Vibroacoustic Monitoring of Screw-Joint Preload in Splinted Full-Arch Implant Prostheses: Patient-Specific Longitudinal Baselines, Separation of Confounders, and Localization of the Affected Joint. A Device Concept and Diagnostic Accuracy Protocol
- Paper type
- Biomedical engineering design concept with diagnostic accuracy study protocol
- Status
- Original Work · Working Paper
Contents
Abstract
Background. Loosening of abutment and prosthetic screws is among the most frequent technical complications of implant restorations. In splinted full-arch prostheses a single loosened screw produces no mobility and may go unnoticed until a screw fractures or the framework is overloaded. Laboratory studies have shown that resonance and acoustic measurements change with screw loosening, including in a splinted six-implant prosthesis, and that machine-learning classifiers separate loosened from intact joints on bench models. The idea of a vibroacoustic signature of joint degradation is therefore not new. What the searches for this manuscript did not find is evidence in patients, a method that separates preload loss from other causes of change, or a study of diagnostic accuracy.
Objective. To refine the question to the part that remains open and to specify a monitoring concept and a protocol to answer it: whether deviation from a patient-specific vibroacoustic baseline detects and localizes preload loss in splinted full-arch prostheses under clinical conditions.
Framework. The screw joint is treated as a preload-dependent contact stiffness in series with the stiffness of the bone–implant interface. Modal sensitivity analysis relates stiffness changes at individual joints to shifts in the natural frequencies of the prosthesis, and sequential change detection is applied to each patient’s own measurements.
Proposed methods. Bench experiments with instrumented screws vary preload and deliberately introduce confounders. A finite element model supplies the sensitivity matrix. A prospective diagnostic accuracy study compares the index measurement with removal torque at scheduled maintenance visits.
Central hypothesis. Deviation from an individual baseline detects loss of preload at a single joint with better accuracy than any population threshold, and nonlinear features distinguish preload loss from loss of bone support.
Conclusions. A series-stiffness analysis suggests that frequency shifts from partial preload loss are small, which places a demanding requirement on repeatability and may limit the method. No device has been built and no measurements have been made.
Keywords: dental implant; screw loosening; preload; resonance frequency analysis; vibration analysis; structural health monitoring; diagnostic accuracy; full-arch prosthesis
1. Introduction
A screw joint holds because the tightened screw is stretched and clamps the parts together. The clamping force is the preload. When preload falls, the joint can slip or open under function, the screw sees larger alternating stress, and loosening or fatigue fracture follows [1–3]. In implant dentistry these events are common. A meta-analysis of implant-supported single crowns estimated a five-year cumulative incidence of screw loosening of 8.8% [4]. For implant-supported fixed dental prostheses, abutment or screw loosening was reported in 5.3% over five years [5]. For fixed complete prostheses in edentulous patients, screw fracture was the most frequent implant-related technical complication, at 10.4% after five years and 20.8% after ten [6]. In a 14-year retrospective study of internal conical connections, screw loosening had occurred at 46.4% of implants [7].
Detection is the weak point. On a single crown, a loose screw eventually announces itself as mobility. In a splinted prosthesis the other screws hold the framework still, and one joint can lose its preload silently. The clinician finds it when the prosthesis is removed for maintenance, when the patient reports a noise or a taste, or when a screw breaks. Radiographs are of little help: in a recent study, experienced dentists classified torque decay from periapical radiographs with accuracy of 42% to 52%, no better than chance, whereas a convolutional neural network reached 88% on an independent test set [8]. Removal and retightening at every recall is the alternative, and it has costs. Each cycle wears the mating surfaces, and access holes must be reopened and restored.
A noninvasive test of joint integrity would therefore have clinical use, and vibration is the obvious candidate. The stiffness of a clamped joint depends on its preload, and stiffness governs resonance. Resonance frequency analysis has been used for decades to assess the stability of implants in bone [9, 10], and structural engineers detect loosened bolts and other damage from changes in vibration response [11, 12].
The idea is plausible enough that it has already been pursued. Miyashita and colleagues showed that resonance frequency measured with a triaxial accelerometer fell as an abutment screw was loosened, in a mandibular model and in rabbit tibia [13]. Guo and colleagues detected loosening of individual prosthetic screws in splinted six-implant prostheses on casts by acoustic resonance analysis [14]. Sammour and colleagues combined triaxial resonance measurements with machine learning across three connection designs [15]. The concept in the title originally assigned to this paper, vibroacoustic fingerprinting of implant–abutment connections, is in substance covered by these studies. This manuscript therefore does not present the concept as new. It asks what remains unanswered after them and proposes how to answer it.
