Protons versus photons for the treatment of chordoma

Abstract

Background

Chordoma is a rare primary bone tumour with a high propensity for local recurrence. Surgical resection is the mainstay of treatment, but complete resection is often morbid due to tumour location. Similarly, the dose of radiotherapy (RT) that surrounding healthy organs can tolerate is frequently below that required to provide effective tumour control. Therefore, clinicians have investigated different radiation delivery techniques, often in combination with surgery, aimed to improve the therapeutic ratio.

Objectives

To assess the effects and toxicity of proton and photon adjuvant radiotherapy (RT) in people with biopsy‐confirmed chordoma.

Search methods

We searched CENTRAL (2021, Issue 4); MEDLINE Ovid (1946 to April 2021); Embase Ovid (1980 to April 2021) and online registers of clinical trials, and abstracts of scientific meetings up until April 2021.

Selection criteria

We included adults with pathologically confirmed primary chordoma, who were irradiated with curative intent, with protons or photons in the form of fractionated RT, SRS (stereotactic radiosurgery), SBRT (stereotactic body radiotherapy), or IMRT (intensity modulated radiation therapy). We limited analysis to studies that included outcomes of participants treated with both protons and photons.

Data collection and analysis

The primary outcomes were local control, mortality, recurrence, and treatment‐related toxicity. We followed current standard Cochrane methodological procedures for data extraction, management, and analysis. We used the ROBINS‐I tool to assess risk of bias, and GRADE to assess the certainty of the evidence.

Main results

We included six observational studies with 187 adult participants. We judged all studies to be at high risk of bias. Four studies were included in meta‐analysis.

We are uncertain if proton compared to photon therapy worsens or has no effect on local control (hazard ratio (HR) 5.34, 95% confidence interval (CI) 0.66 to 43.43; 2 observational studies, 39 participants; very low‐certainty evidence).

Median survival time ranged between 45.5 months and 66 months. We are uncertain if proton compared to photon therapy reduces or has no effect on mortality (HR 0.44, 95% CI 0.13 to 1.57; 4 observational studies, 65 participants; very low‐certainty evidence).

Median recurrence‐free survival ranged between 3 and 10 years. We are uncertain whether proton compared to photon therapy reduces or has no effect on recurrence (HR 0.34, 95% CI 0.10 to 1.17; 4 observational studies, 94 participants; very low‐certainty evidence).

One study assessed treatment‐related toxicity and reported that four participants on proton therapy developed radiation‐induced necrosis in the temporal bone, radiation‐induced damage to the brainstem, and chronic mastoiditis; one participant on photon therapy developed hearing loss, worsening of the seventh cranial nerve paresis, and ulcerative keratitis (risk ratio (RR) 1.28, 95% CI 0.17 to 9.86; 1 observational study, 33 participants; very low‐certainty evidence). There is no evidence that protons led to reduced toxicity.

There is very low‐certainty evidence to show an advantage for proton therapy in comparison to photon therapy with respect to local control, mortality, recurrence, and treatment related toxicity.

Authors' conclusions

There is a lack of published evidence to confirm a clinical difference in effect with either proton or photon therapy for the treatment of chordoma. As radiation techniques evolve, multi‐institutional data should be collected prospectively and published, to help identify persons that would most benefit from the available radiation treatment techniques.

Plain language summary

Proton radiation therapy versus conventional radiation treatment (i.e. photon radiation) for the treatment of chordoma

What is the issue?
Chordoma is a rare tumour that can grow at any location in the spine. Because some of the locations are difficult to reach, some of the tumour may be left behind, even after aggressive surgery. When this happens, radiation to the postoperative area can help reduce the chance of the tumour coming back. Given the closeness to the spinal cord and other important organs, special radiation techniques are required to maximise tumour control, while minimising harm to normal tissues in the body.

Why is this review important?
At the time of this review, there is a lack of evidence to confirm a benefit with proton therapy compared to conventional photon radiation treatment. However, it is important to note that this review found a number of delivery techniques for both proton and photon therapy. As evidence comparing the latest techniques for proton and photon therapy emerges, we will update this review.

The aim of the review:
The aim of this Cochrane Review was to find out whether proton therapy is more effective than photon therapy for improving local control, and reducing mortality, recurrence, and treatment‐related toxicity. We collected and analysed all relevant studies to answer this question.

What are the main findings?
We included six studies that enrolled 187 adults with chordoma, four of which we included in a meta‐analysis. We are uncertain whether proton radiation therapy compared to conventional photon radiation treatment improves local control, because our confidence in the evidence was very low (2 observational studies, 39 participants). We are uncertain whether proton therapy compared to photon therapy reduces mortality (4 observational studies, 65 participants), recurrence (4 observational studies, 94 participants), or treatment‐related toxicity (1 observational study, 33 participants).

Quality of the Evidence:
We found very low‐certainty evidence for all outcomes due to high risk of bias in all the included observational studies, very small sample sizes, and varying ranges in the duration of participants' follow‐up period between studies.

Authors' conclusions:
From the currently available evidence, we do not know if proton therapy, used for adults with chordoma, is associated with clinically appreciable benefits and acceptable harms.

The existing research on this question is solely of non‐randomised design, and is underpowered, with substantial imprecision. Any estimates of effect based on the existing insufficient evidence is very uncertain and is likely to change with future research.

Summary of findings

Summary of findings 1. Protons compared to photons for the treatment of chordoma.

Protons compared to photons for the treatment of chordoma
Patient or population: adults with chordoma Setting: hospitals Intervention: protons Comparison: photons
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with photons	Risk with protons
Local control (follow‐up: median ranged between 45.5 and 66 months)	400 per 1000	935 per 1000 (286 to 1000)	HR 5.34 (0.66 to 43.43)	39 (27 analysed, 2 non‐randomised studies)	⊕⊝⊝⊝ Very lowa,b,c	We are uncertain if proton compared to photon therapy effects local control.
Mortality (follow‐up: median ranged between 45.5 and 66 months)	600 per 1000	332 per 1000 (112 to 763)	HR 0.44 (0.13 to 1.57)	65 (39 analysed, 4 non‐randomised studies)	⊕⊝⊝⊝ Very lowa,d,e,f,g	We reported mortality instead of overall survival for methodological reasons. We are uncertain if proton compared to photon therapy affects mortality.
Recurrence (follow up: median ranged between 3 and 10 years)	600 per 1000	268 per 1000 (88 to 658)	HR 0.34 (0.10 to 1.17)	94 (81 participants, 4 non‐randomised studies)	⊕⊝⊝⊝ Very lowd,f,h	We reported recurrence instead of recurrence‐free survival for methodological reasons. We are uncertain if proton compared to photon therapy effects recurrence.
Treatment‐related toxicity	125 per 1000	160 per 1000 (21 to 1000)	RR 1.28 (0.17 to 9.86)	33 (1 observational study)	⊕⊝⊝⊝ Very lowg,i	We are uncertain if proton compared to photon therapy effects treatment‐related toxicity.
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). The assumed risk in the photon group for local control, mortality, and recurrence outcomes are based on the event rate in Park 2006. CI: confidence interval; RR: risk ratio; HR: hazard ratio
GRADE Working Group grades of evidence High certainty. We are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty. We are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty. Our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect. Very low certainty. We have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect.

^aWe did not include one study in the meta‐analysis, as we could not conduct analysis with only one participant who received photons. We used IPD from the other study for the synthesis. 
^bWe downgraded by two levels due to high risk of bias of the included study.
^cWe downgraded by two levels due to very serious imprecision due to a very small sample size that did not meet the optimal information size, and a very wide confidence interval that crossed the line of no effect and failed to exclude appreciable benefit.
^dWe downgraded by two levels due to high risk of bias in the included studies.
^eWe downgraded by one level as one study compared proton versus photon therapy and the second study compared proton plus photon therapy versus photon therapy alone.
^fWe were unable to include data from one study in the meta‐analysis as we were unable to extract numeric data for outcomes per group. (6 participants per proton therapy group and 2 participants per photon therapy group) (Yasuda 2012) 
^gWe downgraded by two levels due to a very small sample size that did not meet the optimal information size, and a very wide confidence interval that crossed the line of no effect and failed to exclude appreciable harm. 
^hThe included studies varied widely in the duration of follow‐up. Median follow‐up ranged from 12 to 40 months in one study, 66 months in another study, and 26.5 to 29 months in a third. Also, recurrence should be scored independently of further therapeutic intervention (use of second surgery) as defined by Takahashi 2009.
ⁱWe downgraded by one level due to high risk of bias of the included study.

