Introduction
Bone-anchored hearing devices (BAHAs) are an established treatment for conductive, mixed and single-sided sensorineural hearing loss when conventional amplification is ineffective or contra-indicated.Reference Cass and Mudd1, Reference Cantore, Cianfrone, Tauro, Bevilacqua, De Carli and Bianco2 Initially developed for patients with structural ear abnormalities, indications have expanded to include contact dermatitis, external auditory canal contra-indications to hearing mould use and single-sided deafness.Reference Bento, Kiesewetter, Ikari and Brito3, Reference Wazen, Young, Farrugia, Chandrasekhar, Ghossaini and Borik4 Together with advances in device design, this has increased both implant volume and patient diversity.Reference Wazen, Young, Farrugia, Chandrasekhar, Ghossaini and Borik4 Two implantation systems predominate in current practice: percutaneous devices, which transmit sound through a skin-penetrating abutment, and transcutaneous devices, which maintain intact skin via magnetic coupling between internal and external components.Reference Hol, Nelissen, Agterberg, Cremers and Snik5, Reference Sprinzl and Wolf-Magele6 Although transcutaneous systems are thought to reduce soft-tissue morbidity, percutaneous devices remain the most widely implanted bone-conduction systems and the primary setting in which these complications occur.Reference Lagerkvist, Carvalho, Holmberg, Petersson, Cremers and Hultcrantz7
The permanent skin-abutment interface inherent to percutaneous BAHAs creates a lasting vulnerability at the implant site. The abutment functions as a foreign body that traverses the dermal barrier, disrupts local defence mechanisms, promotes a distinct peri-implant microbiome and exposes the surrounding tissue to chronic mechanical and antigenic stress.Reference Costa, Costa, Soares, Feliciano, Almeida, Sousa and Coutinho8 The resulting soft tissue reactions range from mild erythema to granulation tissue formation, dehiscence, skin overgrowth and abscess, which can progress to implant instability and device extrusion.Reference Costa, Costa, Soares, Feliciano, Almeida, Sousa and Coutinho8
Device extrusion is a clinically meaningful yet distinctly under-characterised mode of failure. Device extrusion has been attributed to two primary mechanisms: failure of osseointegration of the titanium fixture, which accounts for the majority of reported implant losses in adults, and traumatic avulsion of the abutment or fixture from direct mechanical impact.Reference Bezdjian, Smith, Thomeer, Willie and Daniel9 Chronic peri-implant inflammation and infection have also been described as contributors to progressive implant instability, providing a plausible biological pathway through which recurrent soft tissue complications may result in device loss.Reference Costa, Costa, Soares, Feliciano, Almeida, Sousa and Coutinho8
The risk factors for soft tissue complications and BAHA device failure have been broadly studied across three domains: patient-level characteristics, including smoking status, diabetes mellitus, obesity, cardiovascular disease and dermatologic co-morbidities; device-level factors, including implant type, abutment size and osseointegration mechanism; and surgical factors, such as soft tissue thickness and incision technique.Reference Teunissen, Kok, Janssen, Hol and Bosman10, Reference Zammit, Chandrasekar, Sweed and Attlmayr11 Associations between post-operative wound complications and eventual device failure are established in the general implant survival literature, with certain co-morbidities conferring substantially increased hazard for implant loss.Reference Zammit, Chandrasekar, Sweed and Attlmayr11 Nevertheless, the influence of wound complication timing and recurrence on the risk of implant extrusion remains poorly understood. Furthermore, prior studies frequently combined heterogeneous device-removal outcomes, failing to distinguish extrusion from elective explantation as distinct clinical endpoints. This limits the ability to identify patients at increased risk for extrusion and to provide evidence-based counselling regarding long-term implant outcomes.
This retrospective cohort study examined the time-adjusted relationship between demographics, medical co-morbidities and post-implantation wound complications with subsequent device extrusion following primary implantation. By modelling wound complications as time-dependent variables within a longitudinal framework, this study aimed to determine whether the timing and cumulative burden of post-operative wound events independently predict implant extrusion beyond established demographic and medical risk factors.
Methods
Study design
The study was a retrospective cohort analysis of a neurotology group practice, including adults aged 18 years or older who underwent primary bone-anchored implantation between December 2003 and January 2025. The study was approved prior to data collection by the Institutional Review Board at Henry Ford Health in Detroit, MI, USA. Patients were included if they: (1) underwent primary implantation in one or both ears; (2) had post-operative audiogram evaluation; (3) had a medical history available; and (4) had at least one post-operative visit. Patients were analysed on a per-ear basis for bilateral implantation.
