Skip to main content

How Much Misfit Is Too Much? A Preload-Consumption Model and Dose–Response Protocol for Implant Prosthesis Misfit

Paper type
Hypothesis and Theory (with study protocol)
Status
Original Work · Working Paper
Keywords:
implant prosthesis misfit · passive fit · screw preload · bolted joint · screw loosening · framework stiffness
Contents
  1. 1Introduction
  2. 2Background and critical appraisal of the evidence
  3. 3Hypothesis and theoretical framework
  4. 4Proposed methodology
  5. 5Competing explanations and how the design discriminates among them
  6. 6Clinical and engineering implications
  7. 7Limitations
  8. 8Conclusions

Abstract

Background. Passive fit is demanded of multi-unit implant frameworks, yet no tolerable misfit threshold has been derived. Human, animal and randomized data have not shown that misfit harms peri-implant bone, while bench studies suggest that misfit lowers screw joint stability.

Hypothesis. Misfit acts on the screw joint, not on bone. Closing a vertical gap g by tightening elastically deforms the framework, which consumes part of the screw preload F_p. The residual clamp force is F_c=F_p−k_fg, where k_f is the framework stiffness in closing the gap. The tolerable gap is therefore not one number: it scales inversely with k_f (span, cross-section, modulus) and directly with the preload margin, and it depends on the connection.

Proposed approach. We state a bolted-joint model with defined symbols and units, including bending and shear compliance of the framework and compliance of the supports in series. An illustrative calculation under stated assumptions gives gaps that consume 25% or 50% of an assumed 380 N preload for cobalt–chromium, titanium, zirconia and poly(methyl methacrylate) (PMMA) segments of 10 to 30 mm. We derive eight falsifiable predictions and specify a prospective multicenter cohort of fixed full-arch frameworks, with digital misfit measurement at delivery, bench stiffness measurement of each delivered framework, and 36-month counts of screw events, with marginal bone change as a negative-control outcome. A sample-size calculation yields 69 events for a hazard ratio of 1.5 per standard deviation of preload consumed.

Expected contribution. The model converts a slogan into stiffness-dependent, testable thresholds and separates the preload-consumption account from confounding by workflow quality.

Conclusions. Bench studies already link misfit to reduced preload or loosening torque; the contribution here is the explicit stiffness-based form, its predictions and a design that can discriminate it from raw-gap and bone-overload accounts. The model is untested and probably overstates consumption when supports are compliant.

1. Introduction

A screw-retained multi-unit framework is expected to seat on all of its abutments without being forced. This expectation, termed passive fit, has been taught for decades as a precondition for implant success. The strict form of the requirement cannot be met: laboratory and clinical procedures are inadequate to deliver an absolutely passive framework, and the reviewers who examined the question judged that the effect of non-passivity on implant success is doubtful [1]. A systematic review of five human and five animal studies concluded that biological harm from misfit was not confirmed, while screw-related complications were observed [2]. The same review described the tolerable amount of misfit as a matter of debate [2].

The numerical limit that clinicians quote, about 150 µm, is usually traced to a 1991 report of failures and complications in 391 consecutive Brånemark prostheses [3]. The retrieved abstract of that report describes success rates and complication patterns at one year and contains no misfit threshold, so the attribution is unverified here. A consensus review of full-arch impression accuracy asked for an objective, verified misfit criterion, reporting trueness differences between digital and conventional techniques of up to 162 ± 77 µm in clinical studies and up to 43 µm in laboratory studies [4]. The requirement is therefore a design rule with a number that nobody has derived.

Two lines of evidence frame the problem. First, studies in which misfit was measured against bone have not found harm. In a cohort of edentulous patients, mean misfit was about 100 µm and did not correlate with marginal bone loss at one or five years [5]. A randomized trial of non-passively fitting superstructures reported that misfit-related strain declined over six months [6]. Animal studies with deliberately misfitting frameworks in rabbits [7] and with ill-fitting prostheses in an immediate versus delayed loading design [8] likewise did not establish bone loss. A primate study cautions that unloaded implants cannot exclude harm under dynamic load [9]. Second, bench studies repeatedly associate misfit with weaker screw joints, but they have measured outcomes without a mechanical account of why a particular gap should matter for a particular framework [10–13].

This manuscript proposes such an account. The central idea is simple. Tightening a screw against a gap makes the screw do work on the framework. The framework pushes back with a force that is proportional to its stiffness in the closing direction and to the gap. That force is paid for out of the clamp that the screw would otherwise exert on the joint. What matters for the joint is the clamp that remains, relative to the separating loads that function will impose. If this is correct, then misfit acts on the joint first, the complications are loosening and fracture of prosthetic screws, and the tolerable gap depends on structure that the threshold literature has not measured: framework bending stiffness, screw preload and connection design.

The contribution is deliberately bounded. We do not claim that nobody has linked misfit to preload; the bench literature has (Section 2). We claim that no analytic model that combines framework stiffness with preload to give a stiffness-dependent tolerable gap was found in our search, that the model yields predictions the existing studies were not designed to test, and that a prospective cohort can test it. Section 3 states the model, Section 4 the protocol, and Sections 5 to 8 the competing explanations, implications, limitations and remaining work.

2. Background and critical appraisal of the evidence

Origin of the passivity requirement and the missing threshold

The conventional rationale is that a rigid multi-unit framework forced onto misaligned abutments would load the bone with a permanent static strain. Reviews have since described the clinical significance of passive fit as questionable while still endorsing efforts to improve it [1]. A systematic review has examined the biomechanical sequelae of implant framework misfit [14]. The thread that connects these papers is that a threshold was assumed, not derived; where numbers were proposed, their provenance cannot be reconstructed from abstracts [3].

Evidence on bone: consistently null, with a caveat

Table 1 lists the pivotal studies and an operational evidence level. The bone evidence is consistent across designs. The patient cohort found no relationship between measured misfit and marginal bone loss [5]. The randomized trial found falling misfit strain over time, which is compatible with bone adaptation or with relaxation of the framework and joint [6]. Rabbit and animal work did not show bone loss from misfit preload [7, 8]. The primate study measured bone around unloaded implants, and its authors recognized that the finding could not be extrapolated to functional loading [9].

