-
Authors' Reply
- John L Moran, John V Peter , Preeta John, Petra L Graham, Ige A George, and Andrew D Bersten (13 May 2008)
-
Glucocorticoid treatment in ARDS: an inexpensive and cost-effective treatment with a favorable risk/benefit profile.
- Gianfranco U Meduri, Paul E. Marik, Djillali Annane (18 May 2008)
-
Authors' Reply
- John L Moran, John V Peter , Preeta John, Petra L Graham, Ige A George, and Andrew D Bersten (26 May 2008)
-
Steroids in ARDS: Interpret with caution.
- Richard Gibbs (2 August 2008)
|
|
|||
|
Matthias Briel, Research Associate Dep. of Clinical Epidemiology & Biostatistics, McMaster University, Hamilton, ON, Canada L8N3Z5, Francois Lamontagne (Intensive Care Fellow)
Send response to journal:
|
We read the systematic review by Peter et al.[1] with great interest and welcome their Bayesian approach to the topic. We have 4 comments: First, the literature search of Peter et al. seem to have missed at least one eligible randomized trial: Laggner et al. 1987 (16 patients with ARDS).[2] Including this trial, however, will not change the overall results. Second, Peter et al. do not report the distribution(s) they put on their non-informative priors. An allegedly non-informative inverse-gamma prior may turn out to be more informative than originally intended.[3] Third, the authors abstained from assessing publication bias due to the small number of studies per subgroup and give Sterne et al.[4] as a reference. We agree with Sterne et al. who in fact argue that “tests for small study effects should routinely be performed in meta-analysis”, although the results of these tests should be interpreted cautiously depending on the number of included studies. Fourth, we think the authors’ claim that “meta-regression…,not subgroup-analysis, is optimal…in the search for predictors of heterogeneity between studies” might be overstated and their reference (Greenland 1994)[5] does not support the claim. Like any regression analysis, meta-regression analysis may be problematic if there are few observations (here: studies) and several predictors, leading to overfitted models.[6] In his 1994 article, Greenland in fact does not argue for meta- regression as opposed to subgroup analysis; he “condemns quality scores” by saying that “use of such scores can seriously obscure heterogeneity sources and should be replaced by stratification or regression analyses of the relation of study results to the items or components of the score.”[5] We would argue that an individual patient data meta-analysis would be ideal to identify predictors of variability between studies, and we agree with the authors that such an approach would be most useful to investigate e.g. the optimal time to start steroids. REFERENCES 1. Peter JV, John P, Graham PL, et al. Corticosteroids in the prevention and treatment of acute respiratory distress syndrome (ARDS) in adults: meta-analysis. BMJ 2008. 2. Laggner AN, Lenz K, Base W, et al. [Effect of high-dose prednisolone on lung fluid in patients with non-cardiogenic lung edema]. Wien Klin Wochenschr 1987;99:245-249. 3. Gelman A. Prior distributions for variance parameters in hierarchical models. Bayesian Analysis 2006;1:515-533. 4. Sterne JA, Gavaghan D, Egger M. Publication and related bias in meta-analysis: power of statistical tests and prevalence in the literature. J Clin Epidemiol 2000;53:1119-1129. 5. Greenland S. Invited commentary: a critical look at some popular meta-analytic methods. Am J Epidemiol 1994;140:290-296. 6. Babyak MA. What you see may not be what you get: a brief, nontechnical introduction to overfitting in regression-type models. Psychosom Med 2004;66:411-421. Competing interests: None declared |
|||
|
|
|||
|
John L Moran, Senior Consultant Department of Intensive Care Medicine, The Queen Elizabeth Hospital, Woodville, SA 5011, Australia, John V Peter , Preeta John, Petra L Graham, Ige A George, and Andrew D Bersten
Send response to journal:
|
