A systematic review of the literature was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement (24). The search was simultaneously conducted on the PubMed, Scopus, and Web of Science online databases, and updated to August 16, 2021. In order to retrieve all the publications dealing with (C)TRA for laryngo-tracheal involvement by TC the query string was composed as (tracheal resection) OR (tracheal involvement) AND (thyroid cancer) OR (thyroid neoplasm) OR (thyroid tumor). The search was conducted by two authors (C.P. and D.L.) who independently assessed the eligibility of the studies by screening article titles and abstracts, and then discussed their inclusion by reading the full-text of the selected works. Discrepancies were clarified by discussion between authors.

The Population/problem Intervention/exposure Comparison, Outcome, and Study design (PICOS) model was adopted for the review (25) ( Table 1 ). The inclusion criteria were as follows: English language, publication from January 1, 1985 to the last day of online search (August 16, 2021), articles including data on (C)TRA for airway involvement by TC and reporting a case series of at least 5 patients. Exclusion criteria were: case reports, case series with less than 5 patients, papers purely describing results of surgical techniques different from (C)TRA (e.g., shaving, window resections, total laryngectomy) or that did not report sufficient data on outcomes and complications and focused on other related issues (e.g., radiological or clinical diagnosis, anesthetic issues, adjuvant treatments). Additionally, papers with duplicated or overlapping data from the same center were excluded, maintaining, when possible, the largest and more recent study among those available. Finally, a case series published by the first author (C.P.) (26), already included in this systematic review, was updated with data of patients treated from the time of the article publication (2016) to date, and their oncologic outcomes updated accordingly.

For each paper included in the systematic review, at the end of the selection process, evaluation of its quality was carried out following the Newcastle-Ottawa Scale (NOS) adapted for cross-sectional, cohort and case-control studies (27). The NOS was considered the evaluation method of choice, based on the recent literature (28, 29). The quality assessment was independently estimated by two different authors (D.L. and M.T.).

Data on study design, number of patients, age, gender, diagnostic work-up, TC histology and degree of airway invasion, length and type of resection, perioperative mortality, surgical complications, rate of patients who remained tracheostomy-dependent after (C)TRA, and oncological outcomes were collected, and a specific database was built.

The primary outcome was proportion of patients who developed a complication, calculated as the number of patients with reported complications divided by the total number of patients treated by (C)TRA for TC. Secondary outcome was the proportion of tracheostomy-dependent patients, defined as the number of patients with long-term tracheostomy dependency divided by the total number of patients treated by (C)TRA.

Meta-analysis of proportions was conducted through a generalized linear mixed model based on logit transformation (30). Pooled analyses are presented as forest plots. For each study, proportions and relative 95% confidence interval (CI) are depicted as gray squares and horizontal lines, respectively. The weight of each study on the overall effect estimate is reported and represented by the square size. The pooled proportion estimate and relative 95% CI, depicted as a diamond, are reported at the bottom of the forest plot. Heterogeneity between studies was assessed with Higgins I² and τ² tests (31), defined as low if I²<25%, moderate if between 25-50%, and substantial if >50% (32).

Publication bias was assessed through funnel plot assessment. Statistical analysis was performed with R (version 4.0.5, R foundation for Statistical Computing, Vienna, Austria). Statistical significance was defined as p<0.05.

The initial literature search yielded 1196 titles (525 records came from the PubMed database, 407 from Scopus, and 264 from Web of Science). Among these, 519 articles were excluded because present in two databases, and 147 due to publication in a language other than English. One article (33) was added from other sources, after being identified through the references of other manuscripts. Three-hundred-seventy-three articles were excluded after review of the title, and 48 by the abstract. From the remaining 110 full-text articles, 73 were excluded because they did not meet the eligibility criteria. Finally, 37 papers (3, 26, 33–67) were considered appropriate for the present systematic review ( Figure 1 , Table 2 ).