These studies have the same limitation. Misfit in the patient cohort was moderate and varied little, so a weak gradient could not be detected, and the maximum was about 275 µm [5]. Animal frameworks and loading may differ from those of a human full arch. The honest summary is that bone harm from misfit has not been shown, which is a weaker statement than showing that it does not exist. For the present argument, that is sufficient: the bone null is the reason to look for another pathway, and the model supplies a negative-control outcome so that the bone hypothesis can be examined in the same cohort.

Evidence on the screw joint: concordant bench signals, no mechanism

Screw joint mechanics in implant dentistry were derived from mechanical engineering principles in the 1990s, including a simple relationship between applied torque and preload [15]. Measurements followed: at recommended torques, preload of gold and titanium abutment screws averaged 468 N and 382 N respectively, with stresses of roughly 56 to 58% of yield, so tightening was within the elastic range [16]. Strain-gauged abutment load cells showed that mean preload varied roughly threefold across commercial systems at recommended torque [17], and preload of gold prosthetic screws was monitored for 15 hours after tightening to test whether it declines with time [18]. Screw material and surface treatment change the rotation needed to generate a given preload [19]. These data supply the scale of F_p used below: a few hundred newtons, with a large spread between systems.

Against this background, misfit studies have used four kinds of outcome.

  • Loosening or detorque after a fixed interval. Dentures cast to passive and non-passive fit levels were tested with titanium and gold screws; passive fit with titanium screws gave better joint stability, and the authors recommended minimizing misfit [10]. In a second mandibular series, metal–ceramic prostheses and misfit each reduced loosening torque [11].
  • Preload after cycling. A study of single crowns with unilateral misfit of 50, 100 and 200 µm found reduced detorque in all groups, and the effect of misfit on preload maintenance was limited to some conditions [12].
  • Strain-gauged preload in multi-implant frameworks. In a four-implant model with a passive framework and three misfit frameworks, the passive framework had the highest preload at two torque levels, and retorquing after 10 minutes did not reliably raise preload under misfit [13].
  • Framework or peri-implant strain. Shims of about 30 µm placed under one of two implants produced a linear rise in framework strain and in peri-implant strain as misfit increased, in titanium and zirconia frameworks [20, 21].

Two further studies bear on mechanism. In cantilevered digitally manufactured frameworks, reverse torque values were lower than the applied torque in all groups, and they fell further after 200,000 cycles in some groups, with misfit values in the tens of micrometers [22]. In zirconia bars on two conical-connection implants, a screw resistance test using a digital torque wrench measured the screw rotation angle at simulated vertical misfits of 50, 100 and 150 µm [23]; that measurement is relevant to the protocol below. Separately, rotational misfit of the hexagon (up to about 15 degrees) correlated with earlier screw joint failure under off-axis cycling, with under 2 degrees the most stable [24]. That is a different geometric quantity from vertical gap, but it shows that fit and screw stability were linked experimentally long before digital misfit metrics.

A finite element study of full-arch titanium frameworks reached the nearest point to the present model from the computational side: it simulated forced seating by prescribed displacement for horizontal, vertical and combined misfits and found higher framework stress for vertical misfit than horizontal misfit of equal size, but it applied no screw preload and no occlusal load [25]. It therefore described the framework side of the interaction (the stored elastic energy) without asking what remained for the screw.

Clinical complication data

Clinical data on screw complications are heterogeneous. A systematic review of single-implant restorations reported about 97% freedom from screw loosening complications at three years [26], and a later review of single and splinted prostheses placed loosening at 7 to 11% [27]. A review of bench preload loss after dynamic fatigue reported losses of 16 to 39% [28]. A recent meta-analysis addressed connection design and cyclic loading in relation to screw stability [29]. None of these separates the contribution of framework misfit from that of connection design, load, screw material and operator. The only human data that correlate measured misfit with an outcome concern bone [5], and the clinical screw-event signal attributed to misfit is limited to retrospective observations summarized in the systematic review [2].

Measurement of misfit is itself unsettled

Digital full-arch workflows have made misfit measurable at scale, but the reference standard is a problem. A meta-analysis of photogrammetry versus intraoral scanning found photogrammetry better in 10 of 13 studies, most of them in vitro [30]. A later meta-analysis of 37 studies found a small difference between digital and conventional impressions, with conventional slightly better for more than four implants [31]. Another meta-analysis described the findings as contradictory, with very high heterogeneity [32]. In a human cadaver model, best-fit deviation metrics underestimated the seating misfit that a virtual-fit assessment revealed by up to 150 µm [33]. Any dose–response study must therefore state which gap it measures: a scan deviation, a virtual-fit gap or a screw-resistance angle. The protocol below takes the last two as primary and secondary measures and treats scan best-fit deviation as a covariate only.

Table 1. Pivotal studies and operational evidence level.
Study (ref.)Design and sampleMain finding relevant hereMain bias or limitLevel
Jemt and Book 1996 [5]Patient cohort; misfit mean about 100 µm, maximum about 275 µmNo correlation of misfit with marginal bone loss at 1 and 5 yearsNarrow misfit range; low power for weak gradientB
Karl and Taylor 2016 [6]Randomized trial of non-passive superstructuresMisfit strain fell over 6 monthsSmall trial; strain surrogateB
Jemt, Lekholm and Johansson 2000 [7]; Duyck et al. 2005 [8]Animal studies of misfit frameworksNo established bone loss from misfit preloadSpecies and loading differ from human full archC
Carr et al. 1996 [9]Primate study, unloaded implantsCautions that dynamic load is untestedUnloaded modelC
Katsoulis et al. 2017 [2]Systematic review, 5 human and 5 animal studiesBone harm not confirmed; screw complications observedHeterogeneous misfit definitionsB
Spazzin et al. 2009 [10]; Bacchi et al. 2013 [11]Bench, mandibular dentures, passive vs non-passiveMisfit reduces screw joint stability (loosening torque)Cast models; 24-h endpoint; stiffness not reportedC
Al-Otaibi and Akeel 2014 [13]Bench, strain-gauged preload, four-implant modelPassive framework had highest preloadOne model; no stiffnessC
Abduo and Lyons 2012 [20]; Abduo and Swain 2012 [21]Bench strain with shims, titanium and zirconiaLinear strain rise with gapStrain, not preload; 10 frameworksC
Burguete et al. 1994 [15]Mechanical derivationTorque–preload relation from engineering principlesNot specific to misfitD
Arikan Kalayci and Guncu 2026 [25]Finite element, forced seatingVertical misfit gave higher framework stress than horizontalNo preload appliedD

Note. Levels are operational categories for this manuscript, not GRADE: A strong, B moderate, C preliminary, D mechanistic or theoretical, E speculative hypothesis.