2. Background and Current Scientific Evidence
Mechanics of the implant screw joint
Tightening torque is converted partly into screw elongation and mostly into friction at the threads and under the head, so that preload for a given torque varies with surface condition and lubrication [2, 3]. After tightening, microscopic asperities on the contacting surfaces flatten and some preload is lost, the settling effect; retightening after about ten minutes is recommended to compensate [16, 17]. Finite element analysis of the dental screw joint has represented the contact nonlinearity explicitly [18]. Joint stability also depends on connection geometry, with external hexagonal, internal, and conical designs differing in how they resist lateral load [19, 20]. Loosening is promoted by bending loads, cantilevers, and parafunction, although reviews note that evidence linking bruxism to specific complications is limited [21]. Misfit between a framework and its implants is thought to introduce static strain and predispose to loosening, but a systematic review found the evidence on clinical consequences insufficient [22].
Vibration methods for implant stability
Resonance frequency analysis was introduced to measure the stability of the implant–tissue interface by attaching a transducer to the implant and recording its resonance, which varies with the stiffness of the surrounding support and the height of exposed implant [9]. The technique became a standard research and clinical tool [10]. The Periotest method assesses damping from the contact time of a tapping rod [23]. The two do not measure the same thing: a three-year prospective cohort found only weak to moderate correlation between them [24]. Quantitative ultrasound has been reported to estimate stability parameters in bone phantoms with errors four to eight times smaller than resonance frequency analysis [25]. All of these methods address the bone–implant interface, and the commercial resonance device is cleared for that purpose [26]. They are applied to the implant or abutment with the prosthesis removed.
Vibration and acoustic detection of screw loosening
Three laboratory studies bear directly on the present question.
Miyashita and colleagues tightened an abutment screw in a mandibular model and loosened it in steps from 0 to 90 degrees. Resonance frequency fell with loosening, and differences from the initial value became significant at 10 to 15 degrees in vitro and between 5 and 20 degrees in rabbit tibia [13]. The accelerometer was mounted rigidly on the abutment.
Guo and colleagues studied screw-retained prostheses on six implants in maxillary and mandibular casts. A tapping stimulus was applied beside each screw and acoustic signals were collected. Deviation from the insertion torque changed the resonance peak within 6 to 9 kHz, peak frequency correlated positively with torque for every screw, and the size of the effect depended on the position of the screw: loosening in posterior and anterior regions influenced the peak more than loosening in the central region [14].
Sammour and colleagues tested conical, internal tri-channel, and external hexagonal connections on six implants in jaw models under seven loosening conditions. A random forest classifier achieved 100% accuracy for intact versus abnormal and 83% to 92% for seven-class classification, and a convolutional autoencoder identified loosening from 5 degrees [15].
These studies establish feasibility on the bench. They share limitations that are visible from their abstracts. The specimens were casts, models, or animal bone. Loosening was imposed as an angle of back-rotation or a reduced torque, and preload itself was not reported as measured. The classifier results come from a small number of implants measured repeatedly, a setting in which accuracy estimates are optimistic unless training and test data come from different specimens. No study reported diagnostic accuracy against a reference standard in patients.
Related work outside dentistry
Vibration analysis has been used during hip arthroplasty to judge stem fixation, in a protocol tested in 83 patients [27], and modal parameters predicted acetabular cup fixation in a composite bone model [28]. For spinal instrumentation, a vibroacoustic method with a sensor on the screw head and a neural network classifier detected simulated pedicle screw loosening in four cadaveric spines with sensitivity of 91.5% and specificity of 91.1% [29]. For temporomandibular joint implants, acoustic emission sensors distinguished loosened from tightened fixation screws only when placed directly on the screw, because activity at other screws masked the signal when the system was observed as a whole [30]. This last observation is a caution for splinted dental prostheses.
Patent records retrieved for this manuscript include a system for evaluating implant stability by vibration [31], a method using torsional vibration resonance [32], and orthopedic devices with embedded sensors for joint stability or for load, wear, and infection [33, 34]. Abstracts or summaries were read; claims were not analyzed.
Summary
Vibration response changes measurably with screw loosening on bench models, including in splinted prostheses, and classifiers can separate the conditions imposed. Evidence in patients, against a reference standard, and in the presence of the sources of variation that clinical measurement introduces was not found.
3. Research Gap
The refined gap has five parts.
- Clinical diagnostic accuracy. No study identified in the searches reports sensitivity and specificity of a vibroacoustic test for screw-joint preload loss in patients against a reference standard.
- Between-patient variability. Resonance of a prosthesis depends on its mass, span, material, and supports. A frequency that indicates loosening in one patient may be normal in another. The bench studies compare a specimen with itself. Whether a threshold transfers between patients is untested, and there is reason to expect that it does not.
- Confounding. Changes in bone support, soft-tissue contact, prosthesis mass after repair or wear, temperature, and sensor placement all change vibration response. None of the three studies was designed to separate these from preload loss.
