Whose Titanium Is It? Source Attribution and Causal Direction of Titanium in Peri-Implant Tissues: A Study Protocol
- Paper type
- Study Protocol
- Status
- Original Work · Working Paper
- Keywords:
- peri-implantitis · titanium particles · titanium dioxide · ICP-MS · source attribution · nested case-control · reverse causation
Contents
Abstract
Background. Titanium is repeatedly detected in peri-implant mucosa, plaque and cytology samples, and several cross-sectional studies associate higher levels with peri-implantitis. Total-titanium assays cannot distinguish implant-derived metal from pigment-grade titanium dioxide in food and oral-care products, or from sampling and instrument contamination. No longitudinal human study has tested whether titanium burden precedes bone loss or follows it.
Objective and hypothesis. We propose that a substantial share of the titanium signal in published case-control contrasts is exogenous or artifactual, and that implant-attributed titanium behaves as a marker of peri-implant inflammation rather than as an independent driver of bone loss. The protocol is designed so that either conclusion can be rejected.
Proposed approach. A prospective maintenance cohort of adults with titanium or titanium-zirconium implants in health or mucositis, with one randomly selected index implant per patient and a contralateral tooth site as within-patient control. Samples are taken under dietary and oral-care titanium dioxide control, with procedural blanks and non-metal instruments, and analyzed by tandem ICP-MS for total titanium, single-particle ICP-MS for particle number and size, and alloy tracers (aluminum and vanadium for Ti-6Al-4V; zirconium for TiZr). Tissue from incident cases is examined for phase by synchrotron absorption spectroscopy. A nested case-control analysis on stored serial samples tests whether baseline implant-attributed titanium predicts incident radiographic bone loss; cross-lagged models test temporal order. Sample-size arithmetic (checked in Python) gives 100 events, 125 cases and 500 controls to detect a hazard ratio of 1.4 per standard deviation, and 226 events to exclude a ratio above 1.25.
Expected contribution. A specified attribution model, pre-registered decision rules separating “driver”, “bystander” and “consequence”, and explicit power requirements.
Conclusions. The protocol is a design, not a result. A registered cross-sectional study already targets related tracer elements, so the contribution is limited to the longitudinal ordering and attribution framework.
1. Introduction
Whether titanium released from dental implants contributes to peri-implantitis is a question about measurement before it is a question about biology. A reader who sees that titanium is higher in plaque or tissue from diseased implants must decide whether the extra titanium came from the implant, from the diet, from toothpaste, or from the laboratory, and, separately, whether it was there before the disease or arrived with it. Existing literature answers neither question with the design it would require.
Systematic evidence agrees that titanium particles are a common finding around implants and that diseased sites tend to show more of them [1]. The largest within-patient tissue study found particles in essentially all peri-implant soft tissue, with differences between patients and implant systems but not between diseased and healthy sites in the same person [2]. Reviews describe an association between biocorrosion, particles and complications, while stating that evidence for a one-way causal relationship is insufficient and that higher titanium in lesions could be a consequence of biofilm and inflammation [3, 4]. A competing narrative treats particles as a key modifier of dysbiosis and immune signaling [5]. These positions are compatible with the same cross-sectional data, which is the problem.
Two methodological gaps sit upstream of the causal debate. First, titanium dioxide is a widespread exogenous source. In a pilot study that controlled intake of titanium dioxide-containing foods and toothpaste, titanium and zirconium were not quantifiable in mucosa around healthy implants, and in participants without implants a titanium-dioxide candy challenge changed the measured gingival titanium in most samples [6]. Second, even a correct total-titanium value says nothing about order in time. The only registered study found that addresses peri-implant metallic particles with a progression framing is described as cross-sectional [7].
This protocol addresses both gaps in one design. Section 2 appraises the evidence. Section 3 states the hypothesis, an attribution model and falsifiable predictions. Section 4 specifies the cohort, analytical chain, controls, sample size and analysis. Later sections cover competing explanations, implications, limitations and the work that must be done before submission. All appraisal is at abstract level; Embase, Scopus and Web of Science were not accessible.
2. Background and critical appraisal of the evidence
Origin of the concept
Titanium debris in peri-implant tissue has been described for decades, and the mechanisms proposed include wear during insertion, tribocorrosion at the implant-abutment connection, and electrochemical dissolution under biofilm and inflammatory conditions [1, 3, 8]. Mapping of titanium particles in peri-implant oral mucosa by laser ablation ICP-MS showed particles in tissue from both test and control sites, which already illustrates the attribution problem [9]. Laboratory work shows that titanium ions can form particles that activate interleukin-1 beta release from lipopolysaccharide-primed macrophages [10], and that mechanical damage to titanium oxide leaves only partial passivation under inflammatory conditions [11]. These mechanistic studies make a driver role plausible; they do not show that it operates at the concentrations and in the proportion found in patients.
Evidence for higher titanium at diseased sites
Submucosal plaque from 20 implants with peri-implantitis and 20 healthy implants (30 participants) showed higher mean titanium in the diseased group after normalization to bacterial DNA mass, analyzed by ICP-MS with generalized estimating equations [12]. A cytology study and a biopsy study point the same way: exfoliative cytology in a pilot [13], and biopsies from 13 peri-implantitis patients compared with 11 periodontitis patients, in whom the comparator tissue had no implant exposure at all [14]. A 2026 report combining clinical samples with an animal model also describes higher titanium at diseased sites [15]. A small microbiome study found titanium in 40% of sampled sites and reported that titanium level, more than disease status, tracked community composition [16].
Evidence against a simple association
In 21 patients, micro-PIXE showed that particles are common, that most lie within about 2 mm of the implant-tissue interface, that implant system but not time in function influenced density, and that occurrence did not differ between diseased and non-diseased sites within the same person [2]. In 39 peri-implantitis biopsies, particles were seen in every lesion but macrophages or giant cells engulfing them were not identified [17]. Peri-implantitis around ceramic implants showed immunohistological features similar to those around titanium implants in a small pilot, so titanium is not necessary for the lesion [18]. In dogs, zirconia implants lost less bone during ligature-induced disease, but not during later spontaneous progression [19]. A microbiome study found differences that were site-specific rather than patient-wide [20].
The attribution problem
Four observations frame the attribution problem.
- Exogenous titanium reaches oral tissue. Food-grade titanium dioxide crosses the buccal mucosa in pigs [21], titanium dioxide particles have been found in human tissues generally [22], and exfoliated oral mucosal cells have been proposed as bioindicators of systemic titanium exposure [23]. In the Geneva pilot, quantifiable titanium and zirconium disappeared around healthy implants when diet and toothpaste were controlled [6].