According to the NOS adapted for cross-sectional studies (range of the scale, 0-9), the scores ranged from 2 to 7 (median, 5). Detailed scores for each article are reported in Table 3 . All included manuscripts were retrospective single institution cross-sectional studies, except for one (42), which was a retrospective bi-institutional case series.

Overall, 656 patients were included in the current systematic review. Gender of patients treated by (C)TRA was detailed in 18 papers (26, 34, 35, 38, 40, 41, 43, 46, 50–53, 55, 57, 58, 62, 64, 66) for a total of 355 patients, of whom 59% were females.

Age of patients was reported in 12 manuscripts (26, 34, 39–41, 50–53, 57, 64, 66) for a total of 121 patients, with a mean of 60 years (range, 20-85).

Seventeen manuscripts reported data on (C)TRA for DTC alone (3, 33, 34, 38, 40, 41, 43–45, 53, 58–63, 66), while 20 for mixed histologies (26, 35–37, 39, 42, 46–52, 54–57, 64, 65, 67). Overall, the distribution of histopathological types in patients treated by (C)TRA was detailed for 376 of them (59%) and their frequency in descending order was as follows: papillary (79%), follicular (7%), poorly differentiated (5%), medullary (2%), Hürtle cell (2%), anaplastic (1%), follicular variant of papillary cancer (1%), metastasis to the thyroid gland from other organs (1%), and rare histotypes such as thyroid squamous cell carcinoma, giant cell carcinoma, and carcinoma with lymphoepithelioma-like pattern (2% all together) (26, 34, 35, 37–40, 42, 44–47, 49–53, 55, 57, 62, 64, 66).

Thirteen articles exclusively described the results of (C)TRA (26, 35, 37, 38, 40, 42, 48, 50–53, 55, 57), while 24 reported data about different treatment strategies also including (C)TRA (3, 33, 34, 36, 39, 41, 43–47, 49, 54, 56, 58–67).

Eight studies did not provide detailed information on the diagnostic work-up employed for detection and assessment of airway invasion by TC (3, 33, 35–37, 42, 45, 51). The remaining 29 manuscripts specified the diagnostic methods utilized (26, 34, 38–41, 43, 44, 46–50, 52–67). Expectedly, neck and chest x-ray for airway invasion assessment were rarely mentioned in studies published after 1996. In contrast, airway endoscopy (either flexible or rigid, under local or general anesthesia) was reported in 100% of the series, computed tomography (CT) in 93%, ultrasonography (US) in 45%, and magnetic resonance (MR) in 32%.

Distinction between types of resection (purely tracheal resection and anastomosis [TRA] or also involving part of the cricoid [CTRA] with consequent thyro-crico-tracheal anastomosis) was reported in 35 papers (3, 26, 33–47, 49–58, 60–67) for a total of 619 patients undergoing 466 (75%) TRA and 153 (25%) CTRA.

Length of resection was reported in 19 papers (26, 33, 34, 37, 38, 40, 42–44, 46, 49–53, 62, 64, 65, 67), usually as range and mean in centimeters or number of removed tracheal rings (for TRA), with associated portions of adjacent cricoid cartilage (for CTRA). In some instances, detailed tables allowed to exactly know the extent of (C)TRA for each patient. However, the length of (C)TRA for 350 patients ranged between 0.5 and 6 cm (mean, 2.5).

The Shin classification was explicitly used to quantify the depth of airway invasion by TC in 10 manuscripts (26, 37, 38, 40, 41, 44, 51, 57, 62, 67), for a total of 148 patients subdivided as follows: Shin I in 12 (8%) patients, Shin II in 35 (24%), Shin III in 39 (26%), and Shin IV in 62 (42%).

Data on perioperative mortality were provided in 36 articles (3, 26, 33–53, 55–67), for a total of 632 patients. The random effects model pooled risk of postoperative mortality was 2.0% (95% CI, 1.0-4.0%), with low heterogeneity. Forest and funnel plots are reported in Figures 2A, B .

