Included studies

Show code 
source("analysis/meta_helpers.R")
ma <- load_ma()

Analytic dataset

All six included studies are observational (retrospective cohorts, or a small case series for Lu2017) of HIV-negative / immunocompromised adults. Each contributes one two-arm comparison: adjunctive echinocandin/caspofungin + TMP-SMX versus TMP-SMX monotherapy, with all-cause mortality at the timepoint shown.

Show code 
ma |>
  transmute(Study = study, Tier = tier, Timing = timing,
            `Mortality timepoint` = timepoint,
            `Combination (deaths/N)` = paste0(e_comb, "/", n_comb),
            `Monotherapy (deaths/N)` = paste0(e_ref, "/", n_ref),
            OR = round(OR, 2),
            `95% CI` = sprintf("%.2f\u2013%.2f", exp(yi - 1.96 * sei), exp(yi + 1.96 * sei))) |>
  knitr::kable()
Study Tier Timing Mortality timepoint Combination (deaths/N) Monotherapy (deaths/N) OR 95% CI
Qi2025 primary mixed in-hospital 18/35 31/79 1.64 0.74–3.66
Xu2025 primary mixed 28-day 33/89 19/31 0.37 0.16–0.86
Lu2017 primary mixed episode/in-hosp 1/5 2/6 0.50 0.03–7.99
Jin2019 sensitivity initial 3-month 8/35 33/91 0.52 0.21–1.28
Qi2023 sensitivity initial 90-day 17/43 10/21 0.72 0.25–2.06
Li2024a sensitivity initial episode (unspec) 5/20 3/18 1.67 0.34–8.26

The primary tier comprises studies reporting in-hospital or ~30-day mortality; the sensitivity tier adds studies reporting longer-timepoint mortality. Treatment timing (initial vs mixed) is classified following Yang et al. (2026); no salvage-vs-monotherapy study is available.

Verification notes

Every count above was checked against the source full text. Key decisions:

Study Note
Qi2025 Full-cohort Table 4 (18/35 vs 31/79). The same PUMCH cohort as Yanmeng2026 (duplicate removed). Prior meta-analyses that report 18/27 used the ventilated subgroup.
Xu2025 Multicentre cohort; the monotherapy arm was sicker (more solid tumour, more aspergillosis), so the crude comparison is subject to confounding by indication. Most influential study (see sensitivity).
Lu2017 Very small case series (n = 5 vs 6), heart-transplant recipients; mortality timepoint not explicitly defined.
Jin2019 3-month mortality; full-cohort arms (not the β-D-glucan subgroups).
Qi2023 90-day mortality; the initial caspofungin arm used (salvage arm excluded from this contrast).
Li2024a Confounded: the combination arm bundles caspofungin with a corticosteroid and a lower TMP dose, so it cannot isolate an echinocandin effect. Retained in the sensitivity tier only, flagged.

Studies not analysed

  • Tian2020 — excluded: an HIV-positive cohort (published in HIV Medicine), outside the population.
  • Wang2019 pilot, Zhang2018 — no TMP-SMX monotherapy comparator arm.
  • Taniguchi (administrative dataset) — mortality events were back-calculated; excluded from the primary synthesis.
  • Seven studies in Yang et al. that we could not retrieve — mostly Chinese-language papers / theses (CNKI/Wanfang); one (Wang-ZG-2019) was obtained and found ineligible (single-arm). See conclusions.
Translate: