Department of Otolaryngology-Head and Neck Surgery, The Newcastle upon Tyne Hospitals National Health Service Foundation Trust, Newcastle upon Tyne, United Kingdom
Consultant Surgeon and Honorary Clinical Senior Lecturer, Department of Otolaryngology-Head and Neck Surgery, The Newcastle upon Tyne Hospitals NHS Foundation Trust, Newcastle Upon Tyne, NE7 7DN, United Kingdom
The role of open conservation (partial) laryngeal surgery in radiorecurrent laryngeal cancers is unclear, and the procedure is not widely accepted or practiced. The objective of this review was to assess the oncologic and functional outcomes of partial laryngectomy in radiorecurrent tumors of the larynx reported in the literature.
The systematic review was performed using independently developed search strategies and included Medline, Embase, Zetoc, conference proceedings, and, when appropriate, a manual search. Inclusion criteria for the articles were set a priori. All included articles were subjected to quality assessment. Pooled estimates of local control at 24 months and of disease-free and overall survival rates were calculated using both a fixed-effects model (inverse square) and a random-effects model (DerSimonian-Laird).
The search identified 401 publications, of which 26 studies satisfied all inclusion criteria. Ten studies had a quality score ≥6 (good), and 16 had a score of 4 or 5 (fair). The pooled estimates of oncologic outcomes using the random-effects model were as follows: The local control rate at 24 months for 560 patients was 86.9% (95% CI, 84%-89.5%), the disease-free survival rate for 352 patients was 91.2% (95% CI, 88.2%-93.9%), and the overall survival rate for 360 patients was 83.1% (95% CI, 79.1%-86.7%). Decannulation of tracheostomy occurred in 95.1% (95% CI, 92.6%-97.2%) of the patients who were analyzed (n = 315), whereas the pooled mean larynx preservation rate was 83.9% (95% CI, 80.7%-87%; n = 502)
Residual or recurrent laryngeal cancer is a difficult clinical problem. Although several treatment options exist for patients with laryngeal cancer at first presentation, options for those with recurrent cancer will be limited based on the initial treatment received. For many early cancers, radiation therapy is a widely used treatment that produces good results. Cancers that recur after radiation therapy often demonstrate aggressive behavior, arise in a field where lymphatic drainage is unpredictable, and are associated with poor control rates. Total laryngectomy is an often recommended option, even for early radiorecurrent cancers, because it is technically easy, and the outcomes are predictable. However, total laryngectomy has far-reaching consequences for the patient in terms of function and quality of life.
Other less often practiced options include open conservation laryngectomy and transoral laser microsurgical resection. The latter is a relatively recent procedure, and its wide use can be limited by the availability of specialized equipment, especially in the developing world, and by the limitations of transoral access. Open partial laryngectomy has a well defined role both in cancer at initial presentation and in the radiorecurrent setting, but it is used less frequently because of the technical complexity of some procedures, the lack of expertise, and the unpredictable functional outcomes.
There have been several reports from different centers about the effectiveness of partial laryngectomy in terms of local tumor control and functional results in residual or recurrent cancer, but it is not widely practiced in many countries. Most articles on partial laryngectomy have been from a few centers, and the results included only a small number of patients. Thus, there is a need to pool results to achieve larger sample sizes and generate confidence in the use of this technique in the radiorecurrent setting. The objective of the current systematic review was to report the pooled outcomes of open partial laryngectomy in all its various forms in the setting of radiorecurrent laryngeal cancer.
MATERIALS AND METHODS
The systematic review was performed by the first author and by a librarian trained in literature searches. We independently developed search strategies and searched the Medline and Embase databases. The search terms included the following in various combinations to maximize the yield: conservation surgery, cricohyoepiglottopexy, cricohyoidopexy (CHP), cricohyopexy, frontolateral, hemilaryngectomy, horizontal, laryngectomy, partial, subtotal, supracricoid, supraglottic, and vertical. The references from all full texts that were selected were checked to identify suitable articles. The Zetoc database and conference proceedings also were searched, and we contacted national experts in the field. The search was performed for the first time on September 12, 2007 and was set to automatically update periodically until September 2009. Inclusion criteria for the articles were set a priori. These are listed in Table 1. The inclusion criteria deliberately were kept wide to encompass as many articles as possible without jeopardizing the validity of the results.
