Systemic corticosteroids in fibrotic lung disease: a systematic review and meta-analysis

Search strategy

We developed a comprehensive search strategy with the help of an experienced medical librarian. We searched Embase, Medline, Cochrane CENTRAL and Web of Science for eligible trials and cohorts from inception to 17 April 2023. We also reviewed previous SR addressing the topic to ensure no studies were missed.1 7 We did not use any language restrictions and included only primary source clinical trial data. We reviewed secondary analyses and post-hoc analyses for subgroup data, as required. Online supplemental eTable 1 presents the search strategy.

Eligibility criteria

We included randomised controlled trials (RCTs) that randomised adult patients (≥18 years old) with a diagnosis of fILD of any aetiology, except sarcoidosis, using any previous diagnostic criteria, to treatment with corticosteroids versus standard care or placebo or no treatment. We also included both retrospective, prospective and mixed cohort studies. We extracted in-study subgroup data if studies reported on both non-fibrotic and fILD patients. If there was no in-study subgroup data, we determined whether to include studies if at least 80% or more of the patients included had fILD. We excluded studies investigating the effectiveness of corticosteroids on acute exacerbations as well as trials predominately investigating non-fILD.

Our two outcomes of interest included change in force vital capacity (FVC) (%-predicted or mL) and all-cause mortality. For both outcomes, we collected data at the longest follow-up or closest to 52 weeks.

Study selection and data extraction

We used COVIDENCE to screen eligible trials.10 Pairs of reviewers, following training and calibration exercises to ensure sufficient agreement, worked independently and in duplicate to screen titles and abstracts of search records and subsequently the full texts of records that were determined potentially eligible at the title and abstract screening stage. Reviewers resolved discrepancies by discussion or, when necessary, by third party adjudication. Similarly, the reviewers worked independently and in duplicate to extract data from eligible trials, and resolved discrepancies by discussion or, when necessary, by third party adjudication.

We collected data on trial characteristics (author, year published, trial registration, country of enrolment), patient characteristics (age, sex, ethnicity, comorbidities, C reactive protein, white cell count, proportion of patients on home oxygen, with previous exacerbations and aetiology of their fibrotic disease), intervention characteristics (type of corticosteroid, dose, duration and baseline treatments) and outcomes of interest.

For dichotomous outcomes, we extracted the number of participants analysed and number of events in each arm. For cohort studies, we collected OR, HRs or relative risks (RR) with event rate. For continuous outcomes, we collected data on mean difference (MD) and SD. When studies report other measures of variability other than SD, we converted them to SD using methods proposed by Hozo et al.11

Risks of bias

We planned to assess the risk of bias for individual RCTs using the Cochrane tool (RoB 2.0).12 13 For cohort studies, we used ROBINS-I to assess for risk of bias.14 When rating the mortality outcome, for a study to be rated at low risk of bias for confounding, they would need to at least adjust for age, sex, smoking status, cointerventions with antifibrotic and other immunomodulators, pretreatment FVC (either in mL or as %-predicted), and duration of disease. For change in FVC (%), a study would have corrected for age, sex, baseline lung function and cointerventions.

Statistical methods

For all outcomes, we performed a random effects MA with the restricted maximum likelihood heterogeneity estimator. We summarised the effects of interventions using RR for dichotomous outcomes and MD for continuous outcomes, both with associated 95% CIs. For FVC, as some studies reported either mL or % predicted, we used standardised MDs as a sensitivity analysis in addition to analysing the predominant measure (% predicted). To facilitate interpretation of dichotomous outcomes, we calculated absolute risk differences per 1000 patients and corresponding 95% CIs. For studies that only reported HRs, we converted them to RR using available total number and event rates. In the absence of these data, we used methods by Short et al to convert HR to RR using established baseline risks.15

We planned to also perform a dose–response analysis for mortality using methods proposed by Orsini and Longnecker; however, there were insufficient data to perform these analyses (see protocol).

We assessed heterogeneity by inspection of forest plots, the I² statistic and the χ² test. We considered heterogeneity ranging from 0% to 40% as potentially unimportant, 30% to 60% as moderate heterogeneity, 50% to 90% as substantial heterogeneity and 75% to 100% as critical heterogeneity.16 For outcomes with 10 or more studies, we assessed for publication bias or small study effects using both visual inspection of funnel plots and the Egger’s test.17

We performed all analyses using STATA V.18.

