Review Papers Against Their Own Results

 Quick Explanation Copied

 Long Explanation

What the paper reports

Petrelli et al. pooled 76 studies (83,614 patients with solid tumours; 2057 references screened) comparing steroid users versus non-users. Steroid use associated with reduced overall survival (HR 1.18, 95% CI 1.10–1.26; p<0.01; 69 studies; I²=67%) and reduced progression-free survival (HR 1.13, 95% CI 1.01–1.26; p=0.03; 40 studies; I²=67%). Harms were concentrated in advanced disease, supportive-care indications, and lung cancer, where death risk reportedly rose ~47%; in a few randomized settings (prostate/early breast cancer), steroids were neutral Meta-analysis core results and subgroups.

Critical appraisal

Confounding by indication is the dominant threat. Steroids were mostly given for supportive/palliative reasons in advanced disease; sicker patients receive steroids, so the observed association plausibly reflects disease severity, not a drug effect. The authors acknowledge selection bias and that randomized studies showed neutral effects Author-acknowledged limitations. Heterogeneity was high (I²=67% for both endpoints), median NOS quality was only 6/9, most analyses were retrospective, and median follow-up spanned 9–252 months — an unusually wide range suggesting highly heterogeneous cohorts. Dose, duration, and steroid type were largely unreported, making dose-response testing impossible.

Internal inconsistency: the Methods state Egger's test was used, while Results report "no statistical heterogeneity" for the funnel-plot analysis — contradictory given I²=67% in the main models; also, the abstract states P<.01 for OS while heterogeneity methodology description is compressed, and the PRISMA-referenced search text oddly describes "immune checkpoint inhibitor" populations inherited from the authors' prior ICI meta-analysis Search-strategy text inconsistency.

Mechanistic plausibility exists but is not proven clinically: the paper cites glucocorticoid induction of SGK1, immunosuppression via CD28 pathway attenuation, and preclinical metastasis promotion Glucocorticoid metastasis preclinical evidence Dexamethasone immunosuppression mechanisms. Conversely, a small retrospective series (n=19) found no clear detriment from escalated immunosuppression after irAEs, and low-dose dexamethasone remains standard in metastatic spinal cord compression — contexts the pooled design cannot resolve irAE immunosuppression case series MSCC low-dose steroid review.

Verdict and what would change it

The association is likely real as a prognostic marker but the causal interpretation is weak. A large prospective randomized trial of steroid-sparing versus standard supportive care in advanced NSCLC, with dose/duration stratification and individual-patient-data meta-analysis, would be the decisive test; if such trials showed no survival difference after rigorous stage/performance-status adjustment, the paper's central conclusion would be falsified. The recommendation to "reduce and avoid steroids" in palliative settings should be treated as hypothesis-generating, not practice-changing.

Author reviews:

 BGPT Paper Review

Study Novelty

Aggregates known observational signals (e.g., steroids with checkpoint inhibitors) into one large pooled estimate; the quantitative cross-tumour synthesis is somewhat new but the hypothesis and subfindings were pre-existing.

Scientific Quality

Large pooled dataset and pre-specified subgroups, but high heterogeneity (I2=67%), predominantly retrospective/univariate studies, curve-extrapolated HRs, a methods section containing leftover text from the authors' prior ICI meta-analysis, and contradictory heterogeneity/publication-bias reporting reduce quality.

Study Generality

Covers multiple solid tumour types, stages, and steroid indications, yielding broadly applicable hypotheses, though subgroup results (NSCLC, palliative setting) are tumour- and context-specific.

Study Usefulness

Useful for flagging a potentially modifiable practice pattern and prioritizing prospective steroid-sparing trials; limited immediate clinical utility given confounding by indication.

Study Reproducibility

Search strings, RevMan 5.3, Tierney methods, ROBINS-I, and NOS are reported and supplementary study tables exist, but full study-level HR data and exact subgroup numbers are partially in supplements, limiting independent replication.

Explanatory Depth

Mechanistic discussion (SGK1, T-cell suppression, adrenal androgen suppression) is superficial and borrowed from prior reviews; no mediation or dose-response analysis to distinguish causation from confounding.