Introduction
In the UK, there are just under 2000 new diagnoses per annum of laryngeal squamous cell carcinoma.1 Early stage squamous cell carcinoma of the glottic larynx (i.e. tumour (T) stage 1 to 2, node (N) stage 0 and metastasis (M) stage 0) is characterised by small tumour volume and a low incidence of regional metastasis. For these reasons, early glottic squamous cell carcinoma has a relatively high chance of cure and a low chance of metastatic spread. The aim of treatment is to achieve local control with preservation of the larynx. The main modalities of treatment are either external beam radiotherapy (RT) or transoral laser microsurgery. Historically, early glottic squamous cell carcinoma has been treated by external beam RT. The choice of RT was based on good oncological results, in the absence of a surgical option with low morbidity rates. Open (or external), conservation laryngeal surgery is associated with equivalent local control but at the expense of a major operation, temporary tracheostomy, temporarily increased aspiration risk and significant dysphonia. However, the popularisation of transoral laser microsurgery in the 1990s has made available a low-morbidity operation that has emerged as an equal alternative to RT for the curative treatment of glottic squamous cell carcinoma.
A Cochrane database systematic review by Dey et al. Reference Dey, Arnold, Wight, MacKenzie, Kelly and Wilson2 of the management of early (i.e. T1 and T2i) squamous cell carcinoma of the glottis found insufficient evidence to guide management decisions regarding the most effective treatment. Only one trial comparing RT and surgery was eligible for inclusion, this dating back to patients treated in 1978 and comparing RT and open surgery only. It can be seen that there is very little evidence for many of the advantages and disadvantages of each approach. A UK trial, EaStER (Early Stage glottic cancer: Endoscopic excision versus Radiotherapy; chief investigator, Professor M Birchall), was approved for funding by Cancer Research UK in 2004, dependent on the results of a feasibility study. This was a randomised trial with patients undergoing one to one randomisation to either transoral laser surgery or RT. Comparatively low numbers of patients were successfully randomised in the feasibility study, and the trial was withdrawn in 2009. The main recruitment problem was the fact that the two arms of the trial were so completely different, posing problems for patients and clinicians (i.e. either the patient or the clinician strongly preferred one modality over the other). Furthermore, a significant proportion of patients were not suitable for surgery. A recent meta-analysis of 26 retrospective case series assessing laser versus RT for early stage glottic cancer found no clear difference in oncological outcomes.Reference Higgins, Shah, Ogaick and Enepekides3
Whilst acknowledging the lack of an evidence base for this subject, the authors of the present review consider the category of ‘early stage’ disease (i.e. T1 to T2) to be too broad to guide clinical decision-making. Indeed, the recent review by Higgins et al. Reference Higgins, Shah, Ogaick and Enepekides3 included carcinoma in situ cases in the analysis, as have other studies. Anecdotally, the perception appears to be that T1a lesions without anterior commissure involvement are equally amenable to transoral laser surgery or RT, whereas T1b lesions may be better treated with RT because of concerns over voice outcomes after transoral laser surgery for such lesions. However, the most recent reviewReference Spielmann, Majumdar and Morton4 of voice outcomes following transoral laser surgery versus RT for early stage tumours did not comment on the differences between T1a and T1b tumours, seemingly because there was a lack of specific outcome data for T1b tumours. Clinical decision-making therefore breaks down early stage disease into individual stages.
Thus, the aim of the present review was to analyse clinical outcomes for T1a and T1b glottic squamous cell carcinoma treated by primary RT or primary transoral laser surgery, albeit with the expectation of reduced sample sizes in comparison with Higgins and colleagues' review.Reference Higgins, Shah, Ogaick and Enepekides3 The present study is the first, and thus far only, review to address T1a and T1b outcomes separately.
Materials and methods
Search strategy
The core search terms were ‘laryngeal neoplasms’ and ‘glottis/vocal cords’. These were combined with either ‘laser’, ‘laser therapy’, ‘microsurgery’ or ‘radiotherapy’. The databases searched were Ovid Medline (1948–2011), Embase (1980–2011), Cochrane Database of Systematic Reviews, and Web of Knowledge. All papers deemed relevant had their bibliographies cross-referenced for any unidentified publications. English abstracts were read; non-English full manuscripts were ordered only if the abstract appeared in English and stated the inclusion criteria. Where appropriate, authors were contacted to provide specific details required for this review.
Data extraction
All three authors performed independent literature searches, with input from a university librarian. The three searches were then combined and full articles read by two authors (JO'H and JJH) to assess suitability for inclusion. When data were felt to have been duplicated by authors in more than one manuscript, the most recent study was included for the current review.
Selection criteria and outcome measures
All publications comprising level four evidence or greater were included. Strict criteria for inclusion were set, as follows.
Firstly, papers needed to show data on local control for either T1a or T1b squamous cell carcinoma of the glottis treated primarily by transoral laser surgery or RT. The local control percentage had to be stated in print or be calculable from published raw data.