- Localization. In a splinted prosthesis the clinical question is which screw. Guo and colleagues tested each screw locally [14]; whether a loosened joint can be identified from a limited set of measurements, and how its position affects detectability, has not been analyzed.
- Repeatability. The size of the change that can be distinguished from measurement noise under clinical conditions is unknown.
These are questions of evidence about an existing idea. They are classified accordingly in Appendix A.
4. Research Objective and Central Hypothesis
The objective is to determine whether longitudinal vibroacoustic measurement, referenced to each patient’s own baseline, detects and localizes preload loss in screw-retained full-arch implant prostheses.
Specific aims are:
- To specify a measurement device and a repositioning fixture that give repeatable excitation and sensing at defined points on a prosthesis.
- To quantify on the bench the relation between measured preload and vibroacoustic features, and the effect of each confounder.
- To derive, from a mechanical model, the sensitivity of each feature to each joint and to use it for localization.
- To estimate repeatability and the minimal detectable change under clinical conditions.
- To estimate diagnostic accuracy in patients against removal torque.
The central hypothesis is that deviation of vibroacoustic features from a patient-specific baseline detects preload loss at a single joint with higher accuracy than a threshold derived from other patients, and that amplitude-dependent features distinguish preload loss from loss of bone support. The first part would be refuted if a population classifier performed as well as the individualized one. The second would be refuted if the features attributed to joint nonlinearity changed equally when bone support was reduced on the bench. A further possibility, which the analysis in Section 7 takes seriously, is that neither approach reaches clinically useful accuracy because the signal is too small.
5. Scientific and Theoretical Framework
The joint as a preload-dependent stiffness
Two rough surfaces pressed together touch at asperities. As the clamping force rises, more asperities come into contact and the interface stiffens. The tangential and normal contact stiffness of a screw joint, k_j, is therefore an increasing function of preload F_p:
k_j=k_j(F_p), dk_j/dF_p>0 (1)
The form of the function depends on surface topography and connection geometry and must be measured for each connection type. Below a critical preload the joint opens or slips during part of each vibration cycle, and its response becomes nonlinear: apparent stiffness falls with amplitude, damping rises, and harmonics of the excitation frequency appear. These features are used in structural monitoring to identify loosened fasteners [12].
The prosthesis as a multi-support structure
A full-arch prosthesis on N implants is modeled as an elastic body with mass matrix M and stiffness matrix K, connected to ground at each implant through two stiffnesses in series: the joint, k_j, and the bone–implant support, k_(b,j). The effective support stiffness is
k_(e,j)=(k_j k_(b,j))/(k_j+k_(b,j)) (2)
Natural frequencies ω_r and mass-normalized mode shapes ϕ_r satisfy Kϕ_r=ω_r^2Mϕ_r. For a small change in the support stiffnesses, first-order perturbation gives
Δω_r/ω_r≈1/2∑_j=1^N s_(rj) (Δk_(e,j))/(k_(e,j)) (3)
where s_(rj) is the fraction of the strain energy of mode r stored in support j. The matrix S=(s_(rj)) is the sensitivity matrix. It states which modes respond to which joints. A support that lies near a node of a mode stores little of that mode’s energy and its loosening leaves that frequency unchanged. The finding that loosening of posterior and anterior screws shifted the resonance peak more than loosening of central screws [14] is what a position-dependent sensitivity of this kind would produce, although the abstract does not allow the modes involved to be identified.
Features
The measured quantities are frequency response functions between an excitation point and a sensing point. From them are extracted natural frequencies and damping ratios of identifiable modes, and, under sinusoidal excitation at two or more amplitudes, the amplitude dependence of frequency and the ratio of second-harmonic to fundamental response. The first two are linear features and respond to any change in stiffness or mass. The last two are nonlinear features and are expected to respond to an opening or slipping contact and not to a uniform reduction in the stiffness of bone. This expectation is a hypothesis.
Confounders
Let y_t be the feature vector at visit t, θ_t the vector of fractional changes in joint stiffness, and c_t the vector of confounders (changes in bone support at each implant, mass, temperature, soft-tissue contact). To first order,
y_t−y_0=A θ_t+B c_t+ε_t (4)
Preload loss can be separated from a confounder only if the corresponding columns of A and B differ. Equation (2) shows the difficulty at once. A low-frequency mode of the prosthesis senses each support only through k_(e,j), so a loss of joint stiffness and a loss of bone stiffness at the same implant produce the same change in every linear feature. They are not separable from prosthesis-level modal data. Separation must come from elsewhere: from nonlinear features, from local high-frequency response near the screw in which the implant’s own inertia participates, or from independent information such as radiographic bone level.