- Tooth sites carry titanium. In 16 participants with titanium implants and zirconia abutments, titanium particles were found at contralateral natural tooth sites as well as at implant sites, although counts were higher at the implant-abutment interface and in adjacent mucosa [24]. Mucosa of people without titanium implants has been reported to contain low titanium [3].
- Tissue from inflamed sites contains more than one titanium chemistry. Synchrotron micro-X-ray absorption near-edge spectroscopy identified predominantly metallic titanium and titanium dioxide in inflamed tissue near titanium implants, with local particle density as high as roughly 40 million per mm^3 [25]. Metallic titanium is hard to attribute to diet. Titanium dioxide, however, is also the composition of the implant’s own passive oxide film, so oxide chemistry alone does not exclude an implant source.
- Analytical agreement is imperfect. An inter-laboratory comparison of titanium in whole blood, with seven laboratories using ICP techniques and spikes of soluble titanium or titanium dioxide particles, was undertaken because accurate exposure analysis requires fit-for-purpose methodology and no certified blood reference materials for natural low titanium levels were available [26]. A 2026 review of detection methods describes heterogeneity in how submicron debris is measured [27], and another 2026 review covers analytical detection in biological matrices [28]. We did not locate dedicated methodological papers on titanium contamination in ICP-MS of oral samples; the controls specified below are therefore stated as validation requirements, not as established practice.
Analytical building blocks that exist
Single-particle ICP-MS has been applied to titanium dioxide particles in foodstuffs and E171 additive [29] and in human urine [30], and to titanium nanoparticles from implantoplasty debris in rat tissue after enzymatic digestion, in a study that also measured aluminum and vanadium [31]. Synchrotron speciation in peri-implant tissue exists [25], as does scanning electron microscopy with energy-dispersive spectroscopy using titanium-free microtome blades [17]. Each component has precedent; their combination in one diseased-versus-healthy or longitudinal design does not, as far as the sources searched show.
Temporality
The cross-sectional studies above cannot order events. Peri-implantitis bone loss is reported to follow a non-linear pattern with an increasing rate over time and to vary between subjects [32], so a cohort must follow patients long enough to capture onset and must treat onset time as interval-censored. Searches of PubMed, Europe PMC and ClinicalTrials.gov found no longitudinal human study relating baseline titanium to later bone loss. A randomized trial of abrasive versus non-abrasive cleaning reported only a non-significant trend toward higher plaque titanium at 8 weeks after abrasive treatment (N = 34) [33]; that trial concerns treatment-induced release, not prediction.
| Study | Design and size | Finding relevant here | Main bias or limitation | Level |
|---|---|---|---|---|
| Safioti 2017 [12] | Cross-sectional, 20 vs 20 implants, plaque ICP-MS | Higher Ti at peri-implantitis after plaque normalization | No exogenous TiO2 control; cross-sectional | C |
| Pettersson 2019 [14] | Biopsy ICP-MS, 13 vs 11 | Ti measured in peri-implantitis vs periodontitis tissue | Comparator had no implant | C |
| Dionigi 2025 [2] | 21 patients, micro-PIXE, within-patient | Particles common; no diseased vs healthy difference | Small; no phase or source analysis | B |
| Cionca 2023 [6] | Three-phase pilot, 41 participants | Ti and Zr not quantifiable once diet and toothpaste controlled | Healthy implants only; tiny subgroups | B |
| Nelson 2020 [25] | Synchrotron speciation, inflamed tissue | Metallic Ti and TiO2 both present | Diseased tissue only; no control | C |
| Tawse-Smith 2017 [24] | Single arm, 16 participants, tooth-site controls | Ti particles also at tooth sites | One prosthetic design; no disease contrast | C |
| Daubert 2018 [16] | Cross-sectional, 15 implants | Ti level tracked microbiome | Tiny; temporal order untestable | C |
| Mombelli 2018 [3] | Critical review | Direction of causality unproven | Narrative review | D |
Note. Evidence levels are operational categories used in this manuscript (A strong, B moderate, C preliminary, D mechanistic or theoretical, E speculative hypothesis), not GRADE ratings.
3. Hypothesis and theoretical framework
Formal statements
The framework separates a measurement hypothesis from a causal hypothesis, because the second cannot be tested until the first is resolved.
H1 (measurement). After control of exogenous titanium dioxide exposure and contamination, the implant-attributed titanium difference between implants that later develop bone loss and implants that do not is smaller than the unadjusted total-titanium difference reported in cross-sectional studies.
H2 (direction). Implant-attributed titanium does not predict incident bone loss after adjustment for time-varying inflammation, but rises after onset of inflammation or bone loss (the “bystander or consequence” hypothesis).
The competing “driver” hypothesis states the opposite for H2: baseline implant-attributed titanium predicts incident bone loss independent of inflammation, and the predictive path from titanium to bone is stronger than the reverse path. The protocol is built so that H2 and its competitor produce different data.
Assumptions
- Titanium at a sampled site is a sum of implant-derived, exogenous and artifactual contributions, and these are additive in mass.
- Exogenous titanium dioxide enters implant and contralateral tooth sites from the same oral exposure, with a site-specific retention factor that can be bounded.
- Implant-derived titanium carries at least one signature that exogenous titanium dioxide lacks: metallic oxidation state, or alloy tracer elements in alloy implants, or zirconium for titanium-zirconium implants.
- Contralateral tooth sites carry no implant-derived titanium, except from adjacent titanium-bearing restorations or orthodontic and prosthetic hardware, which are exclusion criteria.
- Radiographic bone loss beyond measurement error and early remodeling is a valid, if imperfect, proxy for disease progression.
Assumption 3 is weakest. The native oxide of an implant is titanium dioxide, and the synchrotron study found titanium dioxide as well as metallic titanium in inflamed tissue [25]. Particles that are pure oxide with no tracer are therefore ambiguous; the framework retains them as an “indeterminate” class instead of assigning them to either source.