3. Hypothesis and theoretical framework

Formal statement

Hypothesis H. For screw-retained multi-unit frameworks with a vertical gap at one or more joints, (i) the mechanically relevant dose of misfit is the fraction of screw preload consumed in closing the gap, ρ=k_(eff) g/F_p, not the gap g itself; (ii) ρ is associated with the hazard of prosthetic screw loosening and fracture, with the association concentrated at the joints adjacent to the gap; and (iii) after adjustment for ρ, marginal bone change is unrelated to misfit magnitude within the clinically encountered range.

Consequently the tolerable gap is a function g_(tol)(k_f,k_i,F_p,F_(req)) and not a constant.

Model and symbols

Consider one joint at which the framework does not seat by a vertical distance g while the other joints are seated (the quantity measured by a one-screw test). Tightening the screw to its target preload pulls the framework down by g. We make the following assumptions.

  • A1. The screw tension reached by torque tightening on that joint is F_p and is not changed by the framework reaction. This ignores any change in the torque–tension relation caused by the reaction and is the largest simplification in the model.
  • A2. Response is linear and elastic: framework and supports behave as linear springs in the closing direction. Yielding of the framework or screw is excluded.
  • A3. Friction, embedment settling and stress relaxation are treated separately (Section 3.5).
  • A4. External loads are quasi-static, axial and applied to the joint through the framework.

Define the symbols:

  • g: vertical gap before tightening (mm; reported in µm).
  • F_p: screw tension (preload) generated by the tightening protocol on a passively seated joint (N).
  • k_f: stiffness of the framework at the joint in the closing direction, with the remaining joints acting as supports (N/mm).
  • k_i: combined stiffness of the abutment, implant and bone at the joints that resist the closing movement (N/mm).
  • k_(eff): effective stiffness resisting closure (N/mm).
  • F_r: reaction force of the framework on the joint after closure (N).
  • F_c: residual clamp force across the joint interface (N).
  • ρ: preload-consumed ratio (dimensionless).
  • F_e: external separating force at the joint (N); Φ: load factor, the fraction of F_e carried by the screw (dimensionless, 0 to 1).
  • F_(req): clamp force needed to keep the joint closed under the design separating load (N).

Closing the gap deforms the framework and the supports in series, so the same force F_r passes through both springs and the displacements add: g=F_r/k_f+F_r/k_i. Hence

k_(eff)=(k_f k_i)/(k_f+k_i),  F_r=k_(eff) g.

In the free body of the framework at the joint, the screw pulls down with F_p, while the interface contact force and the framework reaction act upward, so

F_c=F_p−k_(eff) g,  ρ=(k_(eff) g)/F_p.

When the supports are rigid (k_i→∞), k_(eff)→k_f and the expression reduces to F_c=F_p−k_fg. Because k_(eff)≤k_f, the rigid-support form is an upper bound on consumption. The reaction is shared with the other joints: they receive an opposite contact force and the bone supports receive the corresponding load, so the gap-side screw is not the only part of the system affected; modeling the other joints is left for the bench program.

Under a separating force F_e at the joint, a standard bolted-joint treatment assigns a fraction Φ to the screw, so the joint interface relieves by (1−Φ)F_e. The joint begins to open when (1−Φ)F_e>F_c, giving the separation load

F_(sep)=F_c/(1−Φ).

Writing the clamp that the design load requires as F_(req)=(1−Φ)F_(e,max), the tolerable gap is

g_(tol)=(F_p−F_(req))/(k_(eff))=(f F_p)/(k_(eff)),  f=1−(F_(req))/F_p,

where f is the fraction of preload that may be consumed. The general engineering reference for such calculations is the VDI 2230 guideline [34], which was not read for this manuscript; the notation here follows standard bolted-joint usage, and our departure is to identify the consumed preload with the elastic work of closing a framework gap and to carry the framework’s stiffness into the clinical threshold.

Framework stiffness

For a framework segment idealized as a prismatic rectangular beam of width b and height h (mm, h in the bending direction) and span L (mm), with Young’s modulus E (N/mm², that is, GPa × 1000) and Poisson’s ratio ν, the compliance is the sum of bending and shear terms (Euler–Bernoulli plus Timoshenko):

1/k_f=(L^3)/αEI+(s L)/κGA, I=(bh^3)/12, A=bh, G=E/(2(1+ν)), κ=5/6.

Here α=3 and s=1 for a cantilever-like segment loaded at its free end (a gap at an end implant, with the rest of the framework acting as a clamped base), and α=48 and s=1/4 for a simply supported span loaded at its center (a gap at a central implant between two seated ones). Real frameworks lie between these supports and are curved, nonuniform and weakened at connectors. The idealization is a scaling device: it shows how g_(tol) should depend on span, section and modulus, and it is not a substitute for measuring k_f on each delivered framework (Section 4). When bending dominates, k_f∝Ebh^3/L^3 and therefore g_(tol)∝L^3/(Ebh^3).

