Local Control following Stereotactic Body Radiation Therapy for Stage I Non-Small Cell Lung Cancer

Abstract

Purpose:

Numerous dose and fractionation schedules have been used to treat medically inoperable stage I Non-small cell lung cancer (NSCLC) with stereotactic body radiation therapy (SBRT) or stereotactic ablative radiotherapy (SABR). We evaluated published experiences with SBRT to determine local control (LC) rates as a function of SBRT dose.

Methods:

One hundred sixty published articles reporting LC rates following SBRT for stage I NSCLC were identified. Quality of the series was assessed by evaluating the number of patients in the study, homogeneity of the dose regimen, length of follow-up time, and reporting of LC. Clinical data including 1, 2, 3, and 5 year tumor control probabilities for T1, T2, and combined T1 and T2 stage as a function of the biological effective dose were fitted to the linear quadratic (LQ), Universal survival curve (USC), and regrowth models.

Results:

Forty-six studies met inclusion criteria. As measured by the goodness of fit χ²/ndf, with ndf as the number of degrees of freedom, none of the models were ideal fits for the data. Of the three models, the regrowth model provides the best fit to the clinical data. For the regrowth model, the fitting yielded an α/β ratio of approximately 25 Gy for T1 tumors, 19 Gy for T2 tumors, and 21 Gy for T1 and T2 combined. In order to achieve the maximal LC rate, the predicted physical dose schemes when prescribed at the periphery of the planning target volume (PTV) are 43 +/− 1 Gy in 3 fractions, 47 +/− 1 Gy in 4 fractions, and 50 +/− 1 Gy in 5 fractions for combined T1 and T2 tumors.

Conclusion:

Early stage NSCLC is radioresponsive when treated with SBRT/SABR. A steep dose-response relationship exists with high rates of durable LC when physical doses of 43–50 Gy are delivered in 3–5 fractions.

Summary

Stereotactic body radiation therapy (SBRT) or stereotactic ablative radiotherapy (SABR) is an effective treatment for medically inoperable early stage NSCLC. The authors quantitatively evaluated published experience with thoracic SBRT for early NSCLC and modeled local control rates as a function of SBRT dose. Early stage NSCLC is radioresponsive when treated with SBRT/SABR. A steep dose-response relationship exists with high rates of durable LC when physical doses of 43–50 Gy are delivered in 3–5 fractions.

1. CLINICAL SIGNIFICANCE

Non-small cell lung cancer (NSCLC) is the 2^nd most common cancer overall (228,190 cases per year in the US), and the leading cause of cancer-related death (159,480 deaths) in both men and women [77]. Stage I disease represents approximately a quarter of the patients diagnosed with NSCLC and accounts for the most curable cohort of the population (SEER 18 2004–2010; AJCC Cancer Staging Manual, 7^th edition). The standard treatment for medically operable stage I NSCLC has historically been an anatomical resection with lobectomy as well as hilar and mediastinal lymph node dissection [1,16,22,63]. However, the majority of patients with NSCLC have a history of chronic tobacco use and a median age of diagnosis of 65–74 years [SEER 18 2004–2010; AJCC Cancer Staging Manual, 7^th edition], and often have cardiopulmonary comorbidities (e.g. cardiac and pulmonary) that make them at high-risk for resection. Some patients are deemed to be medically inoperable [15,82]. The increasing use of screening for lung cancer, based on the National Lung Screening Trial may increase the number of patients with early stage NSCLC appropriate for non-surgical treatments[57].

The historical standard therapy for patients with unresectable early stage NSCLC was conventionally fractionated radiation therapy; e.g. 2–3 Gy per fraction to a dose of ≈ 54–60 Gy. However, the reported long-term local control (≈ 30–70%) and overall survival (≈ 15–30%) rates with this approach are suboptimal[7,28,67]. Advances in imaging, radiation treatment planning, and delivery (e.g. with image-guidance and/or motion management) enable the delivery of “ablative doses” of radiation (e.g. 18–20 Gy times three fractions) to very small targets (often termed SBRT/SABR) that appears to yield better outcomes for early stage NSCLC[50,64,89,90,92,99].