Outcomes
The primary endpoint was device extrusion, defined as the unplanned loss or detachment of the percutaneous abutment and/or the osseointegrated implant fixture not caused by elective surgical removal. Extrusion events were classified as spontaneous or traumatic based on retrospective clinical documentation. Extrusion events were classified as spontaneous when no external precipitating event was documented or traumatic when there was an identifiable and documented inciting event, often involving direct physical impact or trauma. Elective surgical removal was not classified as extrusion.
Prognostic variables included early (<60 days) or recurrent (>3 instances) wound complications, demographics (age at primary implantation, year of implantation, gender and ethnicity), body mass index, medical co-morbidities, including temporal bone radiation, and contralateral (better-hearing) audiometry. Bone-anchored device characteristics were also included, including abutment length and device subtype (percutaneous or transcutaneous). Detailed operative variables describing surgical technique were inconsistent across the study period and could not be reliably incorporated; however, implantation year was used as a surrogate marker for surgical technique.
Pure tone averages (PTAs) for pre-operatively evaluated patients were assessed according to the American Speech-Language-Hearing Association guidelines for audiometry, using the average of 500-, 1000-, 2000- and 4000-Hz pure tone thresholds. The Holgers grading scale was not utilised in the analysis because standardised peri-abutment assessments could not be reliably obtained from retrospective documentation.
Statistical analysis
Descriptive statistics were generated for demographic co-variates, contralateral audiometric data and wound complication characteristics. Time-to-event data were analysed using univariate and multivariate Cox proportional hazards regression, with hazard ratios and 95 per cent confidence intervals (CIs) calculated for device extrusion. Multicollinearity among co-variates in the multivariate model was evaluated using variance inflation factors, with a threshold of 5 used to flag potential collinearity.
To assess the proportional-hazards assumption, the multivariate Cox regression model was evaluated with Schoenfeld residuals and the resulting plots were visually inspected. Most co-variates satisfied the proportional hazards assumptions (p > 0.05), but age at implantation and radiation exposure (p = 0.042 and p = 0.015, respectively) showed evidence of non-proportionality. The plots were visually inspected and showed only minor departures from proportionality; given their clinical significance and the limited number of radiation-exposed patients (n = 5), they were included in the analysis. The multivariate model included only variables associated with extrusion on univariate analysis (p < 0.10) and clinically relevant demographic and device-related factors to minimise model overfitting, given the number of events observed.
Early and frequent wound complications were modelled as time-dependent variables using a counting-process Cox regression framework, given that these features accumulate over the post-operative course rather than being established at the time of surgery. The implantation year was included as a potential confounder, given the evolution in implant technique and technology, and was categorised in a sensitivity analysis (before or after 2014) based on approximate institutional trends in surgical approach. The cumulative incidence of device extrusion over 10 years following primary implantation was estimated using time-dependent Cox regression with appropriate censoring. All statistical analyses were performed in RStudio (R version 4.5.2).12 Significance was set at p < 0.05.
Results
Study cohort demographics
A total of 1090 primary bone-anchored implants from 2003 to 2025 met the inclusion criteria, of which 73 (6.7 per cent) experienced device extrusion over the study period. There were 20 (27.2 per cent) traumatic and 53 (72.6 per cent) spontaneous extrusions. The cohort was predominantly female (61.1 per cent) and White (94.0 per cent), with a median patient age of 54.6 years (interquartile range, 41.6–64.3). Age and gender distributions were broadly comparable between patients who experienced extrusion and those who did not, with the 50–59 years age group accounting for the largest proportion of extrusion events (27.4 per cent). All 73 extrusion events occurred among recipients of percutaneous devices. No extrusions were observed among the 39 patients who received transcutaneous implants (Table 1). Based on a sensitivity analysis of the implantation era, there was no significant difference in extrusion risk between implants placed before and after 2014 (hazard ratio = 1.36; 95 per cent CI = 0.74–2.24; p = 0.23).
