Radiotherapy and Oncology xxx (2017) xxx–xxx
Contents lists available at ScienceDirect
Radiotherapy and Oncology journal homepage: www.thegreenjournal.com
Original article
Hypofractionated radiation therapy for basal and squamous cell skin cancer: A meta-analysis Nicholas G. Zaorsky a,b,⇑,1, Charles T. Lee a,1, Eddie Zhang a, Scott W. Keith c, Thomas J. Galloway a,⇑ a Department of Radiation Oncology, Fox Chase Cancer Center, Philadelphia; b Department of Radiation Oncology, Penn State Cancer Institute, Hershey; and Pharmacology and Experimental Therapeutics, Division of Biostatistics, Sidney Kimmel Medical College at Thomas Jefferson University, Philadelphia, USA
a r t i c l e
i n f o
Article history: Received 2 April 2017 Received in revised form 21 July 2017 Accepted 7 August 2017 Available online xxxx Keywords: Cosmesis Elderly Meta-analysis Radiation therapy Systematic review Skin neoplasms
c
Department of
a b s t r a c t Purpose: To characterize the cosmetic outcomes and local recurrence (LR) rates of various hypofractionated radiation therapy (RT) regimens for skin basal and squamous cell cancers (BCCs/SCCs). Methods: A PICOS/PRISMA/MOOSE selection protocol was performed to identify 344 articles published between 1985–2016 evaluating patients with T1–2 N0 SCCs/BCCs treated with definitive RT. Biologically equivalent doses with a/b = 3 (BED3s) were calculated. The primary endpoint was posttreatment cosmesis. Mixed effects regression models were used to estimate weighted linear relationships between BED3 and cosmetic outcomes. Results: A total of 21 studies were identified detailing the treatment of 9729 skin BCC/SCC patients, across seven countries, with external beam RT (n = 9255) or brachytherapy (n = 474). Median follow-up was 36 months (range: 12–77). Median dose was 45 Gy/11 fractions (interquartile range: 37.5 Gy/6–55 Gy/18) at 4 Gy/fraction (interquartile range: 2.5–6 Gy); most hypofractionated 18.75 Gy/1. There was a trend to decreased ‘‘good” cosmesis with higher total dose: 3.4% ‘‘good” cosmesis/10 Gy BED3, p = 0.01. Similarly, there was a trend to increased ‘‘fair” cosmesis with higher dose: +3.8% ‘‘fair” cosmesis/10 Gy BED3, p = 0.006. At a BED3 of 100 Gy, the expected rate of ‘‘good” cosmesis is 79% (95% confidence interval: 70%, 88%). Hypofractionated schedules produced similar cosmesis to conventionally fractionated schedules, at the same BED3. Fewer than 8% of patients experienced ‘‘poor” cosmesis, independent of dose or fractionation regimen. Conclusion: Hypofractionated RT has favorable cosmesis for patients with skin BCCs/SCCs. We recommend clinicians consider these commonly-used regimens, which all have BED3 of 100 Gy: 50 Gy/15 fractions, 36.75 Gy/7 fractions, or 35 Gy/5 fractions, as they result in ‘‘good” cosmesis in 80% of patients. Ó 2017 Elsevier B.V. All rights reserved. Radiotherapy and Oncology xxx (2017) xxx–xxx
Non-melanoma skin cancer is the most commonly diagnosed malignancy in the US [1], and its incidence increases with age [2]. Of these cancers, basal cell carcinoma (BCC) makes up 75–80% of diagnoses and squamous cell carcinoma (SCC) makes up the Abbreviations: BCC, basal cell carcinoma; BED, biologically equivalent dose; CI, confidence interval; EBRT, external beam radiation therapy; HDR-BT, high dose rate brachytherapy; IQR, interquartile range; PICOS, Population, Intervention, Control, Outcome, Study Design; PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses; RT, radiation therapy; RTOG, Radiation Therapy Oncology Group; SCC, squamous cell carcinoma. ⇑ Corresponding authors at: Department of Radiation Oncology, Penn State Cancer Institute, 500 University Drive, Hershey, PA 17033, USA (N.G. Zaorsky). Department of Radiation Oncology, Fox Chase Cancer Center, 333 Cottman Avenue, Philadelphia, PA 19111-2497, USA (T.J. Galloway). E-mail addresses:
[email protected] (N.G. Zaorsky), thomas.galloway @fccc.edu (T.J. Galloway). URL: http://www.nicholaszaorsky.com (N.G. Zaorsky). @NicholasZaorsky (N.G. Zaorsky) 1 The authors contributed equally to this work.