Background

Description of the condition

Chordoma is a rare primary bone tumour of presumed notochord origin, with an incidence of 0.08 per 100,000 people per year, based on statistics from the Surveillance Epidemiology and End Results Registry (SEER) Program (McMaster 2001). Median age at presentation is approximately 60 years old; although uncommon in individuals younger than 40 years, it may also occur in children and adolescents (Borba 1996). Chordoma is a low‐ to intermediate‐grade malignancy, but with locally aggressive behaviour. It is historically divided into three subtypes based on light microscopic morphology: conventional, chondroid, and de‐differentiated. Conventional chordomas are the most common, and are characterised by a lack of cartilaginous or mesenchymal components (Heffelfinger 1973). Chondroid histology exhibits good long‐term outcomes; de‐differentiated chordoma often demonstrates an aggressive pattern with less favourable prognoses (Casali 2007). More recently, a distinct molecular subtype with a very poor prognosis has been defined, termed poorly differentiated chordoma (Hasselblatt 2016). Chordomas arise mainly along the midline in the skeleton, with a predilection for the base of the coccyx (i.e. sacrococcygeal, 50%) and skull base (i.e. clivus, 30% to 40%) regions. Chordomas are rarely metastatic at first presentation, but distant metastases have been reported in up to 40% of people, usually occurring late in the course of the disease (Catton 1996).

Description of the intervention

Since chordoma is a locally aggressive tumour, with a high propensity for local recurrence, control of the primary disease remains the major therapeutic challenge. For example, a meta‐analysis of 464 people with cranial chordoma reported a recurrence rate of 68%, and a median disease‐free interval of only 23 months (average 45 months (Jian 2010)). Current management involves surgery, radiation therapy (RT), or both. Chemotherapy is generally ineffective, and is used mainly in the advanced (metastatic) stage of disease (Casali 2007). Surgery is the mainstay of treatment, and aims to establish a definitive diagnosis, while attempting maximum resection and decompression. Given the difficult locations and the relationship with surrounding healthy tissues, potential morbidity from surgery is high. Moreover, local recurrence and progression are frequent even in cases of optimal surgery (Walcott 2012). Hence, RT has been used either pre‐ or postoperatively, in an attempt to achieve local tumour control and possibly improve overall survival (OS (Rotondo 2015)). Since the 1970s, conventional radiotherapy with photons has been given as both curative and supportive treatments (Cummings 1983). Due to anatomical locations, radiation oncologists face the same constraints that hamper adequate surgical excision. The dose of RT that surrounding healthy organs can tolerate is frequently below that required to provide effective treatment. These healthy tissue radiotherapy dose constraints vary by original tumour location in the skull base or spine.

How the intervention might work

Photons, created from either x‐rays or gamma rays, deposit their energy along the entire linear path they travel, attenuating, but not stopping within the body of the participant. Hence, the treatment of the tumour area with photons requires the irradiation of healthy tissues both near to and away from the target. As a consequence, most of the early series using photon RT were not able to deliver effective radiation dose levels within the tumour. By contrast, 'particles' (protons, ions, neutrons) behave differently than photons in tissue: they exhibit a moderate entrance dose, give off most of their energy at a defined depth, and then stop. Consequently, it is possible to limit the radiation dose to the healthy tissues beyond the tumour, while delivering a higher dose to the tumour. Theoretically, protons offer advantages over photons in the treatment of chordoma (Walcott 2012), and are a supported option for postoperative adjuvant irradiation in this disease (NCCN Guidelines 2018). However, the development of advanced photon techniques, such as stereotactic body radiotherapy (SBRT), stereotactic radiosurgery (SRS), and intensity‐modulated RT (IMRT) improve the planning capability when delivering photon RT (Amichetti 2012; Baskar 2012; Sahgal 2014; Gatfield 2018). These techniques generally take advantage of three‐dimensional (3D) imaging, such as magnetic resonance imaging (MRI) or computed tomography (CT) to define tumour and normal tissue, and then use multiple beam angles to minimise total dose to surrounding normal tissue. Furthermore, small planning margins have been made possible with photons by utilising kV (kilovoltage) cone‐beam computer tomography (CBCT) to verify the person's positioning with daily RT treatments, compared to planer imaging still in use by most proton centres (Alcorn 2014, Sahgal 2014).

Why it is important to do this review

The low incidence of this malignancy means only a few centres worldwide have experience in its management. Thus, there is little robust evidence to guide the therapeutic strategies, which is summarised by the Chordoma global consensus group (Stacchiotti 2015). Although technical and theoretical advantages support the use of protons, there are few available centres offering this treatment internationally. Moreover, the cost of treatment with protons is significantly more expensive than with photons. Conversely, precision photon RT is cheaper, treatment machines are widespread, and in the past decade, there has been bridging of the technical gap between photons and protons (Sahgal 2014). There is uncertainty about the relative superiority of the different techniques, therefore, a systematic review assessing whether modern photon RT is as effective as particles is needed.

Objectives

To assess the effects and toxicity of proton and photon adjuvant radiotherapy (RT) in people with biopsy‐confirmed chordoma.

Methods

Criteria for considering studies for this review

Types of studies

All included studies were observational. Our search didn't retrieve randomised controlled trials (RCTs), quasi‐RCTs, cluster‐RCTs, or controlled clinical trials (CCTs) of proton versus photon therapy. Given the small number of included studies and small study size, we did not report outcomes according to the tumour location. These criteria were set as the type of surgery and radiotherapy (RT; see Characteristics of included studies). We stated whether the included studies had incomplete or obvious confounders (see Assessment of risk of bias in included studies). We did not restrict the inclusion criteria to studies with at least three years follow‐up, although it has been reported that half of chordoma recurrences occur within the first three years (Weber 2016). Instead, we reported the follow‐up period for each study.

Types of participants

We included adult participants, aged 18 years or older, with pathologically‐confirmed primary chordoma, who were irradiated with a curative intent with protons or high‐energy (megavoltage) photons in the form of fractionated RT, SRS (stereotactic radiosurgery), SBRT (stereotactic body radiotherapy), or IMRT (intensity modulated radiation therapy). In the curative setting, a 60 Cobalt Gray Equivalent (CGE) dose or higher is typically used to achieve local control, and therefore, we restricted inclusion to people treated to at least this dose (Terahara 1999; Igaki 2004; Noel 2005). We included people who had undergone biopsy, subtotal, or gross resection.

We excluded people treated with brachytherapy.

All studies included participants who met the inclusion criteria, so we used the subset of data that was of interest to this Cochrane Review.

Types of interventions

Interventions

We included studies if they assessed any of the following particle interventions.

• Irradiation with protons

• Irradiation with ions

• Irradiation with neutrons

Comparison

• Irradiation with high‐energy (megavoltage) photons

Types of outcome measures

Primary outcomes

• Local control (LC) with proton compared to photon therapy, defined by the authors of the included studies by only included study as the absence of tumor or persistent mass with no evident growth (Park 2006).

• Overall survival (OS) with proton compared to photon therapy, defined by authors of the included studies (5 years (Park 2006 ; Jagersberg 2017) , and 10 years overall survival (Park 2006).

• Progression‐free survival (PFS) with proton compared to photon therapy, defined by authors of the included studies as the period between the time of additional surgery and former surgery (Takahashi 2009).

• Treatment‐related toxicity, using definitions provided by the Common Terminology Criteria for Adverse Events, particularly the frequency of higher severity (grade 2 or above) adverse events (CTCAE 2017).

Secondary outcomes

No secondary outcomes were assessed.

We presented a 'Summary of findings table' reporting the following outcomes listed in order of priority (Atkins 2004):

• Local control (LC).

• Mortality (Instead of OS).

• Progression‐free survival (PFS).

• Treatment‐related toxicity.

Search methods for identification of studies

We conducted a comprehensive, systematic search to identify all relevant studies, regardless of language. Imaging, surgical procedures, and RT techniques before the 1990s are very different from current clinical practice; these differences may significantly affect clinical outcomes and their interpretation (Amichetti 2009). Therefore, we considered studies prior to 1990 separately in any analyses.

