Corticosteroid Therapy for Patients Hospitalized With Community-Acquired Pneumonia: A Systematic Review and Meta-analysis

We read the recent review and meta-analysis by Siemieniuk & colleagues’ (1) published in your journal. The paper states, in light of the evidence with moderate to high certainty, the systemic corticosteroids were helpful in adults with community acquired pneumonia. The effect estimates obtained were significant and precise. We differ in opinion at choosing the methods to draw 95% CI and the variation occurring in interpretation of results thereby in the analysis for estimating the effect of systemic corticosteroids on need of mechanical ventilation in patients hospitalised with CAP. The authors have used Hartung – Knapp – Sidik-Jonkman modification (2) which should be preferred when there were more than five studies included in an analysis. But the analyses, shown in figures 2 and 3 of this meta-analysis (1) had five or less studies. The stated method has a limitations when there were 5 or fewer studies & of varying sample sizes (2). A sensitivity analysis using the DL method was actually required and interpretation of the effect estimates should be done by comparing these two methods.
Using the conventional method (3) & Revman 5.3.5 the confidence intervals become comparatively wider and even cross the line of no effect when demonstrating the effect of systemic corticosteroids on need of mechanical ventilation in patients hospitalized due to severe CAP. Relative risks mentioned in the meta-analysis were RR= 0.54 (95% CI 0.50-0.58) but using the conventional method the estimates obtained were RR=0.54 (95% CI 0.28-1.04). Albeit, the point estimates are same but 95% CIs widen, even for the overall estimate [(RR=0.45 (95% CI 0.26-0.79)] of this analysis which still remain significant [RR=0.45 (95%CI 0.25-0.82)]. Had they have mentioned and done the sensitivity analysis for estimating CIs, better interpretation of results was possible. Similarly the estimates obtained for the analysis done for effects of corticosteroids on development of ARDS in patients hospitalised with CAP were RR= 0.24 (95%CI 0.10-0.56) but the estimates obtained by conventional method were RR = 0.24 (95% CI 0.08 -0.77).

References
1. Siemieniuk RA, Meade MO, Alonso-Coello P, Briel M, Evaniew N, Prasad M, Alexander PE, Fei Y, Vandvik PO, Loeb M, Guyatt GH. Corticosteroid Therapy for Patients Hospitalized With Community-Acquired Pneumonia: A Systematic Review and Meta-analysis. Ann Intern Med. 2015 Aug 11
2. IntHout J, Ioannidis JP, Borm GF. The Hartung–Knapp–Sidik–Jonkman method for random effects meta-analysis is straightforward and considerably outperforms the standard DerSimonian–Laird method. BMC Med Res Methodol. 2014;14:25.
3. DerSimonian R, Laird N. Meta-analysis in clinical trials. Controlled Clinical Trials 1986; 7: 177-188.

We congratulate Siemieniuk and co-workers[1] for demonstrating that adjunctive corticosteroids is safe and beneficial in hospitalized patients with CAP. The major limit of these studies remains the lack of indications on when, how, and why using corticosteroids. Each included RCT reported a different way to administer corticosteroids in terms of route, dose, duration, and choice of the agent. Moreover, the criteria of severity for CAP were not uniform. So, it is not surprising if the Authors failed to indicate if corticosteroids work for any hospitalized CAP, and which treatment regimen is most effective. Corticosteroids by-passed the usual steps to register new indication, no phase II or dose finding trial exist. These old and inexpensive drugs are often used on an individual physician choice basis in daily clinical practice.

On the other hand, CAP is a common infection with a wide spectrum of clinical severity ranging from a self-limiting illness to life-threatening ones. Inflammation in mild CAP may be a healing response without need to be modulated, but when it becomes a dysregulated systemic cascade the use of corticosteroids is justified. So, just after the conclusion of our RCT published in 2005[2] with an impressive efficacy of prolonged low dose steroids for severe CAP, we wondered how to manage CAP in our clinical practice. Pragmatically, from then  to date, we are using steroids for established or ongoing severe CAP, according to the ATS criteria with special focus on PaO2:FiO2≤250 and C-reactive protein (CRP) >80mg/dL.   Keeping in mind the pivotal role of inflammation and the results of Annane[3] and Meduri[4], respectively on sepsis and ARDS, we use pulsed methylprednisolone 20mg i.v. every 8 hours for at least 7 days and then tapering doses according to the CRP levels. To date we treated hundreds of patients.