Individual baselines and change detection
Each patient’s baseline μ_0 and within-patient variance are estimated from repeated measurements made after delivery, once settling has been compensated by retightening. At later visits the standardized deviation z_t of a summary feature is monitored. With frequent measurements a cumulative sum scheme [35] accumulates small consistent deviations:
C_t=max(0, C_(t−1)+z_t−k), signal when C_t>h (5)
with reference value k and decision limit h chosen for a target false-alarm rate. With recall intervals of months the series is short, and detection in practice rests on one or two visits. The cumulative scheme becomes relevant only if measurements can be made far more often than at recall, which would require a patient-operated device and is outside the present protocol.
6. Methodology
Device concept
The device has three parts. The exciter delivers a controlled mechanical input, either a miniature impact of calibrated force or a piezoelectric actuator driven with a swept sine. The sensor is a miniature triaxial accelerometer, with a contact microphone as an alternative to be compared on the bench. The fixture is a rigid jig that seats on the occlusal surfaces of the prosthesis in a unique position and carries the exciter and sensor to defined points beside each screw access. Because full-arch prostheses are digitally designed, the fixture can be produced from the same design file. The fixture addresses the dependence of the measurement on position that is evident in the bench literature [14, 30]. Excitation amplitude is kept well below functional loads so that the test cannot itself disturb the joint, and this is verified on the bench by measuring preload before and after repeated testing.
Stage A: bench experiments
Full-arch frameworks in titanium and zirconia are mounted on four and six implants embedded in bone-analog blocks. Preload truth is obtained from instrumented screws carrying strain gauges, calibrated in a tensile fixture, so that the independent variable is clamping force and not torque or back-rotation. Each joint in turn is taken through a series of preload levels in randomized order while the others remain at full preload. Combinations of two reduced joints are included.
Confounders are then introduced one at a time at constant preload: reduced support stiffness at one implant, by substituting a more compliant embedding material; a mucosal analog in contact with the intaglio surface; added and removed mass representing repair or wear; temperature between room and mouth temperature; removal and reseating of the fixture; and different operators. Each is then combined with preload loss. Connection types include those studied previously [15].
Outcomes are the feature changes per unit preload loss, the feature changes per unit confounder, and their ratio.
Stage B: mechanical model
A finite element model of each bench configuration is built with joints represented as preload-dependent contact, following established practice for the dental screw joint [18], and verified and validated against the bench modal data in a manner consistent with published credibility frameworks [36]. The model supplies the sensitivity matrix of equation (3). Its predicted pattern of frequency shifts for each loosened joint is compared with the measured pattern. For clinical use a reduced model generated from the design file would replace the full model; whether a reduced model is adequate is a question for this stage.
Stage C: clinical diagnostic accuracy study
Design is prospective, with consecutive enrollment and reporting according to the standard for diagnostic accuracy studies [37]. Participants are adults with screw-retained full-arch prostheses in a maintenance program in which the prosthesis is removed at defined intervals as routine care.
Baseline measurements are made at delivery or at the first visit after all screws have been tightened to the manufacturer’s torque and retightened. The index test is repeated at each recall before any screw is touched. The reference standard follows immediately: each screw is removed with a calibrated digital torque device and the peak removal torque recorded. A joint is classified as having lost preload when removal torque is below a prespecified fraction of the installation torque. Operators of the reference standard are blinded to the index result, and the index analysis is completed before unblinding.
Removal torque is an imperfect reference. It reflects friction as well as preload and can be normal in a joint whose preload has partly relaxed. The bench stage quantifies the relation between removal torque and measured preload for the connections used, so that the reference threshold can be set with knowledge of its error.
Covariates are radiographic bone level at each implant, time in function, opposing dentition, reported parafunction, prosthesis material, and any repair since baseline.
Repeatability
Within a visit, the fixture is removed and reseated and the measurement repeated by the same and by a second operator. Between visits, a subset of participants is measured twice one to two weeks apart without intervention. The standard error of measurement and the minimal detectable change are estimated for each feature.
Sample size
The unit of analysis for detection is the joint. To estimate a sensitivity of 0.85 with a 95% confidence interval of half-width 0.10 requires about 49 joints with preload loss by the normal approximation. The number of joints that must be examined depends on the proportion found loose at recall, which is not known for this definition; if it were 5% to 10%, between 490 and 980 joints would be needed, before inflation for clustering of joints within prostheses. With four to six joints per arch this corresponds to roughly 100 to 250 arch-visits. These figures are planning arithmetic based on assumed proportions. The pilot phase would replace the assumptions with estimates.
Analysis
Sensitivity, specificity, and predictive values are estimated at joint level with confidence intervals that account for clustering, and at arch level for the question of whether any joint is affected. Receiver operating characteristic curves compare three decision rules: deviation from the individual baseline, a population threshold on absolute feature values, and a classifier trained on other patients. Localization is scored as the proportion of affected arches in which the joint ranked most likely is the affected one. Any learned model is trained and tested on different patients, never on repeated measurements of the same prosthesis, to avoid the optimistic bias discussed in Section 2.3. Analyses of confounding relate false-positive results to changes in bone level, repairs, and temperature.