Table 1. Inclusion Criteria for Selecting Articles for the Review
1. Articles published from 1980 onward
2. Published series of ≥10 primary partial laryngeal procedures by the same institution/authors
3. Data clearly distinguish results of partial laryngeal procedures if published along with other procedures
4. Partial procedures do not include near-total laryngectomy
5. Clear description of tumor stage and selection criteria
6. Clear description of local control rates over 24 months
The search excluded articles that were published before 1980 for the following reasons. The surgical repertoire until the 1960s was primarily vertical partial laryngectomy (VPL) and supraglottic laryngectomy (SGL). Supracricoid laryngectomy (SCPL), although it was described much earlier, was popularized in the 1970s. We believed that it was pragmatic to include articles that were published on the subject in the next decade, allowing time for the learning curve to be climbed. In addition, few articles were published before 1980, and they reported relatively small number of patients, many of whom underwent extended procedures and complex reconstructions that currently are neither widely practiced nor relevant. The survival and functional results reported also were not uniform, which made them difficult to compare or pool. Non-English-language articles were excluded. For studies in which the results from primary partial laryngectomy were included along with those from radiorecurrent tumors, the results from radiorecurrent tumors were separated out.
The abstracts were analyzed to identify articles that fulfilled inclusion criteria. The abstracts obtained from the multiple database searches identified many duplicates that were removed using Endnote version X2 (Thompson Reuters, Carlsbad, Calif). Data from the identified articles were entered into an Excel spreadsheet (Microsoft Corp., Redmond, Wash) for further analysis.
Given the variation in patient characteristics across the globe, we chose local control rates after primary surgery and salvage surgery (often total laryngectomy) as the single oncologic outcome measure of interest. These were measured after the primary procedure and after salvage surgery. Local control rates at ≥24 months of follow-up were required for the study to be included in the review.
We also assessed other outcomes when they were available, including overall and disease-free survival, salvage laryngectomy rates, and decannulation rates from the articles. Many studies exist outside the remit of this review that focus purely on assessing functional outcomes (voice and swallowing). Given the variable reporting practices, the use of outcome measures that cannot be compared directly, and the need for different search strategies, we did not include functional outcomes in the current review apart from summarizing comparable data in articles that met inclusion criteria. Although most articles discussed the duration of hospital stay, we did not choose to measure this variable, because several local factors weigh heavily in the decision to discharge patients home (eg, the extent of community support).
To our knowledge, no widely accepted measures of quality assessment of case series exist.1 We used one of the quality-assessment forms that has been used by the National Institute for Health and Clinical Excellence2 (Table 2). The maximum possible score for a given study using this scoring system is 8.
Each item was answered either yes (score=1) or no (score=0).
1. Case series collected in more than 1 center (ie, a multicenter study)
2. Is the hypothesis/aim/objective of the study clearly described?
3. Are the inclusion and exclusion criteria (case definition) clearly reported?
4. Is there a clear definition of the outcomes reported?
5. Were data collected prospectively?
6. Is there an explicit statement that patients were recruited consecutively?
7. Are the main findings of the study clearly described?
8. Are outcomes stratified (eg, by disease stage, abnormal test results, patient characteristics)?
The data were entered into Microsoft Excel. The statistical package R Version 2.9 (The R Foundation for Statistical Computing, Vienna, Austria) was used for all analyses. Because most of the proportions were in the range between 80% and 100%, we used an arcsine square-root transform on our data to stabilize the variances. Thereafter, the effect of publication and other bias was assessed visually using a funnel plot, and the I2 statistic was calculated.3 This statistic is a measure of heterogeneity between studies and ranges from 0% to 100%; high percentages indicate greater heterogeneity in the data.
When studies have low heterogeneity (pragmatically, I2 < 25%), the differences between reported outcomes can be explained simply by the observed natural differences between patients. In this case, we can consider that all patients are part of the same larger pool. A fixed-effects meta-analysis is appropriate in which each patient is given approximately equal weight.
However, with high heterogeneity, the studies differ by more than can be explained by intrapatient effects. This implies that there were differences in the patients studied, in the treatment interventions, or in the outcome measures. In this case, a random-effects meta-analysis is appropriate in which each study is given more equal weighting.
Because the range of heterogeneity was between 0% and 60%, the meta-analysis was conducted using both a fixed-effects model (inverse square) and a random-effects model (DerSimonian-Laird). Forest plots were generated, and the pooled estimate of each statistic in the plot were estimated using the random-effects model (DerSimonian-Laird), because the confidence intervals were more conservative.