We thank Briel and Lamontagne for their comments upon our paper (1) and their endorsement of our overall approach. We offer these comments in reply to their “methodological annotations”: Point 1. We indeed would appear to have omitted the Laggner et al study, and thank Briel and Lamontagne for bringing it to our attention. Although the import of the Abstract was patho-physiological, mortality data was provided in their Tabelle 1 (page 247). We also agree that inclusion of this mortality data would not have changed the overall conclusions. Point 2. That the inverse-gamma prior could be informative is possible. However, following Warn et al (our reference 14) we “..replace(d) Smith’s inverse-gamma IG(3,1) prior for tau^2 with a Unif(0,2) for tau, which again covers the range of plausible values...” (2). That is, an inverse- gamma prior was not used in our study. Point 3. The quotation from Sterne et al (3) reads “Tests for small study effects should routinely be performed in meta-analysis. Their power is however limited, particularly for moderate amounts of bias or meta- analyses based on a small number of small studies”. A perusal of the power of the formal tests (note that in our study (1) we said “Publication bias was not formally assessed...”) with study n=5 in the Sterne et al (3) paper (Tables 3-5, pages 1123-1124) suggests a power of < 20% under most scenarios. Sterne et al also suggest that “Statistical tests...are required ...to quantify the evidence for asymmetry”. The utility of such testing with study n=4 and 5 (as was the case with our meta-analysis) remains to be demonstrated. Point 4. We have no disagreement with the recommendation of “stratification or regression analyses” in the quotation from Greenland (our reference 42). The operative word is “or” and we find that in the majority of meta-analyses (especially in the critical care field) a metaregression approach is eschewed. Furthermore, stratification (with respect to a continuous predictor variable) invariably reduces to dichotomisation (presumably a function of small sample size), with the attendant problems to which we referred in our paper. That is, as adroitly put by Panageas et al (our reference 32), “When [where] you look matters” (4). We certainly do not advocate a multivariable approach to metaregression with small study number (the interpretation of such a regression would be problematic and indeed subject to over-fitting); our paper used a pre-specified number of univariate meta-regressions. In this context, the inferential basis of such an overall (regression) relationship of predictor to treatment effect, thus preserving meta- analytic study size (5), seems more assured. We note a recent comment (in an update of the Stata™ meta-analysis module “metan”) by well known contributors to the meta-analytic paradigm: "A major addition to metan is the ability to perform stratified or subgroup analyses. These may be used to investigate the possibility that treatment effects vary between subgroups; however, formal comparisons between subgroups are best performed by using meta-regression...” (6). Reference List (1) Peter JV, John P, Graham PL, Moran JL, George IA, Bersten A. Corticosteroids in the prevention and treatment of acute respiratory distress syndrome (ARDS) in adults: meta-analysis. BMJ 2008; 336(7651):1006-1009. (2) Warn DE, Thompson SG, Spiegelhalter DJ. Bayesian random effects meta-analysis of trials with binary outcomes: methods for the absolute risk difference and relative risk scales. Stat Med 2002; 21(11):1601-1623. (3) Sterne JA, Gavaghan D, Egger M. Publication and related bias in meta-analysis: power of statistical tests and prevalence in the literature. J Clin Epidemiol 2000; 53(11):1119-1129. (4) Panageas KS, Ben-Porat L, Dickler MN, Chapman PB, Schrag D. When you look matters: the effect of assessment schedule on progression-free survival. J Natl Cancer Inst 2007; 99(6):428-432. (5) Fedorov V, Mannino F, Zhang R. Consequences of dichotomization. Pharmaceutical Statistics 2008; DOI: 10.1002/pst.331. (6) Harris RJ, Bradburn MJ, Deeks JJ, Harbord RM, Altman DG, Sterne JA. metan: fixed and random-effects meta-analysis. The Stata Journal 2008; 8(1):3-28. Competing interests: None declared |
|||
|
|
|||
|
Gianfranco U Meduri, Professor of Medicine University of Tennessee HSC -956 Court Avenue Room H316, Paul E. Marik, Djillali Annane
Send response to journal:
|