Electronic searches

We searched the following electronic databases:

• Cochrane Central Register of Controlled Trials (CENTRAL; 2021, Issue 4) in the Cochrane Library (searched 29 April 2021; Appendix 1);

• MEDLINE Ovid (1946 to week 16, 2021; Appendix 2);

• Embase Ovid (1980 to week 16, 2021; Appendix 3).

For databases other than MEDLINE, we adapted the search strategy accordingly.

Searching other resources

We searched the following electronic trial registries and databases to 29 April 2021.

• ClinicalTrials.gov (http://clinicaltrials.gov);

• International Clinical Trials Registry (http://apps.who.int/trialsearch);

• Current Controlled Trials (www.controlled-trials.com);

• National Comprehensive Cancer Network (NCCN) Guidelines (www.nccn.org/professionals/physician_gls/f_guidelines.asp);

• Turning Research into Practice (TRIP; www.tripdatabase.com);

• National Guideline Clearinghouse (www.guideline.gov).

We handsearched the following resources.

• Reference lists of relevant articles to identify additional publications;

• Conference proceedings of the:

  • American Society for Radiotherapy and Oncology (ASTRO; 1990 to April 2021);

  • European Society for Radiotherapy and Oncology (ESTRO; 1990 to April 2021);

  • Particle Therapy Co‐Operative Group (PTCOG; 1990 to April 2021);

  • North American Skull Base Society (NASBS; 1990 to April 2021).

Data collection and analysis

Selection of studies

Four review authors (IS, DT, EL, SD) independently screened the title and abstract of all articles identified as a result of search, for inclusion. We obtained the full‐text articles of those identified as potentially eligible for inclusion, and the available review authors (IS, SD) independently assessed the full‐text articles. We resolved any disagreements by discussion or by consulting a third review author (DT). No potential conflict of interest existed (e.g. no authorship of a potentially included study). We listed all studies excluded after full‐text assessment and their reason(s) for exclusion in a Characteristics of excluded studies table. We illustrated the study selection progress in a PRISMA diagram (Figure 1).

1.

Study flow diagram.

Data extraction and management

We extracted data on the following: publication data (first author's last name, year of publication, institution), study design, sample size, participants' features (mean age, sex ratio, race, country of the population studied, tumour location, type of surgery, the definition of extent of resection), details of interventions (indication for RT, RT technique, time interval of participant accrual, total dose, dose per fraction, target definition), clinical outcomes (outcome definitions, LC, OS, PFS rates, type of toxicity, toxicity rates, length of follow‐up). We considered studies in which at least part of the RT course used particles as particle studies.

For eligible studies, two review authors (IS and SD) independently extracted data using the data extraction form. If more than one publication referred to the same study, or there was more than one paper from the same institution, we used the most recent publication, or that with the most complete reporting. . We resolved any disagreements by discussion or consulted a third review author (EL or DT). We considered studies to have sufficient data if at least one of the listed primary outcomes, and at least one of the selected time points were reported.

None of the included studies required translation as they were all in English.

Assessment of risk of bias in included studies

We based the risk of bias evaluation on the data provided in the included studies. We did not retrieve any randomized controlled trials; all the included studies were observational. Hence, we used the Risk of Bias In Non‐randomized Studies of Interventions (ROBINS‐I) assessment tool (Sterne 2016). We assessed the risk of bias according to the following domains.

• Pre‐intervention bias: due to confounding based on referral pattern and departmental protocol eligibility criteria as well as bias in selection of participants into the study

• At‐intervention bias: in classification of the intervention

• Post‐intervention bias:

  • due to deviations from the intended intervention

  • due to missing data

  • in measurement of outcomes

  • in selection of the reported results

Each of the seven domains of bias contain signalling questions to facilitate judgements of risk of bias. The full signalling question and response framework for each outcome is provided in Sterne 2016.

Important confounders of interest in this Cochrane Review include the following:

• Histology (conventional, chondroid, de‐differentiated, poorly differentiated)

• Extent of resection

• Location (skull base, spine)

• Type of surgery (biopsy, complete, incomplete resection)

• RT total dose

• Timing of RT (prior to surgery, early after surgery (< 90 days), at relapse or progression (> 90 days after surgery)

We resolved any discrepancies about results by discussion and consensus agreement. We included all assessments in a risk of bias graph, and a risk of bias summary figure, in addition to risk of bias tables (Figure 2; Figure 3).

2.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

3.

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Measures of treatment effect

Survival‐type outcomes

The measure of treatment effect chosen for survival‐type outcomes was the hazard ratio with 95% confidence intervals (HR (95% CI)).

Dichotomous outcomes

The measure of treatment effect chosen for treatment related toxicity was the risk ratio (RR) with 95% CI.

Unit of analysis issues

We planed to take into account the level at which randomization occurred. We planed to extract the following information for cluster‐RCTs (Higgins 2011).

• The number of clusters (or groups) randomised to each intervention group

• The outcome data ignoring the cluster design for the total number of individuals

• An estimate of the intracluster (or intraclass) correlation coefficient (ICC)

However, we did not include cluster‐randomised trials, or trials in which participants received more than one intervention

Dealing with missing data

We did not impute for any missing data. We contacted the study authors to request more information. However, we were unable to obtain more information or missing data from study authors, and we conducted available‐case analysis.

Assessment of heterogeneity

We accounted for heterogeneity by visually inspecting the forest plots to detect overlapping confidence intervals, using the Chi² test and the I² statistic (Higgins 2011). The I² statistic represents the percentage of the overall variability due to between‐study (or interstudy) variability, as opposed to within‐study (or intrastudy) variability (Higgins 2011). We interpreted the I² value as follows: 0% to 40% might not be important; 30% to 60% may represent moderate heterogeneity; 50% to 90% may represent substantial heterogeneity; and 75% to 100% may represent considerable heterogeneity (Deeks 2001).

Assessment of reporting biases

We intended to investigate potential publication bias by using funnel plots, but we were unable to, because < 10 studies met the inclusion criteria of the review.

Data synthesis

We analysed the data using Review Manager 5 (Review Manager 2014). Chordoma of skull base and of the spinal axis were not analysed separately due to the small number of studies and small study sizes. We were unable to extract the HRs and their variances for survival‐type data from the articles, as they were not reported. We either estimated the log HR (experimental relative to control) by (O – E)/V, which has a standard error 1/√V, where O is the observed number of events in the experimental intervention, E is the log‐rank expected number of events in the experimental intervention, O – E is the log‐rank statistic, and V is the variance of the log‐rank statistic (Higgins 2011), and computed the log HR from individual patient data (Parmar 1998), from three studies (Park 2006; Takahashi 2009; Jagersberg 2017), or we depicted the log HR and standard error of log HR from the Kaplan‐Meier curve, using an Excel spreadsheet by Tierney 2007 for one study (Colli 2001).

Regarding treatment‐related toxicity, we used the RR as a measure of association, and fixed a higher type I error (α = 0.10, two‐sided (Shadish 2002)).

We meta‐analysed the HRs and RRs by using the generic inverse‐variance method. We used a random‐effects model, as studies were clinically and statistically heterogenous. (DerSimonian 1986).

Subgroup analysis and investigation of heterogeneity

There were too few included studies that reported sufficient details, so we were unable to conduct subgroup analyses for categorical covariates, or meta‐regression for categorical and continuous covariates, according to the following:

• Histology (conventional, chondroid, de‐differentiated, poorly differentiated)

• Location (skull base, spine).

• Age (continuous)

• Gender (male, female)

• Type of surgery (biopsy, complete, incomplete resection)

• RT total dose (continuous) and fractionation (dose per fraction, continuous)

• RT target volume (treatment margin used, if any, in centimetres, continuous)

• Timing of RT (prior to surgery, early after surgery (< 90 days), at relapse or progression (> 90 days after surgery)

Sensitivity analysis

We intended to explore reasons for study heterogeneity, and if necessary, carry out sensitivity analysis by omitting studies rated as high risk of bias' However, we only included a small number of studies, which did not enable us to conduct a sensitivity analysis.

Summary of findings and assessment of the certainty of the evidence

We presented the overall certainty of the evidence for each outcome according to the GRADE approach, which takes into account issues related to internal validity (risk of bias, inconsistency, imprecision, publication bias), and external validity, such as directness of results (Langendam 2013). We created a summary of findings table, based on the methods described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), and using GRADEpro GDT software (GRADEpro GDT). We used the GRADE checklist and GRADE Working Group certainty of evidence definitions (Meader 2014). We downgraded the evidence from high certainty by one level for serious (or by two for very serious) concerns for study limitations.