Cohort demographics and characteristics stratified by the presence or absence of implant extrusion
Table 1 Long description
Comparison of demographics, device type, wound complications, hearing measures, and co-morbidities for 1,090 implants, split into 73 with implant extrusion and 1,017 without. Median age was similar across groups, in the mid 50s, and the age-group distribution was broadly comparable. Gender distribution was similar, with females about three fifths overall and slightly lower in the extrusion group. Race and ethnicity were predominantly White in both groups, with small counts in other categories and some missing data. All extrusion cases used percutaneous devices, while transcutaneous devices appeared only in the no-extrusion group. Frequent wound complications and early wound complications occurred at similar rates in both groups. Hearing differed modestly: the extrusion group had a slightly higher median contralateral pure tone average and a lower share with normal contralateral hearing. Co-morbidities showed larger differences: obesity and active smoking were more common with extrusion, and temporal bone radiation was rare but proportionally higher in the extrusion group; diabetes and hypertension were similar between groups.
* Race and ethnicity were not available for all patients (n = 16). IQR = interquartile range; PTA = pure tone average; dB HL = decibel hearing loss; BMI = body mass index
Obesity was more prevalent among patients who experienced extrusion compared with those who did not (42.5 per cent vs 25.4 per cent), as were active smoking (19.2 per cent vs 13.6 per cent) and temporal bone radiation (2.7 per cent vs 0.3 per cent). Rates of type II diabetes, hypertension and psychiatric mood disorder were similarly distributed across the extrusion and non-extrusion groups, with all having a rate difference of less than 4 per cent nominally. Wound complication burden did not differ meaningfully between groups. Median contralateral PTA was modestly higher in the extrusion group (23.8 decibel hearing loss (dB HL); interquartile range, 13.8–36.2) relative to those without extrusion (20.0 dB HL; interquartile range, 11.2–33.8). As for normal contralateral hearing, those with extrusion had slightly lower rates of normal contralateral hearing than those without extrusion (49.3 per cent vs 58.5 per cent) (Table 1).
Proportional hazards regression
On univariate Cox regression, temporal bone radiation (hazard ratio = 7.74; 95 per cent CI = 1.88–31.82; p = 0.005) and obesity (hazard ratio = 2.10; 95 per cent CI = 1.30–3.39; p = 0.002) were significantly associated with device extrusion. Abutment length was also a significant univariate predictor of extrusion (hazard ratio = 1.15 per mm; 95 per cent CI = 1.02–1.31; p = 0.028), suggesting a modest incremental increase in extrusion risk with longer abutments. Active smoking trended toward significance on univariate analysis (hazard ratio = 1.63; 95 per cent CI = 0.91–2.92; p = 0.10) but did not reach the prespecified threshold. Neither early wound complications (hazard ratio = 1.18, 95 per cent CI = 0.60–2.29; p = 0.64) nor frequent wound complications (hazard ratio = 1.02; 95 per cent CI = 0.37–2.84; p = 0.97) were associated with increased extrusion risk. The total wound complication counts similarly showed no significant association (hazard ratio = 0.88; 95 per cent CI = 0.72–1.09; p = 0.25). Age, gender, racial minority group membership and contralateral PTA were not significant on univariate analysis (Table 2).
Univariate and multivariate hazard ratios for risk of bone-anchored device extrusion after primary implantation
Table 2 Long description
Univariate and multivariate hazard ratios with confidence intervals and p values are reported for factors associated with bone-anchored device extrusion after primary implantation. In both analyses, temporal bone radiation is linked to a markedly higher extrusion risk, with a strong association in univariate results and a still-elevated association after adjustment. Obesity is also consistently associated with about a twofold higher risk in both univariate and multivariate models. Abutment length shows increased risk in univariate analysis and remains suggestive after adjustment, with a p value just under 0.10. Age, male gender, and racial minority status show hazard ratios near one and are not statistically significant in either model. Wound complication measures, including frequent or early complications, do not show clear associations, and the continuous wound complication count is only reported for univariate analysis. Better-ear hearing level (contralateral pure tone average) is not associated with extrusion risk. Active smoking trends toward higher risk, with p values around 0.10 in univariate and below 0.10 in multivariate, but confidence intervals include no effect, so this finding is uncertain.
p values less than 0.10 are shown in bold. CI = confidence interval; PTA = pure tone average; dB HL = decibel hearing loss; BMI = body mass index
In multivariate Cox regression, temporal bone radiation (adjusted hazard ratio = 5.65; 95 per cent CI = 1.35–23.70; p = 0.018) and obesity (adjusted hazard ratio = 2.17, 95 per cent CI = 1.33–3.55; p = 0.002) remained the only independent predictors of device extrusion. Only five implants were exposed to radiation, of which two resulted in extrusion. Abutment length showed a trend towards independent significance that did not meet the prespecified threshold (adjusted hazard ratio = 1.13; 95 per cent CI = 0.98–1.30; p = 0.098). Wound complication-related variables were not significant in either univariate or multivariate analyses.