Mass-balance model
For site s of patient i at time t, the measured titanium mass per sample is
M_(obs,ist)=M_(impl,ist)+M_(exo,ist)+M_(art,ist)
where M is in nanograms per sample and the subscripts denote implant-derived, exogenous and artifact contributions. Artifact mass is estimated from procedural, reagent and field blanks processed identically and is subtracted, with its variance carried forward. At the contralateral tooth site, M_(impl)≈0 by assumption 4, so
M_(exo,impl)=κ (M_(obs,tooth)−M_(art,tooth))
where κ is a dimensionless factor that scales tooth-site exogenous titanium to the implant-site matrix. It is not known. The analysis therefore reports results over a pre-specified range, for example κ from 0.5 to 2, and treats κ=1 as the primary scenario. The implant-attributed titanium is then
Δ_(ist)=M_(obs,impl)−M_(art,impl)−M_(exo,impl)
with negative values retained as measurement noise and not truncated at zero, which would bias the mean upward. Both site masses are normalized to a matrix quantity (bacterial DNA mass or total protein per sample), as plaque mass normalization was used in earlier plaque work [12]. The unit of the exposure variable in the cohort model is the standardized logarithm of Δ after adding a pre-specified offset equal to the limit of quantification, with censored values handled by censored-data likelihood methods.
Alloy tracer fraction
For Ti-6Al-4V implants the nominal composition is about 6 wt% aluminum and 4 wt% vanadium with titanium as the balance. Taking titanium as roughly 90 wt%, the nominal mass ratios are ρ_(Al)=6/90≈0.067 and ρ_V=4/90≈0.044. The manufacturer’s certificate should replace these nominal values. A tracer-based implant fraction for element e is
f_(e,ist)=(C_(e,site)−C_(e,ref))/(ρ_e C_(Ti,site))
where C is concentration in the sample digest and C_(e,ref) is the tracer background from the tooth site of the same patient. If all measured titanium came from the alloy and dissolved or fragmented without fractionation, f_e≈1; if all came from titanium dioxide pigment, f_e≈0. Vanadium is the preferred tracer because it is expected to have a lower environmental background than aluminum, an expectation that the pilot must confirm. In the rat implantoplasty model, aluminum and vanadium were detected in all tissues and, except for vanadium in brain, did not differ between control and experimental animals, so these tracers did not separate experimental from control animals [31]. For titanium-zirconium implants, zirconium plays the same role; in the Geneva pilot, zirconium was found only in participants with zirconia implants [6]. Commercially pure titanium implants have no alloy tracer, so these implants are stratified and attributed through the tooth contrast and speciation only.
Differential dissolution can distort f_e: ions of different elements need not leave the surface in alloy proportion, and particles and ions partition differently between plaque matrix and digest. The tracer fraction is accordingly interpreted as supporting evidence of origin, not as an exact apportionment, and single-particle co-detection of titanium with tracers is preferred where instrument capability allows.
Competing causal structures
Four causal structures are consistent with a cross-sectional association between titanium and disease:
- Driver: implant-attributed titanium raises the hazard of bone loss.
- Consequence: inflammation, acidic or fluoride-containing biofilm conditions and mechanical debridement raise titanium release, which was the direction proposed in the critical review [3] and in the original framing of plaque titanium as a corrosion product [12].
- Common cause: implant system, surface, abutment design, prosthetic wear or maintenance instrumentation increase both titanium and disease risk. Implant system influenced particle density in the within-patient tissue study [2], and abutment material and wear are plausible determinants of release [24].
- Artifact: the association arises from uncontrolled exogenous titanium dioxide or contamination that differs systematically between groups, for example through different diets, oral hygiene products or sampling instruments.
An additional link must be handled: a third variable may mediate. Titanium may alter community composition, as one small study suggested [16] and a later study of crevicular titanium and community structure reported [34]. A mediating role would still count as the driver structure for the purpose of this protocol, but the microbiome is measured so that it can be examined as a mediator.
Numbered, falsifiable predictions
P1. Under dietary and oral-care titanium dioxide control and with blanks, plaque and fluid titanium at healthy implants falls toward the limit of quantification. Refuted if the controlled median remains well above the blank-based limit and the tooth-site median is below it.
P2. Within patients, implant-site and tooth-site titanium are positively correlated (a shared exogenous component). Refuted if rank correlation is near zero in a sample large enough to estimate it.
P3. The implant-attributed difference Δ between later cases and controls is smaller than the difference in raw total titanium. Refuted if the two differences are similar and the attributed difference excludes zero.
P4 (driver). Baseline Δ predicts incident bone loss with a hazard ratio of about 1.4 per standard deviation after adjustment (the design target). Not supported if the lower 95% limit is at or below 1.0. A meaningful driver effect is excluded only if the upper one-sided limit is below 1.25 (Section 4.7).
P5 (temporal order). In cross-lagged models, the path from earlier titanium to later bone level is larger than the reverse path if titanium is a driver, and smaller if titanium is a consequence. Refuted if the two paths are indistinguishable with intervals narrow enough to exclude a difference.
P6 (onset-aligned trajectory). In incident cases aligned on estimated onset, titanium slope after onset exceeds titanium slope before onset if consequence. Refuted if slopes are equal and baseline titanium predicts onset.
P7 (tracer specificity). Tracer-positive particles occur at alloy-implant sites and not at tooth-site controls. Refuted if tracer elements appear in tooth-site samples at similar ratios, which would point to environmental sources or to cross-contamination.
Why a null result would be informative
A cohort result near a hazard ratio of 1.0 with a narrow interval would not prove that titanium is harmless in every setting. It would show that, at the exposure range occurring in maintenance patients under this attribution, a baseline titanium measurement carries no predictive information beyond clinical inflammation. That would shift the burden to experiments showing that a particular titanium exposure alters tissue behavior, and it would stop the use of unattributed titanium as a risk marker.
Boundary with related questions
Insertion-phase shedding, abutment-material tribocorrosion and treatment-induced release are separate questions with their own designs. Particles shed at insertion, for example, would be present at baseline and thus appear in this protocol as an exposure of unknown cause; the tracer and tooth-control strategy can show that it is implant-derived but not why. Abutment material and prosthetic design are recorded as covariates. A tissue study reported titanium together with iron in peri-implantitis lesions, a reminder that non-titanium metals can co-occur and that the tracer panel needs to include iron as a contamination indicator [35].
4. Proposed methodology
Design overview
The design is a prospective multicenter cohort of patients in supportive peri-implant maintenance, with biospecimens stored at each visit and analyzed in a nested case-control sample. A nested design is chosen because the assays are expensive and the samples can be stored, and because analysis of a sampled risk set retains most of the information of the full cohort when several controls are drawn per case [36]. A case-cohort design is an alternative that allows several outcomes to share one comparison subcohort [36]; it is not primary here because the cross-lagged analysis needs repeated measures on cases and matched controls, and the nested design supports this directly.
The cohort has three stages. Stage 0 is a pilot to establish blanks, limits of quantification, washout duration and assay reliability, and to estimate incidence. Stage 1 is the cohort with serial sampling. Stage 2 is a tissue sub-study in incident cases who undergo surgical therapy.
