Silicon Nitride Coatings on Dental and Orthopedic Implants: A Systematic Review of Preclinical Studies

Abstract

Background: Silicon nitride is an established bulk ceramic in spinal fusion, but a coating can delaminate, dissolve or wear through, and preclinical evidence on coatings has not been synthesized. Methods: Databases and trial registries were searched in October 2026. Eligible studies were in vitro, bench or animal studies of silicon nitride coatings on implant substrates. Risk of bias was assessed with tools for animal and laboratory studies; the synthesis was narrative. Results: Of 476 records screened, 44 publications from 19 research groups were included (18 films, seven doped or graded films, 19 composites). Counted descriptively, 21 of 27 cell studies and 12 of 13 bacterial studies favored a coating, but came from only nine and five groups and did not measure effect size. In one bench series of sputtered coatings on cobalt–chromium, wear in serum was similar to bulk silicon nitride for all but one coating; adhesion and dissolution remained concerns. Six animal studies, from two groups, reported more bone with a coating; the one comparison with hydroxyapatite favored hydroxyapatite. All had moderate risk of bias by a descriptive rule. Conclusions: The evidence is preclinical, short-term and clustered in few groups, and does not establish the long-term stability of coated implants.

1. Introduction

Silicon nitride (Si₃N₄) is an established bulk ceramic in spinal fusion cages, and its clinical record is mixed: retrospective series report outcomes similar to other cage materials [1], a long-term case series followed the first lumbar spacers for decades [2], a manufacturer-funded randomized trial reached non-inferiority to polyetheretherketone (PEEK) only after a post hoc change in margin [3], and a recent network meta-analysis suggests higher fusion odds than PEEK [4]. Silicon nitride is also applied as a thin coating on titanium, cobalt–chromium–molybdenum, zirconia and polymer implants, with the aim of improving bone integration, limiting bacterial colonization and reducing wear debris. A coating can delaminate, dissolve or wear through, which a monolithic ceramic cannot, so evidence for the bulk material does not show that a coating is durable [5]. Laboratory studies of bulk silicon nitride reported lower bacterial adhesion than PEEK or titanium [6] and, in rat calvarial defects, more new bone and less infection [7], which is part of the rationale for using it as a coating. The investigators’ own report of the same randomized trial found insufficient evidence of non-inferiority [8]. These clinical data on the bulk material are an indirect context and do not apply directly to coatings.

Silicon nitride is a non-oxide structural ceramic with high strength, fracture toughness and wear resistance [9,10] that has been studied as an orthopedic biomaterial since the 1980s [11]. Early studies evaluated its cytotoxicity and biocompatibility in vitro [12,13,14] and its tissue response in rabbits [11,15]. Later work suggested that silicon nitride is not bioinert and that its surface chemistry and nitrogen content influence osteogenic behavior in vitro [16,17]; in vivo, bone and implant responses were examined in rat calvarial defects [6] and in a murine tibial model in which surface topography influenced both infection and osseointegration [18]. Compared with PEEK and titanium, silicon nitride showed lower bacterial adhesion or activity in vitro [6,19] and caused lysis of Porphyromonas gingivalis [20], and spectroscopic studies relate this to surface chemistry and to the response of Staphylococcus epidermidis [21,22]. A recent report of silicon nitride containing TiC and oxide sintering aids also found fewer Staphylococcus aureus and Escherichia coli colonies than on alumina, titanium alloy and PEEK, but that material is a composite and not a coating [23]. Much of this work comes from groups linked to the manufacturer, and results depend on assay conditions. As for stability, silicon nitride powder dissolves in phosphate-buffered saline [24]; the biological effect of silicon nitride particles depends on particle size, surface composition and donor [25]; and sintered silicon nitride containing strontium, magnesium and silicon has been designed to be biodegradable [26]. Whether these properties are retained by a thin coating, which can delaminate, dissolve or wear through, is the question addressed in this review.

Preclinical studies of silicon nitride coatings are scattered across materials science, tribology and biomedical journals and use different substrates, deposition methods, test models and outcomes [27,28]. No systematic synthesis exists. Systematic reviews of other implant coatings show why one is needed: calcium-phosphate coatings gave no significant gain in bone-to-implant contact over etched titanium in animals [29], strontium coatings gave a gain but with very high heterogeneity and signs of publication bias [30], and titanium nitride and titanium niobium nitride coated knee implants gave no clinical advantage over uncoated implants [31]. For PEEK, a common substrate for silicon nitride coatings, unmodified material is less bioactive than titanium and surface modification appears to help [32,33]. The aim of this review was to identify and appraise all preclinical (in vitro, bench and animal) studies of silicon-nitride-based coatings on implant substrates; to describe their composition, deposition methods, models and outcomes; to assess risk of bias; and to determine whether the evidence supports clinical translation. A companion search for human studies of coated implants found none and is reported through the PROSPERO registration (CRD420261528198). Narrative reviews of silicon nitride ceramics [34] and of surface modification and coating of dental implants [35,36,37,38] describe the field but do not synthesize the preclinical evidence on silicon nitride coatings.

2. Materials and Methods

2.1. Review Question and PICOS Framework

The review question was: in preclinical models, how do silicon-nitride-based coatings on dental or orthopedic implant substrates affect biological, antibacterial, mechanical, tribological and corrosion outcomes compared with uncoated substrates or alternative coatings? The framework of population, intervention, comparator, outcome and study design (PICOS) is given in Table 1. The review is reported according to the PRISMA 2020 statement [39] (checklist in Supplementary File S5).

Table 1. Eligibility criteria based on the PICOS framework.

PICOS	Inclusion Criteria
Population	Substrates relevant to dental or orthopedic implants (titanium and alloys, cobalt–chromium–molybdenum (CoCrMo), zirconia, stainless steel, polyetheretherketone, polyetherketoneketone (PEKK) and carbon-fiber-reinforced polymers)
Intervention	Coating of silicon nitride (SiNx, Si3N4, SiN:H), silicon carbon nitride (SiCN), or a composite coating in which silicon nitride is the active component, applied by any deposition method
Comparison	Uncoated substrate, alternative coating, or a different silicon nitride composition
Outcomes	Cell adhesion, proliferation or differentiation; bone formation or osseointegration; antibacterial activity; hardness, adhesion, wear or friction; corrosion, dissolution or ion and particle release; any clinical outcome if reported
Study design	In vitro, bench or animal studies with full text available in English; no date limit

2.2. Information Sources

Five databases were searched on 3 October 2026 without language or date limits: MEDLINE (Ovid MEDLINE(R) ALL 2026, reached through Journals@Ovid because the MEDLINE [Ovid] interface returned an access error; the search was later reproduced in PubMed, see Section 2.3), Embase (Embase.com, automatic mapping disabled), Scopus (title, abstract and keywords), the Web of Science Core Collection and Cochrane CENTRAL (Cochrane Library, trials only). Three further database searches were made later: a search of MEDLINE through ProQuest on 25 October 2026, the ProQuest strategy re-run in PubMed on 25 October 2026, and a broader search in PubMed on 27 October 2026 (strategies in Supplementary File S1). Other sources were the reference lists of retrieved full texts, targeted searches of PubMed and publisher sites, and trial registries. Trial registries were searched on 22 October 2026—ClinicalTrials.gov, ISRCTN, ChiCTR (China), the Netherlands Trial Register (OMON), the Korean Clinical Research Information Service (CRIS), the Brazilian Clinical Trials Registry (ReBEC) and the Iranian Registry of Clinical Trials (IRCT)—with the term “silicon nitride” and, in some registries, “Si₃N₄” and equivalents in Chinese, Portuguese and Persian (Supplementary File S1). The WHO International Clinical Trials Registry Platform (ICTRP) search portal was searched on 24 October 2026 with “silicon nitride”, “Si₃N₄”, “SiN_x”, “silicon carbon nitride” and “SiCN” (the portal had returned a server error on 22 October). Forward citation searching was not performed; citation searching was limited to scanning the reference lists of the included studies and of relevant reviews.

2.3. Search Strategy

Each search combined three concept blocks—silicon nitride terms (silicon nitride, Si₃N₄, SiN_x), coating or deposition terms (coating, coated, surface modification, sputtering, high-power impulse magnetron sputtering (HiPIMS), laser cladding, physical vapor deposition (PVD), chemical vapor deposition (CVD)) and implant terms (implant, dental implant, osseointegration, arthroplasty, prosthesis)—searched in title, abstract and keyword fields. The strings as run, the platforms, dates, fields and record counts are given in Supplementary File S1. For Cochrane CENTRAL, the three concept lines returned 13, 39,442 and 47,573 records, and their combination returned none. The original Ovid search history was not retained. A later attempt to re-run the search in Ovid exported records labeled Journals@Ovid Full Text, a different database (16 or 17 records, of which 11 were not in the log and none were eligible), and these records were not used. The MEDLINE search was therefore reproduced in PubMed on 27 October 2026 with an equivalent string (Supplementary File S1); it returned 29 records, identical to the 29 records of the Ovid MEDLINE search and all already among the records found, so it added no new records. A second, broader PubMed search was run on the same day by the second reviewer; it added wildcard forms, further coating and deposition terms (thin film, nanofilm, surface layer, electrophoretic deposition, laser sintering, melt bonding, micro-arc oxidation (MAO), plasma-enhanced CVD (PECVD)) and broader implant terms (Supplementary File S1). It returned 81 records, which include those 29; 65 were already among the 328 records of the first searches, and the 16 new records were screened by the reviewers and all excluded (Section 3.1). A search of MEDLINE through ProQuest on 25 October 2026 combined substance name, MeSH and free-text terms for silicon nitride, for coating and film terms, and for implant terms (Supplementary File S1); it retrieved 275 records, but only 48 could be exported, and their 39 new records were excluded at title and abstract. Because the other 227 records could not be exported, the same strategy was re-run in PubMed on 25 October 2026 (Supplementary File S1): it returned 193 records, of which 100 were already among the records found and 93 were new. All 93 were excluded at title and abstract (after the abstracts of three borderline records were read), and none concerned a silicon nitride coating on a dental or orthopedic implant. The 227 unexported ProQuest records were not screened, which is a limitation. The term implant retrieved many ion implantation papers from semiconductor and materials engineering, which were removed at screening. As a retrospective check of sensitivity, the database searches retrieved 41 of the 44 included studies; the other three (laser cladding on polyethylene [40], a study of HiPIMS process parameters [41] and a hot-pressed PEEK surface [42]) were found by reference list scanning and web searches. The strategy requires an implant-related term and may miss studies that describe a coating without one, so it should not be regarded as exhaustive. The broader strategy was run only in PubMed and was not repeated in Embase, Scopus, Web of Science or Cochrane CENTRAL because of time, which is a limitation.

2.4. Eligibility Criteria

Studies were eligible if they reported primary experimental data on a silicon-nitride-based coating on an implant-relevant substrate (Table 1). A silicon nitride coating was defined as a layer, film or surface treatment on an implant substrate in which silicon nitride is the principal active component. Studies were classified as (i) Si₃N₄ or SiN_x films; (ii) silicon carbon nitride (SiCN) and other doped or graded films, including multilayers; and (iii) particulate or composite coatings, including polymer composites and laser-sintered, laser-clad or micro-arc oxidation layers. Studies were excluded in five situations: silicon nitride was a bulk ceramic, a composite material or a substrate rather than a coating; the coating was a different material or the application was outside dental and orthopedic implants (for example neural probes, electronic encapsulation or biosensors); the coating was deposited only on silicon wafers or other non-implant substrates; the paper stated no implant or biomedical application; or the record was a review, book chapter, commentary or conference abstract without a full report. Results are not pooled across categories, and differences in composition, structure, thickness, deposition process and degradation behavior are stated where they affect interpretation.

2.5. Selection Process

Records were de-duplicated by DOI and title. A first pass over titles and abstracts was made with an artificial intelligence (AI) model (Claude Sonnet 5.5, Anthropic, San Francisco, CA, USA) against the eligibility criteria, and its decisions are documented record by record in the screening log. Two reviewers (M.M.A. and N.F.C.S.) then screened all 476 records independently in separate blinded spreadsheets, choosing include for full text, exclude with a reason, or unsure (treated as include). The 328 records of the first searches were screened from the title and abstract; the 148 records added by the later searches (ProQuest and PubMed) were screened after an AI pre-screen, which each reviewer reviewed and confirmed. Agreement over the 476 records was 98.7% (Cohen’s kappa 0.94). Fourteen records on which the reviewers disagreed, or on which both were unsure about a record that the first pass had excluded, were decided by a third reviewer (ALQAHTANI, M.A.) who applied the registered criteria. The arbiter’s entries were drafted with AI assistance and were reviewed and approved by the arbiter. Three of these records had no abstract in the database export, so their PubMed records were consulted. Ten were excluded and four were sent to full-text assessment, where three were included and one was excluded (Supplementary File S3). Two other disagreements were settled by discussion between the reviewers. Full texts of records carried forward were assessed against the criteria, and reasons for exclusion were recorded (Supplementary File S3). Eligibility at full text was first assessed in one AI-assisted pass, which the reviewers verified. Reports found by reference list scanning, targeted searches and papers supplied by the authors were assessed in the same way and are shown in the other-methods arm of the flow diagram. One eligible report (a 1997 report on stainless steel with ion-beam-assisted silicon nitride coatings; Supplementary File S1) could not be retrieved: it was not available for download, and a request to the corresponding author received no reply. One included study (Esmaeili et al., 2024) [43] was available only as a scanned copy that was read by optical character recognition, and its values were checked by the extraction reviewer.

2.6. Data Extraction

Data were extracted with a piloted form: study design, substrate, coating composition, deposition method, thickness and key deposition parameters, comparator, model and sample size, outcomes and time points, main findings, and funding or conflicts of interest. Each study was classified into one of the three categories, and the structure and other phases reported (amorphous or crystalline phase, silicon, silica, oxynitride and other phases), the organism and endpoint of any antibacterial assay, and the test conditions of any bench durability test were extracted. For the 44 included studies, a first draft of the extraction and of the risk-of-bias judgments was prepared by an AI model (Claude Sonnet 5.5); a further reviewer (Alkhelaiwi, M.A.) checked the extracted data against the source articles and rated risk of bias independently of the other reviewers.