Twenty-nine articles (26, 34–40, 42–47, 49–53, 55–57, 59, 61–63, 65–67), including 557 patients, provided data on the proportion of patients suffering from postoperative complications. Complications were mostly bilateral RLN palsy, anastomotic dehiscence, hypoparathyroidism, and pulmonary complications, which are listed in detail in Table 2 . The overall summary estimate of the proportion of patients who developed any complication after (C)TRA for TC was 27.0% (95% CI, 20.0-36.0%) ( Figures 3A, B ). Heterogeneity was high (I^{2 =} 55.0%).

Data on long-term tracheotomy-dependency were reported in 30 studies (26, 33–35, 37–40, 42–53, 55–57, 59, 61–63, 65–67) for a total of 527 patients. The summary estimate of the proportion of patients remaining dependent on tracheotomy after (C)TRA was 4.0% (95%CI, 2.0-8.0%). Heterogeneity of studies was low ( Figures 4A, B ).

Oncological outcomes details were available in 29 articles (3, 26, 34, 35, 37–41, 43–49, 51–55, 57–59, 62–66), as reported in Table 2 . Seventeen studies (3, 26, 35, 43–49, 54, 58, 59, 62, 63, 65, 66) provided survival estimates. Five-year overall survival (OS) rate, reported by 11 studies (3, 26, 44, 48, 49, 54, 59, 62, 63, 65, 66), ranged from 61% (49) to 100% (62), whereas only 5 papers (26, 35, 43, 54, 65) reported the 10-year OS rate [ranging between 78.1% (35) and 42.1% (26)], and two studies (46, 65) also provided the 15-year OS rate (26.0% and 35.9%, respectively). Five manuscripts (26, 45, 47, 54, 58) reported 5- and 10-year disease specific survival estimates, which were in the range of 75.8-90% and 54.5-62.9%, respectively.

Specific data on adjuvant radioactive iodine (RAI) were reported by 12 papers (26, 40, 44, 46, 51–54, 62, 63, 65, 66), while 8 (26, 44, 46, 51–53, 62, 66) contained details on the use of postoperative external beam radiotherapy (EBRT). Most patients (86%; 95%CI, 61-96%; I² = 22%) had been treated with adjuvant RAI, whereas indication to postoperative EBRT was far less common (11%; 95%CI, 4-30%; I² = 18%). No survival analysis according to adjuvant therapies was conducted by any of the included studies.

When the cartilages or inter-cartilaginous ligaments are penetrated by neoplastic cells up to the level of submucosa, the TC spreads along the cartilaginous framework horizontally (following the inter-cartilaginous blood and lymphatic vessels) and vertically, before fungating into the airway lumen. As a consequence, the endoluminal real tumor extension is frequently more important than what can be seen from the outer surface of the organ (8, 19, 37). This, together with the uncertainty in the precise clinical assessment of the in-depth neoplastic extension within the cartilaginous framework, represents the most important pathological basis for justifying (C)TRA when dealing with tumors penetrating through the crico-tracheal axis, and the most evident limiting factor in supporting a window resection (per se based on the evaluation of the neoplastic extent just at the level of the outer airway surface).

In appropriately selected cases (i.e., for short segment airway involvement not beyond the external perichondrium) and with carefully performed surgery, shaving achieves complete resection and offers local control rates as high as 95-100% (58, 61, 68, 69). On the other hand, shaving for TC with deeper airway involvement, where R0 resection cannot be achieved or confirmed by frozen sections of cartilaginous tissues and possible microscopic penetration of the tumor into the airway submucosa via lymphatics piercing the inter-cartilaginous spaces may remain undetected, local failure with dismal outcomes have been reported in the literature (70, 71). Moreover, while several series included in this systematic review related to patients treated primarily by (C)TRA reported 5-year survival figures >80% (34, 41, 44, 45, 47, 48, 55, 57, 58, 62–66), the complication and mortality rates are considerably higher and survival lower when the procedure is applied as a salvage operation for recurrence after more conservative initial surgery (72, 73).