Participants
Adults with at least one titanium or titanium-zirconium implant in function for at least 12 months and at least one natural tooth contralateral to the index implant position are eligible if the index implant is in health or peri-implant mucositis by the 2017 classification [37], with no radiographic bone loss beyond the threshold defined in Section 4.6 at baseline. One index implant per patient is chosen by a computer-generated random draw among eligible implants, which removes within-patient clustering from the primary analysis and prevents selection of the implant with the most visible wear.
Exclusion criteria: peri-implantitis or its surgical treatment at the index implant; implantoplasty at any implant in the same quadrant; titanium-bearing restorations, orthodontic hardware or orthopedic titanium implants adjacent to the tooth control site; use of metal instruments on the index site in the preceding three months; inability to adhere to the exposure-control window; and zirconia implants (outside scope, though zirconium-based attribution could be applied in a separate arm). Implant system, surface, alloy grade, abutment material, connection type, prosthesis type and time in function are recorded. Abutment material is recorded because wear at the abutment interface can release titanium, and tooth-site and implant-site particles were both present in earlier work [24]; a pilot also reported more particles and pro-inflammatory gene expression with zirconia abutments in a very small sample [38].
Exposure control and sampling
Exposure window. Before each sampling visit, participants use a toothpaste and mouthrinse confirmed by ingredient list not to contain titanium dioxide, and avoid foods and supplements listing titanium dioxide or E171, with a diary. The Geneva pilot controlled intake of titanium dioxide and toothpaste and a titanium dioxide candy challenge changed measured values [6]; its abstract does not state the washout duration. We propose 7 days for oral-care products and 72 hours for foods as starting values, to be validated in Stage 0 by a within-participant challenge.
Instruments. All sampling uses sterile polymer or glass instruments verified free of titanium by blank analysis. Metal curettes and ultrasonic tips are not used at the index implant or tooth site on sampling days or during the study, and maintenance at index sites uses non-metal instruments, logged by date. Abrasive metal instrumentation of implants has been tested as a source of titanium release in a randomized trial, where the difference at 8 weeks was a trend [33]; because it remains a plausible source, instrumentation is both restricted and recorded. Paper-based carriers can contain titanium dioxide as a filler, an engineering consideration not supported by a source reviewed here; if paper strips are used for fluid, a paper-lot blank is mandatory.
Samples. At each visit, in fixed order: (1) submucosal plaque from the deepest site of the index implant; (2) plaque from the mirror-image or nearest contralateral tooth site, with matched probing depth category where possible; (3) a field blank, opened and closed beside the patient; (4) a saliva sample for background exposure. The first sample taken at each site is the analytic sample, and sampling order is alternated between implant and tooth sites by randomization. Specimens are placed in titanium-free polymer tubes and frozen.
Analytical chain
The chain is tiered so that each question is answered by the technique that can answer it and each tier has its own controls.
| Tier | Method | Quantity | Key controls | Evidence status |
|---|---|---|---|---|
| 1 | Microwave acid digestion; ICP-MS with collision or reaction cell (tandem instrument preferred) | Total Ti, Al, V, Zr, Fe (ng per sample) | Procedural, reagent and field blanks; spiked recovery; certified or in-house reference material; replicate digests | Tissue digestion for Ti, Al, V used in rat work [31]; low-level Ti analysis shows inter-laboratory bias [26]; interference management to be validated |
| 2 | Enzymatic extraction; single-particle ICP-MS | Particle number, mass per particle, equivalent diameter, dissolved fraction | Particle-free blanks; transport-efficiency calibration; size standards; spiked particle recovery | Applied to E171 and food TiO2 [29], urine [30] and rat tissue after implantoplasty [31] |
| 3 | Tracer elements by Tier 1 and by single-particle co-detection if instrument allows | Al/Ti, V/Ti and Zr/Ti ratios; tracer fraction fe | Tooth-site tracer background; alloy certificate; Fe as contamination indicator | Rat work found Al and V did not differ between control and experimental animals except V in brain [31]; human validation absent |
| 4 | Synchrotron micro-XRF and micro-XANES on tissue sections (Stage 2) | Oxidation state or chemical form; metallic vs TiO2 | Reference spectra for Ti metal, Ti-6Al-4V oxide and pigment TiO2; Ti-free microtome blades | Done once for inflamed tissue [25]; limited beam access |
| 5 | Scanning electron microscopy with EDS on filtered specimens or sections | Morphology, size, elemental composition | Ti-free blades and substrates [17]; blank substrate | Established but semi-quantitative |
Limits. The limit of detection is set from the distribution of blanks, not from instrument noise alone. In the Geneva pilot, limits were 0.18 micrograms per liter for titanium and 0.07 for zirconium, and most values in healthy mucosa were below detection before exposure control [6]. We pre-specify a limit of quantification equal to ten blank standard deviations, with all samples below it treated as censored. We do not substitute fixed fractions of the limit.
Phase. No single measurement separates sources. Metallic titanium is not a pigment constituent and so supports an implant origin; tracer co-occurrence supports an alloy origin; titanium dioxide without tracers is classed “indeterminate” unless the tooth-site contrast shows excess at the implant. Crystal-phase identification of titanium dioxide as anatase or rutile could add information, but we did not find a peri-implant study that uses it to separate implant oxide from pigment, and the phase composition of implant surface oxides versus food-grade pigment would have to be established with reference materials before it could be used. In this protocol phase calls are therefore treated as supporting, not decisive.
Size. Particle size alone cannot discriminate. Implant-related particles have been reported from about 100 nanometers to 54 micrometers [1], which overlaps the size range of pigment particles analyzed by single-particle methods in food [29]. Size and number from Tier 2 are therefore descriptive of burden, not of source.
Stage 2 tissue sub-study
In incident cases who proceed to surgical therapy, granulation tissue is collected with Ti-free instruments and processed for Tier 4 and 5 analyses; periodontitis or implant-free gingiva tissue from patients undergoing unrelated surgery provides a matrix comparator, and a pre-implant gingiva reference follows the approach named in the registered cross-sectional study [7]. This sub-study is descriptive; it informs the attribution classes and cannot test temporality.
Outcomes
Primary outcome. Radiographic marginal bone-level loss of at least 1.0 mm at the index implant relative to the baseline radiograph, measured on standardized parallel-technique radiographs by two blinded readers with a calibration exercise. The threshold is a pre-specified choice; sensitivity analyses use 0.5 mm and 2.0 mm and continuous bone change. Because progression is non-linear and varies between patients [32], radiographs are taken at 12, 24 and 36 months and onset is treated as interval-censored.