Time-to-extrusion analysis
The overall extrusion incidence at 1 year was 3.3 per cent (95 per cent CI = 2.2–4.3 per cent), and at 5 and 10 years was 6.4 per cent (95 per cent CI = 4.7–8.1 per cent) and 10.2 per cent (95 per cent CI = 7.4–12.8 per cent), respectively (Figure 1). Kaplan–Meier analysis demonstrated that device extrusion occurred most frequently in the spontaneous cohort during the 6 months following primary implantation, with an incidence at 1 year of 2.6 per cent (95 per cent CI = 1.7–3.5 per cent) compared with traumatic mechanism at 0.5 per cent (95 per cent CI = 0.2–0.9 per cent), after which the rate of increase slowed but remained linear throughout the follow-up period. When stratified by mechanism, spontaneous extrusion occurred more frequently than traumatic extrusion, with the cumulative incidence curve for spontaneous extrusion remaining consistently above that of traumatic extrusion throughout follow up. Although traumatic extrusions were less common, their cumulative incidence continued to increase gradually over time. Neither spontaneous nor traumatic extrusion reached a plateau during the 10-year follow up (Figure 1).
Kaplan–Meier survival curve for bone-anchored implant extrusion through spontaneous and traumatic mechanisms for the adult cohort.
Figure 1 Long description
A line graph of cumulative incidence of extrusion for an adult cohort, with three step like curves and shaded confidence bands. The horizontal axis label is, Time since implantation left parenthesis years right parenthesis, with tick labels 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10. The vertical axis label is, Cumulative incidence of extrusion left parenthesis percent right parenthesis, with tick labels 0 percent, 2 percent, 4 percent, 6 percent, 8 percent, 10 percent, 12 percent, 14 percent. Legend entries are, Overall, Spontaneous, Traumatic. The Overall curve is the highest across the full time range, the Spontaneous curve stays between Overall and Traumatic and the Traumatic curve is the lowest. All three curves rise in small steps over time and do not flatten by 10 years. Shaded bands surround each curve to show uncertainty. Number at risk table below the plot has the header, Number at risk. Time columns align with the x axis ticks 0 through 10. One row of values is shown as 1090, 884, 756, 659, 556, 464, 391, 312, 255, 214, 169.
Discussion
This retrospective cohort study provides a longitudinal characterisation of device extrusion as a discrete failure endpoint in adults following primary BAHA implantation. This distinction has historically been obscured in the literature by the conflation of extrusion with elective surgical removal. Across 1090 implant ears with over two decades of review, the overall extrusion rate of 6.7 per cent was consistent with previously reported total implant loss estimates. No extrusion events occurred among recipients of transcutaneous devices.
Contrary to the study’s primary hypothesis, neither early (<60 days) nor frequent (>3 instances) wound complications independently predicted device extrusion. Instead, temporal bone radiation and obesity emerged as independent predictors, suggesting distinct clinical risk factors for different failure modes. The present study did not evaluate biological or mechanical mechanisms underlying elective explantation versus extrusion and thus cannot establish the causal factors. Future hypothesis-driven studies are needed to better characterise the local factors that lead to device failure. Elective explantation is a provider-initiated intervention that may be in response to persistent soft-tissue intolerance or other clinical factors, whereas extrusion is the undesired loss of the device.
Temporal bone radiation and obesity were independent predictors of extrusion. Although the radiated subgroup was small (n = 5), this association is biologically plausible but should be interpreted cautiously, as only two of five exposed implants resulted in extrusion. The calculated adjusted effect size (adjusted hazard ratio = 5.65; 95 per cent CI = 1.35–23.70; p = 0.018) was based on sparse data consistent with the resulting CI, indicating uncertainty in the estimate. Ionising radiation damages the bone microvasculature and impairs the cellular repair mechanisms necessary for osseointegration, creating conditions that may not be apparent at the time of implantation but that progressively undermine fixture stability over the years.Reference Omar, Rydén, Wamied, Al‐Otain, Alhawaj and Abuohashish13–Reference Li, Chen, Ren, Chen, Wu and Wang16 Implant instability in irradiated patients has been observed with cochlear implants and dental fixtures.Reference Ahmad, Lovin, Lee, Nader and Gidley17–Reference Mertens, Staudinger, Smielowski, Rückschloss, Schnug and Hoffmann19 In addition, increasing body mass may contribute to implant instability by impairing wound healing, altering tissue perfusion, promoting chronic inflammation or increasing mechanical loading at the implant site.Reference Stevens, O’Connell and Meyer20–Reference Insua, Galindo‐Moreno, Miron, Wang and Monje23 Patients with a history of temporal bone radiation or obesity who are being considered for percutaneous implantation should be counselled on this risk explicitly, and heightened post-operative surveillance is prudent in this population.