In most of the reports using this approach, typical patient selection criteria include co-morbid conditions that preclude a safe oncologic resection, such as poor pulmonary function tests (FEV1 < 1.2 liters, and DLCO < 50%) [30,49,90]. Most patients are staged with a whole body Computed tomography (CT),and/or Positron emission tomography-computed tomography (PET-CT) scan. Patients suspected of having lymph node involvement (interlobar, hilar, or mediastinal) are not candidates for SBRT. More often than not, pathological staging of the mediastinal nodes is not done due to the risk of invasive procedures in this patient population[17,72]. However, tissue diagnosis of the NSCLC subtype using CT-guided or endobronchial ultrasound-guided (EBUS-guided) needle biopsy is recommended [17,78]. Of note, EBUS directed biopsy is usually only appropriate for centrally located primary tumors. Ideally, primary tumor size is restricted to ≤ 5 cm (e.g. T1a – T2a), and thus the optimal patients for SBRT include clinically staged IA and IB medically inoperable NSCLC. Of note, recent ASTRO guidelines conditionally recommend SBRT for tumors larger than five cm that are not suitable for surgical resection with appropriate counseling of patients regarding higher risk of locoregional and distant failures[100]. Nevertheless, the majority of patients in the available literature were treated for lesions ≤ 5 cm and in non-central locations as tumors in central locations have less favorable outcomes with SBRT (see section 7, “Special Situations).

2. ENDPOINTS

The primary endpoint reported in the literature was local tumor control (LC) at the primary site of SBRT. When reported, the actuarial rates of local control, defined as no local progression at the primary tumor site as assessed by CT or PET-CT imaging, at 1, 2, 3, and 5 years were recorded; however the majority of the studies only reported outcomes up to 3 years. Overall survival data is often reported in the literature and was collected in this review. However, this data was not used in the final analysis and modeling due to the lack of consistent reporting of this endpoint in the reviewed literature. Distant failure was not recorded in our review due to minimal observed correlations to models assessing local tumor control probabilities (TCP).

Comparisons with surgical series are challenging since most surgical series define local failure to include failure within the lobe of the lung (in cases of sublobar resection) as well as locoregional or regional failures (failures within hilar and mediastinal lymph nodes). These metrics are not routinely reported in the SBRT literature. Further, patients undergoing SBRT for early stage NSCLC generally have greater competing risks for death from causes other than their lung cancer compared to patients undergoing surgery, as the latter have fewer competing comorbidities[46,107]. Thus, it is possible that reported actuarial local control rates at 1–3 years after SBRT over-estimate the true LC. Since patients dying of intercurrent deaths (death not due to lung cancer during the follow-up period for lung cancer after treatment) are censored, perhaps leaving an “enriched healthier” subset of evaluable patients while those that died of intercurrent illnesses may have had occult local progression prior to death. Indeed, most matched-pair comparisons between SBRT versus surgery report an inferior overall survival with SBRT despite comparable cause-specific survival[66,76].

In most SBRT series, LC was assessed using CT and PET-CT based imaging and applying the Response Evaluation Criteria in Solid Tumors (RECIST) and/or changes in PET-FDG activity. Some studies reported pathological confirmation of tumor recurrence in a subset of the patients. Nevertheless, given that the majority of the recurrences were assessed radiographically, there is certainly some uncertainty in the reported LC rates. SBRT can cause scarring/inflammatory changes that result in tissue distortion making radiographic interpretations difficult[36,37,42]. These changes can mask, and thus delay, the diagnosis of tumor recurrence. Similarly, local inflammation soon (within 1 year) after SBRT often causes an increase in FDG uptake which can make response assessment unreliable and can lead to false positives[79,108]. Thus, data from studies with longer follow-up are likely more accurate in their assessment of LC.