Illustrative computation under stated assumptions — not empirical data

We computed g_(tol) for fixed fractions f with the script misfit_preload_calc.py (saved in the paper folder, run in this session; output in calc_output.txt). Assumed inputs, none of them measured here: F_p=380 N, chosen to be of the order of mean titanium-screw preload reported in a bench measurement [16] and within the spread between systems seen in another [17]; supports rigid (k_i→∞); κ=5/6; handbook-order moduli and Poisson ratios of 200 GPa and 0.30 for cobalt–chromium, 110 GPa and 0.34 for titanium, 210 GPa and 0.30 for zirconia, and 3 GPa and 0.37 for PMMA; cross-section 5 × 5 mm for the three rigid materials and 6 mm wide × 8 mm high for PMMA (assumed, as a bulky monolithic acrylic segment). Spans were 10, 15, 20 and 30 mm. At L/h of 2 the Euler–Bernoulli term alone is inadequate, which is why the shear term was included.

Table 2. Gap (µm) that consumes a stated fraction f of an assumed 380 N preload; rigid supports; computed values under stated assumptions.
Support, fraction, materialL = 10 mmL = 15 mmL = 20 mmL = 30 mm
Cantilever-like, f = 0.25, CoCr3.611.125.583.9
Cantilever-like, f = 0.25, titanium6.620.346.4152.6
Cantilever-like, f = 0.25, zirconia3.510.624.379.9
Cantilever-like, f = 0.25, PMMA62.9171.7373.21178.4
Cantilever-like, f = 0.50, CoCr7.322.351.0167.7
Cantilever-like, f = 0.50, titanium13.340.692.9305.1
Cantilever-like, f = 0.50, zirconia6.921.248.6159.7
Cantilever-like, f = 0.50, PMMA125.8343.4746.52356.7
Simply supported, f = 0.25, CoCr0.30.91.85.6
Simply supported, f = 0.25, titanium0.61.63.310.2
Simply supported, f = 0.25, zirconia0.30.81.75.3
Simply supported, f = 0.25, PMMA8.016.831.585.8

The computation reproduces the expected ordering. Tolerable gaps rise steeply with span: for cobalt–chromium in the cantilever-like case at f=0.25, they rise from 3.6 µm at 10 mm to 83.9 µm at 30 mm. Cobalt–chromium and zirconia segments of equal geometry give nearly equal values, titanium about twice as much, and the 3 GPa PMMA segment tolerates one to two orders of magnitude more, even though its section is larger. Section height matters strongly: for cobalt–chromium, 5 mm wide, 20 mm span, cantilever-like, f=0.25, the gap was 114.6, 49.0, 25.5 and 15.1 µm for heights of 3, 4, 5 and 6 mm. Preload matters linearly: for the same case with F_p of 180, 380 and 640 N (the latter two within the range of the bench measurements cited above), the gap was 12.1, 25.5 and 43.0 µm. A gap of 100 µm in the 20 mm cobalt–chromium cantilever-like case consumed 372.5 N, a ratio ρ of 0.98 of the assumed preload; for titanium the ratio was 0.54 and for the PMMA segment 0.07. For support compliance, the factor 1+k_f/k_i by which the tolerable gap increases was 8.45, 2.86 and 1.37 for assumed k_i of 500, 2,000 and 10,000 N/mm. These k_i values are arbitrary assumptions for illustration; no measured values are used.

An unfavorable result for the model. With rigid supports, the model implies that a gap of 100 µm under a short, stiff alloy segment would exhaust nearly all of a 380 N preload. Mean misfit of about 100 µm has been reported in a patient cohort [5], and published screw-loosening rates are of the order of a few to about ten percent [26, 27], not the near-universal failure that exhausted preload would imply, so the rigid-support parameterization probably overstates consumption. Candidate reasons are compliant supports, which the series factor addresses; a framework that is more flexible at its connectors than the solid beam; a measured gap that is larger than the distance actually closed because the closure is distributed over several joints; and plastic settling at the interfaces. This discrepancy is treated here as the principal calibration target and falsifier of the rigid form: if measured preload loss under known gaps is much smaller than k_fg in all frameworks, then k_(eff) is much lower than k_f and the useful dose variable must be defined empirically.

Time dependence and settling

Screw preload is not constant. Settling of microscopic surface asperities reduces tension in the first minutes and hours after tightening, and preload of gold prosthetic screws was monitored over 15 hours for this reason [18]. Retightening after a delay is a recommended countermeasure, but in a strain-gauged four-implant model it did not reliably raise preload under misfit [13]. We write the effective preload as F_p(t)=F_(p0)−ΔF_s(t) and the reaction as F_r(t)=k_(eff)(t) g. For alloy frameworks k_(eff) should be stable, so settling lowers F_c additively. For resin-bearing frameworks, stress relaxation can reduce F_r over months, and a declining misfit strain was observed in a randomized trial [6]. The model therefore predicts that preload consumption is front-loaded and that its consequences emerge early (P7).

Numbered predictions

Each prediction is stated so that a specific observation would refute it.

  • P1 (dose variable). Screw-event hazard is better explained by ρ than by g alone. Refuted if a model with g alone fits equal or better by information criterion and ρ adds nothing.
  • P2 (stiffness interaction). At equal gap, hazard rises with k_f. Refuted if the g×k_f interaction is null or negative with a confidence interval excluding the predicted direction.
  • P3 (span and section). Short, thick, stiff frameworks (cobalt–chromium or zirconia, short spans) have a lower tolerable gap than long, slender or resin frameworks; g_(tol) scales approximately with L^3/(Ebh^3). Refuted if measured preload loss per micrometer of gap does not scale with measured k_f in bench tests.
  • P4 (preload). Raising F_p raises F_c one for one at a given gap until the screw limit is reached, so a larger preload margin shifts g_(tol) upward in proportion. Refuted if bench preload under a fixed gap is unaffected by torque.
  • P5 (negative control). Marginal bone change is unrelated to ρ and to g within the cohort’s range. Refuted by a monotone relationship that survives adjustment.
  • P6 (location). Screw events cluster at the joint adjacent to the gap, not randomly among joints. Refuted if events are uniformly distributed among screws of the same prosthesis.
  • P7 (timing). For alloy frameworks the hazard attributable to ρ is concentrated in the first 12 months. Refuted if hazard rises later only, which would favor fatigue or load accumulation.
  • P8 (connection). Connection design modifies the relation through F_(req) and Φ, so effect modification by connection type is expected; the direction is not predicted and is to be estimated. Refuted if the association is identical across connection types with adequate power.