Secondly, the follow-up period needed to exceed a mean or median of 36 months in order for data to be included in statistical calculations. Any papers with follow-up data for shorter time periods were included in the list of identified papers but their data were not included in the overall statistical analysis.
Thirdly, papers including local control data for both T1a and T1b tumours for a particular treatment modality were analysed as a separate group, with the expectation that such case series provided ‘stronger’ evidence for comparison.
Treatment definitions
In order for a study to be included in the review, transoral laser surgery or RT must have been the initial treatment option. Transoral laser surgery is frequently performed on more than one occasion; there is often uncertainty over how to interpret the significance of close or involved pathological margins. This may result in a further one or more transoral laser surgery resection procedures to ensure ‘clearance’. The current review did not seek to analyse how many transoral laser surgery resections were required as the data were too varied to draw any meaningful conclusions. The ‘process’ of transoral laser surgery was therefore taken to imply initial treatment with transoral laser surgery, regardless of the total number of resections performed.
Specific RT schedules were not analysed.
The method of tumour staging was not assessed. It was assumed that authors quoting outcomes for T1a and T1b lesions did so in the knowledge of the correct diagnostic definitions.
Statistical analysis
No strict statistical meta-analysis was planned prior to database searching. In the absence of controlled trials, it was subsequently deemed that simple weighted averages of data should be calculated.
Results
Of the 345 studies identified in the search, 36 were included in the review (Figure 1). Tables I to IV show the 36 studies according to treatment type and tumour stage. All 36 studies were retrospective case series (i.e. level four evidence).
Fig. 1 Systematic review searches.
Table I Transoral laser outcomes for stage T1A tumours
Pts = patients; mth = months; yr = years
Table II Transoral laser outcomes for stage T1B tumours
Pts = patients; mth = months; yr = years; Med = median
Table III Radiotherapy outcomes for stage T1A tumours
Pts = patients; yr = years; Med = median
Table IV Radiotherapy outcomes for stage T1B tumours
Pts = patients; yr = years; Med = median
Six studiesReference Schrijvers, van Riel, Langendijk, Dikkers, Schuuring and van der Wal8, Reference Thurnher, Erovic, Frommlet, Brannath, Ehrenberger and Jansen10, Reference Goor, Peeters, Mahieu, Langendijk, Leemans and Verdonck-de Leeuw12, Reference Krengli, Policarpo, Manfredda, Aluffi, Gambaro and Panella19, Reference Spector, Sessions, Chao, Haughey, Hanson and Simpson25, Reference Epstein, Lee, Kashima and Johns27 included data for patients treated with transoral laser surgery or RT, but none allocated treatment according to a trial protocol. Twenty-three studies provided data for transoral laser surgery local control outcomes, and 19 provided data for RT local control outcomes.
Thirty studies provided follow-up data for a mean or median follow-up period of at least 36 months. For T1a tumours treated with RT, 2237 out of 2507 (89.2 per cent) patients had at least 36 months' local control. For those T1a tumour patients treated with transoral laser surgery, 1163 out of 1308 (88.9 per cent) had at least 36 months' local control.
For T1b tumour patients treated with RT, 468 out of 544 (86.0 per cent) had at least 36 months' local control. For T1b tumour patients treated with transoral laser surgery, 154 out of 194 (76.8 per cent) had at least 36 months' local control.
Sixteen studies gave local control figures for both T1a and T1b tumours treated with transoral laser surgery (n = 985 and 187, respectively). Local control rates were 87.7 and 78.1 per cent, respectively.
Fifteen studies gave local control figures for both T1a and T1b tumours treated with RT (n = 2340 and 544, respectively). Local control rates were 89.7 and 86.0 per cent, respectively.
Discussion
Although the treatment of early stage glottic squamous cell carcinoma has been the subject of many review articles, the current study represents the first structured review assessing outcomes for individual stages of disease, namely T1a and T1b tumours. Review findings demonstrated comparable local control rates for T1a tumours treated with either initial transoral laser surgery or RT. Findings also demonstrated a trend towards improved local control rates with RT. There were clearly many more reported cases of glottic squamous cell carcinoma treated with RT in the literature, compared with cases receiving transoral laser surgery. There were also many more reported cases of T1a disease than T1b disease.
There is good reason to view T1a and T1b tumours separately, as the treatment decision-making process appears to differ. There is emerging consensus that voice outcomes are similar for T1a tumours when transoral laser surgery is compared with RT. Certainly, when individual studies were examined in the most recent reviewReference Spielmann, Majumdar and Morton4 of this subject, the vast majority of tumours included were T1a lesions. There is little conclusive evidence assessing voice outcomes for T1b tumours, but there are at least theoretical concerns regarding voice outcomes after transoral laser surgery for these tumours.
Limitations of study
Review findings suggest similar local control outcomes for RT and transoral laser surgery for T1a tumours, and more favourable local control for T1b tumours treated with RT. However, the relatively small numbers, heterogeneity of studies and lack of any formal sample size calculation mean that no firm conclusions can be drawn.