Smoking and longer abutment length showed trends towards increased extrusion risk but did not reach statistical significance. Both findings are biologically plausible given the effects of smoking on microvascular and bone remodelling, and the potential for increased mechanical stress with longer abutments.Reference Ghanem, Abduljabbar, Akram, Vohra, Kellesarian and Javed24–Reference den Besten, Nelissen, Peer, Faber, Dun and de Wolf26
Kaplan–Meier analysis demonstrated that cumulative extrusion incidence rose most steeply in the first four months following implantation, consistent with the known biology of early osseointegration failure, before levelling off to a more gradual accumulation rate across the remainder of follow up. Spontaneous extrusion was more common than traumatic extrusion at all time points and showed a more pronounced early concentration. In contrast, traumatic extrusions accumulated more steadily over time, as one would expect from events tied to unpredictable external forces rather than a defined biological window. Clinically, these findings suggest that post-operative surveillance to detect spontaneous extrusion is most critical early, while monitoring for traumatic events requires sustained attention throughout the device’s lifespan.
The study has several limitations. Because of its retrospective design, causal relationships cannot be established. In addition, the quality and completeness of documentation over the 22-year study period were inherently variable, particularly in characterising extrusion events and their potential precipitating factors. There was no demonstrated association between early (≤2014) and late (>2014) implantation, which was included as a surrogate marker for surgical technique; however, evolution in approach cannot be excluded as a potential confounder. Systematic Holgers grading could not be applied, limiting the ability to objectively link soft-tissue severity to extrusion risk. The cohort was predominantly White (94 per cent), reflecting the institution’s demographics and limiting generalisability to more diverse populations. The temporal bone radiation subgroup was small (n = 5). Although the finding is statistically significant and biologically coherent, it should be interpreted cautiously and replicated in larger samples before drawing firm clinical conclusions. Finally, the classification of individual extrusion events as spontaneous or traumatic relied on chart notation, and misclassification is possible when the precipitating mechanism was not clearly documented at the time of the clinical encounter.
• Bone-anchored hearing aids are an established treatment option for conductive, mixed and single-sided deafness, with percutaneous and transcutaneous device subtypes
• Post-operative wound complications and infections are demonstrated predictors of surgical device explantation, more common in percutaneous implants with a permanent skin-abutment interface
• Previous investigations combined elective explantation and extrusion, failing to distinguish spontaneous and traumatic extrusion as a distinct clinical failure endpoint
• In this large retrospective cohort of 1090 implanted ears, the 10-year cumulative probability of extrusion was 10.2 per cent and occurred exclusively in percutaneous implants, with spontaneous extrusion being the predominant mode of failure
• Temporal bone radiation (adjusted hazard ratio = 5.65) and obesity (adjusted hazard ratio = 2.17) were independent predictors of device extrusion on multivariate modelling, whereas early (<60 days) and frequent (>3 instances) wound complications were not predictive, indicating that extrusion and elective explantation may have distinct clinical risk factors
Conclusion
In this longitudinal cohort of 1090 adults following primary BAHA implantation, the overall device extrusion rate of 6.7 per cent and the 10-year extrusion rate of 10.2 per cent are consistent with prior estimates of total implant loss in the percutaneous BAHA literature. Early and frequent post-operative wound complications, while established predictors of surgical device explantation, were not independently associated with extrusion, suggesting that elective explantation and extrusion may have different clinical risk profiles; however, additional studies are needed to establish the biological and mechanical mechanisms underlying failure. Temporal bone radiation and obesity were independent predictors of extrusion, whereas wound complications were not. Patients with these risk factors may benefit from closer post-operative surveillance.
Data availability statement
Data are available from the authors on request.
Competing interests
SB reports receiving honoraria from Integra, Olympus and Stryker. All other authors: None declared.
Open access