2.7. Risk of Bias Assessment

Animal studies were assessed with the risk-of-bias tool of the Systematic Review Centre for Laboratory Animal Experimentation (SYRCLE) [44], with each of the ten domains rated low, high or unclear. A study was classed as high overall if any domain was high, moderate if no domain was high and at least two were unclear, and low otherwise. In vitro and bench studies were assessed with the risk-of-bias tool for dental materials studies (RoBDEMAT) [45], with nine items rated sufficiently reported, insufficiently reported or not reported across four domains (planning and allocation, specimen preparation, outcome assessment, and data treatment and outcome reporting); item-level judgments are given in Supplementary File S2. The overall category for each in vitro study was recorded as low, moderate or high using a rule fixed retrospectively before the final ratings: high if the control group or standardized testing was not reported; low if all items were sufficiently reported; otherwise moderate. Risk of bias was rated by one reviewer (Alkhelaiwi, M.A.) independently of the other reviewers; the registered protocol specified two raters, so this is a deviation and a limitation. The overall labels are a descriptive categorization of reporting and not a rating of the certainty of the evidence. They were derived from the domain judgments of each tool (SYRCLE) or from the rule above (RoBDEMAT); the two tools use different items and are not comparable, and a label of moderate risk does not mean that studies from different experimental domains are of comparable quality. Item-level judgments for every study are given in Supplementary Files S2 and S4.

2.8. Data Synthesis

A structured narrative synthesis was done, grouped by coating category, substrate, deposition method and outcome, because the heterogeneity of models and outcomes precluded pooling. The unit of synthesis is the publication that reports one or more experiments on a silicon nitride coating, and the 44 included publications are not assumed to be independent. Possible overlap was assessed from authorship (two or more authors in common), coating and model, and publications from the same research group were treated as one group in group-level counts (Supplementary File S4, Overlap groups). Counts of studies with favorable results are descriptive counts of publications, always given with the number of research groups; they ignore sample size, effect magnitude, precision, multiple outcomes and selective reporting, and they are not pooled or comparative effects. Results are kept separate by coating category (films; doped or graded films; particulate or composite coatings) because their exposed surface chemistry and likely failure modes differ. Random-effects meta-analysis was to be considered only if three or more clinically and methodologically similar studies reported the same outcome, which did not occur. No formal GRADE rating was assigned for laboratory evidence. Each synthesis was interpreted with reference to the number of independent studies, risk of bias, consistency, precision and directness for clinical performance.

2.9. Use of Generative AI

Generative AI (Claude Sonnet 5.5, Anthropic, San Francisco, CA, USA) was used to help plan the search strategies, to run supplementary web searches, to process the exported database records and screen them in a first pass, to draft the data extraction and the risk-of-bias judgments for the retrieved full texts, to format references, to plot Figure 1, Figure 2 and Figure 3 from the authors’ counts and classification, and to draft and edit the text. The database searches themselves were run by the authors and library staff. The authors checked the references in four ways: (1) Existence: Each reference was resolved to a record by DOI or PubMed identifier. (2) Bibliographic accuracy: Authors, title, journal, year, volume and pages were compared with the publisher or database record. (3) Content: Each statement attributed to a source was compared with its full text, or with its abstract where no full text was obtained. (4) Numerical accuracy: Values quoted in the text and tables were compared with the source. The AI pre-screen of the 148 records added by the later searches and the first draft of the arbiter’s decisions on 14 records were reviewed and confirmed by the reviewers and the arbiter. The first drafts of the full-text eligibility decisions, the data extraction and the risk-of-bias judgments were prepared with AI assistance and were verified by the reviewers named in Section 2.5, Section 2.6 and Section 2.7. The authors edited all AI-assisted text and take full responsibility for the content.

Figure 1. PRISMA 2020 flow diagram of study selection. * Raw database records from searches run between 3 and 27 October 2026; duplicates removed by DOI and title matching. † Two reviewers screened all records independently, and a third reviewer decided 14 records: non-biomedical, n = 213; different coating material or application, n = 109; review, overview, book chapter or commentary, n = 50; not a coating (bulk silicon nitride, composite or substrate), n = 51. ‡ One report could not be obtained: it was not available for download, and the corresponding author did not reply. § Eligibility at full text was first assessed in one AI-assisted pass and verified by the reviewers. Reports found by reference list scanning, web searches and papers supplied by the authors. Adapted from the PRISMA 2020 flow diagram, licensed under CC BY 4.0. PubMed (expanded search, 27 October 2026) added 81 records, of which 65 were duplicates; MEDLINE via ProQuest (25 October 2026) retrieved 275 records, of which 48 could be exported (9 duplicates); the ProQuest strategy re-run in PubMed (25 October 2026) added 193 records (100 duplicates). Registers: ClinicalTrials.gov, ISRCTN, ChiCTR, OMON, CRIS, ReBEC and IRCT were searched on 22 October 2026 (3 records, of which 1 was a duplicate and 2 concerned bulk silicon nitride spinal cage trials); the WHO ICTRP portal (24 October 2026) returned 3 records, none relevant.

Figure 2. Included studies by coating category and model (n = 44). Categories: (i) Si₃N₄ or SiN_x films; (ii) SiCN and other doped or graded films; (iii) particulate or composite coatings. This figure was plotted in Python (version 3.12.3) with Matplotlib (version 3.10.8) from the authors’ classification.

Figure 3. Summary of RoBDEMAT item judgments for the 43 laboratory studies. This figure was plotted in Python (version 3.12.3) with Matplotlib (version 3.10.8) from the item-level judgments in Supplementary File S2.

2.10. Registration and Protocol

The registered PROSPERO record (CRD420261528198; https://www.crd.york.ac.uk/PROSPERO/view/CRD420261528198 (accessed on 9 November 2026)) covers human studies of silicon-nitride-coated implants and found no eligible studies at title and abstract screening. The extension to preclinical studies was not registered in advance. Its protocol was registered retrospectively on the Open Science Framework on 15 October 2026 (https://doi.org/10.17605/OSF.IO/9TWQZ (accessed on 9 November 2026)), after the database searches (3 October 2026) and a first-pass AI-assisted screening had been done. This registration covers laboratory and animal studies, which PROSPERO does not accept, so the scope differs from that of the PROSPERO record. The registration was updated on 18 October 2026 to correct drafting errors (placeholder text, the Ovid query wording, and the software and model entries); the update did not change the review question, eligibility criteria, outcomes or planned methods. The differences between the registered protocol and the final methods are listed in the OSF amendment and in Supplementary File S4 (Protocol deviations). The registration was retrospective. Searches were added (an expanded PubMed search, ProQuest MEDLINE and WHO ICTRP), and the Ovid search history was not retained. Citation searching was backward only. First passes for screening, full-text decisions and extraction were AI-assisted and verified by the reviewers, and the arbiter’s entries were drafted with AI assistance. One rater assessed risk of bias, and the RoBDEMAT overall rule was fixed retrospectively. One report could not be retrieved. The 227 ProQuest records that could not be exported were not screened individually.

3. Results

3.1. Study Selection

The database searches identified 485 records (MEDLINE 29, Embase 27, Scopus 192, Web of Science 237, Cochrane CENTRAL 0); the broader PubMed search added 81 records (65 already found), the ProQuest MEDLINE search exported 48 records from 275 retrieved (9 already found), and the PubMed re-run of the ProQuest strategy exported 193 records (100 already found). In total, 807 records were identified and 331 duplicates were removed, leaving 476 records. Of the 476 records screened at the title and abstract levels, 423 were excluded: 213 as non-biomedical; 109 for a different coating material or an application outside dental and orthopedic implants; 50 as reviews, overviews, book chapters or commentaries; and 51 because silicon nitride was a bulk ceramic, a composite or a substrate and not a coating. Fourteen records on which the reviewers disagreed, or on which both were unsure about a record that the first pass had excluded, were decided by a third reviewer: ten were excluded and four were sent to full-text assessment. Fifty-three reports were carried forward for full-text retrieval. Full texts of 52 were assessed: 41 met the eligibility criteria and eleven were excluded, five because the coating was deposited on silicon wafers or substrates only, not on an implant-relevant substrate, four because the paper stated no implant or biomedical application, and two because the coating was a different compound (a silicon-containing carbon nitride and a polymer-derived silicon oxycarbonitride, SiCNO). One report (a 1997 report on stainless steel with ion-beam-assisted silicon nitride coatings; Supplementary File S1) could not be retrieved: it was not available for download, and the corresponding author did not reply to a request. Eight further reports from citation searching, web searches and papers supplied by the authors were all assessed: three were included, and five were excluded (bulk silicon nitride; a coating on silica glass only; a bulk PEEK–silicon nitride composite; a study of materials for implantable MEMS devices, outside dental and orthopedic use; and a review). In total, 44 studies are included. A PRISMA 2020 flow diagram is shown in Figure 1 [39].

The sixteen full-text reports excluded (eleven from the database searches and five from other methods) are listed with reasons in Supplementary File S3. Of the 44 included studies, 41 came from the database searches and three from other methods. The PubMed reproduction of the MEDLINE search returned 29 records, all already among the records found. The second reviewer’s broader PubMed search added 16 new records, which were screened and excluded (15 at title and abstract and one, a SiCN coating on silicon wafers, in full text). The ProQuest MEDLINE search added 39 new records among the 48 exported, and the PubMed re-run of its strategy added 93 new records, all excluded at title and abstract. The other 227 records of the ProQuest search could not be exported and were not screened individually. Trial registries. ClinicalTrials.gov retrieved two records, ISRCTN 0, the Netherlands Trial Register 1, and CRIS, ReBEC and IRCT 0 each; ChiCTR retrieved 0 records for the Chinese term for silicon nitride (and 39 for “dental implant”, none concerning silicon nitride). The two ClinicalTrials.gov records (NCT01511445 and NCT01557829, the SNAP trial) and the Dutch record (NL-OMON36245, the SNAP trial as registered in the Netherlands, a duplicate of NCT01557829) concern bulk silicon nitride spinal cages and not coated implants. The WHO ICTRP portal (24 October 2026) returned two trials for “silicon nitride” (the SNAP trial again and a trial of titanium nitride-coated stainless-steel crowns in primary teeth, which is a different coating material), none for “Si₃N₄”, “silicon carbon nitride” or “SiCN”, and one unrelated trial for “SiN_x” (a nasal spray for allergic rhinitis). No registered trial of a silicon-nitride-coated implant was identified in the seven registries or in WHO ICTRP. The human studies protocol (PROSPERO CRD420261528198) found no eligible human study. We therefore distinguish “no published clinical study of a silicon nitride-coated implant was identified in the bibliographic searches” from “no registered trial was identified in the registries searched”; a zero-record result in Cochrane CENTRAL is not evidence of the latter.

3.2. Study Characteristics

The 44 studies were published between 2012 and 2026 (Table 2). Fifteen were bench studies of mechanical, tribological, micromotion, composition, dissolution, debris, flexural, adhesion or corrosion behavior [27,28,41,46,47,48,49,50,51,52,53,54,55,56,57]; twenty-two were in vitro or combined in vitro and bench studies [40,42,43,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76]; six combined in vitro work with an animal experiment [77,78,79,80,81,82]; and one was an animal study [83]. Fourteen studies coated cobalt–chromium alloys [27,28,41,46,47,48,50,51,52,53,55,59,60,76]; nineteen coated titanium or titanium alloys [43,48,54,56,57,58,61,62,63,64,65,66,67,71,72,73,74,79,82] (one study coated both [48]); nine coated polymers or polymer composites (PEEK, polyetherketoneketone (PEKK) or carbon-fiber-reinforced PEKK) [40,42,49,70,77,78,80,81,83]; one coated polyethylene, titanium, zirconia-toughened alumina and zirconia [68]; one coated yttria-stabilized zirconia [69]; and one coated zirconia-toughened alumina [75]. Deposition methods were reactive HiPIMS [27,28,41,46,50,51,53], other sputtering or PVD [55,59,60,65,71,76,79,82], PECVD, plasma CVD or microwave plasma deposition [47,54,61,63,66,70], MAO with silicon nitride [67,73], cathodic arc deposition [64], laser sintering or cladding [40,57,58,68,69], suspension coating with melt binding [77,80], sulfonation with silicon nitride microparticles [78,81], spray coating with compression molding [49], electrophoretic deposition [43,72,74], and hot pressing with salt leaching [42]; the method was not stated in the text assessed for three studies [48,62,83], and in one study, the silicon nitride layer was a buffer under an apatite film grown by induction heating [56]. Seven studies used animals: three used rabbits [77,78,83] and four used rats [79,80,81,82], with follow-up of 4 to 12 weeks. Nine studies had authors affiliated with coating or ceramic manufacturers [40,41,42,46,49,58,68,69,75], co-inventorship of a patent on silicon nitride coatings was declared in two [50,51], and several included studies came from one research program [27,28,41,46,50,51,52,53]. Two included reports may share data or coatings with excluded papers from the same group [54]. By category, 18 studies tested Si₃N₄ or SiN_x films, seven tested silicon carbon nitride or doped or graded films, and 19 tested particulate or composite coatings (Table 3 and Figure 2).

Table 2. Characteristics of the included studies.