majority of the remaining cases [3,4]. Most localized (i.e. T1–2 N0) BCCs and SCCs are destroyed locally (excised, desiccated, frozen); however, a number of factors may preclude surgical extirpation, including patient comorbidities, use of anticoagulants, tumor location in a region that would engender substantial morbidity (e.g. nose, orbit), and subsequent presence of a large uncorrectable surgical defect. Thus, radiation therapy (RT) is typically an alternative for localized tumors [5,6] and has superior outcomes to other approaches (e.g. cryotherapy, topical chemotherapeutics) [7–9]. Within RT, multiple fractionation options exist for small, localized, thin skin tumors, ranging from 6.5 weeks of conventional fractions (64 Gy in 32 fractions) to the extreme hypofractionation of 20 Gy in a single fraction [5,6,10]. For most cancers, including skin BCCs and SCCs, conventional fractionation (i.e. at 2 Gy per day) has long been considered a standard approach that reduces long-term toxicity rates while maintaining a low rate of local
http://dx.doi.org/10.1016/j.radonc.2017.08.011 0167-8140/Ó 2017 Elsevier B.V. All rights reserved.
Please cite this article in press as: Zaorsky NG et al. Hypofractionated radiation therapy for basal and squamous cell skin cancer: A meta-analysis. Radiother Oncol (2017), http://dx.doi.org/10.1016/j.radonc.2017.08.011
2
RT for skin cancer
recurrence [10]. However, given the relative lack of radiosensitive structures in a superficial skin field and regenerative capacity of the skin, hypofractionated approaches are often pursued for superficial skin tumors in a desire to maximize convenience. Current dogma asserts that the trade-off of more convenient treatment is worse cosmesis, however the data supporting this are not strong. Further, prolonged treatment times are inconvenient to elderly patients and are not cost-effective. In modern medicine, it is crucial for primary care physicians and specialists (including oncologists) to work together to provide consistent, accurate information to patients regarding treatment options for indolent skin cancer. The purposes of this work were to characterize the cosmetic outcomes and local recurrence of various RT fractionation regimens, particularly hypofractionated techniques, for SCCs and BCCs of the skin. The results of the work would give general practitioners, dermatologists, radiation oncologists, and surgeons insight into cosmetic results and local control of hypofractionated regimens for elderly patients who cannot tolerate surgery or prolonged fractionation schedules. Methods Evidence acquisition The inclusion criteria for the literature search was defined using the Population, Intervention, Control, Outcome, Study Design (PICOS; Supplementary Table 1) approach [11]. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA; Supplementary Fig. 1) literature selection protocol was used by two authors for article selection. Further, guidelines from the Meta-analysis of Observational Studies in Epidemiology (MOOSE; Supplementary Table 2) were used [12]. Medical literature including clinical trials, clinical studies, evaluation studies, comparative studies, multicenter studies, and case reports published in English at any time up to 2016 was searched in PubMed and MEDLINE. The following terms were used: (‘‘skin” OR ‘‘epidermis” OR ‘‘epidermal” OR ‘‘cutaneous”) AND (‘‘keratinocyte” OR ‘‘squamous” OR basal”) AND (‘‘cancer” OR ‘‘carcinoma” OR ‘‘malignancy”) AND (‘‘radiotherapy” OR ‘‘radiation therapy” OR ‘‘brachytherapy”). Abstracts and other unfinished works were excluded. Minimum follow-up time was set at 12 months because our primary endpoint was to assess long-term cosmesis. Further, skin BCCs and SCC recurrences occur late, with the peak being 4 years out for BCC [13]. Articles that focused on advanced stage or palliative therapy and dosimetry were excluded. Finally, levels of evidence were assigned to each included study based on Centre of Evidence Based Medicine (CEBM) criteria. Outcome measures Cosmesis The primary endpoint was cosmesis. We characterized cosmesis discretely in three categories because most clinicians treating with hypofractionated regimens are interested in preventing ‘‘poor” cosmetic outcome. Most studies reported the presence of moderatesevere toxicities, and those referable to cosmesis were coded as ‘‘poor” for the purposes of our analysis. Certain studies provided a trichotomy of ‘‘good,” ‘‘fair,” and ‘‘poor” toxicity, and these were coded directly into our database. A few studies used the Radiation Therapy Oncology Group (RTOG) grading system (or an analogous system), as provided in Supplementary Table 3. For such cases, cosmesis was coded as ‘‘good” (i.e. Grade 0–1 toxicity), ‘‘poor” (i.e. Grade 3–5), or ‘‘fair” (i.e. the remaining cases, Grade 2). This coding system was devised by the authors to most closely approximate the trichotomy of toxicities. Cosmesis grades were marked for individual fractionation regimens of each study at the latest time of follow-up available. Cosmesis was reviewed and discussed by three of the authors to maintain reporting accuracy.