3. CHALLENGES DEFINING AND SEGMENTING ANATOMIC VOLUMES

Respiration-induced tumor motion is a challenge for target definition. Older series often used breath hold (both deep-exhalation and deep-inhalation) CT’s or used fluoroscopy as a surrogate to define an extreme borders of the target’s motion envelope. Most of the modern studies use 4-dimensional CT (4D-CT) scans to define the target volume, where the images acquired in the same respiratory phase or amplitude are grouped together to reconstruct multiple 3DCTs. The amplitude-based 4DCT reconstruction is preferred because it generates less image motion artifacts. A separate free breathing scan, with or without contrast, is often also obtained. Intravenous (IV) contrast may be useful in settings where the target lesion abuts a large vessel and/or the mediastinal structures[23,34,105].

Some studies have used methods to control the amplitude of tidal volume and thus tumor excursion by simulating the patient with a 4D-CT or fluoroscopy while also using abdominal compression devices[8]. The degree of abdominal compression can be determined by using fluoroscopy and/or imaging implanted radio-opaque fiducial markers within or near the tumor such that the excursion of the marker and tumor is within an acceptable range[29,54]. It is of note that the use of implanted fiducial markers is optional for all respiratory management tools during CT simulation. These motion management strategies at the time of CT simulation are also used to characterize respiratory motion of organs at risk (OARs). Fusion of a PET and a CT scan can help define the tumor borders and is especially helpful when the tumor is adjacent to lung atelectasis, the mediastinum, diaphragm, stomach, liver, etc.[40,68]. Ideally, the PET-CT fusion should be performed in the same respiratory phase (amplitude).

During treatment planning, the contrast enhanced CT as well as the 4D-CT can be used to segment an internal target volume (ITV) using a Boolean operation to account for the motion envelope [41,105]. Various methods including using only the end-expiratory and end-inspiratory phases, or segmenting the tumor in all respiratory phases and using a Boolean operations to combine the contours, as well as generating maximal intensity projection (MIP) images and segmenting the target, have been used to defined the ITV. In addition, based on the ICRU 62 definition, CTV=GTV with no margin[20,53]. Further, due to the high doses per fraction, the doses to the ‘non-target’ tissues immediately adjacent to the PTV receive relatively-high, and likely ‘therapeutic dose’ for potential microscopic disease. Thus, the favorable outcomes reported without a formal CTV expansion should not be taken as proof that there is no microscopic spread. Typical expansions from ITV to PTV are in the range of 3–8 mm in the axial dimensions, and 5–10 mm in the cranio-caudal dimensions with or without respiratory gating or tracking. If respiratory gating is utilized, then the ITV is defined based on the phases selected for treatment. Commonly, near end-expiratory phases (gating phase 30%–70%) are used due to maximal tumor stability and minimal tumor motion in these phases. Alternatively, some studies choose near-inspiratory phases (gating phase 90%–10% or inspiration breath-hold) as the total lung volume is larger, and the percent of lung irradiated to any given dose is likely lowest, in these phases.

Various techniques to control, monitor and/or limit respiratory motion can be used including passive breath-hold with visual or audio feedback to the patient, active breath control (where air movement is restricted by a device), tracking of external (e.g. surface markers) or internal (e.g. implanted markers, diaphragm, or the tumor itself), and abdominal compression[58,80]. Each of these approaches has its own benefits and limitations. Typically, gating and tracking to improve normal tissue sparing are most useful for tumors with relatively-large respiratory excursions[93,95], but these approaches typically increase treatment times. Tumors in the middle and lower lobes are generally more mobile than those in the upper lobes. Caveats in characterizing tumor and organ motion using a 4D-CT include artifacts induced due to patients’ irregular breathing patterns and reproducibility of the breathing patterns at CT simulation compared to treatment days.

Adjusting the window and level on the CT scan will impact target definition[45]. Typically, a ‘lung window’ is best to define a parenchymal tumor as irregular spiculations can be better appreciated. When a tumor abuts another organ composed mostly of soft-tissue (heart, mediastinum, diaphragm/liver, and chest wall), assessing the boundaries of the target at the interface is best done using a mediastinal or soft tissue windows[83].