Recently, we published an observational study[5] performed just after the opening of a new respiratory unit recruiting 1,087 patients with CAP. At that time, only our unit used corticosteroids for CAP in our hospital in contrast with Emergency and Internal Medicine Units. The adjusted odds ratio for hospital death reduction in patients treated with corticosteroids was 5.78 compared to controls. Probably, the identification of molecular predictors for severe sepsis and ARDS will help physicians to identify patients needing early corticosteroid therapy, but just now we are enough confident that our approach is beneficial without be harmful for patients with moderate to severe CAP and sustained systemic inflammation.

 

 

References

1. Siemieniuk RA, Meade MO, Alonso-Coello P, et al. Corticosteroid therapy for patients hospitalized with community-acquired pneumonia: a systematic review and meta-analysis. 2015 Aug 11, doi: 10.7326/M15-0715. [Epub ahead of print]
2. Confalonieri M, Urbino R, Potena A, et al. Hydrocortisone infusion for severe community-acquired pneumonia: a preliminary randomized study. Am J Respir Crit Care Med 2005; 171: 242-248.
3. Annane D, Bellissant E, Bollaert PE, et al. Corticosteroids for severe sepsis and septic shock: a systematic review and meta-analysis. BMJ. 2004;329(7464):480.
4. Meduri GU, Tolley EA, Chrousos GP, et al. Prolonged methylprednisolone treatment suppresses systemic inflammation in patients with unresolving acute respiratory distress syndrome: evidence for inadequate endogenous glucocorticoid secretion and inflammation-induced immune cell resistance to glucocorticoids. Am J Respir Crit Care Med 2002; 165: 983-991.
5. Confalonieri M, Trevisan R, Demsar M, et al. Opening of a respiratory intermediate care unit in a general hospital: impact on mortality and other outcomes. Respiration 2015 Jul 8. [Epub ahead of print]. PMID: 26160422

We were pleased to read a recently published systematic review and meta-analysis by Siemieniuk et al. concerning corticosteroid therapy for patients hospitalized with community-acquired pneumonia (CAP) [1], because the results from their study match those from our analysis published in September 2015 [2].
Both studies similarly suggested that adjunctive corticosteroids do not decrease death in overall population, potentially prevent 60% of deaths in severe CAP subgroups, and shorten length of hospital stay and length to clinical stability by approximately one day.
Strengths of their study included risk evaluations for acute respiratory distress syndrome and for intubation. In addition, they performed rapid analysis and successfully picked up a few studies that were missed in the other systematic reviews.
An essential issue remaining to be clarified is how to select patients with “severe CAP” who will benefit from adjunctive corticosteroids [3]. Traditionally, severity of CAP had been systemically and multi-dimensionally evaluated. Most trials with favorable adjunctive steroid results selected severe CAP with the following criteria: intensive care unit case, American Thoracic Society severe criteria, British Thoracic Society severe criteria, or Pneumonia Severity Index class of V [1,2]. In these trials, adding corticosteroids decreased death approximately 60% [1,2]. We believe that steroid therapy will benefit even “less severe” CAP cases. In the systematic review, Siemieniuk et al. proposed to define severe CAP concerning steroid indication using the Pneumonia Severity Index score of IV or higher, and CURB-65 (Confusion, Urea nitrogen, Respiratory rate, Blood pressure, and age 65 years or older) score of two or higher [1]. Although we do not know how they decided this criteria, we also think these cutoffs are reasonable for assessing indication of adjunctive corticosteroids for CAP.
Another strategy for selecting a steroid-indicative CAP case is using an inflammatory biomarker such as C-reactive protein (CRP) [3,4]. In a trial by Torre et al., CRP > 15 mg/dL was one of key criteria in identifying severe CAP [4]. In Torre’s study, corticosteroids dramatically prevented treatment failure with the small number needed to treatment of four. Probably, we can select a broader population with lower CRP cutoff values, for example CRP > 10 mg/dL, to treat CAP patients with steroids.
Even conducted independently, the results from their and our reviews indicated that adjunctive corticosteroids treatment is a reasonable strategy to treat severe CAP. We hope adjunctive corticosteroids will support millions of patients suffering from severe CAP.
 