Glucocorticoid treatment in ARDS: an inexpensive and cost-effective treatment with a favorable risk/benefit profile. G. Umberto Meduri,1 MD; Paul E. Marik,2 MD; Djillali Annane,3 MD, PhD. 1 University of Tennessee HSC and VA Medical Center, Memphis, TN 2 Thomas Jefferson University, Philadelphia, PA 3 Universite de Versailles Saint-Quentin en Yvelines, Garches, France Dysregulated systemic inflammation is the central pathogenetic process leading to morbidity and mortality in patients with ARDS with glucocorticoids being the most important physiologic modulators of inflammation. The likelihood of a favorable clinical response to glucocorticoids is increased when treatment positively modifies the central pathogenetic process with an acceptable risk margin and is applied to achieve biological resolution of the disease process.1 Recent data suggests that the benefit/risk of glucocorticoids is largely determined by the drug dosage, timing and duration of administration, weaning protocol, and implementation of secondary preventive measures.1 The meta-analysis by Peter and colleagues2 combines randomized trials over a span of 22 years; during this period significant progress was made in our pathogenetic understanding of ARDS and mechanisms of action of glucocorticoids. The experimental model used to justify massive time- limited doses of glucocorticoids in the 1980s relied on the intravenous administration of a lethal bolus of endotoxin that could be offset only by administering equally massive doses of glucocorticoids given before or within an experimentally generated inflammatory window.1 This experimental model did not replicate human sepsis or acute lung injury and was discredited in the early 90s.1 Longitudinal measurement of inflammatory cytokines in ARDS patients have shown that significant systemic and pulmonary inflammation is still present 28 days into the disease process, and that prolonged treatment (28 days) is associated with significant reduction in inflammation and organ dysfunction.3, 4 Glucocorticoid treatment causes homologous downregulation of the glucocorticoid receptor levels in most cell types and suppression of the HPA axis. Slow reduction in glucocorticoid dosage not only forestalls relapse of systemic inflammation but also allows a gradual increase in intracellular GR number and recovery of the HPA axis when it has been suppressed.5 Ample experimental and clinical evidence was previously reviewed6 showing that premature removal of glucocorticoid treatment leads to rebound inflammation and an exaggerated cytokine response to endotoxin. Therefore, while massive time-limited doses of glucocorticoids may not impact on disease pathogenesis, they expose patients to serious morbidity-mortality. We believe that combing randomized trials with marked differences in dosage and duration of treatment, that reflect two separate understanding of disease pathogenesis and glucocorticoids’ mechanisms of action is misleading and historically anachronistic. When we limited our analysis to the recent trials that used prolonged treatment we found a reduction in duration of mechanical ventilation and mortality superior to any investigated intervention in ARDS. In the three trials investigating methylprednisolone of greater than 1 week’s duration (N = 295), mechanical ventilation-free days to day 28 increased by 5.6 days (95% CI 3.49 - 7.68; P < 0.001; I2 8%]; and among patients randomized before day 14 of ARDS (N = 400) we reported a distinct reduction in the RR of death (82/214 (38%) vs. 98/186 (52.5%), RR = 0.78; 95% CI (0.64 – 0.96); P = 0.02, I2 = 0%).6 Until the results of new trials are available, prolonged glucocorticoid treatment (initiated before day 14 of ARDS) in conjunction with secondary prevention measures6 offer an inexpensive and cost-effective treatment with a favorable risk/benefit profile. References 1. Meduri GU. An historical review of glucocorticoid treatment in Sepsis. Disease pathophysiology and the design of treatment investigation. Sepsis 1999; 3:21-38. 2. Peter JV, John P, Graham PL, Moran JL, George IA, Bersten A. Corticosteroids in the prevention and treatment of acute respiratory distress syndrome (ARDS) in adults: meta-analysis. BMJ 2008; 336: 1006-1009. 3. Meduri GU. The host defense response in the progression and outcome of ARDS. Pathophysiological correlations and response to glucocorticoid treatment. Eur Respir J. 1996; 9: 2650-2670. 4. Meduri GU, Yates CR. Systemic inflammation-associated glucocorticoid resistance and outcome in ARDS. Ann. N.Y. Acad. Sci. 2004; 1024: 24-53. 5. Krasner AS. Glucocorticoid-induced adrenal insufficiency. JAMA 1999; 282: 671-6. 6. Meduri GU, Marik PE, Chrousos GP, et al. Steroid treatment in ARDS: a critical appraisal of the ARDS network trial and the recent literature. Intensive Care Med 2008; 34: 61-9. Competing interests: None declared |