Study	Design	Substrate	Coating and Deposition	Comparator	Model
Correa Filho 2019 [28]	Bench (immersion, scratch, wear)	CoCrMo flat discs (21.9 mm) and hip heads (32 and 36 mm)	SiNx (N/Si 0.8 to 1.0) on CrN interlayer; SiCN top layer in one head group; reactive HiPIMS	Uncoated CoCrMo heads (wear); coating variants differing in rotation, heating and power	4 disc coating variants (immersion, scratch n = 3); 3 coated head variants and uncoated (wear, 3 repetitions per head)
Correa Filho 2020 [46]	Bench (scratch, wear)	CoCrMo discs	SiNx-based with N (42.6 to 55.5 at%), C (0 to 25.7), Cr (0 or 12.8), Nb (0 to 24.5); reactive HiPIMS	Coating compositions compared with each other; no uncoated control reported in the abstract	Coating variants on discs; multidirectional wear against UHMWPE pins
Zhou 2024 [47]	Bench (mechanical)	CoCr	SiNx with N/Si 0.65, 1.16 and 1.42; PECVD (NH3 flow varied)	Coating variants compared with each other	Coating variants on CoCr
Zanocco 2020 [58]	In vitro and bench	Anodized commercially pure titanium; commercial Ti-alloy acetabular shells	Si3N4; 3D-additive laser sintering	Uncoated Ti; bulk Si3N4 reference	SaOS-2 cells; S. epidermidis; E. coli (luciferase) on shells
Pettersson 2013 [27]	Bench (mechanical, tribological)	CoCr (ASTM F75) for three coatings and P-doped Si (001) wafers	SixNy and SixCyNz; reactive HiPIMS	Sintered silicon nitride and CoCr (reference)	Coating variants; ball-on-disc in serum
Skjöldebrand 2022 [59]	In vitro and bench	Si wafers and CoCrMo discs	SiN with Cr and Nb gradients (Si 38.6 to 46.9 at%, Nb 2.2 to 4.6, Cr 1.9 to 6.0), with or without CrN interlayer; reactive combinatorial sputtering	With versus without CrN interlayer; composition gradient	Gradient coatings; L929 mouse fibroblasts (indirect cytotoxicity)
Echeverri 2023 [60]	In vitro and bench	CoCrMo discs	SiN with Fe and C additions; Physical vapor deposition (sputtering)	SiN reference coating	Compositional gradients; C8-B4 microglia in 2D extracts and 3D collagen hydrogel
Dai 2019 [77]	In vitro and in vivo (rabbit)	Porous PEEK	Nano silicon nitride coating (CSNPPK); Suspension coating and melt binding	Porous PEEK (PPK); dense PEEK (DPK)	MC3T3-E1 cells (n = 3); 30 male rabbits (3 months, 2.5 ± 0.2 kg) in 3 groups; femoral condyle defects 5 mm × 5 mm
Wu 2019 [78]	In vitro and in vivo (rabbit)	PEKK	Microporous PEKK/SN composite coating (0, 5 and 10 wt% SN: PKS, PKSC5, PKSC10); Sulfonation in sulfuric acid suspension with SN microparticles	PKS (no SN); PEKK	Rat bone mesenchymal stem cells; E. coli ATCC 25922 and S. aureus ATCC 25923; 36 New Zealand white rabbits (18 histology, 18 biomechanics)
Haschke 2020 [48]	Bench (micromotion)	Ti6Al4V and CoCr29Mo6 stem tapers; CoCr or Ti6Al4V heads	SiNx on male taper (two manufacturers); deposition method not stated in the text assessed	Uncoated tapers	Each head and stem combination n = 4; sinusoidal loading of three daily activities; six eddy-current sensors
Huang 2024 [79]	In vitro and in vivo (rat)	Titanium	Si3N4 coating; Non-equilibrium magnetron sputtering	Titanium; sandblasted and acid-etched titanium (SLA-Ti)	hBMSCs (n = 3); S. aureus and E. coli; SD rats (8 weeks, 250 ± 30 g), 15 per time point, three implant subgroups
Tani 2024 [61]	In vitro	Grade 2 titanium disks (15 mm, 1 mm thick)	Silicon nitride film; plasma chemical vapor deposition	Uncoated titanium (control)	Rat bone marrow cells; S. aureus
Komasa 2025 [63]	In vitro	Pure titanium disks (15 mm, 1 mm thick), polished	Silicon nitride coating; Plasma chemical vapor deposition	Uncoated titanium (control)	Rat bone marrow cells from 8-week-old Sprague-Dawley rats; each sample in triplicate
Sharma 2025 [49]	Bench (four-point bending)	Continuous carbon-fiber-reinforced PEKK trauma plates	Submicron Si3N4 powder; spray coating with compression molding	Uncoated specimens	n = 24 per group
Issac 2023 [64]	In vitro and bench	Ti6Al7Nb alloy	Silicon nitride, laser post-treated; cathodic arc deposition	Uncoated alloy; as-deposited coating	MTT cell viability; wear tests
Xiong 2025 [65]	In vitro and bench	Ti-6Al-4V	PEEK/SiNx composite coating; spin coating with PVD	TC4; PEEK-only coatings	n = 3 (cells: 5 parallel samples)
Kaczmarek 2024 [66]	In vitro and bench	Ti6Al7Nb alloy	Multilayer SiN:H and SiCN:H; microwave plasma chemical vapor deposition	Bare alloy	MG-63 cells; 4 replicates
Shen 2022 [67]	In vitro	Commercially pure titanium plates	Si3N4-doped micro-arc oxidation coating	Pure titanium; pristine MAO coating	MC3T3-E1 and angiogenesis assays
Hu 2021 [81]	In vitro and in vivo (rat)	PEKK	Microporous Si3N4/Ta microparticle surface; sulfonation	Sulfonated PEKK; SN-only; Ta-only	rBMS cells; S. aureus, E. coli; 30 rats (infection) and an osseointegration cohort
Marin 2020 [68]	In vitro	Polyethylene, Ti6Al4V, ZTA, zirconia	Si3N4 laser cladding	Uncoated substrates	SaOS-2 cells; S. epidermidis
Filho 2019 [50]	Bench (adhesion, wear)	CoCrMo discs with CrN interlayer	SiNx top layer; reactive HiPIMS	Uncoated CoCrMo	Rockwell indentation; wear vs. Si3N4 balls
Skjöldebrand 2017 [51]	Bench (composition, mechanical)	CoCrMo discs	SiNx; reactive HiPIMS at 200 to 430 °C	Uncoated CoCrMo	Process variants
Pettersson 2016a [52]	Bench (dissolution, electrochemical)	CoCrMo	SiNx with N/Si 0.3, 0.8 and 1.1	Uncoated CoCrMo	Three coating compositions; 60 days
Pettersson 2016b [53]	Bench (wear debris)	Cast CoCr (ASTM F-75)	SiNx; reactive HiPIMS	Ultra-high-molecular-weight polyethylene debris (reference)	Ball-on-disc, 2 h, three replicates
Marin 2020 (Mater. Des.) [69]	In vitro and bench	Yttria-stabilized zirconia	Si3N4 composite coating; laser cladding	Untreated zirconia; stoichiometric Si3N4	SaOS-2 cells; 10 days
Kalisz 2015 [54]	Bench (structural, mechanical, corrosion)	Ti6Al4V	Graphene/SiN/graphene hybrid with a 200 nm SiN film; PECVD	Uncoated alloy; SiN film	Nanoindentation (five per point); corrosion tests in triplicate
Zhu 2024 [62]	In vitro	32 titanium sheets with anodized TiO2 nanotubes (50 and 150 nm)	Quaternized SiCN, SiCN and SiC coatings; deposition method not stated in the text assessed	Uncoated nanotube surfaces; SiC	MC3T3-E1 osteoblasts; 8 groups, n = 4 per group
Dai 2020 [80]	In vitro and in vivo (rat)	PEEK	Nano silicon nitride coating (CSNPK); Suspension coating and melt bonding	PEEK	MC3T3-E1 cells; 20 SD rats (6 months) in 2 groups; skull defects
Enami 2022 [83]	In vivo (rabbit)	Carbon-fiber-reinforced PEEK cylinders	Si3N4 coating (SiN group); coating method not stated in the text assessed	HA-coated CFRP (control, about 20 µm); HA-coated with vancomycin and PVP (HA-VP)	Rabbits (at least 12 months, 3.0 to 3.5 kg); 10 implants per group inserted at random into the femoral diaphysis
Xu 2019 [70]	In vitro	PEEK	Hierarchical amorphous silicon nitride coating (about 500 nm); ICPECVD	Uncoated PEEK	rBMSCs; E. coli and S. aureus; experiments in triplicate
Olofsson 2012 [55]	Bench (tribology, structure)	CoCr discs and Si wafers	SixNy; reactive PVD (magnetron sputtering)	Bulk CoCr; bulk Si3N4	Several coating variants; ball-on-disc wear; nanoindentation
Delgado-López 2013 [56]	Bench (apatite adhesion)	Ti6Al4V	Si3N4 buffer layer under a citrate-apatite film (about 2 µm)	SiC and TiN buffer layers	Buffered substrates of different roughness
Xu 2016 [57]	Bench and in vitro (SBF)	Ti6Al4V	Si3N4 and Si3N4/TCP composite; laser cladding	Bulk material; coatings with and without TCP	Coating groups A, B and C; SBF for 14 and 28 days
Zanocco 2020 [40]	In vitro	Polyethylene	Si3N4 cladding (about 3.5 µm); Nd:YAG pulsed laser in nitrogen	Uncoated polyethylene	S. epidermidis; SaOS-2 cells
Schmidt 2017 [41]	Bench (stress, hardness, adhesion)	Si(001) and CoCrMo disks	SiNx, amorphous; reactive HiPIMS (2 µm on Si, 5 µm on CoCrMo)	Process variants	Residual stress; hardness; Rockwell C adhesion
Yuan 2026 [82]	In vitro and animal	Titanium mesh (selective laser melting)	SiNx nanofilms with three nitrogen contents; reactive DC magnetron sputtering	Bare titanium mesh	RAW264.7 macrophages, BMSCs, S. aureus, E. coli, P. gingivalis; 24 rats, calvarial defect
Wang 2026 [71]	In vitro and bench	Pure titanium (blocks and foils)	a-SiNx films; magnetron sputtering, nitrogen flow 0 to 40 sccm	Uncoated Ti; silicon film	mBMSCs, L929 fibroblasts; hardness, adhesion, corrosion, corrosion-assisted wear
Heydarian 2025 [72]	In vitro and bench	Anodized Ti-6Al-4V (TiO2 nanotubes)	Si3N4 nanoparticles; electrophoretic deposition	Anodized surfaces; uncoated alloy	Cells, S. aureus, SBF immersion; hardness, friction, corrosion
Guo 2023 [73]	In vitro and bench	Commercial pure titanium	MAO coating with Si3N4 nanoparticles (three contents)	Pure Ti; MAO coating without Si3N4	Rat bone marrow mesenchymal stem cells, S. aureus, E. coli; Si release, corrosion
Moskalewicz 2022 [74]	In vitro and bench	Ti-13Nb-13Zr alloy	Si3N4 or CuO nanoparticles in a graphite oxide/hydroxyapatite/alginate coating; electrophoretic deposition	Coatings without nanoparticles; CuO coatings	MG-63 cells, E. coli; adhesion, corrosion, SBF
Pezzotti 2019 [75]	In vitro	Zirconia-toughened alumina	Si3N4 coating (15 µm) by pulsed laser sintering; laser-patterned wells filled with bioglass and Si3N4	Uncoated ZTA; bioglass-only wells	SaOS-2 cells (7 days, n = 3)
Shi 2012 [76]	In vitro and bench	CoCrMo alloy	F:SiCN and SiNO films; DC unbalanced magnetron sputtering at different N2 and CF4 flows	Uncoated CoCrMo; films at other gas flows	Platelet adhesion (n = 6); nanoindentation; pin-on-disk against UHMWPE
Esmaeili 2024 [43]	In vitro and bench	Ti6Al4V	Si3N4/nanodiamond; electrophoretic deposition (20 to 50 V)	Uncoated Ti6Al4V	MG63 cells; nanoindentation and nano-scratch
Boschetto 2021 [42]	In vitro	Bulk PEEK discs	Porous PEEK surface with about 15 vol% beta-Si3N4 particles; hot pressing and salt leaching	Smooth PEEK; NaCl-treated PEEK	SaOS-2 cells (7 days, n = 3); S. epidermidis (12 to 48 h)

Table 3. Classification, composition and structure of the coatings in the included studies. Categories: (i) Si₃N₄ or SiN_x films; (ii) SiCN and other doped or graded films; (iii) particulate or composite coatings. “Not reported” means that the paper did not report the structure or phases.

