Panoramic Radiobiological Modelling of the Contribution of Concomitant Chemotherapy to Biological Effective Dose in Squamous Cell Carcinoma

Objective: Attempts have been made to model the contribution of concomitant chemotherapy to radiotherapy in terms of biological effective dose (BED) for the major squamous cell carcinoma (SCC) sub-sites. Despite SCC’s sharing common aetiology, different concomitant chemoradiotherapy regimens are used in clinical practice. This study aims to compare the contribution of chemotherapy to radiotherapy in terms of BED across the major SCC sub-sites using two different radiobiological models; the intuitive and Poisson methods for its calculation. Methods: Phase 3 trials of radiotherapy versus chemoradiotherapy using conventional fractionation in SCC of the head and neck, lung, cervix, oesophagus and anus were identified. The contribution of chemotherapy (tBEDc) was modeled using both the intuitive and the Poisson model to give a weighted BED in Gray. Results: Weighted tBEDc using the intuitive model were 8.6 Gy10 for head and neck, 6.3 Gy10 for lung, 6.3 Gy10 for cervix and 7.8 Gy10 for anus. The weighted tBEDc using the Poisson model were 1.8 Gy10 for head and neck, 0.9 Gy10 for cervix and 2.1 Gy10 for anus. Conclusion: There is a striking similarity for the value of tBEDc across SCC sub-sites within both models. In head and neck cancer tBEDc derived from the Poisson model is not associated with the same biological effect as the same BED administered as radiotherapy alone. Therefore at this sub-site, where there is good data on radiotherapy dose response in the curative dose range, the Poisson model may be of limited value. However, it may be preferred for sub-sites where such data is lacking.


Introduction
Concomitant radical chemoradiotherapy is used commonly for squamous cell carcinoma (SCC) arising in the head and neck, lung, uterine cervix, oesophagus and anus (Pignon, le Maître, Maillard, Bourhis, & MACH-NC Collaborative Group, 2009;Auperin et al., 2010;Chemoradiotherapy for Cervical Cancer Meta-Analysis Collaboration, 2008;Anonymous, 1996;Herskovic et al., 1992).SCC arising at different anatomical sites may share similar aetiological factors; namely smoking and alcohol misuse.Human papilloma virus (HPV) has been implicated in the pathogenesis of SCC of the oropharynx, cervix and anus (Ang et al., 2010;Bosch, Lorincz, Muñoz, Meijer, & Shah, 2002;De Vuyst, Clifford, Nascimento, Madeleine, & Franceschi, 2009).Where such aetiological similarities occur, it is remarkable to note the difference in radiotherapy and concomitant chemotherapy doses employed in the curative setting to achieve similar local control outcomes seen, for example, 70 Gray in 35 fractions is commonly employed in the radical setting for head and neck SCC, whereas 50.4 Gray in 28 fractions achieves similar rates of local control in anal SCC.Although HPV positive oropharyngeal cancer has been shown to confer a better prognosis compared with HPV negative disease (3-year overall survival 82.4% versus 57.1% p<0.001 for tumours treated with chemoradiotherapy within the RTOG 0129 study) the prognostic implications in anal and cervical cancer are unknown, largely due to the small number of HPV negative tumours associated with these sub-sites (Ang et al., 2010;Zandberg, Bhargave, Badin, & Cullen, 2013).