The search strategy identified 1211 articles in Medline from 1950 to the search date and 802 articles in Embase from 1974 to the search date. Only articles that were published after the 1980s were selected, imported into Endnote, and the duplicates were removed. The removal of duplicates yielded a total of 401 publications. The various stages of systematically assessing the abstracts and reasons for exclusion from the review are described in Figure 1. Financial constraints prevented the inclusion of foreign-language articles that had inadequate information in the abstracts. The process left us with 26 studies that reported on the outcomes of open partial laryngectomy in radiorecurrent tumors and satisfied all inclusion criteria. Ten studies had a quality score ≥6 (good), and 16 studies had a score of 4 or 5 (fair). The majority of articles that satisfied the inclusion criteria were published after the 1990s, which justified our decision to limit the articles to those that were published after the 1980s (Table 3).
Table 3. Various Contributions of the Articles Selected for Review to the Pooled Outcome Figures Generated by the Meta-Analysesa
Some inclusions and exclusions into the review need to be qualified further. Although they were published from the same center, the article by Nibu et al4 and Toma et al5 spanned a period of 15 years and 9 years, respectively, with a 2-year overlap. Given the small number of patients in both studies, we judged that this was an acceptable margin; thus, both articles were included in the current review. Two articles that were included were published by the same authors,6, 7 but those authors clearly indicated that the results were from procedures that were performed at different centers. Two other articles in the review also were published from the same center and spanned similar periods, but they dealt with different open procedures and, thus, were included.8, 9 The study by Holsinger et al10 included 6 patients who underwent transoral cordectomy, the results of which could not be separated from the results from open partial procedures. Four studies11-14 that reported on salvage open partial laryngectomy were excluded, because they did not provide data on local control at 24 months. Others were excluded because they were updates from the same center.15-17
Half of the selected articles were published from Europe6-9, 18-26(n = 13), 7 were from the United States,10, 27-32 3 were from Japan,4, 5, 33 2 were from Australia,34, 35 and 1 was from Mexico.36 In total, 560 patients were identified from the articles that were included. Although all studies described local control rates, this was not the case with other measures. In addition, various numbers of patients underwent different open partial procedures (VPL, SGL, SCPL). Table 3 summarizes the origin of data from the selected articles.
In 165 patients, the separate tumor classification could not be determined from the published data. For the remaining patients, most had T1 tumors (n = 240), the next biggest group had T2 tumors (n = 108), and the smallest group had T3 lesions (n = 18). There were very few T4 lesions (n = 3). The surgical procedures included variations of VPL, SGL, and SCPL with CHP or cricohyoidoepiglottopexy (CHEP). Eight articles each dealt exclusively with or presented data separately for VPL (n = 198)4, 5, 9, 20, 26, 29, 34, 35 and CHP/CHEP procedures (n = 149),6-8, 21-23, 27, 33 respectively. The remaining articles either reported on a mix of procedures or did not specify the type of open partial laryngectomy.
Local control at 24 months
The pooled local control rate obtained from 26 studies with data on 560 patients using both a fixed-effects model (inverse square) and a random-effects model (DerSimonian-Laird) is provided in Table 4. Figure 2 provides the forest plot of studies that contributed to these results and the spread of data with 95% confidence intervals for each study represented by horizontal lines. The funnel plot for this sample of 560 patients provides visual evidence of heterogeneity, as evidenced by an I2 value of 56.1% (Fig. 3). In summary, 6 studies7, 25, 27, 32, 33, 36 reported a 100% local control rate, and 3 studies9, 19, 30 reported local control rates <60%. The latter studies included a significant proportion of patients who underwent SGL. The pooled local control rate for all 560 patients was 86.9% (95% CI, 84%-89.5%)
Table 4. Pooled Outcomes for All Patients who Underwent Open Partial Laryngectomy for Radiorecurrent Disease
No. of Studies
No. of Patients
Heterogeneity: I2, %
Pooled Rates Using Fixed-Effects Model (95% CI), %
Pooled Estimates Using Random-Effects Model (95% CI), %
Disease-free survival data were available from 16 studies (n = 352). The pooled disease-free survival rate exceeded 91% in both models (Table 4, Fig. 4). Moreover, heterogeneity of the studies that were included in this analysis was relatively low with an I2 value of 35.1%. The lowest rate reported in any study was 73%,35 and 2 articles reported a 100% disease-free survival rate.4, 27
Fifteen studies (n = 360) reported overall survival data. Once again, the pooled overall survival rate was identical at 83.1% (95% CI, 79.1%-86.7%) whether a random-effects or a fixed-effects statistical model was used. An I2value of 2.1% confirmed a very low level of heterogeneity between the studies that were included in this subgroup analysis (Fig. 5).