Figure 1 summarizes the pathway.

4. Proposed methodology

The protocol has two components: a bench calibration program that tests whether the model’s mechanics hold, and a prospective cohort that tests whether the clinical dose–response follows. The bench component is a prerequisite because the cohort relies on k_f and F_p being measured accurately. Neither component has been started.

Bench calibration program

Aim. To estimate the slope of clamp loss against gap for frameworks of known stiffness and to estimate k_(eff) relative to k_f.

Specimens. Fixed four-implant (or six-implant) master models with abutment-level connections, with frameworks in cobalt–chromium, titanium, zirconia and a resin-bearing material, each fabricated in at least two section heights and two spans, so that the stiffness range covers the values expected clinically. Each framework is fitted with shims of defined thickness under one joint to create gaps from 0 to about 200 µm, in the manner of earlier shim studies [20, 21], and these gaps are verified by microscopy.

Measurements. (i) k_f of each framework at each joint by a three-point or cantilever bending jig, force through load cell and deflection through displacement transducer; (ii) preload at each joint by a strain-gauged load-cell abutment, as in earlier preload studies [13, 17]; (iii) reverse torque and screw rotation angle, the latter following the digital screw-resistance method [23]; (iv) separation load, determined by applying an axial pull-off force at the joint and recording the load at which the interface first opens. Each framework serves as its own control across the gap series.

Analysis. For each framework, regress clamp force on gap; the slope estimates −k_(eff). Compare the slope with the independently measured k_f; the ratio k_(eff)/k_f is the calibration quantity. Prediction P3 is supported if the slope scales with k_f across frameworks, and refuted if it is flat. A mixed model with framework as a random effect handles repeated gaps. Sample size is set from a pilot, because the variance of the slope is unknown; we do not state a bench n here.

Prospective cohort

Design. Multicenter prospective cohort with 36-month follow-up of patients receiving fixed, screw-retained frameworks on at least four implants (full-arch or long-span). Figure 2 summarizes the design. An observational design is proposed because misfit magnitude cannot be randomized ethically over a clinically relevant range, and because the exposure is a continuous product of laboratory and clinical workflow variables.

Participants. Adults receiving definitive fixed frameworks, with a standardized torque protocol and a screw type recorded. Exclusion: framework revised after delivery for fit during the first month, which would alter the exposure, unless the revision is recorded as a time-varying event; known uncontrolled bruxism is not excluded but is recorded as a covariate, since excluding it would remove the group in which loading is most severe.

Exposure measurement. The primary gap is the virtual-fit gap: the framework is seated on the delivered abutment scans or master and the residual gap at each joint is estimated digitally. The conventional best-fit deviation underestimated seating misfit by up to 150 µm in a cadaver model, so it is not used as the primary exposure [33]. A secondary gap measure is the screw rotation angle recorded with a digital torque wrench during a one-screw test [23]. k_f for each framework is measured on a bench jig before delivery, or computed from the design file by a finite element model that was validated against the bench program. F_p is taken from the calibration of the same screw lot and torque protocol. The primary dose is ρ_(rigid)=k_fg/F_p at the worst joint, and the secondary dose is the calibrated ρ using k_(eff) from the bench program.

Outcomes.

  • Primary: time to first prosthetic screw loosening (recorded clinically, with retightening or replacement) or fracture, at prosthesis level.
  • Secondary: screw-level events, to test location (P6); reverse torque at 3 and 12 months in a subsample, interpreted as an indirect measure of preload since reverse torque is lower than applied torque even in well-fitting frameworks [22]; and framework or abutment fracture.
  • Negative-control outcome: marginal bone level change on standardized radiographs, which the hypothesis predicts to be unrelated to ρ (P5).

Covariates. Connection type, screw material and coating, framework material and section, cantilever length, number and distribution of implants, antagonist, bruxism, center, technician, and the retightening protocol. Operator and technician are covariates and may be random effects, because workflow quality may drive both misfit and screw events (Section 5).

Bias control. Outcome assessors are blinded to g and ρ. The calculation of ρ is performed after outcome assessment by a person who does not see events. The analysis plan is time-stamped and registered before the first measurement; no registration exists at present.

Sample size

We size the cohort on events, using the standard Cox-model approximation for a continuous standardized covariate. With d the number of events, z_(1−α/2) and z_(1−β) the normal quantiles, HR the hazard ratio per standard deviation of ρ, and a variance inflation factor 1/(1−R^2) for correlation of ρ with other covariates:

d=((z_(1−α/2)+z_(1−β))^2)/((lnHR)^2)⋅1/(1−R^2).

Assumptions, chosen by us and not derived from data: two-sided α=0.05, power 0.80, HR=1.5 per standard deviation, and R^2=0.30. The computation (same script) gave 47.7 events without adjustment and 69 with the inflation. For HR of 1.3 and 2.0 the corresponding numbers were 163 and 24. If the 3-year prosthesis-level event probability is 0.10, 0.15 or 0.20, the number of prostheses required for 69 events is 690, 460 or 345, and with 15% attrition 812, 542 or 406. The assumed event probabilities are placeholders informed only by the order of magnitude of published screw-loosening incidence at the implant level [26, 27]; prosthesis-level incidence in multi-unit frameworks is unknown and must be established in a pilot. If both arches of a patient are included, the effective sample size falls by a design effect that depends on the within-patient correlation.

Statistical analysis plan

  • P1. Fit Cox models with shared frailty for center: g alone, k_f alone, g plus k_f, and ρ. Compare by Akaike information criterion and likelihood ratio where models are nested. The test of P1 is whether ρ outperforms g.
  • P2. Fit the interaction g×k_f in a model containing both main effects; report the hazard ratio per 10 µm of gap at the 25th and 75th percentiles of k_f.
  • P3. Stratify by framework material and span class; display the hazard against g within each.
  • P5. Mixed-effects model for bone level change with implant nested in patient, including ρ and g.
  • P6. Conditional logistic regression within prostheses, comparing events at the gap-adjacent joint with other joints.
  • P7. Piecewise-constant hazards at 0 to 12, 12 to 24 and 24 to 36 months with a ρ effect in each interval.
  • P8. Interaction of ρ with connection type.
  • Missing data by multiple imputation; sensitivity analyses excluding frameworks revised within one month; a quantitative bias analysis for unmeasured confounding by workflow quality (E-value).