Secondary outcomes. Incident peri-implantitis by the consensus case definition [37]; bleeding and suppuration on probing; probing depth; plaque index; and subgingival bacterial load. Inflammation measures at each visit enter the models as time-varying covariates.
Sample-size reasoning
All calculations below are closed-form arithmetic re-computed in Python (script samplesize.py in the paper folder). They are Illustrative computation under stated assumptions — not empirical data.
For a standardized continuous exposure in a Cox model, the required number of events is
d=((z_(1−α/2)+z_(1−β))^2)/((lnHR)^2 (1−R^2))
where HR is the hazard ratio per standard deviation of log implant-attributed titanium, R^2 is the proportion of exposure variance explained by the adjustment covariates (variance inflation 1/(1−R^2)), and z values are standard normal quantiles. With two-sided α=0.05 and power 80%, (1.9600+0.8416)^2=7.849.
Assumed values: HR=1.4, R^2=0.3. Then d=7.849/(0.3365^2×0.7)=99.0, so 100 events. For a nested case-control sample with m matched controls per case, the efficiency relative to the full cohort is approximately m/(m+1), a standard approximation whose relative efficiency depends on the number of controls per case [39], so cases required are d(m+1)/m; with m=4 this is 125 cases and 500 controls. This approximation is for the conditional-logistic analysis of risk-set samples and is an assumption of this protocol, to be replaced by simulation once the exposure distribution is known from Stage 0.
For the bystander conclusion, the design must show that an effect larger than HR=1.25 is unlikely. If the true ratio is 1.0 and the one-sided 97.5% upper confidence limit must fall below 1.25 with 80% power, the standard error of the log hazard ratio must be at most ln1.25/(1.9600+0.8416)=0.0796, which gives 158 events unadjusted and 226 events with R^2=0.3, corresponding to 283 cases and 1,132 controls at 1:4.
| Objective | Events | Cases (1:4) | Controls | Cohort at 10% incidence | 15% | 20% |
|---|---|---|---|---|---|---|
| Detect HR 1.4 per SD (80% power) | 100 | 125 | 500 | 1,471 | 981 | 736 |
| Exclude HR above 1.25 (80% power) | 226 | 283 | 1,132 | 3,330 | 2,220 | 1,665 |
Note. HR 1.3 with the same adjustment needs 163 events and HR 1.5 needs 69. With serial samples at four visits and two sites, the first scenario implies about 5,000 assays plus blanks and the second about 11,300.
Measurement error in the exposure attenuates the log hazard ratio roughly in proportion to the reliability λ of the exposure measure, so required events scale approximately as 1/λ^2. A reliability of 0.7 would raise the first scenario from 99 to about 202 events. Stage 0 replicate assays must therefore estimate λ before the cohort size is fixed. The bystander tier is likely to be unattainable at a single center, so the protocol pre-specifies that a smaller study can support only “no driver effect of the hypothesized size” and cannot claim bystander status.
Statistical analysis plan and decision rules
Primary analysis. Conditional logistic regression (or Cox regression with risk-set sampling) of incident bone loss on baseline standardized logΔ with adjustment for implant system, time in function, age, smoking, diabetes, periodontitis history, plaque, bleeding on probing and probing depth at baseline, and instrumentation log. Results are reported over the κ range and for alloy-tracer and commercially pure strata separately.
Temporal analysis. A random-intercept cross-lagged panel model uses repeated titanium and bone-level measures to estimate the path from earlier titanium to later bone level and the reverse path, adjusting for plaque and bleeding. Within cases, onset-aligned slopes of titanium before and after onset are compared (P6). Measurement error is modeled through replicate assays.
Multiplicity and sensitivity. Primary hypothesis P4 is tested at two-sided 0.05; other predictions are secondary. Sensitivity analyses cover alternative thresholds, exclusion of samples with high iron (contamination), exclusion of sites with instrumentation, and multiple imputation for missing visits.
| Pattern | Criteria | Interpretation |
|---|---|---|
| Driver | Baseline Δ HR point estimate at least 1.25 with lower 95% limit above 1.0; dose-response across tertiles; persists after time-varying inflammation; earlier-Ti to later-bone path exceeds the reverse | Supports a predictive, possibly causal role; mechanism not shown |
| Bystander | Upper one-sided limit of HR below 1.25; both cross-lagged paths near zero; total-Ti association disappears after exposure control | No predictive role at the studied exposure |
| Consequence | Baseline HR near 1.0 but positive path from inflammation or bone loss to later Ti; steeper post-onset slope | Ti release follows disease |
| Artifact | Total-Ti difference present but absent in Δ and in tracer fraction | Earlier associations reflect exogenous or procedural Ti |
| Inconclusive | Confidence interval spans both 1.0 and 1.4, or reliability below target | Larger sample or better assays needed |
Bias control
Laboratory staff are blinded to site type, case status and clinical data; radiographic readers are blinded to titanium values. Batches contain samples from cases and controls and from implant and tooth sites, in randomized order, with blanks every ten samples. Selection bias is limited by consecutive enrollment and random index-implant selection. Confounding by maintenance instrumentation is limited by instrument restriction and logging. Loss to follow-up is handled with multiple imputation under stated missing-at-random assumptions, with a tipping-point sensitivity analysis. A prior registered record of a related cross-sectional question [7] and a registered trial of titanium release across non-surgical treatments [40] are acknowledged; this protocol should be registered before enrollment and its decision rules deposited.
Ethics and feasibility
The study requires approval by an institutional review board and informed consent, neither of which has been sought. Sampling is non-invasive at baseline and follow-up; tissue is taken only when surgery is clinically indicated. The principal feasibility risks are the number of patients needed (Table 3), the number of assays, access to a metal-free laboratory and to synchrotron beam time, and the difficulty of maintaining a TiO2-free window across five visits. Stage 0 exists to quantify these before commitment.
5. Competing explanations and how the design discriminates among them
Driver versus consequence. The cross-sectional literature is silent here because both give higher titanium at diseased sites. The cohort discriminates by ordering: a driver predicts that baseline implant-attributed titanium, measured while the site is healthy or mucositic, separates future cases from controls; a consequence predicts no baseline separation but a rise in titanium after inflammation or bone loss starts. Serial samples allow both signatures to be tested in the same patients (P4 to P6).