4. REVIEW OF OUTCOMES DATA

A keyword search for ‘SBRT and lung’ and ‘stereotactic ablative radiotherapy (SABR) and lung’ using PubMed identified 160 studies reporting clinical outcomes from thoracic SBRT for early stage NSCLC published through May 2014. These studies were systematically reviewed. Articles relating to the treatment of oligometastatic disease to the lung were specifically excluded. Each publication was assessed to determine whether the data was collected prospectively or retrospectively, the number of patients, homogeneity of dose prescriptions, the length of follow-up, and whether local control was reported. Studies with fewer than 10 patients, or tumor stage higher than T2 were excluded. Three of the included studies that met the above publication selection criteria had doses in the 3–4 Gy per fraction range (211 patients). Due to the paucity of data in the intermediate dose per fraction range, these studies were included in order to improve model fitting in the shoulder region. The rest of the included studies had dose per fraction ≥ 6 Gy per fraction (3268 patients). Based on this metric for “quality”, 46 studies[3,4,6,9–14,18,19,21,24,26,27,31–33,38,39,44,47,48,52,55,56,59–61,69–71,73,75,81,84,87,88,90,94,97,98,101,103,106,109] were identified and included for data collection and modeling of outcomes. Of note a retrospective series was included if it met all the other criteria for “quality” listed above, as long as it had included ≥ 10 patients (see supplementary material for all included studies and input from each study).

Extracting/comparing dose information:

The reported dose was prescribed to the isocenter in 17 (36%) studies, or to an isodose surface that covered a certain percent of the PTV in 30 (64%) studies. To facilitate the pooling of data from multiple studies for analysis, in the latter studies, reported doses were converted to presumed isocenter doses by the formula: Reported Dose/Reported percent covering isodose. In the 10 cases where percent isodose was not given, authors were contacted directly for this information (3 cases), or 80% isodose coverage was assumed (7 cases). Thus, from each study, we extracted an estimate of the isocenter dose, as well as the dose per fraction, the number of fractions, and an estimate of the total elapsed days of treatment.

In all analyzed studies, treatment plans were based on multiple non-coplanar or coplanar fields delivered with conformal fields or with dynamic conformal arcs or volumetric modulated arc therapy, or with multi-field intensity modulated radiation therapy.

Notably, a variety of tissue inhomogeneity algorithms were used, which confounds the estimation of the dose at the isocenter. Among the various sources of uncertainty, these algorithms may be inaccurate in estimating the PTV dose and distort the dose distribution in a patient-specific manner. Although most protocols recommend prescriptions to the PTV margin[5,91], the dose calculation introduces more uncertainty in this region than at the isocenter[35,48,102,104] or at a suitable calculation point within the tumor for IMRT. In particular, the dosimetric uncertainty at lung-tissue interfaces[35], at the periphery of a PTV[48], and at shallow depths near the skin or surface of a lung tumor[102,104] can be large, up to 120%[102,104], with the less accurate dose calculation methods. In contrast, even for small fields in a heterogeneous environment, dose measurements and Monte Carlo calculations can match within 3%[104] at the isocenter. To resolve the dilemma, the HyTEC dose-response model was constructed using isocenter dose, and for clinical conclusions based on the analysis (Section 8) the results were converted to a PTV margin equivalent using a generic 80% isodose line; the interested reader can convert the results using an applicable isodose line for individual situations.

Most of the studies used linear accelerator-based radiosurgical delivery systems including Cyberknife (Accuray, Sunnyvale, CA), and other more traditional linear accelerator based delivery systems (Truebeam, Triology, Novalis, Artiste, etc.). A few studies utilized either helical tomotherapy or proton beam therapy. Several conventional linac studies noted that only a limited number of non-coplanar beams could be used due to issues relating to patient or couch collision.