5. Competing explanations and how the design discriminates among them

Table 3. Competing explanations and the observations that separate them.
ExplanationWhat it predictsDiscriminating observation in the proposed design
H0: preload consumption (this hypothesis)Events track ρ; stiff frameworks worse at equal gap; events at gap-adjacent joint; bone unrelatedρ model beats g model; positive g×kf interaction; location effect; flat bone gradient
H1: bone overload from misfitBone loss rises with g independent of stiffnessBone-change gradient with g in the negative-control analysis; contradicted by earlier null data [5–8]
H2: misfit is a marker of workflow qualityg associates with events through operator and laboratory, not through mechanicsAssociation persists within technician or center strata and varies with kf as the model predicts; a bare association with g fits H2
H3: raw gap drives risk irrespective of stiffnessHazard rises with g equally across kf strataNull g×kf interaction and no gain from ρ over g
H4: connection geometry, rotational misfit or load dominateEvents track connection design, hexagon fit [24] or occlusal loadρ effect persists after adjustment; effect modification by connection (P8)
H5: settling and friction scatter swamp consumptionPreload at delivery is dominated by friction variabilityBench slope calibration shows keff/kf near zero; cohort shows no ρ effect and wide reverse-torque scatter
H6: measurement error in gAttenuated or noisy gradientAgreement analysis of virtual-fit gap versus screw rotation angle; regression-calibration

Two features of the design carry the discrimination. First, the interaction between g and k_f is the signature of the mechanism: H2 and H3 predict a main effect of g without it, and H1 predicts a bone gradient. A study that only correlated g with events would not separate H0 from H2, and that is why k_f is measured. Second, the negative-control outcome (bone) and the location analysis (P6) are two independent mechanisms checks. If events rise with g but not at the gap-adjacent joint, the explanation by clamp loss at a specific joint is weakened and workflow confounding becomes more plausible.

The design cannot separate H0 from every correlated stiffness variable. Stiffness correlates with framework material, thickness and span, which may also correlate with load transfer, with the dentist’s choices and with patient factors. Adjustment and stratification by material reduce but cannot remove this problem; the bench program provides the controlled counterpart.

6. Clinical and engineering implications

All implications are conditional on validation of the model.

Tolerance would be reported per framework, not per clinic. If P1 to P3 are supported, a single acceptance number such as 100 or 150 µm would be replaced by a stiffness-dependent limit g_(tol) with F_p and F_(req) stated. A short, stiff alloy or zirconia segment would warrant tighter seating verification than a long, slender or resin-bearing one. The illustrative values in Table 2 show why: under the stated assumptions, the same gap would consume very different fractions of preload in different frameworks, though the absolute figures depend on supports and should not be used clinically.

Seating verification becomes a quantitative step. The screw-resistance angle with a digital torque wrench [23] and virtual-fit assessment [33] could be converted into an estimate of ρ if k_f were available from the design file. Digital workflows already store framework geometry and material, so stiffness could in principle be computed at no extra measurement burden.

Torque protocols would interact with fit. Because F_c=F_p−k_(eff)g, a higher preload would offer a larger margin at fixed gap until the screw’s elastic limit; elongation studies suggest tightening at recommended torques stays within the elastic range [16]. That does not justify exceeding manufacturer torque, since the model ignores screw stress from the reaction force and fatigue; it motivates a bench test of the preload margin, not a clinical recommendation.

Design choices that alter k_f. Section height and material modulus enter through Ebh^3, so design software could display k_f at each joint and flag a combination with low tolerance. A deliberately compliant segment or connector could widen tolerance, but this would alter fatigue and fracture behavior of the framework and must be tested rather than assumed.

Impression and scanning accuracy. Consensus reviews ask for a verified misfit threshold because clinical deviations and laboratory deviations differ [4]. A stiffness-dependent threshold would change the acceptance criterion applied to scanning accuracy studies, since a deviation harmless for a flexible framework could matter for a stiff one. The model also explains why meta-analytic comparisons of digital and conventional impressions based only on trueness [30–32] do not translate directly into complication rates.

7. Limitations

The model is a lumped, linear, quasi-static description of a three-dimensional, nonlinear joint. Its limitations are specific.

  • Assumption A1. Torque-controlled tension may change when the framework pushes back, because friction at threads and under the head depends on contact conditions. The preload actually achieved on a gapped joint may differ from that on a passive joint, and the model does not capture it.
  • Rigid-support overstatement. Table 2 assumes rigid supports and idealized beams; Section 3.4 shows that the resulting prediction for stiff alloy segments is implausibly severe. The calibration ratio k_(eff)/k_f is unknown.
  • Unmeasured stiffness in the literature. None of the cited bench or clinical studies reported framework stiffness at the joint; the model’s central variable cannot be tested on existing data.
  • Appraisal depth. Evidence was appraised at abstract level. Several statements rest on truncated abstracts or on the candidate notes, including the content of the systematic reviews and the details of the animal studies, and must be checked against full texts.
  • Preload is not observable in vivo. Reverse torque is an indirect index; it is lower than applied torque even for well-fitting frameworks [22], and its relation to clamp force depends on friction.
  • Outcome definitions. Screw loosening is detected by symptoms, mobility or scheduled checks, so detection depends on follow-up intensity; fracture is rare.
  • Residual confounding. Workflow quality, occlusal load and parafunction cannot be fully measured. The E-value analysis bounds but does not remove this.
  • Misfit definition. Gap measured at delivery by virtual fit is not identical to the gap closed during tightening, and it varies with the scan method [33].
  • Non-axial loading. Bending moments and off-axis loads, which are the usual drivers of loosening in bench fatigue studies, are collapsed into one axial separating force.
  • Negative bone evidence is not proof. Absence of a bone effect in prior data [5, 6] reflects limited power, and the primate caveat [9] remains.
  • Search limits. PubMed and Europe PMC were searched; Embase, Scopus and Web of Science were not accessible. The statement that no analytic misfit–preload model was found is limited to what these sources and the vocabularies used can reveal.