• High‐certainty: we are very confident that the true effect lies close to that of the estimate of the effect.

• Moderate‐certainty: we are moderately confident in the effect estimate. The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different.

• Low‐certainty: our confidence in the effect estimate is limited. The true effect may be substantially different from the estimate of the effect.

• Very low‐certainty: we have very little confidence in the effect estimate. The true effect is likely to be substantially different from the estimate of effect.

Limited available evidence from a small number of included studies didn't permit us to present a table specific to each location.

Results

Description of studies

Results of the search

Our initial search, from 1946 to 2021, yielded 27 references from CENTRAL, 1131 from MEDLINE, and 1592 from Embase. Following de‐duplication across the databases, the combined total yield was 2143 references. After title and abstract screening, we retrieved 156 studies for full‐text review and possible inclusion. We excluded 146 studies that were the wrong study design (N = 40), wrong patient population (N = 3), wrong outcomes (N = 4), no comparison between photon and proton arms (82) or wrong comparator (N = 13), wrong setting (N = 2), paediatric population (1) and no intervention (N = 1) . This resulted in 10 observational studies and four studies were further excluded (see Characteristics of excluded studies; Excluded studies). Six observational studies were included in the review (Included studies). Four of which were ultimately included in the analysis (Figure 1).

Included studies

We extracted data from six studies that met our inclusion criteria including 187 patients. (see Characteristics of included studies). All included studies were observational. The mean age of participants ranged between 40 to 56 years and included both males and females. Participants presented with chordoma and chondrosarcomas in Colli 2001, primary and recurrent sacral chordoma in Park 2006, or cranial and skull base chordoma in Jagersberg 2017, Sheybani 2014, Takahashi 2009 and Yasuda 2012 studies. The mean follow‐up period ranged between 36 months (Takahashi 2009) and 91 months (Park 2006). The extent of tumour resection was either partial, subtotal or radical as determined by MRI and neuroradiological assessment. The average dose of radiotherapy ranged between 60 to 77 Gy, although data were not available for all participants in included studies (Colli 2001; Jagersberg 2017; Park 2006; Sheybani 2014; Takahashi 2009; Yasuda 2012)

Colli 2001 was an observational study for people with chordoma (N = 53) and chondrosarcoma (N = 10) treated with surgery and proton or photon radiotherapy delivered adjuvantly in 33 participants.

• Proton arm: 25 participants

• Photon arm: 8 participants

Jagersberg 2017 was a single‐center retrospective study for cranial chordoma in which 24 surgical treatments were performed and followed by proton or photon radiation.

• Proton arm: 11 participants to a median dose of 74 Gy (E) at 1.8 to 2.0 Gy (E)

• Photon arm: 2 participants

Park 2006 was an observational study in which people were treated with proton or photon radiation alone (N = 6) or combined with surgery (N = 21) for sacral chordoma.

• Proton arm: 22 participants mean 71 Gy (E) in 36 fractions with protons combined with photons

• Photon arm: 5 participants to 60 to 62 Gy

Sheybani 2014 was a retrospective analysis of patients treated at authors’ institution for chordoma (N = 12).

• Proton arm: 11 participants

• Photon arm: 1 participant

Takahashi 2009 was an observation study for people with skull base chordoma who underwent surgery (N = 32), followed by carbon ion, proton, or SRS radiotherapy.

• Proton arm: 14 participants

• Photon arm: 7 participants

Yasuda 2012 was an observation study for people with skull base, cranio‐cervical junction and cervical chordoma who underwent surgery (N = 40). Ten participants received pure protons, 3 received pure photons and 17 received combined protons and photons.

Primary outcomes were assessed in the included studies were as follows:

• Local control: Two studies (N = 39 participants where N = 27 were included in the meta‐analysis) (Park 2006; Sheybani 2014)

• Mortality (instead of overall survival): Four studies (N = 65 where N = 9 were included in the meta‐analysis) (Jagersberg 2017; Park 2006; Sheybani 2014; Yasuda 2012)

• Recurrence (instead of progression‐free survival): Four studies (N = 94 participants where N = 81 were included in the meta‐analysis) (Colli 2001; Park 2006; Takahashi 2009; Yasuda 2012)

• Treatment‐related toxicity: one study (N = 33 participants) (Colli 2001)

Excluded studies

We excluded four reports (see Characteristics of excluded studies). In one, the outcomes were not separated by group; in another, combined treatments (photons plus protons) were not reported for individual people; one was a review rather than a comparison of the effects of two treatment; the fourth one primarily assessed proton therapy and did not report outcomes for the two participants who received photon therapy.

Risk of bias in included studies

Details are provided in the risk of bias table, with the Characteristics of included studies; summaries are provided in Figure 2 and Figure 3. We used the Risk of Bias In Non‐randomized Studies of Interventions (ROBINS‐I) assessment tool (Sterne 2016).

Bias due to confounding:

We determined that all studies were at high risk of bias due to confounding. Study authors did not use an appropriate analysis methods that adjusted for all the important potential confounder.

Bias in selection of participants into the study:

We judged all studies at high risk of bias as selection of participants for both arms based on participant characteristics observed after surgery in Park 2006, Takahashi 2009 and Yasuda 2012, and because participant groups were determined by enrolment time period (older: photons; more recent: protons) for another study (Colli 2001). One study was regarded as unclear risk as reported data for the entire cohort, not by treatment group (Jagersberg 2017). In addition, there was a further imbalance in the number of participants in each arm for all studies.

Bias in classification of interventions:

We judged the risk of bias in classification of interventions as high in all studies due to unavailable information to define intervention groups at the start of the intervention (Colli 2001; Park 2006; Sheybani 2014; Takahashi 2009; Yasuda 2012), except for Jagersberg 2017, which we regarded as unclear risk due to reported data for the entire cohort and not by treatment group.

Bias due to deviations from intended intervention:

Risk of bias due to deviations from intended intervention was unclear in all studies.

Bias due to missing data:

We assessed incomplete outcome data as high risk (Colli 2001; Jagersberg 2017), and low risk (Park 2006; ;Sheybani 2014; Takahashi 2009; Yasuda 2012). Low risk studies included complete reporting for participants in both study arms. High risk studies presented incomplete reports or missing data for: participants from institutions outside the primary institution (Colli 2001); loss to follow‐up of one participant, with no evidence that results were robust in the presence of missing data from a very small study sample (Jagersberg 2017).

Bias in measurement of outcomes:

Risk of bias in measurement of outcomes was low in three studies (Colli 2001; Park 2006; Takahashi 2009), unclear in two studies (Jagersberg 2017; Sheybani 2014), and high risk in one study where we could not extract outcomes for each intervention (Yasuda 2012).

Bias in selection of the reported result:

We were unable to extract data for outcomes separately for each intervention for one study, and determined it at high risk of bias (Yasuda 2012). We judged two studies as unclear risk of bias (Colli 2001; Jagersberg 2017), and three studies at low risk of bias (Park 2006; Sheybani 2014; Takahashi 2009).

Effects of interventions

See: Table 1

See: Table 1.

Primary outcomes

Local control

Two observational studies compared proton with photon therapy and measured local control as an outcome (Park 2006; Sheybani 2014). Park 2006 is an observational study in which people were treated with proton or photon radiation alone (N = 6) or combined with surgery (N = 21) for sacral chordoma with an estimated median follow‐up time of 66 months with a range of 26 months to 261 months. In Sheybani 2014 only one participant was included in the photon group. Median follow‐up time for local control among two studies ranged between 45.5 months and 66 months. The meta‐analysis for local control included 22 participants from one non‐randomised study (Park 2006). We did not include Sheybani 2014 in the meta‐analysis, as there was only one participant who was treated with photons and outcome data were available for all patients and not stratified by treatment. For Park 2006, the estimated log HR was 1.676, and the SE (log HR) was 1.069. Three of the four chordomas that were treated with higher than 73.0 Gy (E) of radiation alone had local control. As this is the only study that assessed local control, it didn't permit us to address the effect of the timing of radiotherapy (RT) (neoadjuvant, adjuvant, or salvage RT) with regard to surgeries. Also, we could not perform a sub‐group analysis on the varying endpoints between studies and the timing of irradiation.

