Complications and Factors Influencing Prosthesis Size Adjustment and Replacement Interval in Post-Laryngectomy Patients: A Single-Center Retrospective Study

1. Introduction

Laryngeal cancer continues to pose a significant global health burden. Recent worldwide estimates indicate that in 2022, laryngeal cancer ranked among the most common malignancies, ranking 18th among cancer-related causes of death and as the 20th most common cancer [1], and these GLOBOCAN 2022 estimates remain the most recent comprehensive global figures available, as such estimates are compiled and updated only periodically. Since the first total laryngectomy, performed by Theodor Billroth in 1873, loss of natural voice has remained one of the most challenging consequences of this surgical treatment [2].

Surgical voice rehabilitation advanced substantially in 1972, when Polish laryngologist Professor Erwin Mozolewski first described a semi-permanent tracheoesophageal voice prosthesis made of silicone with a one-way valve. Although initially published in Polish and long under-recognized internationally, this work predated the widely known Blom-Singer prosthesis introduced in 1980 and laid the foundation for current tracheoesophageal voice restoration techniques [5,6,7]. The subsequent development of tracheoesophageal puncture combined with a one-way valve prosthesis represented a major breakthrough, allowing many patients to regain intelligible speech when other rehabilitation methods proved insufficient [3,4].

Early indwelling devices, followed by more advanced low-resistance models such as Provox, have demonstrated reliable performance — including lower airflow resistance, more consistent phonation, and favorable long-term device survival — in comparative clinical evaluations [8]. The principal factor limiting the long-term function of silicone prostheses is microbial colonization, which leads to valve incompetence and eventual leakage. While transprosthetic leakage is the most frequent trigger for replacement, fistula-related complications -- including tract widening, stenosis, inflammation, or granulation -- often require more complex management strategies [9,10,11].

In our department, routine Provox-based voice rehabilitation was implemented in 2020, accompanied by standardized documentation of every replacement procedure from the beginning. Elective exchanges are typically scheduled every 3–4 months during outpatient follow-up, while urgent procedures are performed on the same day in cases of leakage with aspiration risk. Both elective and urgent replacements are performed as day-case procedures under local anesthesia, with pre-procedural blood tests (complete blood count, INR, APTT) and same-day discharge when no complications arise; general anesthesia is reserved only for the rare secondary interventions described in the Complications section (e.g., re-insertion after fistula closure).

The objective of this study was to examine our six-year experience with Provox voice prosthesis replacement, evaluating (1) factors influencing the time interval between replacements — including the cumulative number of prior procedures and elective versus reactive status — and (2) indications for replacement, modifications in prosthesis size, and the occurrence of intra- and post-procedural complications.

2. Materials and Methods

This retrospective single-center study encompassed all Provox voice prosthesis replacements performed between May 2020 and March 2026 at the Department of Otorhinolaryngology and Oncological Laryngology, Specialist Hospital, Zabrze, Poland. Data were derived from hospital electronic medical records and prospectively completed standardized Provox replacement forms. Only indwelling Provox models were used: Provox Vega (PV), Provox Vega XtraSeal (PVXS), and Provox Vega ActiValve (Atos Medical, Sweden), with shaft lengths from 4.0 to 12.5 mm. Collected parameters included patient demographics (age, sex), prosthesis type and size before and after exchange, indication for replacement (elective vs reactive: valve leakage, perivalvular leak, aphonia, obstruction, fistula issues, other), complications, and clinical notes. All procedures were performed on an ambulatory basis by department staff, including residents and consultants. Urgent replacements occurred on the day of presentation when clinically indicated.

2.1. Study Population and Data Structure

The dataset included 82 patients who underwent a total of 147 voice prosthesis replacement procedures during the study period (December 2024 to February 2026). The number of procedures per patient was limited to recent observations (median: 2; mean: 1.8; range: 1–4), although the number of observations included in individual analyses varied slightly due to missing data for selected variables.