Some of the individual studies were also low in sample size. The authors feel that the quality of data precludes a more rigorous statistical meta-analysis. Indeed, as was decided, it could be argued that no statistical analysis should be performed on such heterogeneous data. The weighted averages used in the analysis were crude, taking no account of variations in patient population between studies. In the absence of a formally conducted trial, one would need to enter individual patient details into a database in order to make more sense of this sort of multi-institutional data and to allow multivariate analysis. Differences in institutional preference for treatment can also be allowed for in a larger study including individual patient data.
Data quality
All the included studies were historical cohort series; no controlled trials were identified. Case series data are always reliant on the quality of recorded notes and database entries, and prone to significant bias. Those studies that included data for both RT and transoral laser surgery did not follow any specific, predetermined treatment allocation technique (i.e. stated in the trial protocol); treatment allocation changed with personnel and individual tumour characteristics, and over time. In one studyReference Goor, Peeters, Mahieu, Langendijk, Leemans and Verdonck-de Leeuw12 (not included in this analysis), RT was preferentially given to patients whose tumours had a reduced mucosal wave on stroboscopy, with transoral laser surgery reserved for tumours with intact mucosal waves. Tumours were all T1a, but clearly RT was given for more invasive patterns of disease. There remained a tendency to ‘cherry-pick’ straightforward T1a tumours for surgery. As such, few studies compared like with like.
It would have been interesting to have assessed other treatment outcomes in addition to local control. However, sample sizes were small and local control was the most consistently documented outcome when analysing T1a and T1b data; thus, other outcome variables were not analysed. Treatment for recurrent or persistent disease occurring after initial therapy varied between studies, and thus could not be adequately assessed. For those tumours treated initially by transoral laser surgery, subsequent treatment options comprised further transoral laser surgery, RT, open partial laryngectomy or total laryngectomy. For those tumours treated initially with RT, further RT was not an option. It would have been interesting to compare laryngectomy rates and laryngectomy-free survival for different treatments; however, any such comparisons were prevented by the small sample numbers and wide practice variation encountered. Furthermore, most studies did not state the number of laryngectomies performed for T1a and T1b tumours.
Several of the included papers distinguished between tumours with and without anterior commissure involvement, in their analyses, and some may argue that this distinction is preferable to categorisation by stage T1a versus T1b. Steiner et al. Reference Steiner, Ambrosch, Rodel and Kron20 found that the five-year local control for T1a patients without anterior commissure involvement was 90 per cent, with laryngeal preservation in 99 per cent, compared with corresponding figures of 84 and 93 per cent when the anterior commissure was involved.
Schrijvers et al. Reference Schrijvers, van Riel, Langendijk, Dikkers, Schuuring and van der Wal8 assessed the laryngeal preservation rates of patients undergoing RT or transoral laser surgery for T1a tumours. In an attempt to make both groups comparable, only tumours with a normal or diminished mucosal wave on stroboscopy were included, while deeper, infiltrating tumours were excluded on the basis of preferential treatment with RT. Whilst local control rates were similar (i.e. 71 per cent for transoral laser surgery and 73 per cent for RT), although possibly lower than those for other studies included in this review, the laryngeal preservation rates were significantly higher in the transoral laser surgery group (95 per cent) than the RT group (77 per cent). Intuitively, this is not too surprising, as failed transoral laser surgery treatment can be salvaged with RT, but failed RT must be treated with salvage surgery. Some may argue that the prevalence of total laryngectomy in the failed RT group in this case series was higher than might be expected, and may note the greater emphasis placed on salvage transoral laser surgery or open partial laryngectomy in other institutions. A greater prevalence of laryngectomy-free survival may well be one advantage of transoral laser surgery, compared with RT; however, this is difficult to assess from the current data due to variation in the management of salvage surgery.
A separate analysis was performed for those papers quoting local control rates for both T1a and T1b tumours for a particular treatment modality. It was felt that such studies may provide a ‘stronger’ comparison, in that staging of T1a versus T1b tumours may have been more accurate. From this analysis, there certainly appeared to be a trend towards improved local control for T1b tumours treated with RT.
To the best of our knowledge, there are no forthcoming comparative studies which address the questions raised in the current review. An alternative approach would be a prospective audit or registration study. However, such studies are very difficult to successfully fund. Although still prone to selection bias at the individual patient level, if conducted on a large enough scale some of this bias could be accounted for using multivariate analysis techniques.
Conclusion
Using the available evidence, this systematic review found no difference between local control rates for T1a or T1b tumours treated by primary transoral laser surgery or RT. However, there did appear to be a trend towards improved local control for T1b tumours treated by RT, albeit derived from a low number of published outcomes. This review was not able to analyse laryngectomy-free survival between the two treatment groups. Further trials or a large scale, prospective audit are necessary to address this important question. Future studies should report separate outcomes for carcinoma in situ, T1a and T1b tumours, and should include total laryngectomy rates.