Study	Category	Coating Composition as Reported	Structure and Other Phases Reported	Substrate; Deposition Method; Thickness
Yuan 2026 [82]	(i)	Amorphous SiNx nanofilms with low, medium and high nitrogen (SiN55, SiN65, SiN75)	Amorphous nanoscale SiNx (TEM: no lattice fringes; SAED: diffuse halo); no other phases reported	Titanium mesh made by selective laser melting from pure titanium powder (0.2 mm thick, 600 µm pores, 60% porosity); reactive DC magnetron sputtering (argon 40 sccm, 0.8 Pa, 4 rpm); nitrogen set by optical emission monitoring at 55, 65 and 75%; Thickness not stated in the text assessed
Wang 2026 [71]	(i)	Amorphous SiNx films; nitrogen content varied with N2 flow rate	Amorphous, Si-N dominated at 20 sccm nitrogen flow; compactness and bonding change at other flows (XRD, FTIR)	Medical-grade pure titanium blocks; Ti foils; Si(001) wafers for characterization; Magnetron sputtering, N2 flow 0, 10, 20, 30 and 40 sccm (Si, SiN-10 to SiN-40); substrate rotated at 4 rpm; About 500 nm (FIB-SEM cross-sections)
Komasa 2025 [63]	(i)	Silicon nitride coating	Crystalline layer on SEM	Pure titanium disks (15 mm, 1 mm thick), polished with SiC papers (800 to 1500 grit); Plasma chemical vapor deposition; 200 nm
Tani 2024 [61]	(i)	Silicon nitride film	Crystalline layers and spherical crystalline accumulations on the coated surface (SEM); XPS: Si, N, O and C; no XRD	Grade 2 titanium disks (15 mm, 1 mm thick); Plasma chemical vapor deposition; 200 nm
Huang 2024 [79]	(i)	Si3N4 coating	Not reported (abstract only)	Titanium; Non-equilibrium magnetron sputtering; Not stated (abstract only)
Zhou 2024 [47]	(i)	SiNx with N/Si 0.65, 1.16 and 1.42	Amorphous at NH3/SiH4 below 1 (no Si3N4 peak); Si reflections at about 54 and 56 degrees at higher ratios (GIXRD)	CoCr; PECVD (NH3 flow varied); About 3 µm (3 h deposition)
Issac 2023 [64]	(i)	Silicon nitride (SiN), laser post-treated	Not reported	Ti6Al7Nb (TiAlNb) alloy, polished; Cathodic arc deposition, then Yb-doped fibre laser treatment; 2 to 3 µm (4 h deposition)
Haschke 2020 [48]	(i)	SiNx on male taper (two manufacturers)	Not reported	Ti6Al4V and CoCr29Mo6 stem tapers; CoCr or Ti6Al4V heads; Not stated in the text assessed; Equal for both manufacturers (value not stated in the text assessed)
Correa Filho 2019 [28]	(i)	SiNx (N/Si 0.8 to 1.0) on CrN interlayer; SiCN top layer in one head group	Higher oxygen content in the lower-density coating	CoCrMo flat discs (21.9 mm) and hip heads (32 and 36 mm); reactive HiPIMS; 4.3 to 6.0 µm
Filho 2019 [50]	(i)	SiNx top layer (N/Si 1.10 to 1.25)	Structure not reported; XPS gives Si, N and O contents of the top layers	Mirror-polished CoCrMo discs with CrN interlayer; reactive DC magnetron sputtering (interlayer, bias 100 to 900 V) then reactive HiPIMS; 1-fold or 3-fold rotation; SiNx top layer about 4.3 to 4.4 µm; CrN interlayer 650 to 810 nm
Xu 2019 [70]	(i)	Hierarchically nanostructured amorphous silicon nitride coating (PKSN)	Amorphous, hierarchical nanostructure	PEEK; Inductively coupled plasma-enhanced chemical vapor deposition (ICPECVD); About 500 nm
Skjöldebrand 2017 [51]	(i)	SiNx (N 23 to 48 at.%)	Amorphous, homogeneous (TEM, electron diffraction); oxygen 1.8 to 5.8 at.% (XPS)	CoCrMo discs; reactive HiPIMS; substrate 200, 350 or 430 °C; N2/Ar 0.06, 0.17 or 0.30; About 7.1 to 7.6 µm
Pettersson 2016a [52]	(i)	SiNx with N/Si 0.3, 0.8 and 1.1	Structure not reported; XPS chemical bonding and N/Si ratio of four SiNx coatings	CoCrMo (substrate and reference); reactive HiPIMS (CemeCon CC800/9 ML); 7.0 to 7.5 µm
Pettersson 2016b [53]	(i)	SiNx	Structure not reported; featureless coating, N/Si about 0.8 (XPS)	Cast CoCr (ASTM F-75, Ra about 5 nm); reactive HiPIMS; 8 ± 0.2 µm
Pettersson 2013 [27]	(i)	SixNy and SixCyNz	Not reported	CoCr (ASTM F75) for three coatings and P-doped Si (001) wafers; reactive HiPIMS; Not stated in the text assessed
Delgado-López 2013 [56]	(i)	Silicon nitride (Si3N4) buffer layer under a citrate-functionalized apatite film (about 2 µm)	Si3N4, SiC and TiN buffer layers not resolved by XRD (only the alpha-Ti substrate reflections, layers too thin); structure of the Si3N4 layer not reported	Ti6Al4V supports; Commercially supplied Si3N4, SiC and TiN buffer layers (VASKO Ltd.); apatite film by induction heating in calcium citrate-phosphate solution at pH 11; Apatite film about 2 µm; buffer thickness not stated
Olofsson 2012 [55]	(i)	SixNy (amorphous or nanocrystalline)	Amorphous or nanocrystalline	CoCr discs and Si wafers; reactive physical vapour deposition (magnetron sputtering); About 0.6 to 1.4 µm depending on the variant (SEM cross-sections)
Schmidt 2017 [41]	(i)	SiNx, X-ray amorphous, Si/N < 1 at N2/Ar ratio 0.28 to 0.3	X-ray amorphous; Si/N below 1	Si(001) wafers and polished CoCrMo disks; reactive HiPIMS (Si targets, N2/Ar); 2 µm on Si; 5 µm on CoCrMo
Shi 2012 [76]	(ii)	Fluorine- and nitrogen-doped silicon nitride films (F:SiCN and SiNO)	Structure not reported; XPS bonding of Si, N, C and F in F:SiCN films and oxygen in SiNO films	CoCrMo alloy (tribology against UHMWPE); other substrates not stated in the text assessed; DC unbalanced magnetron sputtering; CF4 flow 3, 6 and 9 sccm with N2 at 20 sccm (F:SiCN); N2 flow 0, 10, 15 and 20 sccm (SiNO); Measured by step profiler; value not stated (the 50 nm indentation depth was one tenth of the film thickness)
Zhu 2024 [62]	(ii)	Quaternized SiCN, SiCN and SiC coatings	Not reported (quaternized SiCN; no structural analysis found)	32 titanium sheets with anodized TiO2 nanotubes (50 and 150 nm); Not stated in the text assessed; About 10 nm
Kaczmarek 2024 [66]	(ii)	Amorphous multilayer SiN:H and SiCN:H (nitriding steps)	Amorphous multilayer	Ti6Al7Nb alloy; Microwave plasma chemical vapor deposition (multi-stage); 2.48 µm (SiN:H); 2.68 µm (SiCN:H)
Echeverri 2023 [60]	(ii)	SiN with Fe and C additions	Structure not reported; XPS: Si, N, Fe, C and O in the SiFeCN gradients and the SiN reference	CoCrMo discs; Physical vapor deposition (sputtering); SiFeCN about 427 to 524 nm; SiN reference 333 nm (FIB cross-sections)
Skjöldebrand 2022 [59]	(ii)	SiN with Cr and Nb gradients (Si 38.6 to 46.9 at%, Nb 2.2 to 4.6, Cr 1.9 to 6.0), with or without CrN interlayer	Structure not reported; composition by XPS (chromium and niobium additions)	Si wafers and CoCrMo discs; reactive combinatorial sputtering; Not stated
Correa Filho 2020 [46]	(ii)	SiNx-based with N (42.6 to 55.5 at%), C (0 to 25.7), Cr (0 or 12.8), Nb (0 to 24.5)	Amorphous phase gave a smoother surface; low oxygen desirable	CoCrMo discs; reactive HiPIMS; 3.7 to 8.8 µm
Kalisz 2015 [54]	(ii)	Hybrid graphene/SiN/graphene multilayer with a 200 nm silicon nitride film	Not reported	Ti6Al4V (ASTM Grade 5), mirror-polished; PECVD of SiN (120 W, 0.8 Torr, SiH4 150 and NH3 50 mL/min, 350 °C); CVD graphene transferred by the PMMA method; SiN 200 nm
Xiong 2025 [65]	(iii)	PEEK/SiNx composite coating (hierarchical, mineralized-fiber-like)	Amorphous SiNx (per title)	Ti-6Al-4V (TC4); Spin coating of PEEK combined with physical vapor deposition of SiNx (sputtering); SiNx layer about 1 to 2 µm on a PEEK layer of 20 to 30 µm (SEM)
Sharma 2025 [49]	(iii)	Submicron Si3N4 powder (particulate coating)	Not reported for the coating (a roughly 35 nm amorphous silicon oxynitride layer in water is cited from the literature)	Continuous carbon-fiber-reinforced PEKK (CCF-PEKK) trauma plates, three fiber layups, made by fused filament fabrication plus compression molding; Spray coating then embedding by uniaxial compression molding; Not applicable (embedded particulate layer)
Heydarian 2025 [72]	(iii)	Si3N4 nanoparticles on TiO2 nanotubes	XRD identified anatase and rutile TiO2 and the composite coating (15 to 75 degrees); anodized TiO2 initially amorphous; characteristic Si3N4 peaks at about 27 to 52 degrees (Figure 3)	Ti-6Al-4V, anodized (TiO2 nanotubes, 40 V for 1, 2 or 4 h); Electrophoretic deposition on the optimal (4 h) anodized surface; Not stated
Esmaeili 2024 [43]	(iii)	Si3N4/nanodiamond (ND) nanocomposite	XRD: ND and Si3N4 main phases; Ti3Al, VC, TiSi2, V3Si and SiC phases formed on heating	Ti6Al4V sheets polished with silicon carbide papers; Electrophoretic deposition from an aqueous suspension with SDS dispersant; 20, 30, 40 and 50 V compared (40 V best); Thickness not stated
Guo 2023 [73]	(iii)	Micro-arc oxidation coating containing Si3N4 nanoparticles (MSN1, MSN2, MSN3)	Si3N4 nanoparticles amorphous (no crystalline XRD peak); coatings: anatase and rutile TiO2 in MTi and MSN1 to MSN3, no other crystalline peaks	Commercial pure titanium disks (12 × 2 mm); Micro-arc oxidation with 0.01, 0.1 or 0.2 g Si3N4 nanoparticles in the electrolyte; Not stated
Moskalewicz 2022 [74]	(iii)	Multiphase graphite oxide/hydroxyapatite/sodium alginate coating with Si3N4 or CuO nanoparticles (Si3N4/n-GtO/HAp/NaAlg and CuO/n-GtO/HAp/NaAlg)	SAED: amorphous Si3N4 phase; monoclinic CuO; hexagonal hydroxyapatite; graphite oxide flakes	Ti-13Nb-13Zr alloy (2 mm thick polished discs); Electrophoretic deposition, DC 2 to 12 V for 5 min, AISI 316L counter electrode; Si3N4-containing coating 6.7 to 9.3 µm; CuO-containing coating 2.4 to 3.5 µm
Pezzotti 2019 [75]	(iii)	Si3N4 coating (beta-Si3N4 with 6 wt% Y2O3 and 4 wt% Al2O3) and laser-patterned wells filled with bioglass/Si3N4 powder mixtures	Substrate phases by Raman (tetragonal and monoclinic zirconia, alumina); Si3N4 coating confirmed by FTIR; coating structure not otherwise reported	Zirconia-toughened alumina (ZTA) squares, 10 × 10 × 3 mm, polished (Ra 0.22 µm); Pulsed Nd:YAG laser sintering of the Si3N4 coating; laser patterning and powder filling of wells; Coating 15 +/− 5 µm
Boschetto 2021 [42]	(iii)	Porous PEEK surface with beta-Si3N4 particles (about 10 µm) trapped in it (particulate/composite surface)	beta-Si3N4 particles in the porous PEEK surface confirmed by FTIR; structure not otherwise reported	Bulk PEEK discs, 12.7 × 3 mm (ASTM D6262); 24 discs (8 smooth control, 8 NaCl-treated, 8 NaCl-Si3N4); Hot pressing of NaCl grains (150 to 300 µm) with Si3N4 particles at about 10 MPa for 10 min on a 340 °C hot plate, then salt leaching in water; Thickness not stated
Enami 2022 [83]	(iii)	Si3N4 coating (SiN group)	Not reported (abstract-level text only)	Carbon-fiber-reinforced PEEK cylinders; Si3N4 coating (method not stated in the text assessed); Not stated (abstract-level text)
Shen 2022 [67]	(iii)	Silicon nitride (Si3N4)-doped micro-arc oxidation coating (MAO-SN2, MAO-SN4, MAO-SN6)	Phase structure not reported (no XRD); Si and N detected only in the Si3N4-doped coatings (EDS)	Commercially pure titanium plates (1 cm × 1 cm); Micro-arc oxidation with Si3N4 particles in the electrolyte; Not stated
Hu 2021 [81]	(iii)	Microporous surface containing Si3N4 and Ta microparticles (STP)	XRD: PEKK, Si3N4 and Ta characteristic peaks in the microparticle surfaces; FTIR: Si-N-Si band at 689 per cm	Polyetherketoneketone (PEKK); Sulfonation in sulfuric acid suspension with microparticles; Not stated
Dai 2020 [80]	(iii)	Nano silicon nitride coating (CSNPK)	Structure not reported (SEM and EDS only); nano Si3N4 powder used	PEEK; Suspension coating and melt bonding; About 10 µm
Zanocco 2020 [58]	(iii)	Si3N4	Not reported; micrometre Si3N4 layer with metal-ceramic interlocking noted	Anodized commercially pure titanium; commercial Ti-alloy acetabular shells; 3D-additive laser sintering; 15 ± 5 µm
Marin 2020 (Mater. Des.) [69]	(iii)	Composite coating of Si3N4 particles dispersed in nano-crystalline/amorphous silicon	Si3N4 particles in nano-crystalline or amorphous silicon	3 mol% yttria-stabilized zirconia (Kyocera); Laser cladding of Si3N4 powder (three passes); 15 ± 5 µm
Marin 2020 [68]	(iii)	Silicon nitride coating	Partial nitrogen loss on some substrates (silicon-rich phases)	Low-density polyethylene, Ti6Al4V, zirconia-toughened alumina, yttria-stabilized zirconia; Laser cladding of Si3N4 powder; Not stated in the text assessed
Dai 2019 [77]	(iii)	Nano silicon nitride coating (CSNPPK)	Not reported (abstract only)	Porous PEEK; Suspension coating and melt binding; Not stated (abstract only)
Wu 2019 [78]	(iii)	Microporous PEKK/SN composite coating (0, 5 and 10 wt% SN: PKS, PKSC5, PKSC10)	Not reported (abstract only)	PEKK; Sulfonation in sulfuric acid suspension with SN microparticles; Not stated (abstract only)
Xu 2016 [57]	(iii)	Si3N4 and Si3N4/tricalcium phosphate (TCP) composite coatings	Not reported (abstract only)	Ti6Al4V (TC4); Laser cladding with Nd:YAG pulsed, CO2 CW and semiconductor CW lasers; Not stated (abstract only)
Zanocco 2020 [40]	(iii)	Silicon nitride (Si3N4) cladding; Si3N4 particles partly embedded in the polyethylene matrix	Si3N4 on polyethylene confirmed by Raman (Si3N4 and polyethylene bands); crystallinity not reported	Polyethylene; Nd:YAG pulsed laser cladding of Si3N4 powder in nitrogen atmosphere; Lamellar structure, 3.5 ± 0.6 µm

These results should not be read as one consistent signal across the categories. Films, doped films and particulate or composite coatings differ in composition, structure, thickness and degradation, and dissolution and ion release were tuned in the doped films by adding chromium, niobium, iron or carbon [46,59,60]. Few studies report oxygen or silica content or other phases, although silicon nitride reacts with water to form silica and ammonia [23] and dissolves in aqueous media [24]. Where a paper does not report a feature, the table says so. Structure and phase entries were read from the full texts, except for five studies for which only the abstract page was available [57,77,78,79,83].