Historical data prior to the study period were not available in the same structured format. However, some patients had undergone their initial prosthesis insertion before December 2024 and had a history of multiple prior procedures. Procedures were performed by 13 physicians, with three operators accounting for 69% of all procedures during the study period.

See Table 1 for the summary of the baseline characteristics of the study population.

2.2. Statistical Analysis

Statistical analyses were conducted using mixed-effects models with a random intercept for patient to account for repeated observations within individuals. Linear mixed-effects models were applied for continuous outcomes (time between procedures), whereas mixed-effects logistic regression models were used for binary outcomes including complications and procedural characteristics, such as prosthesis size or type change (PV vs PVXS) and procedural status (elective vs expedited). Results are presented as regression coefficients (β) for linear models and odds ratios (ORs) for logistic models, each with corresponding 95% confidence intervals (CIs) and p-values. Between-patient variability was quantified using the intraclass correlation coefficient (ICC). Covariates were selected a priori based on clinical relevance and included patient demographics (age, sex), procedural characteristics (indication for replacement, prosthesis size and type change, procedural status), and treatment history (number of prior procedures, time since initial insertion). Statistical significance was defined as p < 0.05. Analyses were performed in Stata 19.5. Models with convergence issues were interpreted with caution. In cases of perfect or quasi-complete separation, results were summarized descriptively using contingency tables.

3. Results

3.1. Time Interval Between Replacement Procedures

A mixed-effects linear regression model with a random intercept for patient was used to evaluate factors associated with the time interval between procedures. Data on procedure dates were available for the entire historical patient cohort (n = 1059 procedures; 899 intervals from May 2020 to March 2026); in the null model, the average interval between prosthesis replacements was 122.5 days (95% CI: 114.5–130.4), with substantial between-patient variability (variance of the random intercept = 1481.7) and an intraclass correlation coefficient (ICC) of 0.32 (95% CI: 0.24–0.42).

Within the specific study sample analyzed in detail here (147 procedures in 82 patients, December 2024–February 2026), the null model estimated a mean interval of 117.3 days (95% CI: 103.5–131.0), again with substantial between-patient variability (variance of the random intercept = 1462.8) and a higher ICC of 0.55 (95% CI: 0.25–0.82), indicating that approximately 55% of the total variability in replacement interval was attributable to between-patient differences.

In the multivariable model (n = 65 procedures in 50 patients with complete data for all included covariates), only two factors were significantly associated with the interval between procedures (Table 2). A higher number of prior procedures was associated with a shorter interval (β = −3.70 days per procedure; 95% CI: −5.84 to −1.576; p < 0.001), indicating progressively shorter intervals between successive procedures. In contrast, planned procedures were associated with longer intervals (β = 35.47 days; 95% CI: 15.80 to 55.14; p < 0.001).

Between-patient variability remained substantial in the adjusted model (random intercept variance = 796.13), and the likelihood ratio test confirmed that the mixed-effects model provided a better fit than a standard linear model (p = 0.007).

Among the tested but non-significant covariates, gender was not a significant predictor of the interval between events in the mixed-effects model (β = 5.73, p = 0.567), although a simple t-test indicated a significant difference (mean difference = 17.05 days, p = 0.0025); the mixed-effects estimate is considered more appropriate for these repeated-measures data.

3.2. Complications

Complications occurred in 7 of the 147 voice prosthesis replacement procedures (4.8%); these represent intra- or post-procedural adverse events related to the exchange itself, distinct from leakage and aphonia, which were classified as indications for replacement rather than complications. The complications observed included granulation tissue formation (n=2), prosthesis displacement within the fistula (n=1), prosthesis displacement with subsequent tracheoesophageal fistula closure (n=2), and swallowing of the prosthesis (n=2).