We were uncertain if proton compared to photon therapy has any effect on local control (pooled HR 5.34, 95% CI 0.66 to 43.43; 2 observational studies, 39 participants (27 analysed); Analysis 1.1). We assessed the certainty of evidence to be very‐low (Figure 4).

1.1. Analysis.

Comparison 1: Protons versus photons, Outcome 1: Local control

4.

Forest plot of comparison: 1 Protons versus photons, outcome: 1.1 Local control

Overall survival (Mortality)

Four observational studies with 65 participants compared proton with photon therapy and measured mortality rather than overall survival (Jagersberg 2017; Park 2006; Sheybani 2014; Yasuda 2012). The meta‐analysis for mortality included 39 participants across two non‐randomised studies (Park 2006; Jagersberg 2017). Median follow‐up time ranged between 45.5 months and 66 months. Sheybani 2014 was not included in the meta‐analysis as analysis could not be performed on only one participant who was treated with photons with an unknown outcome. We did not include Yasuda 2012 in the meta‐analysis as this paper contained insufficient data to complete data extraction and data clarification was required, we contacted the study corresponding author by email but numeric data couldn't be extracted per each group. We re‐analysed individual participant data (IPD) for Park 2006 and Jagersberg 2017, using the Excel spreadsheet by Tierney 2007. The log HR (‐1.2) and standard error of log HR (1.34) were estimated by re‐analysis of IPD (Tierney 2007) and estimated HR was 0.03, 95% CI 0.02 to 4.19). Similarly, estimated log HR and SE(log HR) were ‐0.698 to 0.736 for Park 2006.

We are uncertain if proton compared to photon therapy has any effect on survival (pooled HR 0.44 to 95% CI 0.13 to 1.57; 4 observational studies, 65 participants (39 analysed); Analysis 1.2 . We assessed the certainty of evidence to be very‐low (Figure 5).

1.2. Analysis.

Comparison 1: Protons versus photons, Outcome 2: Mortality (instead of overall survival)

5.

Forest plot of comparison: 1 Protons versus photons, outcome: 1.2 Mortality (instead of overall survival)

Progression‐free survival (Recurrence)

Four observational studies including 94 participants compared proton with photon therapy and measured recurrence rather than progression‐free survival (Colli 2001; Park 2006; Takahashi 2009; Yasuda 2012). The meta‐analysis for recurrence included 81 participants across 3 non‐randomised studies (Colli 2001; Park 2006; Takahashi 2009). Median follow‐up time ranged between 3 years and 10 years. We did not include Yasuda 2012 in the meta‐analysis as we were unable to extract numeric data for each group. For Colli 2001, recurrence log HR (‐2.37) and standard error of log HR (0.75) were estimated from Kaplan Meier Curve using Excel spreadsheet by Tierney 2007, with estimated HR 0.09, 95% CI 0.02 to 0.40. Re‐analysis was also performed for Park 2006 and estimated log HR, SE(log HR) were ‐0.215 and 0.725 respectively. Similarly, for Takahashi 2009 log HR and SE(log HR) were estimated by re‐analysis of IPD ‐0.71 and 0.67.

We are uncertain if proton compared to photon therapy has any effect on recurrence (pooled HR 0.34, 95% CI 0.10 to 1.17; 4 observational studies, 94 participants (81 analysed); Analysis 1.3; ). We assessed the certainty of evidence to be very‐low (Figure 6).

1.3. Analysis.

Comparison 1: Protons versus photons, Outcome 3: Recurrence (instead of progression‐free survival)

6.

Forest plot of comparison: 1 Protons versus photons, outcome: 1.1 Recurrence (instead of progression‐free survival)

Treatment‐related toxicity

One study was included in the meta‐analysis for treatment‐related toxicity (Colli 2001). It reported that four participants treated with proton therapy developed radiation‐induced necrosis of their temporal bone, radiation‐induced damage to the brainstem, and chronic mastoiditis. One participant treated with photon therapy developed hearing loss, worsening of the seventh cranial nerve paresis, and ulcerative keratitis. The other studies did not report on this outcome.

We are uncertain of the effect of proton compared to photon therapy on treatment‐related toxicity (risk ratio (RR) 1.28, 95% CI 0.17 to 9.86; 1 observational study, 33 participants; Analysis 1.4). We assessed the certainty of evidence to be very low (Figure 7).

1.4. Analysis.

Comparison 1: Protons versus photons, Outcome 4: Treatment‐related toxicity

7.

Forest plot of comparison: 1 Protons versus photons, outcome: 1.4 Treatment‐related toxicity

Toxicity data more than 10 years old were not available, so we didn't report the incidence of chronic or late toxicities according to the RTOG/European Organisation for the Research and Treatment of Cancer Late Radiation Morbidity Scoring Schema (Cox 1995).

Stratified analysis:

We planned but were unable to conduct a stratified analysis by tumour location because all included studies involved cranial chordoma except Park 2006, which included participants with sacral chordoma.

Discussion

We conducted this systematic review to determine if there were differences in outcomes between those who received proton therapy compared to photon therapy for the treatment of chordoma. We included six observational studies in the meta‐analysis but were unable to find evidence in support of one treatment over the other.

Summary of main results

There are insufficient data to show an advantage for proton or photon therapy with respect to local control, overall survival (mortality), progression‐free survival (recurrence), and treatment‐related toxicity.

Overall completeness and applicability of evidence

Overall, we observed no difference between interventions in the outcomes measured. However, we were only able to analyse data from two studies for local control and one for treatment‐related toxicity, areas where the theoretical benefit of protons exist due to their dosimetry. In addition, overall survival is more susceptible to selection bias, as the theoretical benefits of protons could be more often reserved for people with longer life expectancy, due to the limited availability of protons worldwide.

Since the publication of the analysed studies, advancements in proton and photon therapy have continued. For example, more proton centres now implement intensity‐modulated proton therapy (IMPT), a delivery system that results in similar dosimetric advantages as intensity‐modulated radiation therapy (IMRT) over conventional radiotherapy. Advancements in imaging and planning for both proton and photon therapy have also been made in recent decades. Whether these advancements would result in a benefit of one intervention compared to the other for chordoma, warrants future study.

Quality of the evidence

Our review did not identify any randomised studies that compared proton therapy to photon therapy. All studies were retrospective or observation, therefore, did not allow baseline randomisation to either arm. Although the analysed studies were published after the year 2000, the treatment dates span decades, and therefore, some participants' work‐up and treatment would now be considered out of date, such as the lack of MRI use.

We downgraded the level of evidence to very low for all outcomes, due to high risk of bias in included studies; very serious imprecision owing to a very small sample size per study, small number of events, and very wide confidence intervals crossing the line of no effect, which failed to exclude appreciable benefit or harms; included studies for recurrence‐free survival reported variable duration of follow‐up periods. The extent of surgical resection was not consistent among studies and participants. In summary, we considered that all included studies contributed to very low‐certainty evidence, per GRADE methods.

Potential biases in the review process

The search was comprehensive and up to date, allowing us to identify all appropriate studies. We conducted duplicate, independent screening by at least two review authors, of all search results from the title and abstract phase to full‐text screening. Two review authors independently extracted data onto a redesigned data extraction form. We involved a third author to resolve any discrepancy in the extraction process. We tried to contact a study authors to retrieve more details for data extraction. We followed a strict analysis approach to extract log hazard ratio and respective variance, using different methods, and guided by Parmar 1998; Tierney 2007; and Higgins 2011. We assessed the certainty of the evidence for each outcome according to the GRADE approach, and presented a summary of findings table.

Potential limitations of the review were clinical heterogeneity, such as tumour location and completeness of resection, variable duration of follow‐up periods, as well as a limited number of included studies. Small sample sizes led to serious imprecision, hence, downgrading the certainty of the evidence. Some studied were not included in the meta‐analysis due to incomplete available data, or only one participant who was treated with photons. Also, all included studies were observational. Although we identified all relevant studies regardless of language, we did not systematically search all other language databases, therefore, we cannot rule out the possibility that some studies may have been missed.