Safety and ethics
The device contacts the prosthesis and not the tissues. Requirements include electrical safety, cleanability or single-use barriers for the fixture, and demonstration that excitation does not alter preload. The clinical stage requires approval by an institutional review board and informed consent, and adds a measurement to visits that would occur in any case. The investigational device status of the system would have to be determined with the regulator before any clinical use. No approval has been sought and no study has begun.
7. Analytical Development
How large is the signal?
Equation (2) determines how much of a change in joint stiffness reaches the structure. The logarithmic sensitivity of the effective stiffness to the joint stiffness is
(∂lnk_e)/(∂lnk_j)=k_b/(k_j+k_b) (6)
If the joint is much stiffer than the bone support, the support is governed by bone and the structure hardly notices the joint. Table 1 evaluates the change in effective stiffness when joint stiffness is halved, for several ratios of joint to bone stiffness, and the corresponding frequency shift from equation (3) for a mode in which the affected support holds one quarter of the strain energy. Because a halving is not a small change, the stiffness column is computed exactly from equation (2) and the frequency column is the first-order estimate.
| Ratio of joint to bone stiffness before loosening | Change in effective support stiffness | Approximate change in modal frequency |
|---|---|---|
| 1 | −33% | −4.2% |
| 3 | −20% | −2.5% |
| 10 | −8.3% | −1.0% |
| 30 | −3.1% | −0.4% |
The actual ratio for dental implants is not asserted here; it would be measured in Stage A. The table shows the range of possibilities. If a tight joint is an order of magnitude stiffer than the bone support, halving its stiffness shifts a favorable mode by about 1%, and a mode in which the support participates less by a fraction of that. The bench observation that significant frequency changes appeared only after 5 to 15 degrees of back-rotation [13] is compatible with a signal that is small until preload is largely lost, though the abstract does not report the preload that those angles represent.
What repeatability is required?
For a feature with standard error of measurement σ_m, the smallest change between a baseline that is the mean of m measurements and a single later measurement that exceeds noise at the 95% level is
MDC_(95)=1.96 σ_m√(1+1/m) (7)
To detect a shift of 1% with a single baseline measurement (m=1), σ_m must be below 0.36% of the frequency. With three baseline measurements the requirement relaxes to 0.44%. Between-patient variation in the natural frequencies of full-arch prostheses, which differ in span, material, and support, will be far larger than this. The argument for individual baselines follows directly: a population threshold has to contend with between-patient variance, whereas an individual baseline contends only with within-patient measurement error. The same arithmetic shows the risk to the whole approach. If reseating the fixture or a change of two or three degrees in temperature moves the frequency by half a percent, partial preload loss is undetectable by linear features whatever the algorithm.
Localization
Under the assumption that a single joint is affected, localization is a pattern-matching problem. The observed vector of fractional frequency shifts across R modes, d, is compared with each column s_j of the sensitivity matrix, and the joint is chosen that maximizes the normalized inner product:
ȷ=argmaxj(d^Ts_j)/(∥d∥ ∥s_j∥) (8)
Two joints can be told apart only if their columns differ in direction. Joints placed symmetrically about the midline of a symmetric prosthesis have identical columns for symmetric sensing and cannot be distinguished from natural frequencies alone; mode-shape information from the triaxial sensor at several points, or the local measurement beside each screw, is needed to break the symmetry. Without the single-joint assumption, unique identification of N joint changes requires at least N modes with linearly independent sensitivity, which may exceed the number of modes measurable with adequate precision.
Separating joint from bone
As shown in Section 5.4, linear prosthesis-level features cannot separate the two members of the series pair. The protocol tests two candidate discriminators. The first is the nonlinear feature set: a loosened joint is a contact that can open, whereas a compliant bone bed is expected to behave more nearly linearly at the small amplitudes used. The second is local response at higher frequency, where the abutment and screw head move relative to the implant and the joint stiffness enters without being masked by the softer bone. Neither has been demonstrated for dental joints. If both fail, the method could still indicate that a support has changed and leave the cause to be established by removal and radiography, which would retain some clinical value at lower specificity.
8. Discussion
This manuscript began from a concept that turned out, on searching, to have been substantially explored in the laboratory. The honest response is to say so and to move the question to where the uncertainty lies. Bench studies show that a signal exists [13–15]. They do not show that it survives the passage to patients, and the analysis here gives reasons for doubt as well as for hope.
The reasons for doubt are quantitative. The joint is in series with a bone support that may be much more compliant, so that the structure is insensitive to moderate changes in the joint. The measurement must then resolve changes of a percent or less, against repositioning error, temperature, and gradual changes in the prosthesis and its supports. High classification accuracy on bench models does not address this, because on a bench the specimen, its temperature, and its supports are constant. A classifier that reaches 100% accuracy on six implants measured repeatedly [15] has shown that the conditions imposed are distinguishable in that specimen; generalization to other specimens was not part of that design as far as the abstract shows.