Salvage total laryngectomy
In the salvage total laryngectomy group, patients may have undergone total laryngectomy for salvage of a recurrent or residual tumor and/or a nonfunctioning larynx. Most studies reported larynx preservation rates between 70% and 90%. Only 2 articles reported a larynx preservation rate <70%8, 9 and both of those studies focused on supraglottic tumors. The pooled mean larynx preservation rate was 83.9% (95% CI, 80.7%-87%; n = 502).
Reported salvage total laryngectomy rates varied from 0% to 44.4%.5, 6, 8, 9, 20, 21, 25, 27, 29, 30, 32-34 When the 3 studies that reported outcomes—and, consequently, higher rates of salvage total laryngectomy—for supraglottic tumors were excluded, the pooled mean salvage laryngectomy rate was 9.2% (95% CI, 7.4%-11.1%; n = 253). Salvage total laryngectomy for aspiration was reported by 7 authors,5, 20, 21, 23, 24, 27, 35 and the rate ranged from 0% to 14.2%, with a pooled mean of 4.4% (95% CI, 2.2%-7.4%) in 224 patients.
Vertical partial laryngectomy
Eight articles dealt exclusively with or presented data separately for patients who underwent VPL (n = 198).4, 5, 9, 20, 26, 29, 34, 35 The details from these patients are provided in Table 3. Of the 159 patients who had data reported on tumor classification, 124 had T1 tumors, and 35 had T2 tumors. The pooled local control, disease-free survival, and overall survival rates for this procedure and the sample size on which these rates were based are provided in Table 4. The mean larynx preservation rate was 85%, and decannulation occurred in 96.3%. The salvage laryngectomy rate ranged from 13.1% to 16.6%.5, 9, 20, 34 Only Kooper et al20 reported that 1 of 61 patients underwent a laryngectomy for functional reasons.
Supracricoid laryngectomy with cricohyoidoepiglottopexy or cricohyoidopexy
There were 8 reports of patients who underwent SCPL and CHP/CHEP procedures (n = 149).6-8, 21-23, 27, 33 The pooled local control, disease-free survival, and overall survival rates for this procedure, and the sample size on which these were based are provided in Table 4. The larynx was preserved in 83.1% (n = 149), and decannulation occurred in 96%. Total laryngectomy for aspiration was required in 2.6% of patients21, 23, 27, 33 (n = 149).
Fifteen articles reported decannulation outcomes.4, 6-9, 20-23, 25, 27, 29, 32, 33, 36 The pooled mean decannulation rate was 95.1% (95% CI, 92.6%-97.2%) based on data reported for 315 patients.
The time from surgery to decannulation could not be extracted reliably from the articles because of variable reporting methods. Eleven studies reported a mean or median time to decannulation that ranged between 7 days and 31 days. However, those measures were not fully reliable, because a few studies excluded the patients who had delayed decannulation from the analysis.
The incidence of laryngeal stenosis requiring further surgery to enhance airway patency and to facilitate tracheostomy decannulation was reported in 4 articles,5, 21, 24, 31 and the rates varied from 0% to 9%. The pooled mean rate of laryngeal stenosis was 5.3% (95% CI, 2.6%-8.8%) based on data reported for 194 patients.