Agreements and disagreements with other studies or reviews

This study primarily differs from other reviews, because in our efforts to reduce bias, we only considered studies that included people treated with proton or photon therapy. This is different from the Amichetti 2009 systematic review, in which they included single arm proton studies, and authors ultimately concluded that proton therapy had better results compared to conventional photon radiation. Another systematic review summarised proton outcomes without direct comparison to photon treatments (Alahmari 2019). Because these studies showed better outcomes with proton therapy, the Chordoma Global Consensus Group prefers the use of proton therapy, and considers conformal photon irradiation a viable alternative only when similar radiation doses to targets and sparing of normal tissue can be achieved (Stacchiotti 2015). The recommendation is appropriately scored as level of evidence V (studies without control group, case reports, and experts' opinions), consistent with our literature search.

Authors' conclusions

Implications for practice.

Currently, the evidence is very uncertain about the effects of proton therapy compared to photon therapy for people with chordoma, on local control, mortality, recurrence, or treatment‐related toxicity.

Implications for research.

Given the rarity of this tumour type and limited availability of proton therapy, it is unlikely a randomized trial will be performed comparing protons versus photons. With this limitation, we encourage researchers to prospectively document important standardised endpoints, such as local control, treatment‐related toxicity, and quality of life, so that as proton availability increases, informed decision‐making with future people with chordoma will be improved.

History

Protocol first published: Issue 12, 2018

Appendices

Appendix 1. CENTRAL search strategy

#1 MeSH descriptor: [Chordoma] this term only
#2 chordoma* or chordocarcinoma* or chordoepithelioma* or notochordoma*
#3 #1 or #2
#4 MeSH descriptor: [Radiotherapy] explode all trees
#5 Any MeSH descriptor in all MeSH products and with qualifier(s): [radiotherapy ‐ RT]
#6 radiotherap* or irradiation* or radiation* or radiosurgery#7 MeSH descriptor: [Photons] this term only
#8 MeSH descriptor: [Ions] explode all trees
#9 MeSH descriptor: [Protons] this term only
#10 photon* or ion* or proton* or particle* or hadron* 
#11 #4 or #5 or #6 or #7 or #8 or #9 or #10
#12 #3 and #11

Appendix 2. MEDLINE Ovid search strategy

1 Chordoma/
2 (chordoma* or chordocarcinoma* or chordoepithelioma* or notochordoma*).mp.
3 1 or 2
4 exp Radiotherapy/
5 radiotherapy.fs.
6 (radiotherap* or irradiation* or radiation* or radiosurgery).mp.
7 Photons/
8 exp Ions/
9 Protons/
10 (photon* or ion* or proton* or particle* or hadron*).mp.
11 4 or 5 or 6 or 7 or 8 or 9 or 10
12 3 and 11

key:
mp=title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier
fs=floating subheading

Appendix 3. Embase Ovid search strategy

1 chordoma/
2 (chordoma* or chordocarcinoma* or chordoepithelioma* or notochordoma*).mp.
3 1 or 2
4 exp radiotherapy/
5 rt.fs.
6 (radiotherap* or irradiation* or radiation* or radiosurgery).mp.
7 photon/
8 exp Ion/
9 proton/
10 (photon* or ion* or proton* or particle* or hadron*).mp.
11 4 or 5 or 6 or 7 or 8 or 9 or 10
12 3 and 11

key:

mp=title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier

fs=floating subheading

Data and analyses

Comparison 1. Protons versus photons.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Local control	1		Hazard Ratio (IV, Random, 95% CI)	Totals not selected
1.2 Mortality (instead of overall survival)	2	39	Hazard Ratio (IV, Random, 95% CI)	0.44 [0.13, 1.57]
1.3 Recurrence (instead of progression‐free survival)	3	81	Hazard Ratio (IV, Random, 95% CI)	0.34 [0.10, 1.17]
1.4 Treatment‐related toxicity	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Colli 2001.

Study characteristics
Methods	Study design: people with chordoma and chondrosarcoma were consequently recruited in the study. We were unable to obtain further information about the study design. Study date: September 1990 to May 2000 Duration of participation (follow‐up): overall clinical or neuroimaging (or both) follow‐up period ranged from 1 to 150 months (mean 46.1, SD 40.2, median 38 months), from 1 to 150 months (mean 49.9, SD 39.5, median 40 months) in people with chordomas and 1 to 125 months (mean 28.3, SD 38.7, median 12 months) in those with chondrosarcomas
Participants	Total number: 63 (25 protons and 8 photons) Age (mean ± SD): 40.7 years ±15.5 Gender: 35 females/28 males, ratio 1.25:1 Tumour location: chordoma (41 typical and 12 chondroid) and 10 participants harboured chondrosarcomas Type of surgery: 15 participants with chordoma had undergone 41 operations elsewhere, and one with chondrosarcoma had undergone two previous operations elsewhere. Combining the number of previous surgeries performed for all participants were 136 procedures (2.3 per participant). Procedures included: cranioorbitozygomatic/extended cranioorbitozygomatic, unilateral/bilateral maxillotomy, transsphenoidal/extended transsphenoidal, transcondylar, transoral, anterior cervical decompression and fusion, suboccipital craniotomy, retromastoid–petrosal, and frontotemporal craniotomy Extent of resection: determined using MRI by a neuroradiologist and classified as follows: radical (absence of residual tumour or presence of a small questionable area) in 31 participants (49.2%); subtotal (> 90% resection) in 18 participants (28.6%); and partial (< 90% resection) in 14 participants (22.2%)
Interventions	Radiotherapy involved 36 participants with chordomas. Adjuvant proton‐beam radiotherapy for 25 participants and conventional radiation therapy for 8 participants, protons to photons 3.12:1. Others received previous radiotherapy. Proton‐beam radiotherapy: mean 70.53 CGE, median 72 CGE, SD 5.235 CGE, and range 60 CGE to 77.4 CGE (dose available for 12 participants only)
Outcomes	Overall survival for all participants and subgroup by type: typical, chondroid chordoma or chondrosarcoma; extent of resection and karyotype analysis (normal versus abnormal) Overall mortality Complications by tumour type Recurrence free survival (RFS) stratified for protons and photons groups. RFS log HR and standard error (SE) of log HR were estimated from Kaplan Meier Curve using Excel spreadsheet by Tierney 2007 Estimated log HR, SE (log HR): ‐2.37, 0.75; estimated HR (0.09, 95% CI 0.02 to 0.40)
Notes	Study author email: bocolli@fmrp.usp.br
Risk of bias
Bias	Authors' judgement	Support for judgement
Bias due to confounding	High risk	Study authors did not use an appropriate analysis method that adjusted for all important confounding domains
Bias in selection of participants into the study	High risk	Participant groups determined by enrolment time period (older: photons; more recent: protons)
Bias in classification of interventions	High risk	The intervention groups were not clearly defined; no information used to define intervention groups that was recorded at the start of the intervention.
Bias due to deviations from intended interventions	Unclear risk	Not specified
Bias due to missing data	High risk	Not specified, follow‐up data from outside institutions
Bias in measurement of outcomes	Low risk	Most outcome measures were objective; were not influenced by knowledge of the intervention received
Bias in selection of the reported result	Unclear risk	Progression‐free survival reported for both arms

Jagersberg 2017.