The same holds true for window resections, which are advocated by some in case of deeper tumor infiltration into the airway (20, 21). Such a surgical procedure is limited in its width and length of resection due to constraints in terms of airway stability, especially if the surgeon attempts a primary closure of the tracheal defect. As a consequence, if pedicled or revascularized myofascial/myoperichondral or skin flaps for tracheo-cutaneous fistula closure (74) are not employed, an R0 resection with negative margins by such a technique is less probable than after (C)TRA.

In case of incidental intraoperative discovery of TC invading the airway deeper than its external perichondrium, with a shaving procedure likely resulting in an R2-R1 resection, the general consensus is to convert the procedure into (C)TRA if it can be safely performed by the surgeon during the same intervention, or abort the procedure and refer the patient to a tertiary center with adequate experience in airway management (22). Clearly, in the best-case scenario, such a therapeutic option should be anticipated by performing the appropriate preoperative diagnostic work-up (including airway endoscopy and detailed cross-sectional imaging studies), referring the patient preoperatively to another team if adequate surgical expertise for (C)TRA is not available.

Invasion of the prevertebral fascia, major cervico-mediastinal blood vessels and/or massive involvement of the thoracic trachea are situations which make TC unresectable, and are categorized as T4b (8, 75). On the other hand, the only contraindications to (C)TRA include: 1) cranio-caudal extent exceeding 5.5 cm (i.e. 11 tracheal rings or cricoid arch plus 9 tracheal rings) (76) even in young patients (while for older ones, shorter airway resections may already represent an issue, with 4 cm of length being demonstrated to significantly increase the incidence of anastomotic dehiscence) (36, 77, 78); 2) major full-thickness esophageal/hypopharyngeal involvement requiring more than shaving of the external muscular layer or limited full-thickness resection with primary closure; and 3) tumor reaching the glottic plane either anteriorly through the crico-thyroid membrane or posteriorly at the level of the crico-arytenoid joint(s) (26). All these factors should be adequately assessed in the preoperative setting, considering that the proximal and distal airway cuts should be performed one tracheal ring above and below the macroscopic invasion site as appreciated from the adventitial side and checked from inside at the level of the airway lumen, with confirmatory mucosal frozen sections as needed. Clearly, these are contraindications for (C)TRA, but such tumor extensions are amenable to more radical surgical procedures, such as pharyngo-laryngo-esophagectomy for extensive invasion of the larynx and/or esophagus. Similarly, more extensive tracheal invasion can be resected making a mediastinal tracheostomy. In such clinical scenarios, a balanced preoperative counseling that may guide patients along the difficult path of choosing between a better quality of life against a higher chance of perioperative complications should be taken into account.

Preoperative bilateral RLN palsy does not represent per se an absolute contraindication to (C)TRA since an R0 resection with postoperative permanent tracheotomy (and usable voice) is always better than both persistent airway disease or total laryngectomy with/without tracheoesophageal voice prosthesis insertion (26). Of note, one should consider that, in case of anterior cricoid arch resection, the ensuing bilateral loss of vocal cords tension for lack of crico-thyroid muscles, if associated with bilateral RLN palsy, may make posterior cordotomy with/without arytenoidectomy useless in the attempt of getting a patent airway without tracheostomy. Last but not least, placing a tracheostomy below the anastomotic line after (C)TRA with bilateral RLN palsy may significantly increase the risk of postoperative complications such as anastomotic dehiscence, stenosis or tracheo-innominate fistula.

Radical comprehensive approaches like (C)TRA, able to maintain a good quality of life, are strongly recommended in patients with DTC even in the presence of a limited burden of asymptomatic distant metastases (8, 26). However, general health status (age, comorbidities, compliance) and willingness to undergo surgery play a prominent role in selecting patients amenable for such a major surgical procedure.