Local recurrence (LR) LR was defined based on the guidelines at each treating institution. In practice, a biopsy is not always performed. We chose to analyze LR because this would be a major outcome of interest for patients being treated with RT, although LR is relatively rare. Other outcomes (e.g. lymph node metastases, distant metastases) are uncommon in BCCs and superficial SCCs, unless patients are immunocompromised and/or recurrent. Thus, cancer specific mortality is also uncommon. Moreover, patients receiving RT for skin cancer are generally elderly and are much more likely to die of other causes (e.g. secondary cancers; non-cancer causes) [14]. Statistical analysis Calculation of the biologically equivalent dose (BED) is described in Supplementary Text 1 [15–19]. For long-term toxicity/cosmesis and dose–response analysis, weighted mixed effects regression models were used to estimate linear relationships between BED3s of each study and the observed percentages of patients experiencing cosmetic outcomes. Meta-analysis models with mixed effects are sometimes preferable to those with fixed effects in the context of patient care decision making [20,21]. In general, if the estimates are consistent between trials then a fixed effects model might be appropriate. Our chosen approach, which includes random and fixed effects, tends to provide more accurate variances for meta-estimates when the heterogeneity among studies is high and somewhat inflated variance estimates when the heterogeneity is low. As shown in the tabled data and figures, we saw clear evidence of heterogeneity for certain cosmetic results and less so for others. Having meta-analyzed a considerable number of outcome variables here, we chose to assume heterogeneity and apply one consistent robust approach which may have been somewhat conservative for some outcomes where the data were highly consistent. This choice likely did not have strong or adverse impacts on the conclusions we have drawn. The width of the confidence intervals generated by the mixed modeling approach accounts for the degree of heterogeneity we see in the data and appropriately represents our level of uncertainly in parameter estimates. If a study had multiple treatment groups with different fractionation regimens, different BED3s were calculated for each group and included in the regression models. For reference, the following fractionation schedules all approximately equal a BED3 of 105 Gy: 64 Gy/32 fractions, 55 Gy/20 fractions, 50 Gy/15 fractions [22]. Analyses were performed in Stata 12.0 (College Station, Texas, USA). Percentages of ‘‘good”, ‘‘fair”, and ‘‘poor” cosmetic outcomes weighted by number of patients were set as dependent variables and compared to BED3 which was the independent variable. A pvalue of T2 (48 patients total) [13,26,33]. No particular study focused on very elderly or immunocompromised patients. The mean follow-up was 36 months (12–77 months) and mean age was 73 years (62–84 years old). The total dose ranged from 18.75 to 60 Gy for EBRT and 36 to 77.5 Gy for HDR-BT. The lowest number of fractions for EBRT was 1, at 18.75–22.5 Gy, and the highest number of fractions was 40, at 1.5 Gy per fraction. The lowest number of fractions for HDR-BT was 6, at 7 Gy per fraction, and the highest number of fractions was 43, at 1.8 Gy per fraction. The mean BED3 was 125 Gy (range, 60–579 Gy) for EBRT. A BED3 outlier of 579 Gy was seen in the study by Zagrodnik et al. For HDR-BT, the mean BED3 was 113 Gy (72–140 Gy). Across all treatment groups of the included EBRT and HDR-BT studies, the median dose was 45 Gy. Cosmesis There were 2,659 patients whose long-term cosmesis was evaluated. The vast majority of patients had a ‘‘good” cosmetic outcome for any fractionation regimen included in the metaanalysis: the median % of