The 44 publications come from 19 research groups. Five groups account for 30 publications (a titanium-coating group in China, 4; a Japanese titanium-coating group, 2; the Uppsala group, 12; a Chinese group working on PEEK, PEKK and titanium, 6; and a group linked to the manufacturer, 6), and 14 publications come from groups that do not share authors with the others (Supplementary File S4, Overlap groups). Publications from one group may share coatings, controls or cell models, so they are not treated as independent evidence in the group-level counts below. Borderline inclusions (a silicon nitride buffer under an apatite film, silicon nitride as one of two nanoparticle additives, oxygen-containing doped films, and laser-clad coatings with partial nitrogen loss) were kept in their categories and are listed with reasons in Supplementary File S4 (Borderline inclusions).

3.3. Risk of Bias in the Included Studies

All 44 studies were rated at moderate risk of bias overall (Table 4 and Table 5). None reported a sample-size rationale, and none of the in vitro experiments reported blinding of outcome assessment or randomization of specimens. Six of the seven animal studies reported age and weight or strain, but only three stated that animals or implants were randomly allocated [79,82,83]; one of them [82] described the method of sequence generation (a random number table) and blinded outcome assessment, and none described allocation concealment, random housing or blinding of caregivers. Control or comparator groups were present in all but three studies, in which coating variants were compared only with each other [41,46,47]. The judgments were rated independently by one human reviewer after an AI first draft, and no study was excluded on the basis of risk of bias. These labels describe how fully experimental details were reported; they are not ratings of the certainty of the evidence, they come from two different tools, and the uniform label hides differences between laboratory and animal studies.

Table 4. Risk of bias in the included studies (RoBDEMAT items: SR, sufficiently reported; IR, insufficiently reported; NR, not reported; item-level judgments for all studies are in Supplementary File S2; the animal-study domains are in Table 5).

Study	Tool	Control Group	Randomization	Sample-Size Rationale	Blinding	Overall
Correa Filho 2019 [28]	RoBDEMAT	SR	NR	NR	NR	Moderate
Correa Filho 2020 [46]	RoBDEMAT	IR	NR	NR	NR	Moderate
Zhou 2024 [47]	RoBDEMAT	IR	NR	NR	NR	Moderate
Zanocco 2020 [58]	RoBDEMAT	SR	NR	NR	NR	Moderate
Pettersson 2013 [27]	RoBDEMAT	SR	NR	NR	NR	Moderate
Skjöldebrand 2022 [59]	RoBDEMAT	SR	NR	NR	NR	Moderate
Echeverri 2023 [60]	RoBDEMAT	SR	NR	NR	NR	Moderate
Dai 2019 [77]	RoBDEMAT (in vitro); SYRCLE (animal)	SR	NR	NR	NR	Moderate
Wu 2019 [78]	RoBDEMAT (in vitro); SYRCLE (animal)	SR	NR	NR	NR	Moderate
Haschke 2020 [48]	RoBDEMAT	SR	NR	NR	NR	Moderate
Huang 2024 [79]	RoBDEMAT (in vitro); SYRCLE (animal)	SR	NR in vitro; animals randomly divided (method not reported)	NR	NR	Moderate
Tani 2024 [61]	RoBDEMAT	SR	NR	NR	NR	Moderate
Komasa 2025 [63]	RoBDEMAT	SR	NR	NR	NR	Moderate
Sharma 2025 [49]	RoBDEMAT	SR	NR	NR	NR	Moderate
Issac 2023 [64]	RoBDEMAT	SR	NR	NR	NR	Moderate
Xiong 2025 [65]	RoBDEMAT	SR	NR	NR	NR	Moderate
Kaczmarek 2024 [66]	RoBDEMAT	SR	NR	NR	NR	Moderate
Shen 2022 [67]	RoBDEMAT	SR	NR	NR	NR	Moderate
Hu 2021 [81]	RoBDEMAT; SYRCLE	SR	NR	NR	NR	Moderate
Marin 2020 [68]	RoBDEMAT	SR	NR	NR	NR	Moderate
Filho 2019 [50]	RoBDEMAT	SR	NR	NR	NR	Moderate
Skjöldebrand 2017 [51]	RoBDEMAT	SR	NR	NR	NR	Moderate
Pettersson 2016a [52]	RoBDEMAT	SR	NR	NR	NR	Moderate
Pettersson 2016b [53]	RoBDEMAT	IR	NR	NR	NR	Moderate
Marin 2020 (Mater. Des.) [69]	RoBDEMAT	SR	NR	NR	NR	Moderate
Kalisz 2015 [54]	RoBDEMAT	SR	NR	NR	NR	Moderate
Zhu 2024 [62]	RoBDEMAT	SR	NR	NR	NR	Moderate
Dai 2020 [80]	RoBDEMAT (in vitro); SYRCLE (animal)	SR	NR	NR	NR	Moderate
Enami 2022 [83]	SYRCLE	SR	Unclear (stated random; method not reported)	NR	NR	Moderate
Xu 2019 [70]	RoBDEMAT	SR	NR	NR	NR	Moderate
Olofsson 2012 [55]	RoBDEMAT	SR	NR	NR	NR	Moderate
Delgado-López 2013 [56]	RoBDEMAT	SR	NR	NR	NR	Moderate
Xu 2016 [57]	RoBDEMAT	SR	NR	NR	NR	Moderate
Zanocco 2020 [40]	RoBDEMAT	SR	NR	NR	NR	Moderate
Schmidt 2017 [41]	RoBDEMAT	IR	NR	NR	NR	Moderate
Yuan 2026 [82]	RoBDEMAT	SR	NR	NR	NR	Moderate
Esmaeili 2024 [43]	RoBDEMAT	SR	NR	NR	NR	Moderate
Boschetto 2021 [42]	RoBDEMAT	SR	NR	NR	NR	Moderate
Wang 2026 [71]	RoBDEMAT	SR	NR	NR	NR	Moderate
Heydarian 2025 [72]	RoBDEMAT	SR	NR	NR	NR	Moderate
Guo 2023 [73]	RoBDEMAT	SR	NR	NR	NR	Moderate
Moskalewicz 2022 [74]	RoBDEMAT	SR	NR	NR	NR	Moderate
Pezzotti 2019 [75]	RoBDEMAT	SR	NR	NR	NR	Moderate
Shi 2012 [76]	RoBDEMAT	SR	NR	NR	NR	Moderate

Table 5. Risk of bias of the seven animal studies with the SYRCLE tool. Domains: 1, sequence generation; 2, baseline characteristics; 3, allocation concealment; 4, random housing; 5, blinding of caregivers and investigators; 6, random outcome assessment; 7, blinding of outcome assessors; 8, incomplete outcome data; 9, selective outcome reporting; 10, other sources of bias. L, low risk; U, unclear risk.

Study	1	2	3	4	5	6	7	8	9	10	Overall
Dai 2019 [77]	U	L	U	U	U	U	U	U	L	L	Moderate
Wu 2019 [78]	U	L	U	U	U	U	U	U	L	L	Moderate
Huang 2024 [79]	U	L	U	U	U	U	U	U	L	L	Moderate
Dai 2020 [80]	U	L	U	U	U	U	U	U	L	U	Moderate
Hu 2021 [81]	U	U	U	U	U	U	U	U	L	L	Moderate
Enami 2022 [83]	U	L	U	U	U	U	U	U	L	L	Moderate
Yuan 2026 [82]	L	U	U	U	U	U	L	U	U	U	Moderate

3.4. Results of Individual Studies

The main findings of each study are given in Table 6.

Table 6. Main reported findings of the included studies.

Study	Main Reported Findings
Correa Filho 2019 [28]	One-fold rotation gave denser, more stable coatings; three-fold coatings dissolved by 1 to 3 weeks; coated 32 mm heads gave somewhat lower polyethylene wear and 36 mm somewhat higher; best resistance with SiCN top layer and process heating
Correa Filho 2020 [46]	Low-hardness coatings failed earlier by chemical reaction and dissolution accelerated by tribological contact; high-purity, high-density coatings (low oxygen, high nitrogen) were desirable; no definite correlation between composition and wear
Zhou 2024 [47]	Highest hardness (13.19 ± 1.29 GPa) and modulus (132.76 ± 9.32 GPa) at N/Si 1.42 (NH3 30 sccm); roughness adequate; adhesion needs optimization before wear can be evaluated
Zanocco 2020 [58]	Coating met the 20 N scratch threshold; antibacterial and osteogenic behavior comparable to bulk Si3N4; resisted S. epidermidis and E. coli adhesion on acetabular shells
Pettersson 2013 [27]	Most SixNy coatings reached hardness near 18 GPa and modulus near 200 GPa; all but one SixNy coating had wear resistance similar to bulk silicon nitride and higher than CoCr; SixCyNz coatings less wear resistant
Skjöldebrand 2022 [59]	Interlayer reduced delamination; composition did not affect adhesion, roughness or mechanical properties; Co release low; higher Cr (4.3 to 6.0 at%) gave lower Si, Cr and Nb release and better fibroblast viability
Echeverri 2023 [60]	Fe and C reduced ion release compared with the SiN reference; coating extracts were compatible with microglial viability
Dai 2019 [77]	Higher roughness, compressive strength and water absorption; cell responses promoted; more new vessels and bone and higher push-out loads than PPK
Wu 2019 [78]	PKSC10 had antibacterial activity, stimulated osteogenic differentiation and promoted osteogenesis and osseointegration in vivo
Haschke 2020 [48]	Micromotion up to 8.4 ± 0.8 µm; Ti6Al4V stems showed more micromotion than CoCr; coating had no influence; pull-off forces did not differ; coating not damaged in contact areas
Huang 2024 [79]	Best corrosion resistance and antibacterial activity; superior cell responses; more new bone and osseointegration in vivo than Ti and SLA-Ti
Tani 2024 [61]	Improved hydrophilicity, initial cell adhesion and hard tissue differentiation; significant antibacterial activity against S. aureus; no cytotoxicity
Komasa 2025 [63]	Crystalline silicon nitride layer (200 nm); alkaline phosphatase (ALP) and osterix (OSX) expression higher on silicon-nitride-coated titanium than on uncoated titanium at 3 to 7 days
Sharma 2025 [49]	Layup affected flexural modulus and strength; embedded Si3N4 particles had no effect on flexural properties or crack behavior
Issac 2023 [64]	Laser treatment raised microhardness more than 11-fold, reduced residual stress and improved wear resistance
Xiong 2025 [65]	Modulus 45 to 65 GPa; better wear and corrosion resistance; improved cell adhesion, proliferation and differentiation
Kaczmarek 2024 [66]	More than 70% cell survival; SiCN:H-based layers boosted ALP activity; lower friction coefficient than the bare alloy
Shen 2022 [67]	Better osteogenic and angiogenic response and corrosion resistance than titanium and the pristine MAO coating
Hu 2021 [81]	Antibacterial activity, higher rBMS proliferation and differentiation, anti-infective effect in vivo, and more bone and gap-free osseointegration at week 12 than sulfonated PEKK
Marin 2020 [68]	Cladding enhanced osteoconductivity and antibacterial properties; nitrogen partly lost on some substrates
Filho 2019 [50]	Adhesion improved at low interlayer bias voltage (100 V); wear rates lower than or similar to CoCrMo
Skjöldebrand 2017 [51]	Amorphous, homogeneous coatings; hardness 12 to 26 GPa and modulus 173 to 293 GPa rising with nitrogen content; roughness similar to uncoated CoCrMo
Pettersson 2016a [52]	Dissolution 0.2 to 1.4 nm/day, lower with higher N/Si; similar to or lower than CoCrMo (0.7 to 1.2 nm/day); coatings reduced metal ion release
Pettersson 2016b [53]	Debris round, median 40 nm (10 to 500 nm), agglomerates up to 3.2 µm; model particles dissolved in simulated body fluid
Marin 2020 (Mater. Des.) [69]	Coating adhered, roughness rose; mineralized bone tissue comparable to healthy bone; proliferation similar to zirconia and lower than on stoichiometric Si3N4
Kalisz 2015 [54]	Hardness 23 GPa; corrosion current three orders lower than the alloy in acidic fluoride medium; hybrid became porous and soft after corrosion; poor SiN adhesion to graphene
Zhu 2024 [62]	QSiCN coatings significantly increased osteoblast proliferation (p < 0.005)
Dai 2020 [80]	Higher roughness, hydrophilicity and protein adsorption; slow Si ion release; improved cell adhesion, proliferation, differentiation and gene expression; more bone regeneration and osseointegration in vivo
Enami 2022 [83]	SiN group had lower pull-out strength and less bone formation than HA; HA-VP equivalent to HA
Xu 2019 [70]	Dense hierarchical coating; higher roughness, hydrophilicity and protein adsorption; Si ion release; greater antibacterial activity and better rBMSC adhesion, proliferation and osteogenic differentiation than PEEK
Olofsson 2012 [55]	Amorphous or nanocrystalline coatings; hardness and friction against Si3N4 similar to bulk Si3N4; low wear; adhesion to be improved
Delgado-López 2013 [56]	Apatite film most plastic, compact and adherent on the Si3N4 buffer, then SiC and TiN; adhesion related to buffer roughness
Xu 2016 [57]	Compact coating–substrate bonding with suitable laser parameters; microhardness higher than bulk; good bioactivity when TCP was added
Zanocco 2020 [40]	Coating significantly reduced living S. epidermidis and increased hydroxyapatite formed by SaOS-2 cells compared with uncoated polyethylene
Schmidt 2017 [41]	Dense, hard amorphous coatings; compressive stress of up to 2.1 GPa reducible to 0.2 GPa by process tuning; adhesion on CoCrMo assessed by Rockwell C
Yuan 2026 [82]	All three films improved hydrophilicity, corrosion resistance and antibacterial activity compared with titanium; adhesion 12.58, 17.70 and 15.8 N; the SiN65 film promoted M2 macrophage polarization, osteogenic differentiation and bone regeneration in rat calvarial defects
Wang 2026 [71]	The film made at 20 sccm nitrogen flow had the highest hardness (11.31 GPa) and the best corrosion and corrosion-assisted wear resistance, and promoted mBMSC osteogenic differentiation and L929 fibroblast migration
Heydarian 2025 [72]	The 4 h anodized surface with Si3N4 nanoparticles improved antibacterial activity, cell viability and hydroxyapatite formation in simulated body fluid
Guo 2023 [73]	Cell proliferation, osteoblastic differentiation and bactericidal rate rose with the Si3N4 content of the micro-arc oxidation coating (S. aureus bactericidal rate 26.4% for MSN1 versus 0% on pure Ti)
Moskalewicz 2022 [74]	Si3N4-containing coatings adhered strongly to Ti-13Nb-13Zr, were hydrophilic, protected against corrosion and formed hydroxyapatite in SBF; only CuO-containing coatings were clearly antibacterial, and adding Si3N4 or CuO reduced MG-63 viability
Pezzotti 2019 [75]	The Si3N4 coating and the bioglass/Si3N4-filled wells activated osteoblasts compared with uncoated zirconia-toughened alumina
Shi 2012 [76]	Hardness up to 26.5 GPa, lowest friction against UHMWPE, and platelet adhesion linked to surface energy; the SiNO film is a candidate antithrombogenic coating on CoCrMo
Esmaeili 2024 [43]	The coating made at 40 V was best; cell survival 92% versus 86% on uncoated Ti6Al4V; hardness 85 versus 124 and elastic modulus 1.14 versus 3.55 GPa in nanoindentation, and lower abrasion in nano-scratch testing
Boschetto 2021 [42]	Significant increase in osteogenic and antibacterial properties compared with pristine PEEK; the effect is chemical and not only morphological