Numerous attempts have been made to model the contribution of chemotherapy to radiotherapy in terms of the biological effective dose (BED) (Geh, Bond, Bentzen, & Glynne-Jones, 2006;Hartley, Sanghera et al., 2010;Jones, & Sanghera, 2007).Calculating the additional BED provided by chemotherapy is important in attempting to predict toxicity and allows comparison of different regimes.Jones and Dale refer to many of these studies as using the "intuitive" or "rule of thumb" method (Jones & Dale, 2005).Briefly the observation that approximately a 1% increase in local control is seen with a 1% increase in BED is employed to calculate the chemotherapy contribution to local control (tBEDc).In sub-sites such as head and neck cancer, where there are several randomized trials of altered fractionation versus conventional radiotherapy alone, more extensive modeling is possible and weighted values for this dose gradient can be calculated (Fowler, Harari, Leborgne, & Leborgne, 2003).Jones and Dale describe a second method employing the Poisson model for tumour control probability (TCP) (Jones & Dale, 2005).
Given the similar aetiology of SCC sub-sites, this study aims to compare the contribution of synchronous chemotherapy to radiotherapy in terms of BED calculation across the major SCC sub-sites using both the intuitive and Poisson radiobiological models.
The following ratio was derived from a radiobiological study of head and neck cancer (Hartley, 2011): S t = the ratio of the percentage increase in local control to the percentage increase in tBED = 1.2.
Phase 3 prospective randomised controlled trials of conventionally fractionated radiotherapy versus chemoradiotherapy in SCC of the head and neck, lung, uterine cervix, oesophagus and anus were identified.
Trials were included in this study if the total dose (D), dose per fraction (d) overall treatment time (T) and local control rates at 3 years were published.Trials that reported their results as complete response, partial response, stable disease or progressive disease were excluded.Studies were included if the concomitant agent was cisplatin, carboplatin, mitomycin-C, 5-flurouracil or a vinca alkaloid, either as a single agent or in combination.Trials were excluded if a different radiotherapy dose was employed between the two arms of the trial or if they were not published in English.Trials were then grouped by tumour sub-site.A list of excluded trials of conventionally fractionated radiotherapy versus conventionallyfractionated radiotherapy plus synchronous chemotherapy is provided in appendix 1.
An additional 'boost' of radiotherapy was historically administered in anal cancer trials after a gap of 6 weeks (Anonymous, 1996;Bartelink, 1997).An analysis of the United Kingdom Coordinating Committee on Cancer Research (UKCCCR) ACT I trial failed to find evidence that such boosts improved local control after a 6 week gap (Glynne- Jones et al., 2011).Therefore, for the purposes of this study, delayed anal cancer boosts are not taken into account in the calculations.
The contribution of chemotherapy (tBEDc) was modeled using two different methods.
For the intuitive method: tBED for the common radiotherapy components of both arms of the studies was calculated using equation 1.The percentage difference (Δ%) in tBED was obtained by dividing the absolute observed percentage difference in local control by S t (1.2) (equation 2).tBEDc was then obtained by multiplying the radiotherapy component tBED by the percentage difference in tBED expressed as a decimal fraction (equation 3).
For the Poisson method: The overall cytotoxic drug related cell kill (E c ) was calculated using 3 year local control rates as tumour control probabilities for the radiotherapy alone and chemoradiotherapy arms of the trial using equation 4. The tBEDc was obtained by dividing E c by the α value of 0.3 Gy -1 (equation 5).
E c = ln (ln TCP 1 /ln TCP 2 ) Equation 4tBEDc = E c / α Equation 5 Where E c = the overall cytotoxic drug related cell kill (including all cycles of concomitant chemotherapy), ln = natural log, TCP 1 = Tumour control probability from the radiotherapy component (3 year local control rate), TCP 2 = Tumour control probability from the chemoradiotherapy component (3 year local control rate).