patients with ‘‘good” cosmesis was 92% (IQR: 75%–100%). The median % of patients with ‘‘fair” cosmesis was 2% (IQR 0%–15%). There were 675 patients treated to 30.6 Gy in 10.2 Gy/fraction in a single study; of these, 50% developed ‘‘fair” cosmesis [23], and this value skewed the low incidence of ‘‘fair” cosmetic results of other patients in the current meta-analysis. The median % of patients with ‘‘poor” cosmesis was 2% (IQR 0%– 5%). There was one outlier of a patient receiving 40 Gy in 3.5 Gy fractions who experienced ‘‘poor” cosmesis [30]; almost all other patients had ‘‘good” or fair” cosmesis. At a BED3 of 100 Gy, the expected rate of ‘‘good” cosmesis was 79% (95% confidence interval: 70%, 88%). The most common late toxicities noted were hyperpigmentation and telangiectasias; ulceration and necrosis were uncommon and ranged from 0% to 16% of patients treated in studies which reported these complications [8,13,23,26,27,30,33–35]. Only one study specifically reported the incidence of osteonecrosis which was T2 tumors. Moreover, patients were treated across the world (including Europe, North America, South America, and Australia). Patients were most likely to be treated with EBRT (Table 1), as is common in many cancers that do not have a dedicated HDR-BT unit (Table 2). The fractionation schedules used in these studies were mostly hypofractionated: the median Gy per fraction was 4 (IQR 2.5–6), and this is typical in elderly patients. The studies using EBRT were more comprehensive in reporting outcomes and toxicities, while only one HDR-BT study reported outcomes at 5 years. There was no apparent benefit in outcomes or toxicities of HDR-BT vs. EBRT,
Please cite this article in press as: Zaorsky NG et al. Hypofractionated radiation therapy for basal and squamous cell skin cancer: A meta-analysis. Radiother Oncol (2017), http://dx.doi.org/10.1016/j.radonc.2017.08.011
5
N.G. Zaorsky et al. / Radiotherapy and Oncology xxx (2017) xxx–xxx
Good cosmesis
A
Fair cosmesis
B
24 Gy / 4 fractions
C
40 Gy / 10 fractions
36.75 Gy / 7 fractions 64 Gy / 32 fractions 55 Gy / 20 fractions 50 Gy / 15 fractions
Proportion of patients
Proportion of patients
p = 0.013
35 Gy / 5 fractions
30.6 Gy / 3 fractions
p = 0.006
20 Gy / 1 fraction
D
Poor cosmesis
Cosmesis type per BED3
1
Proportion of patients
Proportion of patients
0.9
p = 0.195
0.8 0.7 0.6 0.5 0.4 0.3
%Poor
0.2
%Fair
0.1
%Good 147
135
147
130
133
128
119
120
110
109
108
107
105
105
101
98
90
87
83
83
73
72
68
60
0
BED3
Fig. 1. Cosmesis as a function of BED. (A–C) The rates of ‘‘good,” fair,” and ‘‘poor” cosmesis are plotted vs. BED3 using mixed effects regression models (upper left, upper right, lower left plots), with the 95% confidence interval (shaded region) representing the area of the anticipated average cosmetic outcome at a particular dose, and a p-value < 0.05 representing an association between cosmetic outcome and BED3. To help compare fractionation regimens, the BED3s of several fractionation regimens are labeled in the upper left plot. Hypofractionated schedules produced similar cosmesis to conventionally fractionated schedules, at the same BED3; further, no regimens to produce ‘‘poor” toxicity were identified. (A) There was a trend to decreased ‘‘good” cosmesis with higher total dose: 3.4% ‘‘good” cosmesis/10 Gy BED3, p = 0.01. (B) Similarly, there was a trend to increased ‘‘fair” cosmesis with higher dose: +3.8% ‘‘fair” cosmesis/10 Gy BED3, p = 0.006. (C) Fewer than 8% of patients experienced ‘‘poor” cosmesis, independent of dose or fractionation regimen. (D) The rates of cosmesis are also co-plotted as a function of BED3 in the lower right plot. Taken together with the low LR rates, the results suggest that hypofractionated regimens are safe and efficacious for elderly patients with BCCs and SCCs of the skin.