The first endoscopic examination to be performed in every TC patient should include a flexible videolaryngoscopy, even in the absence of an appreciable hoarseness: in fact, finding a unilateral RLN palsy should prompt to the request of more targeted investigations (such as tracheoscopy and cross-sectional imaging studies) to determine and precisely evaluate potential airway involvement, both quantifying its radial (depth) and cranio-caudal extents (22). This also applies to other common signs and symptoms of advanced TC such as hemoptysis, dyspnea, dysphagia, thyroid fixation or clinically enlarged lymph nodes: even though infrequent, these findings should prompt adequate imaging to exclude aerodigestive tract invasion and quantify it for appropriate surgical treatment planning.

Transcutaneous US can detect the depth of airway invasion, reliably distinguishing superficial (Shin I-II) vs. deeper (Shin III-IV) infiltration with a diagnostic accuracy potentially reaching 93% (79–81). However, US is generally considered highly operator-dependent and less reliable for tumors larger than 4 cm or with major intralesional calcifications, as well as with significant retrosternal extension.

CT is considered superior to US and definitely more reproducible for precise three-dimensional assessment of airway invasion (22), with a mean sensitivity, specificity, and accuracy in detecting tracheal invasion of 59.1%, 91.4%, and 83.2%, respectively (82). It is important to emphasize that the CT should be performed with contrast, to give the most precise information. In particular, the most quoted CT diagnostic criteria are tumor in contact for 180° or more of tracheal circumference, deformity of the airway lumen (i.e. indentation due to pressure effect) at the level of such a contact, focal irregularity, thickening or bulging of the mucosal lining and, finally, presence of tumor within the tracheal lumen (82).

MR seems to have lower diagnostic accuracy than US and CT, with a tendency to overestimate the actual depth of airway invasion (83). Others report superior outcomes with MR compared to other imaging techniques (84). However, a tumor-airway contact exceeding 135° of the tracheal circumference seems to efficiently predict some degree of cartilaginous involvement (85).

Laryngo-tracheoscopy allows appreciation of airway invasion when the airway submucosa is reached (Shin III-IV), thus appearing as a subtle localized or diffuse mucosal redness, with elevation, edema, presence of telangiectasias and vascular engorgement, with focal erosions or endoluminal vegetations in the most obvious scenario (41). This is in line with the experience of the first author, who missed Shin II tracheal invasion in 11% of his series by endoscopy and imaging (26). The sensitivity of this tool for tracheal invasion evaluation is, in fact, reported to be around 85%, with a mean underestimation of the actual cranio-caudal tumor extent of an average of 0.8 (maximum 2) tracheal rings compared to postoperative histopathologic specimens (86).

Endobronchial US (EBUS) is the latest imaging technique for assessment of the presence and degree of airway invasion by TC. Recent reports highlight an accuracy significantly higher than those reported by CT and/or MR, with a sensitivity and specificity of 92% and 83%, respectively (84). However, EBUS is still relatively infrequently used in most medical facilities due to some inherent drawbacks such as increased invasiveness, high cost, and limited utility in evaluating tumors infiltrating at the level of the thyroid upper lobe (82).

It would therefore appear that the most adequate diagnostic algorithm for advanced TC with suspicious airway invasion should be based on careful endoscopy of the larynx and trachea, with US and subsequent CT or MR depending to the local facilities and expertise.