Immobilization devices included stereotactic body frames, customized cradles, and evacuated vacuum cushions; a few studies used no special immobilization. Respiratory motion management strategies varied and included free-breathing techniques with a 4D CT from simulation defining an ITV with or without tracking of implanted metallic fiducials, abdominal compression with fluoroscopic assessment of diaphragmatic motion as surrogate for tumor motion, and no (or unreported) motion management. There were too many permutations related to patient set-up, immobilization, respiratory motion management strategies, and target definition to conduct a meaningful analysis relating outcomes to any of these variables. Thus, data were pooled from studies independent of how these issues were addressed, which leads to further uncertainty. Similarly, there was not enough specificity in the available reports to consider factors such as histology, absolute tumor volume (although T1 vs. T2 tumors were evaluated as variables for the regrowth model), total treatment duration, minimum PTV dose, dose heterogeneity within the tumor, fraction number, tumor location, and irradiation technique (e.g. margin, immobilization, set-up considerations, delivery method, etc.). Additional factors that were not addressed that may affect outcome include molecular mutation status, invasive versus in-situ disease, and staging work-up requirements (invasive or non-invasive mediastinal nodal staging).

Extracting/comparing outcome information:

From each study, the reported local control rates at 1, 2, 3, and 5 years were extracted (as able). For the studies where results were presented in Kaplan-Meier/actuarial figures, the corresponding data were extracted from the figures.

Due to the heterogeneity of data reporting as well as insufficient reporting of outcomes other than LC (i.e. overall survival), outcome modeling was limited to the effect of tumor size (T1 vs. T2 disease for regrowth model), total dose prescribed to the isocenter, and number of fractions on LC (Table 2).

Table 2:

Required physical doses (Gy) at isocenter and covering PTV with the 80% isodose line to reach the maximum TCP, calculated from the three models with the parameters determined in Section 6.

Isocenter Dose (Gy)		3 fractions	4 fractions	5 fractions
Regrowth	T1	52±1	57±1	60±1
	T2	56±1	62±1	66±1
	T1+T2	54±1	59±1	63±1
LQ	T1+T2	55±1	59±1	63±1
USC	T1+T2	55±1	59±1	63±1
PTV Dose (Gy)		3 fractions	4 fractions	5 fractions
Regrowth	T1	42±1	46±1	48±1
	T2	45±1	50±1	53±1
	T1+T2	43±1	47±1	50±1
LQ	T1+ T2	44±1	47±1	50±1
USC	T1+T2	44±1	47±1	50±1

Pooled crude data:

Figure 1 presents TCP of 3-, 4-, and 5-fraction SBRT for combined T1- and T2-stage NSCLC as a function of estimated physical isocenter dose.

Fig.1.

Tumor control probability of 3- (top), 4- (middle) and 5- (bottom) fraction SBRT for T1- and T2-stage NSCLC as a function of physical dose at isocenter. A large error bar for a data point represents a small number of patients associated with that data point.

5. FACTORS AFFECTING OUTCOMES

Using data from the reviewed literature, we studied the dependence of LC on total isocenter dose and number of fractions. We converted dose to biological effective doses using the three models described below. The available large pool of clinical data shows a steep dose-response for local tumor control for SBRT for early stage lung cancer (Figure 1). An additional factor that affects LC is tumor size based on T stage. To achieve a maximum TCP, T2 lesions consistently required a higher physical dose at the isocenter than T1 lesions (approximately 1.3 Gy per fraction higher based on the regrowth model) (Table 2).