Driver versus common cause. Implant system and abutment design may raise both titanium and disease risk. Adjustment for implant system and recording of abutment and prosthesis type address measured common causes. Stratification by alloy grade also helps: if a tracer-defined fraction predicts disease within a single implant system, a system-level confounder is less likely. Unmeasured common causes, such as surgical trauma and implant positioning, remain.
Driver versus artifact. The attribution chain is the discriminating device. If the total-titanium difference between cases and controls vanishes once exogenous titanium dioxide is controlled and blanks subtracted, the earlier associations are artifactual. The Geneva pilot already showed that dietary control can remove quantifiable titanium around healthy implants [6]; the present design asks whether the same happens in the contrast between future cases and controls.
Titanium as mediator of dysbiosis. If titanium shifts community composition before disease, then bacterial composition measured in the same plaque aliquot should differ by baseline titanium before bone loss. Two small studies reported a link between titanium and community structure at single time points [16, 34], and a review argues that titanium particles modify dysbiosis [5]. A site-specific rather than patient-wide microbiome signal [20] would be consistent with a local, not a systemic, source. The design can test this link but is not powered for it; it is secondary.
Ceramic implants and non-titanium disease. Peri-implantitis occurs around ceramic implants [18], so titanium cannot be necessary. The protocol’s question is narrower: whether, among titanium implants, attributed titanium adds predictive information.
What the design cannot discriminate. A predictive association is not a mechanism. A baseline hazard ratio above 1 would be consistent with titanium causing inflammation, but also with titanium marking a site-level wear process or a patient-level susceptibility, for example a hyper-inflammatory host that corrodes implants faster. Experimental work, such as exposing microcosm biofilms or macrophage cultures to measured in vivo titanium concentrations, would be required, and the ionic and particulate forms in such experiments should reflect the attributed fraction found here.
6. Clinical and engineering implications
If the artifact pattern is observed, then titanium measurements in plaque or tissue without exposure control and blanks should not be used as evidence of implant degradation, and cross-sectional reports would need to be read in that light. If the consequence pattern is observed, titanium would be better regarded as a marker of ongoing disease activity than as a target of prevention, and efforts directed at surface chemistry would not be justified by titanium alone. If the driver pattern is observed, then prospective evaluation of abutment materials, insertion protocols and non-metal maintenance instruments would be justified, and baseline implant-attributed titanium might be examined as a risk stratifier, though nothing here supports a diagnostic claim. For engineering, the alloy-tracer approach implies that manufacturers’ certificates of composition and lot traceability would become useful for attribution; and for particle-focused research, the distinction between implant oxide and pigment would need to be resolved with reference materials.
7. Limitations
- The evidence base was appraised from abstracts only; full texts may qualify several statements, especially details of sampling and blanks.
- Embase, Scopus and Web of Science were not accessible; the search was PubMed, Europe PMC and ClinicalTrials.gov. Methodological literature on titanium contamination in ICP-MS was not located through the queries used and may exist.
- Several elements are untested in this context: alloy-tracer fractions in human plaque, plaque as a matrix for single-particle ICP-MS, the washout duration, and the retention factor κ.
- Implant oxide and pigment titanium dioxide share composition; the protocol relies on metallic state, tracers and tooth-site contrast, and a share of the signal will remain indeterminate.
- Tooth sites differ biologically from implant sites (periodontal ligament, pocket ecology, different sampling geometry), so κ is uncertain and the contralateral estimate of exogenous titanium is approximate. Tooth sites in patients with periodontitis history may differ in plaque load.
- Commercially pure titanium implants lack alloy tracers; this may be the majority of the implant market, which would reduce the information from Tier 3.
- The incidence of the primary outcome over three years is assumed, not estimated, and the sample sizes in Table 3 could be wrong by a large factor. The conditional-logistic efficiency factor is an approximation.
- Radiographic bone loss is an imperfect, late proxy for biological disease activity and has measurement error; the 1.0 mm threshold is a protocol choice.
- Baseline sampling at healthy sites by plaque scraping measures dissolved or loosely bound titanium better than titanium embedded in tissue, and the tissue sub-study cannot test temporality.
- A multicenter cohort of 700 to 3,300 patients with serial specimens and a metal-free laboratory is a large undertaking and may not be fundable.
- Results from maintenance patients may not generalize to patients with untreated periodontitis or poor hygiene.
8. Conclusions
Titanium in peri-implant samples cannot be interpreted as implant degradation without a source model, and nothing in the retrieved literature orders titanium and bone loss in time. We propose a cohort with a tiered analytical chain, contralateral tooth-site controls, exposure control and pre-specified decision rules. The arithmetic implies that a driver effect of 1.4 per standard deviation can be detected with about 125 cases, whereas excluding 1.25 needs about 283. The protocol has not been tested, and its outcome could be any of the patterns in Table 4.
Cite this paper
Jain S, Gupta A. Whose Titanium Is It? Source Attribution and Causal Direction of Titanium in Peri-Implant Tissues: A Study Protocol. Working paper. Center for Implant Dentistry; 2026. Available from: https://bayareaimplantdentistry.com/clinician-research-and-publications/peri-implant-titanium-source-attribution
References
- Suárez-López Del Amo F, Garaicoa-Pazmiño C, Fretwurst T, Castilho RM, Squarize CH. Dental implants-associated release of titanium particles: A systematic review. Clin Oral Implants Res. 2018;29(11):1085-1100. doi:10.1111/clr.13372 (opens in a new tab). PMID: 30280418 (opens in a new tab).
- Dionigi C, Nagy G, Derks J, Ichioka Y, Tomasi C, Larsson L, et al. Titanium micro-particles are commonly found in soft tissues surrounding dental implants. Commun Med (Lond). 2025;5(1):78. doi:10.1038/s43856-025-00756-3 (opens in a new tab). PMID: 40102654 (opens in a new tab).
- Mombelli A, Hashim D, Cionca N. What is the impact of titanium particles and biocorrosion on implant survival and complications? A critical review. Clin Oral Implants Res. 2018;29 Suppl 18:37-53. doi:10.1111/clr.13305 (opens in a new tab). PMID: 30306693 (opens in a new tab).
- Ivanovski S, Bartold PM, Huang YS. The role of foreign body response in peri-implantitis: What is the evidence?. Periodontol 2000. 2022;90(1):176-185. doi:10.1111/prd.12456 (opens in a new tab). PMID: 35916872 (opens in a new tab).
- Kotsakis GA, Ganesan SM. Microbial Dysbiosis, Titanium Release, and Peri-implantitis. J Dent Res. 2025;104(5):473-480. doi:10.1177/00220345241307939 (opens in a new tab). PMID: 39953673 (opens in a new tab).