3.5. Results of Syntheses

Because models, substrates, coating compositions and outcomes differed, no quantitative synthesis was attempted, and the results are summarized narratively by outcome.

3.5.1. Mechanical and Tribological Performance

Hardness and elastic modulus were reported for coatings by HiPIMS and PECVD, with some coatings in one HiPIMS series approaching the hardness of sintered silicon nitride (about 18 GPa) [27] and a PECVD coating reaching 13.19 GPa [47]. Wear resistance depended on coating quality: in one series of sputtered coatings on cobalt–chromium, all but one silicon nitride coating had a wear resistance similar to bulk silicon nitride and higher than cobalt–chromium in ball-on-disc tests in serum [27], which applies only to the compositions and conditions of that series (ball-on-disc tests in serum after 1000 and 10,000 revolutions) and does not show equivalent wear performance across silicon nitride coating systems or long-term performance. Low-hardness coatings failed earlier by chemical reaction and dissolution accelerated by contact [46], and coated femoral heads gave somewhat lower polyethylene wear at 32 mm and somewhat higher wear at 36 mm than uncoated heads [28]. A graphene/silicon nitride/graphene multilayer on Ti6Al4V reached a hardness of 23 GPa and reduced the corrosion current by three orders of magnitude in an acidic fluoride medium, but it became porous and soft after corrosion [54]. Hardness and modulus scaled with the nitrogen content of HiPIMS coatings (12 to 26 GPa and 173 to 293 GPa) [51]. Adhesion improved when the chromium nitride interlayer was deposited at a low bias voltage and wear rates were lower than or similar to cobalt–chromium [50]. A coating on modular hip tapers did not change micromotion [48]. Adhesion of a PECVD coating needed optimization [47]. Silicon nitride particles embedded in carbon-fiber PEKK plates did not change flexural modulus, strength or crack behavior [49]. Laser treatment of a cathodic arc silicon nitride coating on a titanium alloy raised microhardness more than 11-fold and reduced residual stress [64], multilayer SiN:H and SiCN:H coatings on Ti6Al7Nb gave a lower friction coefficient than the bare alloy [66], and a PEEK/silicon nitride composite coating reached a modulus of 45 to 65 GPa with better wear and corrosion resistance than uncoated titanium alloy [65]. Silicon nitride coatings on cobalt–chromium made by physical vapor deposition reached hardness and friction against Si₃N₄ similar to bulk Si₃N₄ with low wear, but adhesion to the substrate remained to be improved [55]. A silicon nitride/tricalcium phosphate coating laser-cladded on Ti6Al4V had a higher microhardness than the bulk material [57], and among buffer layers under an apatite film on Ti6Al4V, silicon nitride gave the most adherent, compact and plastic film, ahead of SiC and TiN [56]. A study of HiPIMS process parameters found that compressive residual stress of up to 2.1 GPa could be reduced to 0.2 GPa by tailoring the process, and adhesion on cobalt–chromium–molybdenum disks was assessed by the Rockwell C test [41]. An electrophoretically deposited Si₃N₄/nanodiamond coating on Ti6Al4V raised hardness and elastic modulus in nanoindentation and lowered the abrasion rate in nano-scratch testing [43]. A sputtered amorphous SiN_x film made at a nitrogen flow of 20 sccm had the highest hardness (11.31 GPa) and the best corrosion and corrosion-assisted wear resistance, including in fluoridated artificial saliva, whereas films made at higher nitrogen flows were looser and oxygen-rich and performed worse [71]; anodized Ti-6Al-4V with electrophoretically deposited Si₃N₄ nanoparticles was assessed for hardness, elastic modulus, friction and corrosion [72].

3.5.2. Dissolution and Ion Release

Dense coatings were more stable in serum, and less dense coatings dissolved within one to three weeks [28]. Over 60 days in simulated body fluid, silicon nitride coatings dissolved at 0.2 to 1.4 nm per day, more slowly with higher nitrogen content and at a rate similar to or lower than that of cobalt–chromium–molybdenum, and they reduced metal ion release from the substrate [52]. Wear debris from a silicon nitride coating was round, with a median size of 40 nm, and formed agglomerates up to 3.2 µm, and model particles dissolved in simulated body fluid with a time constant of about 10 days [53]. Adding chromium reduced the release of silicon, chromium and niobium ions [59], and adding iron and carbon reduced ion release compared with a silicon nitride reference [60]. A chromium nitride interlayer reduced delamination in scratch testing [59].

The tests used for coatings address different failure modes, and none alone shows long-term retention (Table 7). A single scratch test measures adhesion under one monotonic load, and fatigue or cyclic adhesion, which is closer to service, was not reported in the studies assessed. Immersion in serum or simulated body fluid measures chemical dissolution, wear tests against polyethylene or silicon nitride measure wear-through under sliding, tribocorrosion combines the two, and micromotion tests at a modular taper address interface motion and not coating retention. Where dissolution was tested, it was in flat discs for weeks, and it was not tested whether the results hold for hip heads. A scratch load of 20 N [58] shows that a coating can resist one scratch, but not that it will remain attached for years, and the fate of detached or dissolved coating debris was characterized in only one study [53]. Thickness, density and oxygen content differed between studies, so results cannot be ranked across studies, and a coating should not be called stable on the strength of a single scratch threshold.

Table 7. Bench tests of coating durability in the included studies and the failure mode each addresses.