A priori subgroup analysis

We planned subgroup analysis for the following moderators: IPF versus non-IPF fILD, high-dose versus low-dose steroid, where high was defined as methylprednisolone-equivalent of >1 mg/kg, and steroid molecule (ie, prednisone vs dexamethasone). We hypothesised that there would be no difference in these subgroups. For randomised trials, we assessed the credibility of statistically significant subgroups using the Instrument for assessing the Credibility of Effect Modification Analyses (ICEMAN) tool.18 For cohort studies, we applied similar principles in lieu of a validated tool.

Certainty of the evidence

For all outcomes, reviewers, working independently and in duplicate, assessed the certainty of the evidence using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach.19 20 We judged the certainty for each outcome as high, moderate, low or very low, based on considerations of risk of bias, inconsistency, indirectness, imprecision and publication bias.

To make judgements regarding imprecision, we used a minimally contextualised approach, which considers only whether CIs include a minimally important effect and does not consider the magnitude of plausible effects, captured by confidence intervals.21 For mortality, we used a minimal clinical important difference (MCID) based on consensus of the authors and considered any difference important. For change in FVC, we used the MCID provided by the updated ATS/ERS/JRS/ALAT (American Thoracic Society/European Respiratory Society/Japanese Respiratory Society/Asociación Latinoamericana de Tórax (Latin American Thoracic Association)) Clinical Practice Guideline of 5%.22

We described our results using guidance from the GRADE Working Group, based on the certainty of evidence and the magnitude of the effect (eg, corticosteroids reduce mortality (high certainty), corticosteroids probably reduce mortality (moderate certainty), corticosteroids may reduce mortality (low certainty) and the effect of corticosteroids on mortality is very uncertain (very low certainty)).20

Study characteristics

Table 1 presents the study characteristics of included cohort studies.5 6 24–30 The age of patients in the included studies ranged from 56.2 to 71.7 years old, and most patients were male (61%). Six studies included patients with IPF and five included patients with non-IPF fILD. There were two cohorts of fHP, two of systemic sclerosis-associated ILD and one of idiopathic non-specific interstitial pneumonia. Two cohorts were from the USA, four from Japan and one each from Belgium, China, the UK and South Korea. All studies were published in English except one in Mandarin.26

Seven studies provided intervention details on dosing and type of corticosteroids. All of them reported using prednisolone or prednisone. The dosing varied between 22.5 mg/day to 53.25 mg/day of prednisolone equivalent on average, with varying tapering schedules.

Risk of bias

All studies were identified as being at risk of bias across multiple domains. Three cohorts were at critical risk of bias due to confounding and all were at risk of bias due to selection of reported results. Online supplemental eTable 2 presents more detail on the risk of bias assessments.

Mortality

Ten cohort studies (1265 patients) reported mortality as an outcome, with MA showing that corticosteroids have an uncertain effect on mortality as compared with no treatment (RR 1.03 (95% CI 0.85 to 1.25); very low certainty evidence). We rated the certainty down due to risk of bias (twice) and imprecision (once). We did not find evidence of publication bias on inspection of the funnel plot or Egger’s test (p=0.26) (online supplemental eFigure 1). Figure 2 presents the forest plot and table 2 presents the summary of findings.

Request permission

Change in FVC

Four cohort studies (439 patients) reported change in FVC as an outcome, with MA showing that corticosteroids have an uncertain effect on the rate of decline in FVC (% predicted) as compared with no treatment (MD 4.29% (95% CI −8.26% to 16.83%); very low certainty evidence). We rated down twice due to risk of bias and once due to imprecision. Table 2 presents the summary of findings and figure 3 present the forest plot.

Request permission

Subgroups and sensitivity analyses

There was no statistically significant subgroup difference in mortality when comparing patients with a diagnosis of IPF vs non-IPF fILD (p=0.36) (online supplemental eFigure 2). There was also no association of dose with mortality in meta-regression (p=0.66) (online supplemental eFigure 3). There was no variability in findings based on the type of corticosteroid used. There were insufficient data to compare high-dose versus low-dose corticosteroid regimens.

Subgroup analysis showed that corticosteroids may reduce the rate of decline of FVC as compared with no treatment in patients with non-IPF fILD (MD 10.89% (95% CI 5.25% to 16.53%); low certainty evidence), but not in IPF (MD −3.80% (95% CI −8.94% to 1.34%); very low certainty evidence). There was a statistically significant subgroup difference between the two groups (p<0.001) as described in the next section.

One study reported change in mL instead of percent predicted for change in FVC. We included this study and pooled the results using standardised MD to facilitate sensitivity analysis (online supplemental eFigures 4 and 5). Using baseline FVC (%) as a moderator, there as no significant difference in mortality (online supplemental eFigure 6). There was no difference in effect and the subgroup difference remained the same with and without inclusion of this study. We took this into account when not rating down for imprecision for this outcome.