and this finding should reinforce the use of the more efficient and less costly EBRT instead of inappropriate use of HDR-BT which tends to have higher reimbursement [43]. Since the LR rates are relatively similar across fractionation regimens, cosmesis is an important endpoint of measure, particularly for single-fraction treatments [16]. We found no apparent correlation between dose or fractionation with cosmesis. These results are in line with many of the studies included in this meta-analysis that reported ‘‘good” cosmesis in greater than 75% of patients [13,25,29–31,33,37]. Van Hezewijk et al. demonstrated that a dose of 44 Gy delivered by EBRT in 4 Gy/fraction intervals resulted in 67% good cosmesis and 33% fair [36]. Lovett et al. demonstrated that doses between 40 and 50 Gy delivered in > 4 Gy/fraction resulted in an outstanding 95% good cosmesis although there were only 19 patients in that treatment group [30]. With BT, doses of 40 Gy and 50 Gy delivered in 5 Gy/fraction intervals resulted in 97.9% and 100% good cosmesis, respectively [38]. Studies using BT similarly had excellent LR and cosmesis [40], though this was typically delivered in 2 Gy fractions up to 60 Gy. However, several EBRT studies did report that increasing field size (5 cm2) and increasing dose per fraction correlated with incidence of radionecrosis. In the study by Abbatucci et al., 2.7% of patients developed skin necrosis of the forehead, nose, and chin following local trauma, most of whom with field size 5 cm2 [23]. Schulte et al., also reported the effect of increasing field size
effect incidence of skin necrosis (6.4%) in their study [33]. In the study by Chan et al. which investigated single-fraction regimens, 16% of patients developed radionecrosis which was significantly higher in the group receiving 22.5 Gy fractions [27]. Additionally, Silva et al., reported that fraction sizes 3 cm where a dose of 26 Gy was able to achieve good control after being delivered in a single fraction. These findings are consistent with the plateau of the sigmoidal dose–response curve seen in other disease sites
Please cite this article in press as: Zaorsky NG et al. Hypofractionated radiation therapy for basal and squamous cell skin cancer: A meta-analysis. Radiother Oncol (2017), http://dx.doi.org/10.1016/j.radonc.2017.08.011
6
RT for skin cancer
Fig. 2. An illustration of recommended RT fractionation options for indolent skin cancers. UPPER PANEL: The two principal fractionation options are conventional fractionation (2 Gy daily fractions, up to a total dose of 64 Gy) and hypofractionation (>2 Gy fractions). Based on the current meta-analysis, we recommend clinicians consider these commonly-used regimens: 50 Gy in 15 fractions, 36.75 Gy/7 fractions, 35 Gy/5 fractions. The fractionation options listed here have similar BED3s and result in LR rates of 4 cm. This was reciprocated by Lovett et al., who additionally reported improved ‘‘good” cosmesis with superficial X-rays compared to electrons, mixed beams, and megavoltage photons [30,40]. In contrast, van Hezewijk et al., did not find any correlation between cosmesis and size of the irradiated area or location [36]. Lastly, the median follow-up time for any fraction arm in this study was 36 months (IQR 30–43), we cannot comment if the same cosmetic results would be expected at a much longer time (e.g. 5– 10 years). Late toxicities from RT tend to increase over time in any disease site (e.g. breast [48] and prostate [22] cancer patients treated with >2 Gy/fraction), and the same would be expected for RT for skin cancer after 10 years of follow-up. Nevertheless, since outcomes and toxicities of more conventionally fractionated regimens are similar to those of hypofractionated regimens with respect to the median follow-up time, we recommend clinicians to use hypofractionated regimens for patients with localized BCC/SCC and apply clinical judgement to individual cases.
This work has limitations. First, we based toxicity evaluations based on clinician evaluations of patients, using a simple scoring system. In general, most acute reactions (e.g. erythema, edema, itch) resolve within 3 months; late reactions (e.g. fibrosis, ulceration, necrosis) start to develop after 3 months, with a rising incidence over time. Thus, cosmesis depends on the median follow-up time of each study. Intra- and inter-observer differences may be present. In some studies, assistants were able to help grade toxicity [13,36]. Additionally, observer bias and publication bias may have played a role in the favorable outcomes of the hypofractionated regimens. However, given the paucity of published research on hypofractionation for skin cancer, a quantitative assessment of bias was not done. We did not blindly evaluate photographs of the skin after treatment; we did not evaluate ethnic subpopulations or patientreported outcomes, which have been shown to be important in prostate and breast cancer [49–51]. We do not have comorbidity data of these patients; those with peripheral arterial disease and diabetes would be expected to have worse toxicity [52]. We strongly encourage future investigators to provide detailed reports of toxicities among their patients, and for basic scientists to identify biomarkers of RT toxicity (e.g. TGF-b, IL-6 in lung cancer) [53]. Further, the BED equation may not adequately characterize extremely hypofractionated regimens (e.g. >8 Gy/fraction) [54], cellular death due to different modes surrounding mitotic catastrophe (e.g. necroptosis) [55], effects on stroma/vasculature (e.g. pericytes) [56], or molecular pathways behind recurrence (e.g. vasculogenesis) [57]. The BED equation also does not take into account target volume, treatment field sizes, presence of hot spots, or prescription method (e.g. to a volume vs. an isodose line) [58]. We have a limited follow-up time of