References:
1. Siemieniuk RA, Meade MO, Alonso-Coello P, Briel M, Evaniew N, Prasad M, et al. Corticosteroid Therapy for Patients Hospitalized With Community-Acquired Pneumonia: A Systematic Review and Meta-analysis. Ann Intern Med. 2015 Oct 6;163(7):519-28.
2. Horita N, Otsuka T, Haranaga S, Namkoong H, Miki M, Miyashita N, et al. Adjunctive Systemic Corticosteroids for Hospitalized Community-Acquired Pneumonia: Systematic Review and Meta-Analysis 2015 Update. Sci Rep. 2015 Sep 16;5:14061.
3. Restrepo MI, Anzueto A, Torres A. Corticosteroids for Severe Community-Acquired Pneumonia: Time to Change Clinical Practice. Ann Intern Med. 2015 Oct 6;163(7):560-1.
4. Torres A, Sibila O, Ferrer M, Polverino E, Menendez R, Mensa J, et al. Effect of corticosteroids on treatment failure among hospitalized patients with severe community-acquired pneumonia and high inflammatory response: a randomized clinical trial. JAMA. 2015 Feb 17;313(7):677-86.

The widely used DerSimonian and Laird (DL) method for summarizing random effects [1] has been criticized for sometimes having inappropriately narrow confidence intervals and high type I error rates [2,3]. The Hartung-Knapp-Sidik-Jonkman (HKSJ) method, which is based on a t-distribution, is a popular alternative because it is straightforward and does not require familiarity with advanced statistical programs [3].

We agree with Singh and Jaiswal that the HKSJ 95% confidence intervals are implausibly narrow in the mechanical ventilation subgroup analysis, especially in the subgroup of studies enrolling patients with more severe pneumonia [4]. This led to a statistically significant interaction between pneumonia severity and the corticosteroid effect (P=0.01), which would not otherwise have occurred using the DL method (P=0.18). Nevertheless, as we discuss in the manuscript and regardless of this interaction, we believe that the apparent subgroup effect is spurious and that corticosteroids are likely beneficial across the range of pneumonia severities. Our findings and our confidence in them, summarized in the GRADE summary of findings table, are the same for all outcomes using either the DL and HKSJ approaches.

We have noted the implausibly narrow confidence intervals using the HKSJ approach and conducted some preliminary analyses to determine when this phenomenon occurs. Our analyses suggest that implausibly narrow confidence intervals with the HKSJ approach occurs when effect sizes are highly consistent between the primary studies. While one could argue that this is more likely to occur by chance with a small number of studies (as it did in this case), we observed the same phenomenon irrespective of the number of studies.

Profile likelihood, Bayesian, and the new modified Knapp-Hartung approaches have also been suggested as alternatives to the DL approach for random effects [2,5]. However, all of these alternatives can generate implausibly wide confidence intervals, especially when there are a small number of studies [2,5].

Emerging evidence therefore indicates that there is no single random effect approach that is ideal for all situations, and that blanket rejection of the DL approach is unwarranted. Investigators should be aware of the limitations of each approach, and in any particular situation choose an approach that yields the most sensible confidence interval.

 
References:

1. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials. 1986; 7:177-88.
2. Cornell JE, Mulrow CD, Localio R, Stack CB, Meibohm AR, Guallar E, et al. Random-effects meta-analysis of inconsistent effects: a time for change. Ann Intern Med. 2014; 160:267-70.
3. IntHout J, Ioannidis JP, Borm GF. The Hartung-Knapp-Sidik-Jonkman method for random effects meta-analysis is straightforward and considerably outperforms the standard DerSimonian-Laird method. BMC Med Res Methodol. 2014; 14:25.
4. Siemieniuk RA, Meade MO, Alonso-Coello P, Briel M, Evaniew N, Prasad M, et al. Corticosteroid Therapy for Patients Hospitalized With Community-Acquired Pneumonia: A Systematic Review and Meta-analysis. Ann Intern Med. 2015; 163:519-28
5. Röver C, Knapp G, Friede T. Hartung-Knapp-Sidik-Jonkman approach and its modification for random-effects meta-analysis with few studies. BMC Med Res Methodol. 2015; 15:99.

Reed A.C. Siemieniuk, MD; Pablo Alonso-Coello, MD, PhD; Gordon H. Guyatt, MD, MSc

Drs. Siemieniuk and Guyatt: Department of Clinical Epidemiology & Biostatistics, McMaster University, 1280 Main Street West, Hamilton, Ontario L8S 4K1, Canada.

Dr. Alonso-Coello: Centro Cochrane Iberoamericano, Instituto de Investigación Biomédica Sant Pau-CIBER de Epidemiología y Salud Pública (CIBERESP-IIB-Sant Pau), Sant Antoni Maria Claret 171, 08041 Barcelona, Spain.