Main findings

Our review updates and expands on previous literature investigating the evidence for corticosteroids in IPF with the important addition of non-IPF fILD. We found that corticosteroids may have a beneficial effect on lung function in non-IPF fILD, although the evidence is of low certainty. This contrasts with IPF, where we found no evidence of a beneficial effect of corticosteroids on lung function and a mean effect that was in the direction of harm. The effect on mortality remains very uncertain in either group, with substantial heterogeneity across studies. The clinical significance of corticosteroids reducing FVC decline in patients with non-IPF fILD is unclear and further highlights the importance of high-quality studies to be done in the future. However, it may indicate an important clinical difference between these types of fILD and drive future treatments and research endeavours.

We identified no randomised trials that compared corticosteroid monotherapy to placebo or standard care despite FDA approval and continued prescriptions for corticosteroids in patients with fILD.31 We did, however, identify multiple trials that compared corticosteroid therapy with other immunomodulatory treatments,2 32–36 with results that ranged from possible benefit to significant harm.2 34

The lack of RCTs comparing corticosteroid monotherapy to placebo or standard care is a significant gap in the literature. This need has been unmet for decades, as the first use of corticosteroids in fILD was in the 1950s.4 There are opportunities for future studies to also investigate specific endotypes to help us identify patients with specific immune responses to epithelial cell injury that need corticosteroids and develop a precision medicine approach to these patients.37 The optimal dosing and type of corticosteroid remain unanswered questions. Future research should aim to address these gaps in our knowledge to provide more definitive guidance for the use of corticosteroids in fibrotic lung disease. While corticosteroids continue to be prescribed for patients with fibrotic lung disease, the evidence supporting their use is limited and of low certainty. Further high-quality research is needed to better understand the role of corticosteroids in the treatment of fibrotic lung disease.

A previous Cochrane review was also unable to identify any RCTs or other available evidence for the use of corticosteroids in IPF.7 Network MAs, which allow for indirect treatment comparisons, have previously reported on the efficacy of corticosteroids as compared with placebo.1 One study estimated, with very low certainty evidence, no difference in mortality but a possible 14% decrease in FVC as compared with placebo. This review only included patients with IPF and not other types of fILD.

Strengths and limitations

Our review is the most updated and rigorous review of the evidence related to the use of corticosteroids in fILD. We performed a systematic search of available evidence and provided quantitative synthesis of the available data using rigorous methodology. Furthermore, we assessed the certainty of the evidence using the most updated GRADE guidance and used principles from the ICEMAN tool to assess the credibility of subgroup effects.

There are considerable limitations to the available evidence incorporated in our review. These include the lack of RCTs, which limit interpretations and conclusions about the efficacy of corticosteroids on important patient outcomes.

All cohort studies were at risk of residual confounding and there is also potential for selective reporting, which may have significant bias. Our assessment of the risk of bias included cohort studies demonstrates this limitation clearly. We accounted for this limitation when assessing the certainty of the evidence. Another limitation is the heterogeneity in outcomes across included cohorts. For example, although most studies in the IPF subgroup reported no effect in reducing all-cause mortality, the effect estimated ranged from significant benefit (RR 0.80 (95% CI 0.68 to 0.93)) to significant harm (RR 1.39 (1.13 to 1.53)). Furthermore, our subgroup analysis comparing IPF and non-IPF fILD is limited based on the few studies available for analysis. This limits major conclusions from the data.

The diagnostic criteria, changing standards of care and residual confounding obfuscate the true effect of corticosteroids in this population.

Furthermore, we were limited in the number of cohorts studied, as we only included studies that compared corticosteroid monotherapy with no treatment, potentially causing publication bias as many cohorts did not report on outcomes of interest or were not clear enough to decide about the population or outcome.

Impact and future directions

Our review highlights a significant gap in evidence synthesis for patients with fILD. Oral corticosteroids are often prescribed to patients with non-IPF fILD and, unfortunately, there are insufficient data to provide informed information about the benefits and harms of this medication. Assessing the effectiveness of corticosteroids in fILD is an important unmet clinical need. Robust evidence is needed to fill this gap. Additional study(ies) is/are needed to compare the relative efficacy of other immunosuppressants with and without corticosteroid in non-IPF fILD. RCTs in fILD are challenging on several fronts but are needed in this research space. The current REMAP-ILD platform trial aims to assess multiple treatment interventions simultaneously in this population group and may be able to close the evidence gap. Our review provides evidence for the rationale to consider the inclusion of corticosteroids within the platform.