The reasons for hope are structural. Each prosthesis can serve as its own control. The geometry is known exactly from the design file, so a mechanical model of the individual structure is available at no additional imaging cost. A fixture can be made that repositions the sensor far more reproducibly than a hand-held probe. And the clinical need is specific: the clinician does not need a preload value, only a reliable indication of whether removal is warranted and which screw to check.
Alternatives should be weighed fairly. Image analysis of radiographs by a neural network achieved 88% accuracy for torque decay in a pilot study [8]; it needs no new hardware but involves radiation and has not been tested on splinted prostheses. Sensors embedded in implants or screws would measure preload directly; orthopedic patents describe embedded sensing [33, 34], but miniaturization, power, and regulatory burden are substantial. Scheduled removal and retightening remains the default and is the comparator against which any monitoring method must show a net benefit.
Within this portfolio, the present work is confined to the sensing modality. A companion paper on failure-tolerant design asks what a loosened joint does to the remaining structure, and another on longitudinal digital twins describes how repeated measurements of any kind could update a patient-specific model. The vibroacoustic measurement, if it proved reliable, would be one input to such a model.
9. Clinical Significance
If the method reached useful accuracy, recall visits for full-arch prostheses could be stratified. Prostheses with no detected change would be left in place, sparing access restorations and wear of the joints, and those with a suspected joint would be removed with attention directed to a specific screw. Early retightening might prevent the screw fractures that are the most frequent technical complication in this population [6]. These benefits are conditional on accuracy that has not been shown. A test with poor specificity would increase removals, and one with poor sensitivity would give false reassurance. Until diagnostic accuracy is established, the method has no clinical role.
10. Limitations
No device exists and no measurement has been made. The stiffness ratios in Table 1 are assumptions spanning a range, and the true values may lie outside it. The first-order frequency estimate is approximate for large changes. The reference standard, removal torque, is an imperfect measure of preload. The individual-baseline strategy requires a clean baseline, which is unavailable for prostheses already in service with unknown joint status. Changes in the prosthesis over years, including wear, repair, and calculus, may make the baseline obsolete. The nonlinear discriminators are untested hypotheses. The literature was assessed from abstracts; details of the excitation and sensing methods of the closest studies, which matter for the device concept, were not available. Engineering databases, where much of the relevant vibration literature is indexed, were not searched.
11. Future Validation
The sequence is fixed by the logic of the problem. Stage A must first show that the signal from partial preload loss exceeds the repeatability achievable with a fixture; if it does not, the clinical stage should not proceed. Stage B must show that a model built from the design file predicts the pattern of shifts. Only then is Stage C justified. Later questions include whether monitoring changes outcomes in a randomized comparison with scheduled removal, whether the approach extends to cemented or angled-channel restorations, and whether a patient-operated device is feasible.
12. Conclusion
Vibroacoustic detection of implant screw loosening has been demonstrated on laboratory models and is not a new concept. Whether it works in patients is unknown. This manuscript refines the question to detection and localization of preload loss against a patient-specific baseline, in the presence of confounders, and specifies the bench, modeling, and clinical stages needed to answer it. A series-stiffness analysis indicates that the signal may be small and that repeatability will decide feasibility. No data are reported.
Cite this paper
Jain S, Gupta A. Vibroacoustic Monitoring of Screw-Joint Preload in Splinted Full-Arch Implant Prostheses: Patient-Specific Longitudinal Baselines, Separation of Confounders, and Localization of the Affected Joint. A Device Concept and Diagnostic Accuracy Protocol. Working paper. Center for Implant Dentistry; 2026. Available from: https://bayareaimplantdentistry.com/clinician-research-and-publications/vibroacoustic-implant-connection-monitoring
References
- Jörnéus L, Jemt T, Carlsson L. Loads and designs of screw joints for single crowns supported by osseointegrated implants. Int J Oral Maxillofac Implants. 1992;7(3):353-9. PMID: 1289261 (opens in a new tab).
- McGlumphy EA, Mendel DA, Holloway JA. Implant screw mechanics. Dent Clin North Am. 1998;42(1):71-89. PMID: 9421671 (opens in a new tab).
- Bickford JH. An introduction to the design and behavior of bolted joints. 3rd ed. New York: Marcel Dekker; 1995.
- Jung RE, Zembic A, Pjetursson BE, Zwahlen M, Thoma DS. Systematic review of the survival rate and the incidence of biological, technical, and aesthetic complications of single crowns on implants reported in longitudinal studies with a mean follow-up of 5 years. Clin Oral Implants Res. 2012;23 Suppl 6:2-21. doi:10.1111/j.1600-0501.2012.02547.x (opens in a new tab). PMID: 23062124 (opens in a new tab).