Voice and speech
Voice and speech outcomes were not assessed systematically in most of the articles. Descriptions included satisfactory speech and swallowing in 76% of patients,24 serviceable speech in 87.5% of patients, little to no useful voice in 12.5% of patients,10 preserved voice in 86% of patients,4 and a significantly rough and breathy voice with satisfactory intelligibility.23
Descriptions of swallowing function also were sketchy in all articles. Examples include reports that all patients had swallowing without aspiration,4, 5 that all patients recovered swallowing function,22 and that there was a 12.5% incidence of postoperative dysphagia.10 Various protocols for enteral feeding and swallowing rehabilitation and a lack of consistent reporting methods for the duration of enteral feeding made it difficult to draw conclusions about the duration of nutritional support. Although some authors preferred prophylactic gastrostomy for all patients,27 the majority used nasogastric tube feeding. Three studies reported that 1 patient needed a permanent feeding tube for nutrition.21, 27, 31
The pathologic basis for primary open conservation laryngectomy in properly selected tumors is well documented. In the radiorecurrent setting, open procedures have been used less commonly because of concerns about unpredictable spread and postoperative function in the radiated organ and a higher risk of complications. Transoral laser-assisted resections are feasible alternative options but may not be applicable to some patients because of problems with surgical access. This background, combined with the paucity of large case series with partial procedures in the radiorecurrent setting, was the main impetus for undertaking the current systematic review. Although other narrative reviews have identified good control rates, to our knowledge, this study is the first systematic review and meta-analysis to examine the oncologic efficacy of partial laryngectomy in the setting of radiorecurrent cancer.
There are several strengths to the current study. Explicit inclusion criteria were set a priori to ensure that reliable results could be pooled. We filtered the studies to ensure that only data from centers that had published on at least 10 partial laryngectomies were included in the review; this was done as a quality-assurance measure, because several case series in the literature have published the results from a small numbers of patients spanning several years. Open partial laryngectomy needs specific expertise to ensure reproducible results, because there are several unique challenges in patient selection, technique, and postoperative care that may not be included as part of training in many head and neck surgical programs.
We believe that local control will be the single most important consideration in choosing the treatment modality for the primary lesion. Other oncologic outcomes have other determinants that play a major role in the outcome. For example, disease-specific survival results will reflect not only local control but also regional control and distant metastases, whereas overall survival results also will depend on the presence of comorbidities, synchronous/metachronous second primaries, and other diseases. This was the primary reason why we included only studies that reported at least 24 months of local control results. The results revealed very good control rates that approached 90% at 24 months and disease-free survival rates that exceeded 91%, reinforcing the effectiveness of the procedures in appropriately selected patients. It must be noted that this is not necessarily at the cost of the salvage total laryngectomy, because the pooled salvage total laryngectomy rate was approximately around 15%. The local control rates offered by open partial laryngectomy are at least comparable, if not higher, than those observed after transoral laser-assisted resection.37, 38
Our systemic review can be criticized for excluding studies that did not mention local control; however, we believe that this was an important inclusion criterion, because there is strong evidence in the literature that overall survival alone is not a suitable endpoint for comparing outcomes across centers worldwide. Like all systematic reviews, there may be a publication bias with respect to centers publishing good outcomes. The nature of the procedure induces a patient selection bias, because patients who undergo open partial laryngectomy are likely to have small-volume recurrences. This cohort of patients also is likely to have good pretreatment performance status, because good pulmonary function is a prerequisite for partial laryngectomy. Thus, the results of our current review cannot be extrapolated to management by other techniques.
One drawback of the current study is the inability to synthesize functional outcomes after partial laryngectomy. This is accounted for in part by poor reporting and in part by the lack of established, validated outcome measures for swallowing and voice after surgery. The significant heterogeneity between studies in some scenarios of the current meta-analysis may limit applicability of the findings.
There are several recommendations that we can make based on the current study. It is evident that open partial laryngectomy for glottic recurrences using any appropriate technique, including VPL or SCPL, produces repeatable outcomes for achieving a cure and local control. These results should encourage more centers to offer open partial laryngectomy rather than total laryngectomy to carefully selected patients. It must be noted that this is a selected group of patients with very few recurrences classified as T3 and T4, and care should be taken not to extrapolate the results to all recurrences.
We recommend that, when reporting the results from partial laryngectomy, future studies should ensure comprehensive reporting of the following outcomes: local control, disease-free survival, overall survival, performance status, comorbidity status, decannulation rate, proportion of patients who require further treatment for laryngeal stenosis, time taken for oral intake, and pretreatment and post-treatment voice and swallowing subjective measures. Clearly, further studies are needed to document functional outcomes after this procedure.
In conclusion, partial laryngectomy procedures appear to be effective for the treatment of recurrent laryngeal cancer. However, it should be noted that most of the evidence available is for earlier stage lesions (T1/T2), thus highlighting the selected nature of this group of patients and the importance of correct selection. When partial laryngectomy is undertaken, both vertical and supracricoid laryngectomy procedures appear to be effective.
CONFLICT OF INTEREST DISCLOSURES
This work was supported by the George Titley Memorial Fund, Freeman Hospital.