Granulation tissue formation (n=2) required surgical removal under local anesthesia followed by systemic antibiotic therapy (amoxicillin-clavulanate for 7 days). Prosthesis displacement was noted in three cases. In one case the prosthesis was successfully removed and replaced with a new one during the same outpatient procedure. In the remaining two cases, displacement led to closure of the tracheoesophageal fistula, necessitating referral for secondary voice prosthesis insertion under general anesthesia. Swallowing of the prosthesis occurred in two patients. Immediate posteroanterior and lateral chest X-ray showed no foreign body in the gastrointestinal tract or airways. No life-threatening complications were recorded. The low complication rate (4.8%) supports the high safety profile of outpatient voice prosthesis exchange in a clinical ENT department.

All observed complications occurred in male patients following a change in prosthesis size and were associated with a PV 10 mm prosthesis at removal, suggesting a potential pattern; however, these findings should be interpreted with caution given the small number of events. No factors were significantly associated with prosthesis size change in the mixed-effects logistic regression analyses. Among the reasons for unscheduled (reactive) replacement, central leakage accounted for approximately 50%, peripheral leakage for 25%, and prosthesis dysfunction without phonation for 19%, with other causes accounting for the remainder (<6%). Planned (elective) procedures were significantly less likely to require a change in prosthesis size (OR = 0.33, 95% CI: 0.12–0.92, p = 0.034).

5. Discussion

This retrospective study of 82 laryngectomized patients undergoing a total of 147 voice prosthesis replacements reveals two clinically important and potentially modifiable patterns in long-term tracheoesophageal voice rehabilitation. Taken together, these results address both central aims of the study: identifying modifiable factors that prolong device lifetime, and characterizing the complications and prosthesis adjustments associated with reactive versus planned replacement. Each additional prior replacement was independently associated with a progressive shortening of the interval to the next exchange by 3.7 days (β = −3.70 days; 95% CI −5.84 to −1.56; p < 0.001). In contrast, planned elective replacements significantly extended device lifetime by more than one month (β = +35.47 days; 95% CI 15.80–55.14; p < 0.001) and markedly reduced the need for intraoperative size or type change (OR = 0.33; 95% CI 0.12–0.92; p = 0.034).

When placed in the broader context of our entire institutional experience (1059 procedures and 899 intervals with a mean device lifetime of 122.5 days), these findings provide compelling longitudinal evidence for a paradigm shift from reactive, leakage-driven management toward scheduled prophylactic replacement protocols.

Tracheoesophageal voice restoration has evolved significantly since its introduction , and early comparative studies confirmed the clinical superiority of modern prostheses over older designs [8]. Our observations align closely with and meaningfully extend the recent literature advocating proactive strategies. Okła et al. demonstrated that systematic prophylactic replacement every three months reduced tracheoesophageal fistula-related complications from 47% to 6% (p < 0.001), corresponding to a 69% relative risk reduction compared with reactive replacement [12]. The mean interval in our extended cohort (122.5 days) falls squarely within the 3–6 month range reported in major international series [11,13,14], yet the progressive acceleration of replacement frequency in patients with multiple prior procedures identifies a distinct “frequent-replacer” phenotype.

In the present series, central (transprosthetic) leakage accounted for approximately 50% of unscheduled prosthesis exchanges, peripheral (periprosthetic) leakage for 25%, and prosthesis dysfunction without adequate phonation for 19%. These findings are broadly consistent with the literature, in which transprosthetic leakage due to biofilm formation and valve deterioration remains the leading indication for replacement. However, the 25% rate of peripheral leakage observed in our cohort aligns closely with the 20–30% incidence of periprosthetic leakage reported by Lorenz et al. in their 20-year retrospective analysis of 232 laryngectomized patients. Notably, Lorenz documented a cumulative periprosthetic leakage rate of 35.7%, with substantial fistula enlargement requiring multiple interventions occurring in 12.5% of cases [15]. Although our data are expressed as proportions of exchanges rather than cumulative patient incidence, the substantial contribution of peripheral leakage underscores that fistula-related complications—particularly tract widening and granulation—continue to represent a clinically significant burden, often necessitating more complex management strategies beyond simple device replacement.