8. Conclusions

The assumption that misfit damages bone has not been supported by human, animal or randomized data in the studies appraised, and no tolerable misfit threshold has been derived. A bolted-joint model offers an alternative mechanism: closing a gap consumes preload in proportion to the stiffness of the framework, so the tolerable gap depends on span, section, modulus, preload and support compliance. The illustrative computation shows the expected ordering and also shows that a rigid-support version overstates consumption, which the calibration program is designed to correct. The hypothesis predicts screw events with a dose–response to preload consumed, a stiffness interaction, a location effect and a flat bone gradient. These predictions are testable in a cohort that measures framework stiffness, which no existing study has done. Until that work is performed, the model remains a hypothesis.

Cite this paper

Jain S, Gupta A. How Much Misfit Is Too Much? A Preload-Consumption Model and Dose–Response Protocol for Implant Prosthesis Misfit. Working paper. Center for Implant Dentistry; 2026. Available from: https://bayareaimplantdentistry.com/clinician-research-and-publications/prosthesis-misfit-preload-model

References

  1. Sahin S, Cehreli MC. The significance of passive framework fit in implant prosthodontics: current status. Implant Dent. 2001;10(2):85-92. doi:10.1097/00008505-200104000-00003 (opens in a new tab). PMID: 11450418 (opens in a new tab).
  2. 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).
  3. Jemt T. Failures and complications in 391 consecutively inserted fixed prostheses supported by Brånemark implants in edentulous jaws: a study of treatment from the time of prosthesis placement to the first annual checkup. Int J Oral Maxillofac Implants. 1991;6(3):270-6. PMID: 1813395 (opens in a new tab).
  4. Rutkūnas V, Gedrimienė A, Mischitz I, Mijiritsky E, Huber S. EPA Consensus Project Paper: Accuracy of Photogrammetry Devices, Intraoral Scanners, and Conventional Techniques for the Full-Arch Implant Impressions: A Systematic Review. Eur J Prosthodont Restor Dent. 2023. doi:10.1922/EJPRD_2481Rutkunas12 (opens in a new tab). PMID: 37314199 (opens in a new tab).
  5. Jemt T, Book K. Prosthesis misfit and marginal bone loss in edentulous implant patients. Int J Oral Maxillofac Implants. 1996;11(5):620-5. PMID: 8908860 (opens in a new tab).
  6. Karl M, Taylor TD. Bone Adaptation Induced by Non-Passively Fitting Implant Superstructures: A Randomized Clinical Trial. Int J Oral Maxillofac Implants. 2016;31(2):369-75. doi:10.11607/jomi.4331 (opens in a new tab). PMID: 27004282 (opens in a new tab).
  7. Jemt T, Lekholm U, Johansson CB. Bone response to implant-supported frameworks with differing degrees of misfit preload: in vivo study in rabbits. Clin Implant Dent Relat Res. 2000;2(3):129-37. doi:10.1111/j.1708-8208.2000.tb00003.x (opens in a new tab). PMID: 11359257 (opens in a new tab).
  8. Duyck J, Vrielinck L, Lambrichts I, Abe Y, Schepers S, Politis C, et al. Biologic response of immediately versus delayed loaded implants supporting ill-fitting prostheses: an animal study. Clin Implant Dent Relat Res. 2005;7(3):150-8. doi:10.1111/j.1708-8208.2005.tb00059.x (opens in a new tab). PMID: 16219245 (opens in a new tab).
  9. Carr AB, Gerard DA, Larsen PE. The response of bone in primates around unloaded dental implants supporting prostheses with different levels of fit. J Prosthet Dent. 1996;76(5):500-9. doi:10.1016/s0022-3913(96)90008-6 (opens in a new tab). PMID: 8933440 (opens in a new tab).
  10. Spazzin AO, Henriques GE, de Arruda Nóbilo MA, Consani RL, Correr-Sobrinho L, Mesquita MF. Influence of prosthetic screw material on joint stability in passive and non-passive implant-supported dentures. Open Dent J. 2009;3:245-9. doi:10.2174/1874210600903010245 (opens in a new tab). PMID: 20148135 (opens in a new tab).
  11. Bacchi A, Paludo L, Ferraz Mesquita M, Schuh C, Federizzi L, Oro Spazzin A. Loosening torque of prosthetic screws in metal-ceramic or metal-acrylic resin implant-supported dentures with different misfit levels. J Biomech. 2013;46(7):1358-62. doi:10.1016/j.jbiomech.2013.01.023 (opens in a new tab). PMID: 23477791 (opens in a new tab).
  12. Assunção WG, Barão VA, Delben JA, Gomes ÉA, Garcia IR Jr. Effect of unilateral misfit on preload of retention screws of implant-supported prostheses submitted to mechanical cycling. J Prosthodont Res. 2011;55(1):12-8. doi:10.1016/j.jpor.2010.05.002 (opens in a new tab). PMID: 20627771 (opens in a new tab).
  13. Al-Otaibi HN, Akeel RF. The effects of two torque values on the screw preload of implant-supported prostheses with passive fit or misfit. Int J Oral Maxillofac Implants. 2014;29(5):1058-63. doi:10.11607/jomi.3419 (opens in a new tab). PMID: 25216130 (opens in a new tab).
  14. Abduo J, Judge RB. Implications of implant framework misfit: a systematic review of biomechanical sequelae. Int J Oral Maxillofac Implants. 2014;29(3):608-21. doi:10.11607/jomi.3418 (opens in a new tab). PMID: 24818199 (opens in a new tab).
  15. Burguete RL, Johns RB, King T, Patterson EA. Tightening characteristics for screwed joints in osseointegrated dental implants. J Prosthet Dent. 1994;71(6):592-9. doi:10.1016/0022-3913(94)90443-x (opens in a new tab). PMID: 8040822 (opens in a new tab).