|||
|
|
|||
|
John L Moran, Senior Consultant Department of Intensive Care Medicine, The Queen Elizabeth Hospital, Woodville, SA 5011, Australia, John V Peter , Preeta John, Petra L Graham, Ige A George, and Andrew D Bersten
Send response to journal:
|
We thank Meduri, Marik and Annane for their discussion ("Glucocorticoid treatment in ARDS: an inexpensive and cost-effective treatment with a favorable risk/benefit profile", 18th May 2008) of our paper (1) and offer these comments by way of response: 1. The efficacy of glucocorticoids in ARDS is to be adduced at both the patho-physiological and clinical trial level. Steroids have a biological treatment rationale, for which we were at pains to argue, and we would agree that there is “little evidence supporting high dose steroids in ARDS” (1). However, the assessment of clinical efficacy is located at the trial level (witness the career of mono-clonal antibodies in sepsis) and the proper resolution of meta-analytic controversy is at the level of the conduct of analysis. Thus, 2. the estimators of treatment efficacy are presumed to be appropriate for the data-at-hand and we have argued that a Bayesian approach is apposite in the current context 3. these Bayesian estimates both accounted for between-trial- heterogeneity and were subject to potential revision by meta-regression. Within the limits of a non individual-patient-data (IPD) meta-analysis, we were unable to define an effect on outcome of time or dose of steroid therapy and, of interest, year of study completion. We were able to elucidate a trend towards an increased number of patients developing new infections as steroid dose increased, consistent with patho-physiological principles of steroid action 4. We have argued in some detail (1) the problematic status of (a) the estimates of mechanical ventilation-free days and the relative risk of death which Meduri, Marik and Annane quote (2) and (b) the interpretation of treatment response rates on the basis of data dependent time cut-off points, and we will not reproduce these arguments here What we do note is the elision between "In the three trials investigating methylprednisolone of greater than 1 week's duration (N = 295), mechanical ventilation-free days to day 28 increased by 5.6 days (95% CI 3.49 - 7.68; P < 0.001; I2 8%)" and "among patients randomized before day 14 of ARDS (N = 400) we reported a distinct reduction in the RR of death (82/214 (38%) vs. 98/186 (52.5%), RR = 0.78; 95% CI (0.64 - 0.96); P = 0.02, I2 = 0%)". The three trials generating the latter estimates are from Figure 3, middle panel “02 Larger studies” in Meduri et al (2): Annane 2006 (3), Steinberg 2006 (4) and Meduri 2007 (5). As documented in Peter et al (1), the timing, duration and dose of steroid (methyl-prednisolone equivalent) varied markedly between these three trials, obfuscating a consistent therapeutic approach: 0.6 mg/kg/day for 7 days with no tapering (Annane 2006), 1 mg/kg/day for 14 days and tapered (Meduri 2007) and 2 mg/kg/day for 14 days (Steinberg 2006) and tapered. Furthermore, in Annane 2006 (3) hydrocortisone was used, which presupposes that all glucocorticoids are equivalent (6) 5. The concept (“Glucocorticoid treatment in ARDS”) never carries the case and we find that the current data do not confirm the proposition. Thus “We cannot … dismiss a null effect” (1) 6. We agree that further trials need be performed and advocate an IPD meta-analysis to clarify current uncertainties Reference List (1) Peter JV, John P, Graham PL, Moran JL, George IA, Bersten A. Corticosteroids in the prevention and treatment of acute respiratory distress syndrome (ARDS) in adults: meta-analysis. BMJ 2008; 336(7651):1006-1009. (2) Meduri G, Marik P, Chrousos G, Pastores S, Arlt W, Beishuizen A et al. Steroid treatment in ARDS: a critical appraisal of the ARDS network trial and the recent literature. Intensive Care Med 2008; 34(1):61-69. (3) Annane D, Sebille V, Bellissant E, Ger I. Effect of low doses of corticosteroids in septic shock patients with or without early acute respiratory distress syndrome. Crit Care Med 2006; 34(1):22-30. (4) The NHLBI Acute Respiratory Distress Syndrome (ARDS) Clinical Trials Network. Efficacy and Safety of Corticosteroids for Persistent Acute Respiratory Distress Syndrome. N Engl J Med 2006; 354(16):1671-1684. (5) Meduri GU, Golden E, Freire AX, Taylor E, Zaman M, Carson SJ et al. Methylprednisolone Infusion in Early Severe ARDS: Results of a Randomized Controlled Trial. Chest 2007; 131(4):954-963. (6) Rhen T, Cidlowski JA. Antiinflammatory Action of Glucocorticoids -- New Mechanisms for Old Drugs. N Engl J Med 2005; 353(16):1711-1723. Competing interests: None declared |