Study characteristics
Methods	Study design: a retrospective clinical study. Data were collected from participants' files, the institutional radiological imaging data base, and standardised telephone interview plus Short Form‐36 Health Surveys (SF‐36) at the start of data acquisition. Study date: 2005 to 2015 Duration of participation (follow‐up): mean follow‐up period 78 ± 54 months, median follow‐up time 64 ± 41 months The study was approved by the local ethics committee board (ID 2016–01789).
Participants	Total number: 13 ( 11 protons and 2 photons) Age (mean ± SD): 56 ± 12 years Gender: 6 females/7 males, ratio 1:1.16 Tumour location: cranial chordoma Type of surgery: 24 surgical treatments overall (30 surgical procedures). Surgery involved solitary, combined surgery (e.g. first transnasal‐trans‐sphenoidal and then pterional; or first transoral transpharyngeal and then lateral suboccipital), resection complete and resection incomplete Extent of resection: participants were grouped into incomplete or gross total resection based on MRI and interdisciplinary tumour board Inclusion criteria: all people surgically treated for histologically proven chordoma at the authors’ institution between 2005 and 2015
Interventions	Eleven participants (85%) received proton beam therapy and two participants (15%) received photon beam therapy (one fractioned and one radiosurgical administration); protons to photons 5.5:1 Proton‐beam radiotherapy: median dose of 74 Gy relative biological effectiveness (RBE), with daily fraction of 1.8 GyRBE to 2.0 GyRBE
Outcomes	Overall survival for all participants (OS): OS Log HR and standard error of log HR were estimated by re‐analysis of IPD from table 2 in the study, and by using the Excel spreadsheet by Tierney 2007; estimated log HR, SE (log HR): ‐1.2, 1.34; estimated HR (0.03, 95% CI 0.02 to 4.19) Progression‐free survival (PFS): 5‐year progression‐free survival rate was 53% (95% CI 26% to 79%); mean progression‐free survival was 64 ± 50 months; data were not available to allow us to calculate PFS per treatment Overall complications: complications from radiotherapy (other than moderate asthenia, alopecia, moderate mucosa problems) were unilateral useful hearing loss in two participants, vision loss under 50% in one participant, and oculomotor deficit in one participant
Notes	Study author email: max.jaegersberg@hcuge.ch
Risk of bias
Bias	Authors' judgement	Support for judgement
Bias due to confounding	High risk	Study authors did not use an appropriate analysis method that adjusted for all important confounding domains.
Bias in selection of participants into the study	Unclear risk	Reported data for entire cohort, not by treatment group
Bias in classification of interventions	Unclear risk	Reported data for entire cohort, not by treatment group
Bias due to deviations from intended interventions	Unclear risk	Not specified
Bias due to missing data	High risk	Only one participant was lost follow‐up, but there was no evidence that results were robust in the presence of missing data, in light of very small study sample.
Bias in measurement of outcomes	Unclear risk	Reported data for entire cohort, not by treatment group
Bias in selection of the reported result	Unclear risk	Reported data for entire cohort, not by treatment group

Park 2006.

Study characteristics
Methods	Study design: an observational study, participants with primary (N = 16), and recurrent (N = 11) chordoma were treated by proton/photon radiation therapy, either alone (6 participants), or combined with surgery (21 participants) For participants who could not return to be seen, data were collected from reports from the physicians and nurses, and by questionnaires or telephone interviews of the participants or their families Study date: November 1982 to November 2002 Duration of participation (follow‐up): mean follow up (SD): 91.11(62.23) months
Participants	Total number: 27 ( 22 protons and 5 photons) Age: median age was 56 years (range 35 to 81 years) in 14 primary, and 55 years (range 52 to 60 years) in 7 participants with recurrent chordoma Gender: 10 females/17 males, female/male ratio 0.588 Tumour location: primary and recurrent sacral chordoma Type of surgery and extent of resection were unclear
Interventions	Proton therapy: 18 participants were treated by surgery combined with protons, photons, or both; 4 participants received only proton combined with photon 73.0 Gy to 77.4 Gy Proton therapy dose: dose was 73.0 Gy to 77.4 Gy in proton combined with photon; mean doses were 71 Gy (E) in 36 fractions for primary, and 77 Gy (E) in 41 fractions for recurrent chordomas Photon therapy: 2 participants received photons only (60 Gy and 62 Gy) Mean duration of treatment was 80 days (range 30 to 172 days) for primary, and 73 days (range 38 to 112 days) for recurrent chordomas
Outcomes	Local control (LC) rate was 71.7%, continuous disease‐free survival rate was 60.5%, and overall survival (OS) rate was 82.5% at 5 years; local control rate was 57.5%, continuous disease‐free survival rate was 48.6%, and overall survival rate was 62.5% at 10 years We re‐analysed data from study tables 1, 2, and 3 to assess LC, OS, and recurrence per group OS: estimated OS log HR, SE (log HR): ‐0.698, 0.736 Local control: estimated log HR, SE (log HR): 1.676, 1.069 Recurrence: estimated log HR, SE (log HR): ‐0.215, 0.725
Notes	Study author e‐mail: lilypark@hotmail.com
Risk of bias
Bias	Authors' judgement	Support for judgement
Bias due to confounding	High risk	Study authors mentioned potential confounders (location, dose and extent of resection) but did not use appropriate analysis method that adjusted for them in context of association between type of intervention and outcomes.
Bias in selection of participants into the study	High risk	Selection of participants for both arms based on participant characteristics observed after surgery.
Bias in classification of interventions	High risk	No information used to define intervention groups was recorded at the start of the intervention.
Bias due to deviations from intended interventions	Unclear risk	Not specified
Bias due to missing data	Low risk	Complete outcomes reported for both arms.
Bias in measurement of outcomes	Low risk	Most outcome measures were objective; would not be influenced by knowledge of the intervention received.
Bias in selection of the reported result	Low risk	Complete outcomes reported for both arms.

Sheybani 2014.

Study characteristics
Methods	Study design: a retrospective analysis using people treated at authors’ institution for chordoma Study date: July 1999 to December 2010 Duration of participation (follow‐up): mean follow‐up was 55 months, estimated median was 45.4 months from analysis of IPD (17 to 134 months).
Participants	Total number: 12 participants (11 protons and 1 photons) Age (median (range)): 47 years (7 to 69 years) Gender: 7 females/5 males Tumour location: skull base chordomas Type of surgery: subtotal resection in 9 participants, multiple surgeries prior to radiation treatment in 2 participants; and biopsy was performed for one patient and was referred for definitive treatment. Exclusion criteria: people with non‐skull base lesions, people who did not have radiation therapy, and those who were treated palliatively
Interventions	Eleven participants (91.6%) received proton beam therapy; and 1 participant (8.4%) received photon beam therapy Proton‐beam radiotherapy: 11 participants with available data had median (81 Gy (46.7 Gy to 85.7 Gy)) and mean (38 Gy (6.6 Gy to 58.1 Gy))doses to the brainstem; 2 clinical target volumes (CTV) were created for each participant, which brought the total tumour dose to 81 Gy in 64 fractions, completed in 32 days
Outcomes	Overall survival for all participants (OS) Local control analysis could not be performed on data from only one participant who received photons and outcome data were available for all patients not stratified by treatment. Overall complications: clinically significant, late toxicity occurred in only one participant, with progressive hearing loss 3 years after his treatment; another three participants developed performance decline after treatment, due to disease progression; we were unable to retrieve any further information per group.
Notes	Study author e‐mail: arshin‐sheybani@uiowa.edu
Risk of bias
Bias	Authors' judgement	Support for judgement
Bias due to confounding	High risk	Only one participant received photons, with inappropriate control of confounders in analysis.
Bias in selection of participants into the study	High risk	Only one participant received photons.
Bias in classification of interventions	High risk	Only one participant received photons. The information used to define intervention groups was not recorded at the start of the intervention
Bias due to deviations from intended interventions	Unclear risk	Not specified
Bias due to missing data	Low risk	Reported data for entire cohort, not by treatment group
Bias in measurement of outcomes	Unclear risk	Most outcome measures were objective; would not be influenced by knowledge of the intervention received
Bias in selection of the reported result	Low risk	Reported data for entire cohort, not by treatment group

Takahashi 2009.