- Cionca N, Meyer J, Michalet S, Varesio E, Hashim D. Quantification of titanium and zirconium elements in oral mucosa around healthy dental implants: a case-control pilot study. Clin Oral Investig. 2023;27(8):4715-4726. doi:10.1007/s00784-023-05099-8 (opens in a new tab). PMID: 37270723 (opens in a new tab).
- ClinicalTrials.gov. NCT07792499: Effect of Metallic Particles on the Progression of Peri-Implantitis (observational, not yet recruiting). U.S. National Library of Medicine; record accessed 17 November 2026. https://clinicaltrials.gov/study/NCT07792499 (opens in a new tab)
- Delgado-Ruiz R, Romanos G. Potential Causes of Titanium Particle and Ion Release in Implant Dentistry: A Systematic Review. Int J Mol Sci. 2018;19(11). doi:10.3390/ijms19113585 (opens in a new tab). PMID: 30428596 (opens in a new tab).
- Flatebø RS, Høl PJ, Leknes KN, Kosler J, Lie SA, Gjerdet NR. Mapping of titanium particles in peri-implant oral mucosa by laser ablation inductively coupled plasma mass spectrometry and high-resolution optical darkfield microscopy. J Oral Pathol Med. 2011;40(5):412-20. doi:10.1111/j.1600-0714.2010.00958.x (opens in a new tab). PMID: 20969628 (opens in a new tab).
- Pettersson M, Kelk P, Belibasakis GN, Bylund D, Molin Thorén M, Johansson A. Titanium ions form particles that activate and execute interleukin-1β release from lipopolysaccharide-primed macrophages. J Periodontal Res. 2017;52(1):21-32. doi:10.1111/jre.12364 (opens in a new tab). PMID: 26987886 (opens in a new tab).
- Kotsakis GA, Xie L, Siddiqui DA, Daubert D, Graham DJ, Gil FJ. Dynamic assessment of titanium surface oxides following mechanical damage reveals only partial passivation under inflammatory conditions. Npj Mater Degrad. 2024;8(1):98. doi:10.1038/s41529-024-00514-1 (opens in a new tab). PMID: 39583186 (opens in a new tab).
- Safioti LM, Kotsakis GA, Pozhitkov AE, Chung WO, Daubert DM. Increased Levels of Dissolved Titanium Are Associated With Peri-Implantitis - A Cross-Sectional Study. J Periodontol. 2017;88(5):436-442. doi:10.1902/jop.2016.160524 (opens in a new tab). PMID: 27858551 (opens in a new tab).
- Olmedo DG, Nalli G, Verdú S, Paparella ML, Cabrini RL. Exfoliative cytology and titanium dental implants: a pilot study. J Periodontol. 2013;84(1):78-83. doi:10.1902/jop.2012.110757 (opens in a new tab). PMID: 22414261 (opens in a new tab).
- Pettersson M, Pettersson J, Johansson A, Molin Thorén M. Titanium release in peri-implantitis. J Oral Rehabil. 2019;46(2):179-188. doi:10.1111/joor.12735 (opens in a new tab). PMID: 30325523 (opens in a new tab).
- Chen L, Bao S, Chen Y, Qian Y, Chen Y, Luo H, et al. Potential role of titanium particles in peri-implantitis: Evidence from an integrated clinical, in vivo, and in vitro study. Biomaterials. 2026;331:124068. doi:10.1016/j.biomaterials.2026.124068 (opens in a new tab). PMID: 41723901 (opens in a new tab).
- Daubert D, Pozhitkov A, McLean J, Kotsakis G. Titanium as a modifier of the peri-implant microbiome structure. Clin Implant Dent Relat Res. 2018;20(6):945-953. doi:10.1111/cid.12676 (opens in a new tab). PMID: 30255621 (opens in a new tab).
- Rakic M, Radunovic M, Petkovic-Curcin A, Tatic Z, Basta-Jovanovic G, Sanz M. Study on the immunopathological effect of titanium particles in peri-implantitis granulation tissue: A case-control study. Clin Oral Implants Res. 2022;33(6):656-666. doi:10.1111/clr.13928 (opens in a new tab). PMID: 35344630 (opens in a new tab).
- Fretwurst T, Müller J, Larsson L, Bronsert P, Hazard D, Castilho RM, et al. Immunohistological composition of peri-implantitis affected tissue around ceramic implants-A pilot study. J Periodontol. 2021;92(4):571-579. doi:10.1002/JPER.20-0169 (opens in a new tab). PMID: 32839977 (opens in a new tab).
- Roehling S, Gahlert M, Janner S, Meng B, Woelfler H, Cochran DL. Ligature-Induced Peri-implant Bone Loss Around Loaded Zirconia and Titanium Implants. Int J Oral Maxillofac Implants. 2019;34(2):357–365. doi:10.11607/jomi.7015 (opens in a new tab). PMID: 30716147 (opens in a new tab).
- Ghensi P, Heidrich V, Bazzani D, Asnicar F, Armanini F, Bertelle A, et al. Shotgun Metagenomics Identifies in a Cross-Sectional Setting Improved Plaque Microbiome Biomarkers for Peri-Implant Diseases. J Clin Periodontol. 2025;52(7):999-1010. doi:10.1111/jcpe.14121 (opens in a new tab). PMID: 40467108 (opens in a new tab).
- Vignard J, Pettes-Duler A, Gaultier E, Cartier C, Weingarten L, Biesemeier A, et al. Food-grade titanium dioxide translocates across the buccal mucosa in pigs and induces genotoxicity in an in vitro model of human oral epithelium. Nanotoxicology. 2023;17(4):289-309. doi:10.1080/17435390.2023.2210664 (opens in a new tab). PMID: 37194738 (opens in a new tab).
- Peters RJB, Oomen AG, van Bemmel G, van Vliet L, Undas AK, Munniks S, et al. Silicon dioxide and titanium dioxide particles found in human tissues. Nanotoxicology. 2020;14(3):420-432. doi:10.1080/17435390.2020.1718232 (opens in a new tab). PMID: 31994971 (opens in a new tab).
- Domingo MG, Nalli GA, Tasat DR, Olmedo DG. Exfoliated oral mucosa cells as bioindicators of short- and long-term systemic titanium contamination. J Trace Elem Med Biol. 2023;76:127114. doi:10.1016/j.jtemb.2022.127114 (opens in a new tab). PMID: 36516572 (opens in a new tab).