Predictors of survival in advanced TC involving the airway may be patient-related (age, gender), tumor-related, and treatment-related. Among tumor-related factors, micro- vs. macroscopic extrathyroidal extension, limited to one vs. multiple organs has been recently demonstrated to play an important role (7). Strap muscles (T3b) and RLN invasion (T4a) have no prognostic influence on survival, but they do affect recurrence in contrast to laryngeal, tracheal, and esophageal involvement which heavily impact both local recurrence and survival rates (9, 87). Tracheal and esophageal invasion (T4a) present no prognostic differences when all tumor tissue can be removed within negative margins (R0 resection). By contrast, invasion of the larynx (T4a) reflects a more aggressive behavior of disease (47), even though no clear distinction is usually made in the literature with respect to the specific anatomic site(s) of TC infiltration. Intuitively, anterior cricoid involvement has a very limited impact on radicality of tumor resection and possibility of organ preservation compared to lateral and/or posterior cricoid infiltration (26). The same holds true when considering invasion of the inferior border of the thyroid laminae compared to transgression of their lateral edges in close proximity with the piriform sinus, or when dealing with a superficial (external perichondrium) vs. a full-thickness thyroid cartilage invasion.

General consensus has been reached on preserving the RLN whenever it is preoperatively functioning, even though encased by tumor, as long as it is not directly infiltrated by TC and all gross disease is removed, adding postoperative adjuvant therapy in the form of RAI or EBRT as indicated (22). Sacrifice of the RLN is generally only justified when preoperatively already non-functioning or when its preservation would inevitably leave behind gross residual disease (R2 resection) (88). In this case, RLN reinnervation by direct suture of healthy stumps at frozen sections, interposition of a nerve graft or suturing the ansa hypoglossi to the distal RLN stump may be considered to maintain vocal muscle tone and improve functional outcomes for voice rehabilitation (89–91).

Tumor histopathological type is a strong predictor of survival in TC invading the aerodigestive tract: the 5-year survival rate in DTC is around 75%, while it declines below 60% in medullary TC, and to 20% in undifferentiated tumors (26, 47).

Among procedure-related prognosticators, the most powerful seems to be R status: a number of series have confirmed better 5-year survival in R0 compared to R2 resections (90-78% vs. 50-35%) (35, 92, 93). Even R1 resections, while apparently presenting similar 5-year survivals (71, 93), in the long run are invariably associated with a higher rate of recurrence (46, 70, 71, 94). Moreover, when comparing immediate R0 resection with R1 resection followed by delayed radicalization, even considering non-organ sparing surgery, 10- and 20-year disease-free survival decrease from 67% and 50% to 7% and 0%, respectively (46). However, the absence of high-quality prospective data does not allow to solve the discrepancy between the above mentioned data and those reported by others, with no statistically significant survival differences between R0 and R1 resections (54, 61).

It should be emphasized at the outset that postoperative RAI and/or EBRT do not replace adequate surgery with R0 resection due to the high failure rates of these adjuvant therapies in controlling residual R2 disease, especially at the level of the airway. Adjuvant therapy by RAI after (C)TRA for T4a DTC is widely used whenever sufficient iodine uptake is demonstrated. However, response to RAI often cannot be established before surgery and is not uniform especially for microscopic residual disease (R1 resection) on the airway surface (95).

Special consideration should be paid to patients who have already received RAI in the past and experience further disease progression, since tolerance to and further potential benefit from RAI is questionable in such a clinical scenario. Moreover, EBRT as initial mode of therapy in TC invading the airway deeper than its cartilaginous framework should not be offered given the limitation it places on wound healing in case of a subsequent (C)TRA (26) and the low probability of appropriate response of bulky TC invading the airway (5). Still controversial is the potential role of adjuvant EBRT after segmental R0 airway resection for DTC, especially when the laryngo-tracheal axis was the only site of macroscopic extrathyroidal extension. Since no survival analysis was available regarding the impact of both RAI and EBRT as adjuvant therapies after (C)TRA, no conclusion on their prognostic role can be withdrawn from the present systematic review.

The most notable limits of our paper are the retrospective nature and relatively low number of patients included in each case series, per se potentially flooded by selection biases and a wide array of geographic, therapeutic, and epidemiologic differences. Lack of details concerning the histotypes and use of adjuvant treatment protocols reduces the possibility to infer their impact on prognosis of patients treated by (C)TRA. Moreover, the evolution in the diagnostic and therapeutic strategies occurred during the long time span of our systematic review must be taken into proper consideration.