Study	Test and Outcomes	Model and Conditions; Duration	Thickness	Main Result
Wang 2026 [71]	Composition (XPS, FTIR); hardness (nanoindentation); adhesion (nano-scratch); corrosion; corrosion-assisted wear (reciprocating sliding); hydrophilicity; ALP activity; cell migration	mBMSCs (osteogenic differentiation, n = 3); L929 fibroblasts (migration, triplicate, 0 and 24 h); electrochemistry in 3.5 wt% NaCl and fluoridated artificial saliva; ALP staining on day 7; migration at 24 h	About 500 nm (FIB-SEM cross-sections)	SiN-20 gave a dense Si-N amorphous structure, hardness 11.31 GPa, the best corrosion and corrosion-assisted wear resistance and promoted mBMSC osteogenic differentiation and L929 migration; higher N2 flow gave a loose oxygen-rich network and poorer performance
Heydarian 2025 [72]	Roughness; hardness and elastic modulus; friction; corrosion (OCP, EIS); FESEM, TEM, EDS, XRD; antibacterial activity; cell viability; hydroxyapatite formation in SBF	S. aureus (antibacterial assay); cytotoxicity (cell type not stated in the text assessed); SBF immersion (duration not stated); not stated	Not stated	4 h anodizing gave the highest roughness and bone-like hardness and modulus; Si3N4 association improved antibacterial activity and cell viability and promoted hydroxyapatite formation in SBF (the 86% to 92% MG63 survival figure in the introduction belongs to an earlier study)
Guo 2023 [73]	Morphology (SEM, EDS, XRD); hydrophilicity and surface energy; Si ion release in PBS (1 to 28 d); corrosion in SBF (n = 5); cell proliferation (CCK-8), ALP and alizarin red; bactericidal rate	Bone marrow mesenchymal stem cells from male Sprague-Dawley rats (1 month old; CCK-8 on days 1, 3 and 5; ALP); S. aureus and E. coli, 1 × 106 CFU/mL, 1 day; Si release 1 to 28 d; bacteria 1 day	Not stated	Si3N4 incorporation changed nano-topography, hydrophilicity and surface energy and improved cell proliferation and osteoblastic differentiation and bactericidal rate with increasing Si3N4 content (S. aureus: pTi 0%, MTi 5.9 +/− 3.2%, MSN1 26.4 +/− 5.6%; MSN3 highest)
Esmaeili 2024 [43]	SEM with EDS, XRD, nanoindentation (hardness, modulus), nano-scratch, cell survival and adhesion	MG63 osteoblast-like cells (survival and adhesion, ex vivo); the coating step was repeated three times per sample; not stated	Thickness not stated	The 40 V coating was best; cell survival 92% on the coated surface versus 86% on uncoated Ti6Al4V; hardness 85 versus 124 and elastic modulus 1.14 versus 3.55 GPa (nanoindentation, 5 mN) and lower abrasion in nano-scratch testing
Moskalewicz 2022 [74]	Microstructure; roughness (profilometry, AFM); adhesion; wettability; corrosion in Ringer solution; apatite formation in SBF; cell viability; antibacterial activity	MG-63 osteosarcoma cells (Alamar Blue); E. coli (antibacterial); each coating type analysed in triplicate; SBF soaking 3 weeks; biofilm grown 24 h; Alamar Blue incubated 5 h	Si3N4-containing coating 6.7 to 9.3 µm; CuO-containing coating 2.4 to 3.5 µm	Coatings with Si3N4 had outstanding adhesion, were hydrophilic, protected the alloy from corrosion and formed hydroxyapatite in SBF; only the CuO-containing coatings were clearly antibacterial against E. coli; combining GtO with CuO and Si3N4 reduced MG-63 viability (more with CuO)
Shi 2012 [76]	XPS; AFM; contact angle and surface energy; nanoindentation; pin-on-disk against UHMWPE in fetal bovine serum; platelet adhesion	Platelet adhesion (n = 6 disks per sample); platelet adhesion time not stated	Measured by step profiler; value not stated (the 50 nm indentation depth was one tenth of the film thickness)	Hardness up to 26.5 GPa and modulus up to 235.6 GPa; films gave the lowest friction against UHMWPE; platelet adhesion related to the polar to dispersion ratio of surface energy; the SiNO film is a candidate antithrombogenic coating for CoCrMo/UHMWPE implants
Correa Filho 2020 [46]	XPS composition; roughness; hardness and modulus (nanoindentation); scratch adhesion; wear	Coating variants on discs; multidirectional wear against UHMWPE pins; 2 million cycles; 25% bovine serum, 37 °C, contact pressure about 2.1 MPa	3.7 to 8.8 µm	Low-hardness coatings failed earlier by chemical reaction and dissolution accelerated by tribological contact; high-purity, high-density coatings (low oxygen, high nitrogen) were desirable; no definite correlation between composition and wear
Correa Filho 2019 [28]	XPS composition; roughness; scratch critical load; friction coefficient; UHMWPE specific wear rate; FIB cross-sections	4 disc coating variants (immersion, scratch n = 3); 3 coated head variants and uncoated (wear, 3 repetitions per head); immersion 0, 1, 3, 6 weeks; wear 10,000 cycles	4.3 to 6.0 µm	One-fold rotation gave denser, more stable coatings; three-fold coatings dissolved by 1 to 3 weeks; coated 32 mm heads gave somewhat lower polyethylene wear and 36 mm somewhat higher; best resistance with SiCN top layer and process heating
Filho 2019 [50]	Morphology; N/Si ratio (XPS); roughness; adhesion (Rockwell); specific wear rate	Roughness three measurements per sample; Rockwell indentation at three locations; reciprocating wear against Si3N4 balls; 10,000 cycles at 37 °C in 25 vol.% fetal bovine serum	SiNx top layer about 4.3 to 4.4 µm; CrN interlayer 650 to 810 nm	Roughness 16 to 33 nm; adhesion improved with a low 100 V bias voltage for the CrN interlayer; specific wear rates lower than or similar to CoCrMo
Skjöldebrand 2017 [51]	Composition (XPS); morphology and microstructure (SEM, TEM); roughness; hardness and modulus (nanoindentation)	Process variants; one process with five samples; not applicable	About 7.1 to 7.6 µm	Coatings homogeneous and amorphous; roughness similar to uncoated CoCrMo (p = 0.25); hardness 12 ± 1 to 26 ± 2 GPa and modulus 173 ± 8 to 293 ± 18 GPa, rising with nitrogen content
Pettersson 2016a [52]	Dissolution rate (electrochemical, ICP, vertical scanning interferometry); surface chemistry (XPS); metal ion release	Three coating compositions versus CoCrMo; two coated samples per time point; 60 days in simulated body fluid with FBS and PBS; 60 days in simulated body fluid with FBS and PBS	7.0 to 7.5 µm	Dissolution rate 0.2 to 1.4 nm/day, lower with higher N/Si ratio; similar to or lower than CoCrMo (0.7 to 1.2 nm/day); highest nitrogen coating mainly Si-N bonds; coatings reduced metal ion release from the substrate
Pettersson 2016b [53]	Debris morphology and size (SEM, EDS); particle dissolution rate (ICP)	Ball-on-disc, 10 mm stroke, 3 Hz, 2 h, three replicates; model SiNx particles used for dissolution because too little debris was produced; 2 h wear test; dissolution measured over time	8 ± 0.2 µm	SiNx debris round, median size 40 nm (10 to 500 nm), forming agglomerates up to 3.2 µm; model particles dissolved in simulated body fluid with c(t) = 39.45 [1 − exp(−1.11 × 10−6 t)] mg/L
Pettersson 2013 [27]	Hardness and modulus (nanoindentation); micro-scratch adhesion and cohesion; ball-on-disc wear	Coating variants; ball-on-disc in serum; 1000 and 10,000 revolutions	Not stated in the text assessed	Most SixNy coatings reached hardness near 18 GPa and modulus near 200 GPa; all but one SixNy coating had wear resistance similar to bulk silicon nitride and higher than CoCr; SixCyNz coatings less wear resistant
Olofsson 2012 [55]	Surface appearance, chemical and phase composition, structure, hardness, coefficient of friction, wear rate	Several coating variants; nanoindentation (30 indents of 50 nm depth per coating); ball-on-disc wear tests against Si3N4; 1000 and 10,000 revolutions	About 0.6 to 1.4 µm depending on the variant (SEM cross-sections)	Coatings amorphous or nanocrystalline with hardness and coefficient of friction against Si3N4 similar to bulk Si3N4; low wear rates; adhesion to the substrates remains to be improved
Schmidt 2017 [41]	Residual stress (wafer curvature by XRD); composition (XPS); morphology (SEM); hardness (nanoindentation); adhesion (Rockwell C)	Process variants; Rockwell C adhesion test on CoCrMo disks; not applicable	2 µm on Si; 5 µm on CoCrMo	Dense, hard SiNx coatings; compressive residual stress of up to 2.1 GPa reduced to 0.2 GPa by tailoring process parameters; excellent adhesion on CoCrMo by the Rockwell C test (ISO 0, HF1)
Zhou 2024 [47]	Hardness; elastic modulus; roughness; adhesion	Coating variants on CoCr; not applicable	About 3 µm (3 h deposition)	Highest hardness (13.19 ± 1.29 GPa) and modulus (132.76 ± 9.32 GPa) at N/Si 1.42 (NH3 30 sccm); roughness adequate; adhesion needs optimization before wear can be evaluated
Haschke 2020 [48]	Micromotion; pull-off force; surface microscopy	Each head and stem combination n = 4; sinusoidal loading of three daily activities; six eddy-current sensors; not applicable	Equal for both manufacturers (value not stated in the text assessed)	Micromotion up to 8.4 ± 0.8 µm; Ti6Al4V stems showed more micromotion than CoCr; coating had no influence; pull-off forces did not differ; coating not damaged in contact areas
Kalisz 2015 [54]	Raman, SEM and AFM structure; hardness and Young’s modulus before and after corrosion; polarization curves (icorr, Ecorr)	Nanoindentation, five indentations per data point; each electrochemical measurement repeated three times (most representative shown); 1 h open-circuit exposure before polarization; corrosion in 0.5 M NaCl, 2 g/L KF, pH 2	SiN 200 nm	Hardness 23.0 GPa (hybrid) vs. 22.4 GPa (SiN) vs. 5.94 GPa (alloy); icorr 3.3E-08 (hybrid), 5.8E-08 (SiN) vs. 6.1 × 10−5 A/cm2 (alloy); after corrosion the hybrid became porous (hardness 0.62 GPa in porous regions) and the SiN degraded; SiN adhered poorly to graphene
Kaczmarek 2024 [66]	EDS and FTIR chemistry; hardness; Young’s modulus; friction coefficient; specific wear rate; wettability; Alamar Blue viability; flow cytometry; ALP activity	MG-63 cells; ALP measured in 4 replicates; experiments repeated at least twice; 7 days of cell culture	2.48 µm (SiN:H); 2.68 µm (SiCN:H)	Layers were biocompatible (more than 70% cell survival); SiCN:H-based layers boosted ALP activity; low friction coefficient compared with the bare alloy (0.32)
Echeverri 2023 [60]	Roughness; ion release (Si, Fe, Co); thickness reduction (dissolution); microglial viability	Compositional gradients; C8-B4 microglia in 2D extracts and 3D collagen hydrogel; 14 days leaching	SiFeCN about 427 to 524 nm; SiN reference 333 nm (FIB cross-sections)	Fe and C reduced ion release compared with the SiN reference; coating extracts were compatible with microglial viability
Issac 2023 [64]	Microhardness; residual stress; friction and wear; wear mechanisms; cell viability	MTT cell viability assay (72 h; cell line not stated in the text assessed); microhardness repeated 5 times; wear tests with wear map; MTT after 72 h	2 to 3 µm (4 h deposition)	Laser treatment raised microhardness more than 11 times versus the alloy, reduced residual stress and improved wear resistance; cell viability assessed by MTT (result not quantified in the text assessed)
Skjöldebrand 2022 [59]	Adhesion (scratch); roughness; hardness and modulus; dissolution and ion release; fibroblast viability	Gradient coatings; L929 mouse fibroblasts (indirect cytotoxicity, MTT, triplicate wells); viability on days 1, 2 and 3 after exposure to extracts; 14 days leaching	Not stated	Interlayer reduced delamination; composition did not affect adhesion, roughness or mechanical properties; Co release low; higher Cr (4.3 to 6.0 at%) gave lower Si, Cr and Nb release and better fibroblast viability
Zanocco 2020 [58]	Scratch resistance (ASTM C1624-05; 20 N threshold); osteoconductivity; bacterial adhesion	SaOS-2 cells; S. epidermidis; E. coli (luciferase) on shells; not stated	15 ± 5 µm	Coating met the 20 N scratch threshold; antibacterial and osteogenic behavior comparable to bulk Si3N4; resisted S. epidermidis and E. coli adhesion on acetabular shells
Xiong 2025 [65]	Young’s modulus; affinity to SBF; wear and friction; corrosion; cell adhesion, proliferation, ALP activity and differentiation	Mean ± SD for n = 3; five parallel samples per group for cell experiments; MC3T3-E1 cells; cells: 1 and 3 days	SiNx layer about 1 to 2 µm on a PEEK layer of 20 to 30 µm (SEM)	Young’s modulus 45 to 65 GPa; better wear and corrosion resistance; greater affinity to SBF; improved cell adhesion, proliferation and differentiation than uncoated TC4 and PEEK-only coatings
Xu 2016 [57]	Surface morphology, cross-sectional microstructure, microhardness, bioactivity in simulated body fluid	Coating groups A, B and C; Vickers microhardness at 1 kg for 15 s; SBF soaking; SBF soaking for 14 and 28 days	Not stated (abstract only)	Compact bonding interface between coating and substrate with appropriate laser parameters; dense coatings; microhardness significantly higher than bulk; good biological activity (apatite formation) when TCP was added to the composite
Delgado-López 2013 [56]	Chemical–physical features, compactness, plasticity and adhesion of the apatite film; surface roughness	Substrates with different buffer layers and roughness; plaques 6 × 3 mm and cylinders 3.5 × 14.2 mm; number per group not stated; not applicable	Apatite film about 2 µm; buffer thickness not stated	Best plasticity, compactness and adherence of the apatite layer grown on the Si3N4 buffer, followed by SiC and TiN; adhesion related to roughness of the buffered substrates

3.5.3. Biological Response In Vitro

Twenty-seven studies tested cell responses. Counted descriptively, 21 reported more favorable adhesion, proliferation, osteogenic or angiogenic response, osteoconductivity or survival on silicon-nitride-based coatings than on an uncoated or alternative comparator [40,42,43,61,62,63,65,66,67,68,70,71,72,73,75,77,78,79,80,81,82]; one reported osteogenic behavior comparable to bulk silicon nitride [58], three compared coating variants or reported compatibility or no difference [59,60,69], one did not quantify its result [64], and one reported reduced viability [74]. The 21 favorable publications come from nine research groups, and the counts take no account of sample size, effect magnitude, precision, multiple outcomes or selective reporting, so they should not be read as an overall effect. A study of coating extracts reported compatibility with microglial viability [60] and a study of fibroblasts reported better viability at optimal chromium content [59]. A laser-cladded silicon nitride coating on zirconia produced mineralized bone tissue of good quality, but cell proliferation was similar to uncoated zirconia and lower than on stoichiometric silicon nitride [69]. A silicon nitride/tricalcium phosphate coating on Ti6Al4V showed good bioactivity in simulated body fluid when tricalcium phosphate was added [57]. The study with reduced viability used a graphite oxide, hydroxyapatite and alginate coating on a titanium alloy, in which adding Si₃N₄ or CuO nanoparticles lowered the viability of MG-63 cells, more so with CuO [74]. In blood compatibility tests of fluorine- and nitrogen-doped silicon nitride films on Co-Cr-Mo, platelet adhesion was related to the polar-to-dispersion ratio of the surface energy, and the SiNO film was proposed as a candidate antithrombogenic coating [76]. In the titanium mesh study, a silicon nitride nanofilm shifted macrophages from the pro-inflammatory M1 to the pro-regenerative M2 phenotype and enhanced osteogenic differentiation in co-culture [82].

3.5.4. Antibacterial Activity

Thirteen studies tested bacteria, and twelve reported antibacterial activity of silicon-nitride-based coatings (Table 8); the twelve come from five research groups, and the counts are descriptive. A laser-sintered coating resisted adhesion of S. epidermidis and E. coli on acetabular shells [58]. Laser-clad silicon nitride reduced bacterial counts of S. epidermidis on four substrates [68] and on polyethylene [40]. Sputtered, plasma-deposited or composite coatings showed antibacterial activity against S. aureus and E. coli [61,70,78,79,81]. Nitrogen-graded nanofilms on titanium mesh reduced S. aureus, E. coli and P. gingivalis counts [82]. A porous PEEK surface with embedded silicon nitride particles reduced the growth of S. epidermidis [42]. Silicon nitride nanoparticles on anodized or micro-arc-oxidized titanium improved antibacterial activity, with a bactericidal rate against S. aureus that rose with the Si₃N₄ content [72,73]. In a graphite oxide, hydroxyapatite and alginate coating on a titanium alloy, only the CuO-containing coatings showed significantly improved antibacterial properties against E. coli [74]. The studies differ in organism, inoculum, culture conditions, exposure time and endpoint (adhesion, viability or colony counts), and for several, the assay and endpoint were not stated in the text assessed (Table 8). Reduced bacterial adhesion in a short in vitro assay is not the same as prevention of biofilm or treatment of infection; only one study tested a coating in an infection model, in rats [81], and none used a clinically relevant multispecies biofilm. Eight of the twelve favorable studies came from two research groups (one linked to the manufacturer), so they are not independent replications. The findings therefore do not show effectiveness against peri-implantitis or prosthetic joint infection.

Table 8. Included studies that tested bacteria. Categories: (i) Si₃N₄ or SiN_x films; (ii) SiCN and other doped or graded films; (iii) particulate or composite coatings. “Not stated” means that the text assessed did not report the feature.