Study characteristics
Methods	Study design: an observational study; recruited people with chordoma who underwent surgery Study date: November 1996 to August 2007 Duration of participation (follow‐up): maximum 10 years; mean follow‐up period since initial visit to the outpatient clinic was 36.3 months (range 3 months to 93 months); mean follow‐up period since carbon ion radiotherapy was 29.6 months (range 7 months to 55 months); and follow‐up since SRS was estimated as 26 months (range 1 month to 56 months)
Participants	Total number: thirty‐two participants received surgery, 21 of whom also received adjuvant radiotherapy Age (mean, range): 41.4 years (range 10 to 75 years). Mean age of participants who received carbon ion radiotherapy at initial presentation was 41.4 years (range 12 to 75 years) Gender: 20 females/12 males; carbon ion therapy: 6 females/3 males Tumor location: skull base chordoma extended to the clivus (81.3%); 21 (65.6%) of tumours had invaded the posterior fossa, and 14 (35.7%) were parasellar Each participant underwent 1 to 6 surgical procedures (average 1.8 per participant); total 59 surgical procedures Type of surgery: 59 operations were performed on the 32 participants. Surgery involved: anterior trans‐petrosal, trans‐oral, extended trans‐sphenoidal, trans‐sphenoidal, orbito‐zygomatic , zygomatic trans‐petrosal, lateral suboccipital, pterional, posterior approach to C2, transcondylar, Le Fort I, subtemporal approach, foramen magnum decompression, and transmaxillar. Extent of resection: subtotal resections (> 90%) in 24 (40.7%) operations; partial (< 90%) resections of the lesion in the remaining operations The mean KPS of the 9 participants treated with adjunctive carbon ion radiotherapy was 83.3 before surgery, 80 after surgery, 80 at 6 months after radiotherapy, and 81.1 at the last follow‐up visit
Interventions	Radiotherapy was administered to 21 participants: carbon ion radiotherapy to 9 participants, proton ion radiotherapy to 5, and stereotactic radiosurgery (SRS) to 7 participants Type of adjunctive radiotherapy was selected according to the distance between the tumour and the optic nerve or brainstem, the amount and location of the residual tumour, availability of the type of adjunctive therapy, and participant's financial status. Dose: 57.6 to 60.8 (average dose 60.4 Cobalt Grey equivalent (CGE) for the 9 participants who received Carbon ion radiotherapy
Outcomes	Overall survival for all participants, and the subgroup carbon ion radiotherapy versus other radiotherapy (protons and SRS combined) and no radiotherapy Overall recurrence‐free survival (RFS) and the subgroup carbon ion radiotherapy versus other radiotherapy (protons and SRS combined) and no radiotherapy The 3‐year recurrence‐free survival rates for the carbon ion therapy treated group was 70.0%, and was higher than that of the radiotherapy treated group (57.1%; proton therapy plus SRS; P = 0.6648). We estimated the RFS log HR and standard error of log HR by re‐analysing IPD from the study results, using an Excel spreadsheet by Tierney 2007. Estimated RFS log HR, SE (log HR): ‐0.71, 0.67; estimated HR 0.49, 95% CI 0.13 to 1.83 Adverse events reported by participants who received carbon ion radiotherapy: tympanitis (2 participants), brain oedema (1 participant), and cranial nerve palsy (as severe adverse event) in one participant, for a total 4/9 (44.4%)
Notes	Study author e‐mail: kawase@sc.itc.keio.ac.jp
Risk of bias
Bias	Authors' judgement	Support for judgement
Bias due to confounding	High risk	Study authors mentioned potential confounders (location), but did not use an appropriate analysis method that adjusted for them in context of association between type of intervention and outcomes.
Bias in selection of participants into the study	High risk	Selection of participants for both arms based on participant characteristics observed after surgery.
Bias in classification of interventions	High risk	The information used to define intervention groups was not recorded at the start of the intervention.
Bias due to deviations from intended interventions	Unclear risk	Not specified
Bias due to missing data	Low risk	Outcomes reported for both arms.
Bias in measurement of outcomes	Low risk	Most outcome measures were objective; would not have been influenced by knowledge of the intervention received.
Bias in selection of the reported result	Low risk	Outcomes reported for both arms.

Yasuda 2012.

Study characteristics
Methods	Study design: observational study for the aim of this review; recruited people with chordoma who underwent surgery Study date: January 2001 to August 2008 Duration of participation (follow‐up): 3 to 99 months for all included participants (mean 46.1 months, median 56.5 months). The mean follow‐up period for the eight participants who received additional radiotherapy was 63.9 months.
Participants	Total number: 40, 23 participants underwent surgery for the first time (primary surgery: PS group). 17 participants had already undergone a surgical resection at least once before, and had another operation during this period for recurrence (recurrence surgery: RS group). Age (mean 45.1, median 46.5) years old; range: 11 to 68 years. Age range of the PS group was 11 to 65 years (mean 40.9 years, median 42.0 years); RS group was 25 to 68 years (mean 50.6 years, median 49.0 years) Gender: 32 females/18 males Tumour location: 17 at the skull base (42.5%), 13 at the cranio‐cervical junction (CCJ; 32.5%), and 10 at the cervical spine (25.0%) Type of surgery: anterolateral, transsphenoidal (5 endoscopic and 2 microscopic), transoral, subfrontal, posterolateral, pterional, subtemporal, and transfacial Extent of resection: extensive radical resection in 17 participants (42.5%), subtotal resection (more than 95% of the tumour reduction) in 19 participants (47.5%), partial resection (80% to 95% of the tumour) in 4 participants (10.0%) Inclusion criteria: pathological diagnosis was confirmed by light microscopic findings and immunohistochemical staining Exclusion criteria: participants with a sarcomatous type of chordoma
Interventions	22 participants without previous history of irradiation, 4 participants received proton therapy (74 CGE), with mean follow‐up of 62.3 months, and 1 participants received photon therapy (55 Gy) 18 participants had a previous history of radiation, 8 of whom had additional radiotherapy after the present surgery; pure proton irradiation (6 participants, mean dose 71 CGE), pure photon radiation (2 participants, mean 55 Gy) Total number of participants who received pure protons = 10, pure photons = 3, and 17 participants received combined protons and photons
Outcomes	Progression‐free survival (PFS) time and overall survival (OS) time were evaluated for each participant Among 4 participants who received proton therapy with previous irradiation, one participant (25%) experienced local relapse, but nobody died. The participant who received photons developed a recurrence, and died at 33‐month follow‐up period. From both groups with additional radiotherapy (N = 8), two participants developed local recurrence (25%) after an average of 27 months, and died an average of 48 months after the treatment.
Notes	Study author email: myasuda@aichi‐med‐u.ac.jp
Risk of bias
Bias	Authors' judgement	Support for judgement
Bias due to confounding	High risk	Study authors mentioned potential confounders (location), but did not use an appropriate analysis method that adjusted for them in context of association between type of intervention and outcomes.
Bias in selection of participants into the study	High risk	Selection of participants for both arms based on participant characteristics observed after surgery.
Bias in classification of interventions	High risk	The information used to define intervention groups was not recorded at the start of the intervention.
Bias due to deviations from intended interventions	Unclear risk	Not specified
Bias due to missing data	Low risk	No missing data among participants who received radiotherapy
Bias in measurement of outcomes	High risk	Most outcome measures were objective; would not be influenced by knowledge of the intervention received. However, we could not extract outcomes for each intervention.
Bias in selection of the reported result	High risk	Outcomes could not be extracted for each intervention.

SD: standard deviation; HR: hazard ratio; MRI: magnetic resonance imaging; IPD: individual participant data; SRS: stereotactic radiosurgery; KPS: Karnofsky performance status; CGE: Cobalt Grey equivalent

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
McDonald 2013	We could not extract numeric data for outcomes per group. Study tables did not give any information that could be stratified per group. There were results for overall survival, overall local control, and toxicities, without subgroup analyses by protons and photons.
Nowakowski 1992	Out of 52 participants with tumours, 24 participants were treated for paraspinal chordoma or chondrosarcoma. Data for treatment with charged particles therapy and photon therapy were unavailable for participants with chordoma or chondrosarcoma,
Van Gompel 2015	The study is mainly an overview of presenting characteristics, imaging, pathology, and overview of different treatment modalities for chordoma, rather than comparing efficacy between protons and photons.
Weber 2005	The study mainly assesses the local control and progression‐free survival of proton therapy. Only two cases of photon therapy and no available data for those two cases.

Differences between protocol and review

We stated in the protocol that eligible studies could include early versus delayed treatment with either particles or photons. However, for the aim of this review, and to accurately assess the effects and toxicity of particle versus photons, we restricted selection to only studies that compared particles versus photons. Furthermore, the screening process revealed several studies with people treated with one modality, but not both. These were restricted in an attempt to reduce bias. All included studies used protons, therefore we changed the name to reflect this; the comparison was protons versus photons.

Contributions of authors

I El Sayed, D Trifiletti, and S Dutta completed the screening of abstracts and full texts.
I El Sayed and S Dutta conducted methodological assessment using the ROBIN‐I tool.
I El Sayed conducted data synthesis, analysis and the GRADE assessment to create a summary of findings table, using GRADEpro GDT software.
I El Sayed, D Trifiletti, E Lehrer, TN Showalter, and S Dutta completed the final version of the review.

Sources of support

Internal sources

• None, Other

External sources

• None, Other

Declarations of interest

I El Sayed: nothing to declare
D Trifiletti: funding from Novocure for unrelated clinical trial research
E Lehrer: nothing to declare
T Showalter: nothing to declare
S Dutta: nothing to declare

New