  16. Haack JE, Sakaguchi RL, Sun T, Coffey JP. Elongation and preload stress in dental implant abutment screws. Int J Oral Maxillofac Implants. 1995;10(5):529-36. PMID: 7590997 (opens in a new tab).
  17. Tan KB, Nicholls JI. Implant-abutment screw joint preload of 7 hex-top abutment systems. Int J Oral Maxillofac Implants. 2001;16(3):367-77. PMID: 11432656 (opens in a new tab).
  18. Cantwell A, Hobkirk JA. Preload loss in gold prosthesis-retaining screws as a function of time. Int J Oral Maxillofac Implants. 2004;19(1):124-32. PMID: 14982365 (opens in a new tab).
  19. Martin WC, Woody RD, Miller BH, Miller AW. Implant abutment screw rotations and preloads for four different screw materials and surfaces. J Prosthet Dent. 2001;86(1):24-32. doi:10.1067/mpr.2001.116230 (opens in a new tab). PMID: 11458261 (opens in a new tab).
  20. Abduo J, Lyons K. Effect of vertical misfit on strain within screw-retained implant titanium and zirconia frameworks. J Prosthodont Res. 2012;56(2):102-9. doi:10.1016/j.jpor.2011.09.001 (opens in a new tab). PMID: 22093362 (opens in a new tab).
  21. Abduo J, Swain M. Influence of vertical misfit of titanium and zirconia frameworks on peri-implant strain. Int J Oral Maxillofac Implants. 2012;27(3):529-36. PMID: 22616045 (opens in a new tab).
  22. Altuwaijri SM, Alotaibi HN, Alnassar TM. The effect of the digital manufacturing technique of cantilevered implant-supported frameworks on abutment screw preload. J Adv Prosthodont. 2022;14(1):22-31. doi:10.4047/jap.2022.14.1.22 (opens in a new tab). PMID: 35284054 (opens in a new tab).
  23. Rutkunas V, Dirse J, Kules D, Mischitz I, Larsson C, Janda M. Misfit simulation on implant prostheses with different combinations of engaging and nonengaging titanium bases. Part 2: Screw resistance test. J Prosthet Dent. 2024;131(2):262-271. doi:10.1016/j.prosdent.2022.04.027 (opens in a new tab). PMID: 35981927 (opens in a new tab).
  24. Binon PP. The effect of implant/abutment hexagonal misfit on screw joint stability. Int J Prosthodont. 1996;9(2):149-60. PMID: 8639238 (opens in a new tab).
  25. Arikan Kalayci H, Guncu MB. Biomechanical Effects of Horizontal, Vertical, and Combined Misfits in Full-Arch Implant-Supported Titanium Frameworks: A Three-Dimensional Finite Element Analysis. J Funct Biomater. 2026;17(9). doi:10.3390/jfb17090478 (opens in a new tab). PMID: 42783725 (opens in a new tab).
  26. Theoharidou A, Petridis HP, Tzannas K, Garefis P. Abutment screw loosening in single-implant restorations: a systematic review. Int J Oral Maxillofac Implants. 2008;23(4):681-90. PMID: 18807565 (opens in a new tab).
  27. Katsavochristou A, Koumoulis D. Incidence of abutment screw failure of single or splinted implant prostheses: A review and update on current clinical status. J Oral Rehabil. 2019;46(8):776-786. doi:10.1111/joor.12817 (opens in a new tab). PMID: 31074882 (opens in a new tab).
  28. Pardal-Peláez B, Montero J. Preload loss of abutment screws after dynamic fatigue in single implant-supported restorations. A systematic review. J Clin Exp Dent. 2017;9(11):e1355-e1361. doi:10.4317/jced.54374 (opens in a new tab). PMID: 29302289 (opens in a new tab).
  29. Sammour SR, Ziqi X, Ogawa T, Yoda N. Impact of implant abutment connection designs and cyclic loading on screw stability in dental implants: A systematic review and meta-analysis. J Prosthet Dent. 2026;135(2):317.e1-317.e14. doi:10.1016/j.prosdent.2025.09.035 (opens in a new tab). PMID: 41102040 (opens in a new tab).
  30. Pozzi A, Arcuri L, Carosi P, Laureti A, Londono J, Wang HL. Photogrammetry Versus Intraoral Scanning in Complete-Arch Digital Implant Impression: A Systematic Review and Meta-Analysis. Clin Implant Dent Relat Res. 2025;27(3):e70059. doi:10.1111/cid.70059 (opens in a new tab). PMID: 40481748 (opens in a new tab).
  31. Alfaraj A, Alqudaihi F, Khurshid Z, Qadiri O, Lin WS. Comparative analyses of accuracy between digital and conventional impressions for complete-arch implant-supported fixed dental prostheses-A systematic review and meta-analysis. J Prosthodont. 2026;35(3):252-274. doi:10.1111/jopr.14094 (opens in a new tab). PMID: 40660725 (opens in a new tab).
  32. Pesce P, Nicolini P, Caponio VCA, Zecca PA, Canullo L, Isola G, et al. Accuracy of Full-Arch Intraoral Scans Versus Conventional Impression: A Systematic Review with a Meta-Analysis and a Proposal to Standardise the Analysis of the Accuracy. J Clin Med. 2024;14(1). doi:10.3390/jcm14010071 (opens in a new tab). PMID: 39797155 (opens in a new tab).
  33. DeFee M, Renne W, Shazib MA, Queiroz A, Mennito A, Borbola D, et al. Virtual-fit assessment of full-arch implant photogrammetry accuracy in a human cadaver model. J Dent. 2026;170:106668. doi:10.1016/j.jdent.2026.106668 (opens in a new tab). PMID: 41932641 (opens in a new tab).
  34. Verein Deutscher Ingenieure (VDI). VDI 2230: Systematic calculation of highly stressed bolted joints. Guideline; edition year not verified; located through a web search listing of its table of contents and not read. Available from: https://m.vdi.de/fileadmin/pages/vdi_de/redakteure/richtlinien/inhaltsverzeichnisse/2360550.pdf (opens in a new tab)