Study	Coating (Category)	Organism	Endpoint and Exposure	Comparator	Result
Tani 2024 [61]	SiN film, 200 nm (i)	S. aureus	Antibacterial activity; assay and time not stated	Uncoated titanium	Significant antibacterial activity
Huang 2024 [79]	Sputtered Si3N4 (i)	S. aureus; E. coli	Antibacterial activity; assay not stated	Titanium; sandblasted acid-etched titanium	Best antibacterial activity
Hu 2021 [81]	Si3N4/Ta microparticles on PEKK (iii)	S. aureus; E. coli (n = 3); rat femoral infection model	Antibacterial activity in vitro; infection in vivo	Sulfonated PEKK; Si3N4-only; Ta-only	Antibacterial activity and an anti-infective effect in vivo
Zanocco 2020 [58]	Laser-sintered Si3N4 (iii)	S. epidermidis; E. coli (luciferase)	Bacterial adhesion on acetabular shells; time not stated	Uncoated titanium; bulk Si3N4	Resisted adhesion; comparable to bulk Si3N4
Marin 2020 [68]	Laser-clad Si3N4 on four substrates (iii)	S. epidermidis	Bacterial counts and optical density; 24 and 48 h	Uncoated substrates	Uncoated Ti6Al4V had a higher bacterial count
Wu 2019 [78]	PEKK/SN composite (iii)	E. coli ATCC 25922; S. aureus ATCC 25923	Antibacterial activity; assay not stated	PKS without Si3N4; PEKK	The 10 wt% composite was antibacterial
Yuan 2026 [82]	SiNx nanofilms, three nitrogen contents (i)	S. aureus; E. coli; P. gingivalis, 1 × 106 CFU/mL	Plate colony counts; 12 h co-culture	Bare titanium mesh	Antibacterial activity of all three films was better than titanium
Boschetto 2021 [42]	Porous PEEK surface with Si3N4 particles (iii)	S. epidermidis (ATCC 14990)	Optical density and CFU counts; 12, 24 and 48 h	Smooth PEEK; NaCl-treated PEEK	Significantly greater antibacterial effect than pristine PEEK
Heydarian 2025 [72]	Si3N4 nanoparticles by EPD on anodized Ti-6Al-4V (iii)	S. aureus	Antibacterial activity; assay not stated	Anodized surface without Si3N4	Improved antibacterial activity
Guo 2023 [73]	Si3N4 in a micro-arc oxidation coating on titanium (iii)	S. aureus; E. coli, 1 × 106 CFU/mL	Bactericidal rate; 1 day	Pure Ti; MAO coating without Si3N4	S. aureus bactericidal rate 26.4 ± 5.6% (MSN1) versus 5.9 ± 3.2% (MAO) and 0% (pure Ti); highest with the most Si3N4
Moskalewicz 2022 [74]	Si3N4 nanoparticles in a GtO/HA/alginate coating by EPD on Ti-13Nb-13Zr (iii)	E. coli; MG-63 cells	Antibacterial activity against E. coli (biofilm grown for 24 h)	CuO-containing coating; coating without nanoparticles	Only the CuO-containing coatings significantly improved antibacterial properties against E. coli; no antibacterial effect is reported for the Si3N4 coating
Xu 2019 [70]	Amorphous Si3N4 on PEEK (i)	E. coli ATCC 25922; S. aureus ATCC 25923; 1 × 106 CFU/mL	Antibacterial activity; 24 h	Uncoated PEEK	Greater antibacterial activity
Zanocco 2020 [40]	Laser-clad Si3N4 on polyethylene (iii)	S. epidermidis, 1 × 107 bacteria	Counts of living bacteria; 12, 24 and 48 h	Uncoated polyethylene	Significantly fewer living bacteria

3.5.5. In Vivo Bone Response

Seven studies included animals (Table 9). Six reported greater bone formation or osseointegration with silicon-nitride-based coatings than with uncoated or porous polymer or titanium controls [77,78,79,80,81,82], at follow-up times of 4 to 12 weeks, and one also reported an anti-infective effect in rats [81]; the six come from two research groups. The only direct comparison of a silicon nitride coating with hydroxyapatite was made in a rabbit femoral model [83]. Carbon-fiber-reinforced PEEK cylinders were coated with hydroxyapatite (HA), with hydroxyapatite plus vancomycin and polyvinylpyrrolidone (PVP) (HA-VP), or with Si₃N₄ (SiN); ten implants per group were placed in the femoral diaphysis and retrieved at 12 weeks. Periprosthetic fractures led to exclusion of one, one and two implants, respectively, so nine, nine and eight implants were scanned by micro-computed tomography and eight, eight and seven were tested in pull-out. Pull-out strength was lower for SiN than for HA (p = 0.044) and than for HA-VP (p = 0.041), and bone volume around the implant was lower for SiN than for HA (p < 0.001), whereas HA and HA-VP did not differ (p = 0.861). The SiN group received no antibiotic or antithrombotic agent, so the comparison of SiN with plain HA is not confounded by those agents, but the two coatings differ in fabrication method. This is one small-animal model with one substrate, and its result should neither be dismissed as a single negative finding nor generalized beyond that model.

Table 9. Animal studies of silicon nitride coatings. “Not stated” means that the text assessed did not report the feature.

Study	Species and Sample	Coating and Comparator	Follow-Up	Outcomes	Main Result
Dai 2019 [77]	Rabbit; 30 males in 3 groups; femoral condyle defects	Nano-SiN on porous PEEK versus porous and dense PEEK	4 and 8 weeks	Micro-CT; histology; push-out	More new vessels and bone and higher push-out loads than porous PEEK
Yuan 2026 [82]	Rat; 24 males, 5 mm calvarial defect, n = 6 per group (random number table)	SiNx nanofilms (three nitrogen contents) on titanium mesh versus bare titanium mesh	8 weeks	Micro-CT BV/TV and histology (blinded evaluator)	The SiN65 film promoted bone regeneration
Wu 2019 [78]	Rabbit; 36 (18 histology, 18 biomechanics)	PEKK/SN composite (0, 5, 10 wt%) versus PEKK	Not stated	Histology; push-out	The 10 wt% composite promoted osteogenesis and osseointegration
Huang 2024 [79]	Rat; 15 per time point; three implant subgroups	Sputtered Si3N4 versus titanium and sandblasted acid-etched titanium	4 and 8 weeks	Micro-CT BV/TV; histology	More new bone and osseointegration
Dai 2020 [80]	Rat; 20 in 2 groups; skull defects	Nano-SiN on PEEK versus PEEK	4 and 8 weeks	Micro-CT BV/TV and BMD; histology	More bone regeneration and osseointegration
Hu 2021 [81]	Rat; 30 males (femoral nail infection model) plus an osseointegration cohort	Si3N4/Ta microparticles on PEKK versus sulfonated PEKK, Si3N4-only and Ta-only	Up to 12 weeks	Infection; micro-CT BV/TV and BMD; histology; bone implant contact	Anti-infective effect; more bone and gap-free osseointegration at week 12
Enami 2022 [83]	Rabbit; 10 implants per group (HA, HA-VP, SiN), femoral diaphysis	Si3N4 on CFRP versus HA and HA with vancomycin and PVP	12 weeks	Pull-out; micro-CT; histology	SiN lower than HA for pull-out (p = 0.044) and bone volume (p < 0.001)

3.5.6. Summary by Coating Category

Table 10 summarizes the cell, antibacterial and animal results for each coating category, with the number of research groups in parentheses. The counts are of publications and are descriptive; they are not pooled or comparative effects, and publications from one group are not independent.

Table 10. Descriptive summary of results by coating category (counts of publications, with research groups in parentheses; not pooled effects).

Coating Category	Publications (Research Groups)	Cell Tests: Favorable of Tested (Groups)	Antibacterial Tests: Favorable of Tested (Groups)	Animal Tests: Favorable of Tested (Groups)	Surface Chemistry and Failure Issues Reported
Films: Si3N4 or SiNx	18 (7)	6 of 7 (3)	4 of 4 (3)	2 of 2 (1)	Amorphous or sub-stoichiometric films with varying nitrogen and oxygen content; limited adhesion, dissolution of the film and of wear debris
SiCN, doped or graded films	7 (5)	2 of 4 (2); the other two compared coating variants or reported compatibility	0 tested	0 tested	Doping (Cr, Nb, Fe, C) and interlayers lowered ion release or delamination; oxygen-containing films
Particulate or composite coatings	19 (10)	13 of 16 (6)	8 of 9 (3)	4 of 5 (1)	Silicon nitride particles in PEEK, PEKK, polyethylene, hydroxyapatite or oxide layers; laser-clad coatings with partial nitrogen loss; behavior depends on the matrix; one composite lowered cell viability; hydroxyapatite outperformed silicon nitride in the one direct comparison
All categories	44 (19)	21 of 27 (9)	12 of 13 (5)	6 of 7 (2)	Groups can contribute to more than one category, so group counts do not add up

3.6. Reporting Biases

No quantitative assessment of publication bias was possible. All animal and in vitro studies that reported biological outcomes reported favorable findings for the coating, except the comparison with hydroxyapatite and the reduced cell viability seen with one composite coating [74], and nine included studies had manufacturer-affiliated authors [40,41,42,46,49,58,68,69,75]. One eligible report could not be retrieved, so selective retrieval could influence these findings. One included study shares its group, coating and cell model with another [61,63], and possible overlap of data cannot be excluded.

4. Discussion

4.1. Principal Findings

In the 44 publications assessed (19 research groups), most laboratory studies of silicon-nitride-based coatings on dental and orthopedic implant substrates reported favorable cell, antibacterial and corrosion-related behavior; these counts are descriptive and take no account of effect size or of the independence of publications from the same group. In bench tests, selected sputtered silicon nitride coatings on cobalt–chromium approached bulk silicon nitride in hardness and wear under the compositions and conditions studied (ball-on-disc tests in serum) [27,55]. This does not show equivalent wear performance across silicon nitride coating systems, nor long-term performance in the body. Dissolution and adhesion, which are failure mechanisms specific to coatings, remained engineering concerns and were addressed by compositional changes and interlayers [59,60]. In animals, six studies from two research groups reported more bone than with uncoated controls, but the one comparison with hydroxyapatite favored hydroxyapatite [83].

4.2. Interpretation

These findings agree with the narrative evidence for the bulk material but do not show durability under load in the body. Test durations were short, wear tests were run against polyethylene for up to 2 million cycles in serum [46], and animal follow-up did not exceed 12 weeks. Indirectness for clinical stability is therefore high.

The three categories were not treated as one intervention. Films differ from each other in nitrogen and oxygen content and are limited mainly by adhesion and dissolution of the exposed film. Doped or graded films were designed to lower ion release and delamination through added elements and interlayers [59,60]. Particulate or composite coatings expose silicon nitride particles in a matrix of polymer, ceramic or oxide, so their behavior depends on the matrix and on particle release, and one composite lowered cell viability [74]. Results from these categories were therefore kept separate and were not pooled (Table 10).

Other systematic reviews of coated implants give context. A review of calcium-phosphate-coated titanium implants in animals found no significant difference in bone-to-implant contact from conventionally etched surfaces [29], whereas a review of strontium-coated titanium in osteoporotic animals reported about 19% more bone-to-implant contact but with very high heterogeneity (I² 97%) and a funnel plot suggesting publication bias [30]. In joint replacement, a systematic review and meta-analysis of titanium nitride and titanium niobium nitride coated knee implants found survival, complication rates and clinical scores comparable with uncoated cobalt–chromium implants, so a laboratory advantage had not translated into a clinical one [31]. Because several silicon nitride coatings were applied to PEEK or PEKK, reviews of PEEK are also relevant: unmodified PEEK was less bioactive than titanium in vitro and in animals, with hydroxyapatite-coated PEEK reaching bone-to-implant contact comparable to coated titanium [32], and surface modification of PEEK appeared to improve cell adhesion, proliferation and osteogenic properties [33]. Together these reviews suggest that favorable laboratory findings for a coating should be read with caution.

Beyond antibacterial effects, aqueous suspensions of silicon nitride powder inactivated SARS-CoV-2 [84,85,86], and the same line of work reported inactivation of influenza A virus, enterovirus and feline calicivirus [86]. Antifungal-type evidence is limited to a plant pathogen, the oomycete Plasmopara viticola [87]. None of the included studies tested viruses or fungi, all of these studies used powders or suspensions rather than coatings, and most came from groups linked to the manufacturer, so they cannot be taken as evidence for coated implants.

4.3. Strengths and Limitations

Strengths include a multi-database search, an explicit eligibility framework with an operational definition and classification of silicon nitride coatings, structured risk-of-bias tools, and a protocol registered on the Open Science Framework. The main limitations concern the search, the selection and the assessment. Search limitations are as follows. The strategy requires an implant-related term and may miss studies that describe a coating without one. MEDLINE was searched through Ovid because the standard interface was unavailable, so the Ovid strategy could not be reported line by line, although a PubMed reproduction returned the same 29 MEDLINE records. The broader PubMed strategy was not repeated in the other databases, and only 48 of the 275 records of the ProQuest search could be exported, and the other 227 were not screened individually. In some registries only the term silicon nitride was used, so a trial that does not use it would be missed. Forward citation searching was not performed, so recent papers that cite the included studies may have been missed, and three of the 44 included studies were not retrieved by the database searches. Limitations of selection and assessment are as follows. One of the 53 eligible reports (a 1997 report on stainless steel with ion-beam-assisted silicon nitride coatings; Supplementary File S1) could not be retrieved. The full-text eligibility, extraction and risk-of-bias judgments were first prepared with AI assistance before verification, risk of bias was rated by a single human reviewer, and the 148 records from the later searches were screened after an AI pre-screen. The preclinical scope was registered retrospectively. Publications from the same research groups may not be independent, the vote counts are descriptive, and the risk-of-bias labels are descriptive categories and not certainty ratings. Much of the evidence on silicon nitride, and several included studies, come from groups linked to the manufacturer. Findings may change if the report that could not be retrieved is obtained.

4.4. Implications

Seven priorities would help to establish the long-term stability of silicon nitride coatings. First, test conditions for adhesion under cyclic loading, tribocorrosion and dissolution in physiological media should be standardized and fully reported, with coating thickness and density stated. Second, silicon nitride coatings should be compared directly with established coatings such as hydroxyapatite in animal models lasting at least 12 weeks, with adequate sample sizes and reporting against the ARRIVE items, including blinding, randomization and sample-size reasoning. Third, antibacterial testing should use biofilm and in vivo infection models that separate anti-adhesion, bactericidal and anti-biofilm effects. Fourth, independent groups without commercial ties should replicate the findings. Fifth, coating-derived debris and its biological fate should be characterized. Sixth, any coated implants used clinically should undergo retrieval analysis. Seventh, any clinical study of a silicon-nitride-coated implant should be registered prospectively, so that the evidence gap can be tracked.

5. Conclusions

In the 44 preclinical studies assessed, silicon-nitride-based coatings showed favorable laboratory and short-term animal results, with unresolved questions about adhesion, dissolution and comparison with hydroxyapatite, and the findings differ between films, doped films and particulate or composite coatings. No published clinical study of a silicon-nitride-coated implant was identified in the bibliographic searches and no registered trial of a silicon-nitride-coated implant was identified in the seven trial registries or in WHO ICTRP. The evidence is preclinical, short-term, and at moderate risk of bias, and it does not establish the long-term stability of coated implants.