This distribution strongly suggests cumulative effects of biofilm colonization and subtle tracheoesophageal fistula instability that can be effectively interrupted by elective scheduling. The low overall complication rate of 4.8%, with all events confined to expedited procedures, further underscores the safety benefit of shifting to planned replacement. Moreover, the substantial between-patient variability (ICC 0.55) indicates that while standardized protocols are feasible at the population level, individualized timing remains important -- a balance best achieved through structured prophylactic regimens rather than ad-hoc decision-making. Strengths of this study include the application of mixed-effects modeling to properly account for repeated measures and inter-individual heterogeneity, the integration of recent data with extensive historical records spanning nearly six years, and the provision of robust real-world evidence from a high-volume Polish tertiary center. These aspects address notable gaps in the predominantly Western literature.

Limitations

The retrospective design and relatively modest size of the recent cohort limited statistical power for the analysis of rare complications. Detailed information on radiotherapy dose, gastroesophageal reflux, and specific prosthesis characteristics was not uniformly available, factors known to influence device lifetime.

Furthermore, our analysis treated Provox prostheses as a broadly homogeneous group; expected replacement intervals, however, differ substantially by device model — Provox Vega ActiValve, for example, is designed for markedly longer wear than the standard Vega or Vega XtraSeal models that predominated in our cohort — and device type was not formally tested as a predictor of replacement interval in the present analysis. Future studies should stratify outcomes by prosthesis model to clarify this heterogeneity.

Nevertheless, the consistency of our key findings with recent prospective and multicenter data supports their broader applicability.

In summary, this analysis provides strong real-world evidence that scheduled prophylactic voice prosthesis replacement can effectively break the cycle of increasingly frequent unscheduled interventions while preserving excellent safety. Adoption of such protocols represents a simple, low-cost opportunity to optimize long-term voice rehabilitation and quality of life for laryngectomized patients worldwide, where laryngeal cancer continues to represent a major oncological burden [16]. Future multicenter prospective trials should now focus on determining the optimal replacement interval and quantifying patient-reported outcomes under standardized prophylactic regimens. While tracheoesophageal voice prostheses remain the gold standard for surgical voice rehabilitation after total laryngectomy, several promising innovations are currently under investigation. These include 3D-printed patient-specific prostheses [17,18], and next-generation high-performance indwelling devices [19]. Recent advancements in surface electromyography (sEMG)-based silent speech interfaces offer a promising alternative for patients experiencing frequent complications with traditional voice prostheses. Raiff et al. (2024) demonstrated that continuous pitch modulation derived from muscle activity produces significantly more natural-sounding speech compared to monotonic electrolarynx contours, with listeners (including laryngectomized patients and their partners) prioritizing sound quality, intelligibility, and prosody. Overall, our data support the safety and potential benefits of shifting from reactive to proactive voice prosthesis management. However, the current body of evidence remains limited by small sample sizes, short follow-up periods, and a lack of standardized outcome measures. While innovative approaches such as sEMG-based silent speech systems show promise in restoring natural prosody [20], and scoping reviews emphasize the ongoing burden on voice-related quality of life [21], there is a clear need for large-scale, multicenter studies that directly compare different rehabilitation strategies, including scheduled prosthesis exchange versus emerging non-invasive technologies. Addressing these gaps will be essential to develop evidence-based guidelines that optimize both functional and psychosocial outcomes for patients after total laryngectomy.

6. Conclusions

This single-center study demonstrates that a higher cumulative number of prior voice prosthesis replacements is associated with progressively shorter intervals to the next exchange (β = −3.70 days; p < 0.001), while planned elective replacements significantly prolong device lifetime (β = +35.47 days; p < 0.001) and reduce the need for prosthesis size change (OR = 0.33; p = 0.034). These findings strongly support the implementation of scheduled, proactive replacement protocols to interrupt the cycle of frequent unscheduled procedures and improve long-term tracheoesophageal voice rehabilitation.

Despite a low overall complication rate of 4.8%, the substantial between-patient variability in replacement intervals highlights the importance of individualized yet protocol-driven management. Future multicenter prospective studies are needed to define the optimal replacement schedule and confirm the clinical benefits of routine planned exchanges.