- Pjetursson BE, Thoma D, Jung R, Zwahlen M, Zembic A. A systematic review of the survival and complication rates of implant-supported fixed dental prostheses (FDPs) after a mean observation period of at least 5 years. Clin Oral Implants Res. 2012;23 Suppl 6:22-38. doi:10.1111/j.1600-0501.2012.02546.x (opens in a new tab). PMID: 23062125 (opens in a new tab).
- Papaspyridakos P, Chen CJ, Chuang SK, Weber HP, Gallucci GO. A systematic review of biologic and technical complications with fixed implant rehabilitations for edentulous patients. Int J Oral Maxillofac Implants. 2012;27(1):102-10. PMID: 22299086 (opens in a new tab).
- Yi Y, Heo SJ, Koak JY, Kim SK. Mechanical complications of implant-supported restorations with internal conical connection implants: A 14-year retrospective study. J Prosthet Dent. 2023;129(5):732-740. doi:10.1016/j.prosdent.2021.06.053 (opens in a new tab). PMID: 34481672 (opens in a new tab).
- Liu G, Li J, Chen X, Lin J, Liu R, Chen Z, et al. AI-based evaluation of implant abutment screw torque decay: A periapical radiograph pilot study. J Periodontol. 2026. doi:10.1002/jper.70127 (opens in a new tab). PMID: 41972958 (opens in a new tab).
- Meredith N, Alleyne D, Cawley P. Quantitative determination of the stability of the implant-tissue interface using resonance frequency analysis. Clin Oral Implants Res. 1996;7(3):261-7. doi:10.1034/j.1600-0501.1996.070308.x (opens in a new tab). PMID: 9151590 (opens in a new tab).
- Sennerby L, Meredith N. Implant stability measurements using resonance frequency analysis: biological and biomechanical aspects and clinical implications. Periodontol 2000. 2008;47:51-66. doi:10.1111/j.1600-0757.2008.00267.x (opens in a new tab). PMID: 18412573 (opens in a new tab).
- Doebling SW, Farrar CR, Prime MB. A Summary Review of Vibration-Based Damage Identification Methods. The Shock and Vibration Digest. 1998;30(2):91-105. doi:10.1177/058310249803000201 (opens in a new tab).
- Farrar CR, Worden K. An introduction to structural health monitoring. Philos Trans A Math Phys Eng Sci. 2007;365(1851):303-15. doi:10.1098/rsta.2006.1928 (opens in a new tab). PMID: 17255041 (opens in a new tab).
- Miyashita M, Ogawa T, Naito H, Shibamoto A, Wang AS, Shobara K, et al. Evaluation of implant screw loosening by resonance frequency analysis with triaxial piezoelectric pick-up: in vitro model and in vivo animal study. Clin Oral Investig. 2018;22(5):2129-2134. doi:10.1007/s00784-018-2467-1 (opens in a new tab). PMID: 29705965 (opens in a new tab).
- Guo J, Wang X, Liu Z, Zhou J, Wang L. Acoustic resonance frequency analysis for evaluating prosthetic screw stability in splinted implant-supported fixed dental prostheses: An in vitro study. J Prosthet Dent. 2025;133(2):542.e1-542.e9. doi:10.1016/j.prosdent.2024.10.028 (opens in a new tab). PMID: 39550228 (opens in a new tab).
- Sammour SR, Naito H, Shigemitsu R, Xie Z, Kimoto T, Sasaki K, et al. Anomaly detection of screw loosening of different implant-abutment connection designs using resonance frequency analysis: An in vitro study. J Prosthet Dent. 2026;135(5):954.e1-954.e14. doi:10.1016/j.prosdent.2025.12.007 (opens in a new tab). PMID: 41484042 (opens in a new tab).
- Winkler S, Ring K, Ring JD, Boberick KG. Implant screw mechanics and the settling effect: overview. J Oral Implantol. 2003;29(5):242-5. doi:10.1563/1548-1336(2003)029<0242:ISMATS>2.3.CO;2 (opens in a new tab). PMID: 14620687 (opens in a new tab).
- Siamos G, Winkler S, Boberick KG. Relationship between implant preload and screw loosening on implant-supported prostheses. J Oral Implantol. 2002;28(2):67-73. doi:10.1563/1548-1336(2002)028<0067:TRBIPA>2.3.CO;2 (opens in a new tab). PMID: 12498448 (opens in a new tab).
- Sakaguchi RL, Borgersen SE. Nonlinear contact analysis of preload in dental implant screws. Int J Oral Maxillofac Implants. 1995;10(3):295-302. PMID: 7615325 (opens in a new tab).
- Binon PP. Implants and components: entering the new millennium. Int J Oral Maxillofac Implants. 2000;15(1):76-94. PMID: 10697942 (opens in a new tab).