|||
|
|
|||
|
Richard Gibbs, Specialist Registrar Intensive Care Medicine Derriford Hospital, Crownhill, Plymouth, PL6 8DH, UK
Send response to journal:
|
Peter et al (1) are to be congratulated for attempting to shed further light in what has become a contentious topic in the critical care community. As acknowledged by both the authors and in the accompanying editorial (2) the considerable heterogeneity in both trial design and dosing regimens mean this Bayesian analysis should be interpreted with caution. Firstly all the trials assessing the prophylactic use of corticocorticoid therapy for ARDS have used high dose steroids. These trials provide a high level of evidence that high doses of corticosteroids should not be used to prevent ARDS in patients at risk of the syndrome, however low and moderate dose corticosteroids have not been rigorously studied in this role. With low dose corticosteroids now a widely accepted treatment modality in septic shock (and given that sepsis is the commonest cause of ARDS), one would hope that a low-dose regimen of corticosteroids does not increase the incidence of ARDS in this at risk population! A considerable amount of the positive mortality data for corticosteroids in established ARDS comes from two studies by Meduri et al (3, 4) and a post hoc sub-group analysis by Annane et al (5). Conclusions based on data from these three trials must intepreted in context. Aside from the methodological issues addressed by Peter et al which included early cessation, the two trials conducted by Meduri et al incorporated a crossover design whereby non- responders in the treatment groups were crossed over to later corticosteroid therapy. This compromises the control group and thus any conclusions about long term outcomes. The only prospective trial of high methodological quality where the control groups were not crossed over to steroid therapy (allowing mortality data intepretation) was the ARDS network trial (6). Although probably underpowered, this study showed no clear mortality benefit of corticosteroid therapy despite improvements in secondary outcomes, and suggested harm in a predefined subgroup of patients who received GCT after 14 days of ARDS. Due to this lack of clear evidence recent ACCCM guidelines give a weak recommendation for corticosteroid use in patients with early severe ARDS or with ARDS of less than 14 days duration (7). However, as the authors state, further well-conducted trials are required to delineate which ARDS populations will benefit from corticosteroid therapy and what are the optimal strategies for dosing, duration and timing of therapy. 1. Peter JV, John P, Graham PL, Moran JL, George IA, Bersten A. Corticosteroids in the prevention and treatment of acute respiratory distress syndrome (ARDS) in adults: meta-analysis. BMJ. 2008;336:1006-1009. 2. Adhikari NK, Scales DC. Corticosteroids for acute respiratory distress syndrome. BMJ. 2008;336:969-970. 3. Meduri GU, Golden E, Freire AX et al. Methylprednisolone infusion in early severe ARDS: results of a randomized controlled trial. Chest. 2007;131:954-963. 4. Meduri GU, Headley AS, Golden E et al. Effect of prolonged methylprednisolone therapy in unresolving acute respiratory distress syndrome: a randomized controlled trial. JAMA : the journal of the American Medical Association. 1998;280:159-165. 5. Annane D, Sébille V, Bellissant E, Group G-I-S. Effect of low doses of corticosteroids in septic shock patients with or without early acute respiratory distress syndrome. Critical care medicine. 2006;34:22-30. 6. Steinberg KP, Hudson LD, Goodman RB et al. Efficacy and safety of corticosteroids for persistent acute respiratory distress syndrome. The New England journal of medicine. 2006;354:1671-1684. 7. Marik PE, Pastores SM, Annane D et al. Recommendations for the diagnosis and management of corticosteroid insufficiency in critically ill adult patients: Consensus statements from an international task force by the American College of Critical Care Medicine. Crit Care Med. 2008 Competing interests: None declared |
|||