Materials and methods

This systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [23] and the Meta-Analysis of Observational Studies in Epidemiology (MOOSE) statements [24]. The study protocol was prospectively registered on the Open Science Framework (OSF; 10.17605/OSF.IO/VDUSW).

Research question and eligibility criteria

This review aimed to determine whether micro-CT is applicable for the detection, characterization, and potential estimation of shooting distance based on GSR in biological and non-biological forensic samples.

Eligible studies were those employing micro-CT to investigate GSR in the aforementioned specimens, without restrictions regarding language or publication date. Reviews, editorials, expert opinions, conference abstracts, case reports, and very small case series (< 5 cases) were excluded. When multiple publications reported potentially overlapping experimental datasets, the study with the largest sample size or the most comprehensive data was retained.

Studies with limited sample sizes (5 ≤ n < 10 cases) were included exclusively in the qualitative synthesis to ensure maximal coverage of the available evidence, while acknowledging the inherent limitations associated with reduced statistical precision.

Search strategy

A systematic literature search was conducted on PubMed, Web of Science, Scopus, Embase, Google Scholar, and Open Access Theses and Dissertations (OATD). For Google Scholar and OATD, the first 100 retrieved records were screened. Additionally, reference lists of all eligible studies were manually examined to identify further relevant publications. The searches were conducted on October 15, 2025, and updated on January 7, 2026.

The search strategy combined controlled vocabulary and free-text terms as follows: (“micro-CT” OR “microCT” OR “micro computed tomography” OR “micro-computed tomography” OR “X-ray microtomography” OR “microtomography” OR “computed micro-tomography” OR “computed microtomography”) AND (“gunshot residue” OR “GSR” OR “shot residue” OR “shotgun residue”).

Study selection

Study selection was independently performed by two reviewers (C.A.V-J. and R.B.V.) using a two-stage screening process. Duplicate records were identified and removed using the Rayyan platform [25]. Titles and abstracts were subsequently screened, followed by full-text assessment of potentially eligible studies. Discrepancies were resolved by consensus or, when necessary, by consultation with a third reviewer (L.M.).

Data extraction

Data extraction was conducted independently by two reviewers (C.A.V-J. and R.B.V.) using a standardized form. Extracted information included: authorship, year of publication, country, sample size, sex and age range (when applicable), firearm type, caliber, projectile characteristics, firing setup, gunshots per distance, firing distances, and outcomes. Detailed information on micro-CT acquisition and processing parameters was also collected, including scanner model, voltage, current, filter, voxel size, field of view, rotation parameters, scan duration, volume of interest, reconstruction and analysis software, Hounsfield unit calibration, and GSR segmentation thresholds. All outcomes related to micro-CT performance were recorded.

Risk of bias assessment

The methodological quality and risk of bias of included studies were assessed independently by two. reviewers (C.A.V-J. and R.B.V.) using the Joanna Briggs Institute (JBI) Critical Appraisal Checklist for Analytical Cross-Sectional Studies (https://jbi.global/critical-appraisal-tools). Although most included investigations were experimental or bench-based ballistic studies, this tool was selected to provide a structured and transparent appraisal of key methodological domains.

The assessment addressed eight criteria: clarity of inclusion criteria; adequacy of sample and setting description; validity and reliability of exposure measurement; objectivity of outcome assessment; identification of confounding factors; strategies to address confounding; validity and reliability of outcome measurement; and appropriateness of statistical analysis. Each item was rated as yes, no, unclear, or not applicable. Disagreements were resolved by consensus or third-party adjudication (P.R.M-F.).

It was necessary to adapt the original checklist for experimental ballistic studies. Accordingly, interpretative modifications were applied as follows. Sample inclusion was assessed in terms of whether biological and non-biological target materials were clearly described, including their provenance and suitability for the experimental model. Study participants and setting were defined by the firearm used, the type and lot of ammunition, and the experimental setup. Exposure measurement was evaluated as the placement of samples in the same environment as the firearm discharge. The measurement of the condition was assessed in terms of whether objective and standardized criteria were applied for the detection of GSR by micro-CT. Confounding factors and strategies to address them included potential pre-existing contamination, since micro-CT detects only Hounsfield units and cannot identify specific particles, the time elapsed before sample collection, shooting conditions (distance, caliber, ammunition type), and image acquisition noise. Results assessment referred to the presence and morphology of GSR detected by micro-CT.

Data analysis

Data related to GSR detection by micro-CT were organized into a spreadsheet dataset and analyzed using R software (version 4.5.1; R Foundation for Statistical Computing, Vienna, Austria). The primary outcome was the mean percentage of detectable GSR, derived from quantitative micro-CT measurements. Variance estimates were derived based on the reported dispersion measures and sample sizes, allowing appropriate weighting in subsequent analyses.

Random-effects meta-analyses

Four random-effects meta-analyses were conducted: one overall analysis pooling data across the available shooting distances (5, 15, 23, 30, and 40 cm) and three distance-specific analyses (5, 15, and 30 cm). Stratified analyses were restricted to distances represented in multiple studies, excluding the 23 and 40 cm conditions. Sensitivity analyses were also conducted by sequentially excluding the 23 and 40 cm distance conditions, individually and simultaneously, to assess their impact on the pooled estimates and heterogeneity measures. Between-study heterogeneity was modeled using a random-effects framework with restricted maximum likelihood (REML) estimation. Pooled estimates of the mean percentage of GSR detected were accompanied by τ², I², and Cochran’s Q statistics [26, 27]. Study weights were calculated as the inverse of the adjusted variance and normalized to sum to 100%. Statistical significance was assessed at a 5% level.

Quadratic mixed-effects meta-regression

A quadratic mixed-effects meta-regression was fitted to examine the continuous relationship between shooting distance and the mean percentage of GSR detected. All available firing distances were included in this analysis, including distances represented by a single study, since the objective was to model the continuous distance-response relationship instead of deriving distance-specific pooled estimates. Analyses were conducted on the original percentage scale, which was retained throughout model fitting and prediction. The model included a random intercept for study identifier to account for within-study dependence, with remaining heterogeneity modeled at the study level. Fixed effects consisted of centered shooting distance and its quadratic term. Parameter estimation was performed using REML [26].

Uncertainty in both the fitted values and the estimated vertex of the quadratic function was quantified using 2,000 parametric bootstrap replications. Bootstrap samples were generated by drawing from a multivariate normal distribution parameterized by the REML fixed-effects estimates and their variance-covariance matrix [28]. For each replication, predicted values were recomputed and the vertex (turning point) of the curve was analytically derived from the quadratic parameters. Empirical bootstrap distributions were then used to obtain 95% confidence intervals. Results were visualized as scatter plots overlaid with the fitted quadratic curve and corresponding bootstrap confidence bands, with point sizes proportional to observational precision (inverse square root of the variance) and colors indicating sample-stage categories.

Gaussian fits with bootstrap confidence intervals

As an exploratory complementary analysis, weighted nonlinear Gaussian models were fitted to characterize the decay pattern of GSR detection as a function of shooting distance, stratified by sample-stage category. Model fitting was performed using the Levenberg–Marquardt algorithm implemented in the nlsLM function [29, 30], with inverse-variance weighting. Parameters representing the peak detection amplitude and decay rate were estimated, and uncertainty was assessed through 1000 nonparametric bootstrap resamplings [28], yielding 95% confidence intervals. Descriptive R² values were used to summarize the agreement between observed data and the fitted Gaussian profiles. Values close to 1 indicate a close correspondence between model and data, whereas negative values indicate that the fitted curve does not reduce residual variance relative to a constant (mean-only) model.