- Kourtis S, Damanaki M, Kaitatzidou S, Kaitatzidou A, Roussou V. Loosening of the fixing screw in single implant crowns: predisposing factors, prevention and treatment options. J Esthet Restor Dent. 2017;29(4):233-246. doi:10.1111/jerd.12303 (opens in a new tab). PMID: 28556604 (opens in a new tab).
- Shafiee E, Nourizadeh A. Bruxism in implant-supported rehabilitations: a narrative review of clinical complications and management strategies. BMC Oral Health. 2025;25(1):1586. doi:10.1186/s12903-025-07005-y (opens in a new tab). PMID: 41068723 (opens in a new tab).
- Katsoulis J, Takeichi T, Sol Gaviria A, Peter L, Katsoulis K. Misfit of implant prostheses and its impact on clinical outcomes. Definition, assessment and a systematic review of the literature. Eur J Oral Implantol. 2017;10 Suppl 1:121-138. PMID: 28944373 (opens in a new tab).
- Schulte W, Lukas D. The Periotest method. Int Dent J. 1992;42(6):433-40. PMID: 1286926 (opens in a new tab).
- Reynolds I, Winning L, Polyzois I. A three-year prospective cohort study evaluating implant stability utilising the Osstell® and Periotest™ devices. Front Dent Med. 2023;4:1139407. doi:10.3389/fdmed.2023.1139407 (opens in a new tab). PMID: 39916929 (opens in a new tab).
- Vayron R, Nguyen VH, Lecuelle B, Haiat G. Evaluation of dental implant stability in bone phantoms: Comparison between a quantitative ultrasound technique and resonance frequency analysis. Clin Implant Dent Relat Res. 2018;20(4):470-478. doi:10.1111/cid.12622 (opens in a new tab). PMID: 29770562 (opens in a new tab).
- U.S. Food and Drug Administration. 510(k) summary K033689 (Osstell Mentor), 21 CFR 872.4200, product code EKX. Silver Spring (MD): FDA.
- Pastrav LC, Jaecques SV, Jonkers I, Perre GV, Mulier M. In vivo evaluation of a vibration analysis technique for the per-operative monitoring of the fixation of hip prostheses. J Orthop Surg Res. 2009;4:10. doi:10.1186/1749-799X-4-10 (opens in a new tab). PMID: 19358703 (opens in a new tab).
- Henyš P, Čapek L. Impact Force, Polar Gap and Modal Parameters Predict Acetabular Cup Fixation: A Study on a Composite Bone. Ann Biomed Eng. 2018;46(4):590-604. doi:10.1007/s10439-018-1980-3 (opens in a new tab). PMID: 29340934 (opens in a new tab).
- Seibold M, Sigrist B, Götschi T, Widmer J, Hodel S, Farshad M, et al. A new sensing paradigm for the vibroacoustic detection of pedicle screw loosening. Med Biol Eng Comput. 2025;63(4):1001-1011. doi:10.1007/s11517-024-03235-4 (opens in a new tab). PMID: 39560918 (opens in a new tab).
- Jacob E, William K, Yani S, Ramachandran RA, Walkosz J, Didem O, et al. Acoustic emission-based diagnostic tool for temporomandibular joint implant fixation screw loosening. J Mech Behav Biomed Mater. 2025;171:107137. doi:10.1016/j.jmbbm.2025.107137 (opens in a new tab). PMID: 40737763 (opens in a new tab).
- University of Washington, applicant. Methods and systems to measure and evaluate stability of medical implants. United States patent application publication US 2020/0060612 A1. Priority 2017 Mar 10; published 2020 Feb 27.
- Implant stability measurement by torsional-vibration resonance frequency. United States patent US 9,949,644.
- Orthosensor Inc, original assignee. Muscular-skeletal joint stability detection and method therefor. United States patent US 9,271,675 B2.
- UT-Battelle LLC; University of Tennessee Research Foundation, assignees. In-vivo orthopedic implant diagnostic device for sensing load, wear, and infection. United States patent US 7,097,662 B2. Priority 2004 Aug 25; granted 2006 Aug 29.
- Page ES. Continuous Inspection Schemes. Biometrika. 1954;41(1/2):100. doi:10.2307/2333009 (opens in a new tab).
- American Society of Mechanical Engineers. Assessing credibility of computational modeling through verification and validation: application to medical devices. ASME V&V 40-2018. New York: ASME; 2018.
- Bossuyt PM, Reitsma JB, Bruns DE, Gatsonis CA, Glasziou PP, Irwig L, et al. STARD 2015: an updated list of essential items for reporting diagnostic accuracy studies. BMJ. 2015;351:h5527. doi:10.1136/bmj.h5527 (opens in a new tab). PMID: 26511519 (opens in a new tab).