- Tawse-Smith A, Ma S, Duncan WJ, Gray A, Reid MR, Rich AM. Implications of Wear at the Titanium-Zirconia Implant-Abutment Interface on the Health of Peri-implant Tissues. Int J Oral Maxillofac Implants. 2017;32(3):599–609. doi:10.11607/jomi.5014 (opens in a new tab). PMID: 28170452 (opens in a new tab).
- Nelson K, Hesse B, Addison O, Morrell AP, Gross C, Lagrange A, et al. Distribution and Chemical Speciation of Exogenous Micro- and Nanoparticles in Inflamed Soft Tissue Adjacent to Titanium and Ceramic Dental Implants. Anal Chem. 2020;92(21):14432-14443. doi:10.1021/acs.analchem.0c02416 (opens in a new tab). PMID: 32970419 (opens in a new tab).
- Koller D , Bramhall P , Devoy J , Goenaga-Infante H , Harrington CF , Leese E , et al. Analysis of soluble or titanium dioxide derived titanium levels in human whole blood: consensus from an inter-laboratory comparison. Analyst. 2018;143(22):5520-5529. doi:10.1039/c8an00824h (opens in a new tab). PMID: 30295302 (opens in a new tab).
- Kleinsorgen F, Torres O, Ferreira F, Henriques B, Ramos A, Souza JCM. Detection of submicron titanium debris associated with peri-implantitis. Discov Nano. 2026;21(1). doi:10.1186/s11671-026-04741-0 (opens in a new tab). PMID: 42377749 (opens in a new tab).
- Woźniak Ł, Antonowicz B, Mierzejewska ŻA, Kosicka E, Lechien JR, Borys J. Titanium Degradation Products from Dental Implants: Mechanisms of Release, Biodistribution, and Analytical Detection in Biological Matrices. Int J Mol Sci. 2026;27(16). doi:10.3390/ijms27167346 (opens in a new tab). PMID: 42653350 (opens in a new tab).
- Bastardo-Fernández I, Chekri R, Noireaux J, Givelet L, Lambeng N, Delvallée A, et al. Characterisation of titanium dioxide (nano)particles in foodstuffs and E171 additives by single particle inductively coupled plasma-tandem mass spectrometry using a highly efficient sample introduction system. Food Addit Contam Part A Chem Anal Control Expo Risk Assess. 2024;41(8):867-884. doi:10.1080/19440049.2024.2359532 (opens in a new tab). PMID: 38833436 (opens in a new tab).
- Salou S, Larivière D, Cirtiu CM, Fleury N. Quantification of titanium dioxide nanoparticles in human urine by single-particle ICP-MS. Anal Bioanal Chem. 2021;413(1):171-181. doi:10.1007/s00216-020-02989-8 (opens in a new tab). PMID: 33123763 (opens in a new tab).
- Pompéu de Moraes D, González-Morales S, Toledano-Serrabona J, Sánchez-Garcés MÁ, Bettmer J, Montes-Bayón M, et al. Tracking soluble and nanoparticulated titanium released in vivo from metal dental implant debris using (single-particle)-ICP-MS. J Trace Elem Med Biol. 2023;77:127143. doi:10.1016/j.jtemb.2023.127143 (opens in a new tab). PMID: 36871433 (opens in a new tab).
- Fransson C, Tomasi C, Pikner SS, Gröndahl K, Wennström JL, Leyland AH, et al. Severity and pattern of peri-implantitis-associated bone loss. J Clin Periodontol. 2010;37(5):442-8. doi:10.1111/j.1600-051X.2010.01537.x (opens in a new tab). PMID: 20507368 (opens in a new tab).
- Kotsakis GA, Botto A, Jazaeri MS, Palaiologou A, Lee E, Daubert DM. Impact of abrasive versus non-abrasive cleaning on peri-implantitis titanium release: A randomized clinical trial. J Periodontol. 2026. doi:10.1002/jper.70158 (opens in a new tab). PMID: 42316309 (opens in a new tab).
- Casarin RCV, Bonilha GM, Paz H, Cerquera TB, Stolf CS, Pimentel SP, et al. Unraveling the Microbial Mosaic and Inflammatory Core of Peri-implantitis. J Dent Res. 2026;105(2):267-275. doi:10.1177/00220345251344585 (opens in a new tab). PMID: 40676912 (opens in a new tab).
- Fretwurst T, Buzanich G, Nahles S, Woelber JP, Riesemeier H, Nelson K. Metal elements in tissue with dental peri-implantitis: a pilot study. Clin Oral Implants Res. 2016;27(9):1178-86. doi:10.1111/clr.12718 (opens in a new tab). PMID: 26508041 (opens in a new tab).
- Ohneberg K, Wolkewitz M, Beyersmann J, Palomar-Martinez M, Olaechea-Astigarraga P, Alvarez-Lerma F, et al. Analysis of Clinical Cohort Data Using Nested Case-control and Case-cohort Sampling Designs. A Powerful and Economical Tool. Methods Inf Med. 2015;54(6):505-14. doi:10.3414/ME14-01-0113 (opens in a new tab). PMID: 26108707 (opens in a new tab).
- Berglundh T, Armitage G, Araujo MG, Avila-Ortiz G, Blanco J, Camargo PM, et al. Peri-implant diseases and conditions: Consensus report of workgroup 4 of the 2017 World Workshop on the Classification of Periodontal and Peri-Implant Diseases and Conditions. J Clin Periodontol. 2018;45 Suppl 20:S286-S291. doi:10.1111/jcpe.12957 (opens in a new tab). PMID: 29926491 (opens in a new tab).
- Olander J, Barkarmo S, Hammarström Johansson P, Wennerberg A, Stenport VF. Inflammatory Gene Profile and Particle Presence in Peri-Implant Mucosa: a Pilot Study on 9 Patients. J Oral Maxillofac Res. 2023;14(3):e2. doi:10.5037/jomr.2023.14302 (opens in a new tab). PMID: 37969950 (opens in a new tab).
- Bertke S, Hein M, Schubauer-Berigan M, Deddens J. A Simulation Study of Relative Efficiency and Bias in the Nested Case-Control Study Design. Epidemiol Methods. 2013;2(1):85-93. doi:10.1515/em-2013-0007 (opens in a new tab). PMID: 26345580 (opens in a new tab).
- ClinicalTrials.gov. NCT07811921: Titanium Particle Release Across Non-Surgical Peri-Implantitis Therapies (interventional, recruiting). U.S. National Library of Medicine; record accessed 17 November 2026. https://clinicaltrials.gov/study/NCT07811921 (opens in a new tab)