6. MATHEMATICAL/BIOLOGICAL MODELS

Three biophysical models were considered and fit to the collected clinical data: the linear-quadratic (LQ) model, the universal survival curve (USC) model[65], and the regrowth model[85]. These were chosen due either to common usage (LQ and USC), or having the best fit to the data (regrowth). For the LQ model, BED is expressed: $BE{D}^{LQ}=D(1+\frac{d}{\alpha /\beta })$, where α and β characterize the intrinsic radiosensitivity of cells and D and d are the total and fractional doses, respectively. For the USC model, $BE{D}^{USC}=\{\begin{array}{ll}D\left(1+\frac{d}{\alpha /\beta }\right),\hfill & d<d_T\hfill \\ \frac{1}{\alpha D_0}\left(D-n{D}_q\right),\hfill & d\ge d_T\hfill \end{array}$ where −1/D₀ and D_q are the slope and x-intercept of the logarithm survival curve, n is the number of fractions, and d_T is the transition dose where the LQ model smoothly transitions to the terminal asymptote of the multitarget model. For these two models, TCP can be expressed as: $=e^{-K_0*e^{-\alpha *BED}}$, where K₀. is the number of clonogenic cells at the beginning of radiotherapy.

The regrowth model links the population averaged TCP and biologically effective dose (BED) as $\text{TCP}=1-\frac{1}{\sqrt{2\pi }}{\int }_{-\infty }^t{e}^{-\frac{x^2}{2}}dx$, where $t=\frac{K-K_{cr}}{{\sigma }_k}$, $K=K_0{e}^{-[\alpha BED-{(\frac{ln2}{T_d}(\tau -T))}^{\delta }]}$, and $BED=D\left(1+\frac{d}{\alpha /\beta }\right)-\frac{ln2}{T_d}\frac{T}{\alpha }$; α and β are radiobiological parameters[86]; τ is follow-up time starting from the beginning of radiation treatment; T is the elapsed treatment time for the SBRT treatment course; T_d is the potential tumor doubling time; δ is a fitting parameter characterizing the speed of tumor cell regrowth after SBRT; D and d are the total and fractional doses, respectively; K₀ is the number of clonogenic cells at the beginning of radiation; K_cr is the critical clonogenic cell number that defines control of an individual tumor;σ_k is the Gaussian width of the distribution of tumor cell numbers. The independent model parameters (α, α/β, T_d, K_cr/K₀, σ_k/K₀, and δ) were determined from a simultaneous fit to the 1-, 2-, 3-, and 5-year actuarial or Kaplan-Meier TCP data. Clinical data including 1, 2, 3, and 5 year TCP for T1, T2, and combined T1 and T2 stage as a function of the biological effective dose to isocenter were fitted to each of the above models allowing all model parameters to freely float to achieve the best of fit. The studies with fractional doses of greater than 3 Gy were considered [3,4,6,9–14,18,19,21,24,26,27,31–33,38,39,44,47,48,52,55,56,59–61,69–71,73,75,81,84,87,88,90,94,97,98,101,103,106,109]. The TCP data were separated for Stage T1 and T2 tumors if data were available; otherwise analysis was performed for mixed stages. The elapsed treatment time of 7/5 times the number of fractions was used if not reported. The least chi-square (χ²) method was used to fit the data with a single set of parameters for all data and two sets for T1 and T2 separately. The goodness of fit was measured by χ²/ndf, where ndf is the number of degrees of freedom, defined as the total number of data points minus the number of free parameters in the fit. Table 1 presents the goodness fit and the model parameters determined from the data fitting. Notice that the regrowth model leads to a lower χ²/ndf value, compared to the LQ and USC models. The fitting with the regrowth model extracted the model parameters separately for T1 and T2. This is because the regrowth model considers the tumor regrowth after the treatment, allowing fitting of all the TCP data collected at different follow-up times. In contrast, the simpler versions of the LQ and USC models that are more often used clinically do not account for post-treatment regrowth and can only be used to fit the TCP data for a given follow-up time, which could not lead to a convergent fitting for T1 and T2 separately.

Table 1:

The goodness of fit (χ2/ndf) and model parameters determined from simultaneous fits to 1−, 2−, 3−, and 5-year TCP data for stage T1 and T2 lung tumors using the regrowth model and from the fits to the 3-year TCP data for combined stages T1 and T2 tumors using the LQ and USC models. The α/β value in the USC model equals to 34.1, which was calculated by $\alpha /\beta =\frac{4\alpha D_0{D}_q}{{\left(1-\alpha D_0\right)}^2}$.