Results
Randomised controlled trials of radiotherapy versus chemoradiotherapy using conventional fractionation are listed in Table 1 for head and neck, cervix, anus and lung.No oesophageal trials meeting the criteria for calculation and therefore inclusion were identified.BEDs were calculated using the equations (  Table 2 lists the same trials with the tBEDc derived using Poisson Modeling.BEDs were calculated using the equations (Pignon et al., 2009;Anonymous, 1996;Herskovic et al., 1992;Ang et al., 2010) described in the methods above.The weighted tBEDc using the Poisson model were: head and neck 1.8 Gy 10 , cervix 0.9 Gy 10 and anus 2.1 Gy 10 .A weighted tBEDc for lung could not be calculated using the Poisson model as a local control rate of 0% was seen at 3 years intrials otherwise meeting the inclusion criteria for calculation by the intuitive method.Appendix 1 lists the excluded phase 3 trials using conventionally fractionated radiotherapy and the reasons for ineligibility.

Discussion
Although the results obtained from the two models differed significantly, the similarity of the magnitude of the contribution of synchronous chemotherapy in terms of BED (tBEDc) across anatomical sub-sites within each of the two models is striking.Taking the example of the intuitive method, tBEDc ranged from its lowest value of 6.3 Gy 10 in lung and cervical cancer to the highest value of 8.6 Gy 10 in head and neck cancer.This difference of 2.3 Gy 10 BED is approximately equivalent to a single 2 Gy fraction.Based on these results it appears that for SCC arising in the head and neck, anus, uterine cervix and lung, synchronous chemotherapy adds between 4.5 and 6.8 Gy in 2 Gray fractions (EQD2) (Lee, Forey, & Coombs, 2012).However, there are many limitations to this appealing yet simplistic analysis.
In the intuitive method, radiosensitivity and repopulation parameters have been assumed to be identical for each of the anatomical sub-sites.This is unlikely to be the case given not only the heterogeneity of tumours within each sub-site but also the different tumour micro-environments at the varied anatomical locations.Although these tumours do share common aetiological factors, smoking remains the predominant risk factor for SCC of the lung whereas Human Papilloma Virus is the more significant factor for anal and cervical SCC (Zandberg et al., 2013;Lee et al., 2012).
In addition, the dose response gradient (S t has been modeled as a constant of 1.2% increase in local control for a 1% increase in BED.This value was derived from a previous study of randomized trials of head and neck cancer where altered fractionation radiotherapy alone schedules were randomized against conventionally fractionated radiotherapy again delivered as a sole modality (Hartley et al., 2010).The use of this value derived from head and neck cancer can be criticized on the basis of tumoural, environmental and aetiological heterogeneity as above.Furthermore, the above constant was derived from radiotherapy data in the range of 62.1 to 76.8 Gy 10 BED.In the current study the radiotherapy dose range was much wider from 45.4 Gy 10 in anal cancer to 77.5 Gy 10 in cervical cancer.The absence of trials comparing radiotherapy alone schedules in non-head and neck cancer SCC makes the derivation of an appropriate dose gradient in anal, cervical and lung cancer currently impossible.However, the Poisson based model may be used for tumour sites where there is no derivable dose gradient from radiotherapy alone studies.
In head and neck cancer the Poisson model may be of limited value.For example, if we take the trial of Dennis et al. ( 2004) in head and neck cancer a 26% increase in local control was seen for the addition of 2.5 Gy 10 BED of chemotherapyccording to the Poisson model.Given the radiotherapy alone component of the treatment contributes 61.7 Gy 10 BED, the tBEDc of 2.5 Gy 10 represents a 4% increase in BED.The Poisson model suggests, therefore, a 6.5% increase in local control for every 1% increase in BED.In head and neck cancer from radiotherapy alone trials we know this gradient in practice is 1.2 for a value of α of 0.3 Gy -1 A further criticism is that synchronous chemotherapy agents have been considered together with no attempt to account for their possible differing potency or dose intensity.To consider individual agents was impossible given the small number of trials that met the eligibility criteria.Previous modeling work has attempted to derive regime specific tBEDc.For example synchronous platinum doublets were found to have a higher tBEDc than synchronous single agents in head and neck cancer (Pettit et al., 2013).
It is important to note that numerous phase 3 randomised trials identified here were accepted for publication in major journals without the documentation of basic radiotherapy details including radiotherapy fractionation, overall treatment time and the endpoint of local control.Given the localised nature of SCC it is imperative that studies report such data.Furthermore prospective trials should have appropriate radiotherapy quality assurance to permit more reliable modeling.Appendix one provides further details of excluded trials.A further limiting factor for eligibility of studies was the choice of 3 year local control as an endpoint as this excluded many lung and oesophageal studies from the analysis due to the poor prognosis associated with these sub-sites.
In conclusion, remarkable similarities in the values of tBEDc are seen within each model across SCC sub-sites The Poisson model may be preferred for sub-sites where the dose response gradient is not established to avoid reliance on parameters extrapolated from squamous cell carcinoma of the head and neck.
Appendix 1. Excluded randomised trials of conventionally fractionated radiotherapy versus chemoradiotherapy in squamous cell carcinoma

a
arm one excluded (induction chemotherapy), b = personal correspondence.N = number of patients in study; Gy = Gray; OTT = overall treatment time of radiotherapy; d = dose per fraction; tBED = biologically effective dose when considering local control from radiotherapy; TCP = Tumour control probability; RT = Radiotherapy; CRT = Chemoradiotherapy; E c = overall cytotoxic drug related cell kill 5-FU = 5-flurouracil; MMC = mitomycin-C; Δ% = difference in percentage; SCCHN = Squamous cell carcinoma of the head and neck; SCC = squamous cell carcinoma, LDR = low dose rate.Rows in bold refer to weighted result for each subsite.

Table 1 .
Pignon et al., 2009;Auperin et al., 2010; Chemoradiotherapy for Cervical Cancer Meta-Analysis Collaboration, 2008;Ang et al., 2010)described in the methods above.The weighted tBEDc using the intuitive model were: head and neck 8.6 Gy 10 , lung 6.3 Gy 10 , cervix 6.3 Gy 10 and anus 7.8 Gy 10 .Lung cancer trials included SCC and other non-small cell histological types.The percentage of patients with SCC histology in each trial is shown below Table1.Derivation of Biologically Effective Dose contribution to tumour local control by chemotherapy (tBEDc) using phase 3 randomised controlled trials of conventionally fractionated radiotherapy versus chemoradiotherapy in squamous cell carcinoma of the head and neck, lung, cervix and anal cancer using the intuitive method

Table 2 .
Derivation of Biologically Effective Dose contribution to tumour local control by chemotherapy (tBEDc) using phase 3 randomised controlled trials of conventionally fractionated radiotherapy versus chemoradiotherapy in squamous cell carcinoma of the head and neck, cervix and anus using the Poisson model