Parameters		χ2/ndf	α (Gy-1)	α/β (Gy)	Various parameters
Regrowth	T1	3.8	0.129±0.004	24.8±1.9	Td=47.1±16.2 days,
					δ=0.267±0.041,
					Kcr/K0=0.010±0.002
					σK/K0=0.005±0.001
	T2	3.8	0.110±0.004	19.3±2.3	Td=95.1±31.0 days
					δ=0.278±0.035
					Kcr/K0=0.012±0.001
					σK/K0=0.007±0.001
	T1+T2	3.8	0.123±0.007	20.7±1.0	Td=63.8±5.8 days
					δ=0.253±0.025
					Kcr/K0=0.008±0.003
					σK/K0=0.004±0.002
LQ	T1+T2	7.1	0.163±0.010	32.5±3.5	K0=(1.07±0.07)×104
USC	T1+T2	7.2	0.163±0.005		D0=1.7±0.1 Gy
					Dq=16.1±1.2 Gy
					K0=(1.06±0.07)×104

Figure 2 presents examples of fitting the TCP data with the regrowth, LQ, and USC models. The fittings of combined T1 and T2 data yield large α/β values (> 20 Gy) based on the regrowth and LQ models (Table 1). The results indicate that TCP has a steep dose response, reaching the maximum TCP at BED ≥ 90 and 110 for T1 and T2 tumors, respectively. As indicated by the value of χ²/ndf in Figure 2, the regrowth model yields slightly better fitting than the LQ and USC models (4.9 vs. 7.1 and 7.2) to the TCP data. Details on the methods and results for the TCP modeling have been reported separately[43].

Fig.2.

Fitting tumor control probability data of SBRT for T1- and T2-stage NSCLC with the regrowth, USC, and LQ models: (a) fitting 1-, 2-, 3-, and 5-year TCP data simultaneously using the regrowth model; fitting 3-year TCP data with the (b) regrowth model, (c) - LQ model, and (d) USC model.

7. SPECIAL SITUATIONS

The summarized data and model-based results are derived entirely from patients with T1 and T2 medically inoperable early stage NSCLC treated with 3–5 fractions of SBRT, without prior radiation. The degree to which these data are applicable to operable patients, those with metastatic disease to the lung from other primary sites (e.g. breast, colorectal, sarcoma), those with previous lung radiation therapy, or those treated with hypofractionation schedules less than three or greater than five fractions, is not well known. For example, data from patients with lung metastases from primary colorectal tumors suggests an inferior local control[62]. There may be meaningful clinical endpoints beyond local control that were not addressed (e.g. recurrence rate in the same lobe, hilar and/or mediastinal lymph nodes, distant metastases and overall survival). The heterogeneity of the published data and lack of consistent reporting for these end-points precludes a rigorous assessment of this relationship at present.

Currently, there are few studies reporting local control for large tumors (e.g. > 5 cm). Additional analyses will be necessary to characterize the relationship between local control as a function of tumor size while accounting for the BED.

Data suggest that the therapeutic ratio is likely different for centrally located tumors[12,51] and thus the selection of optimal dose, schedule, technique, and treatment volume might differ in this setting. For example, in a study that included the risks of late toxicity and complications from SBRT in early stage NSCLC by tumor location, there appeared to be an 8-fold increase in grade 3 or higher complications (bacterial pneumonia, radiation pneumonitis, tracheal-bronchial fistula formation, etc.) when the tumor was within or touched a volume within 2 cm of the tracheal-bronchial tree [88] treated in a dose-escalated manner in 3 fractions. This study set the current definition for a centrally located tumor. The finding of unacceptable toxicity for central tumors has been further investigated, including in a cooperative group setting such as RTOG 0813, a phase I/II study designed to determine the maximal tolerated dose and efficacy of SBRT in a dose-escalated 5-fractions regimen from 10 to 12 Gy per fraction[5]. Early reported results showed reasonable overall rates of toxicity even at the highest dose level allowed by the protocol (60 Gy in 5 fractions) which was associated with a 7.2% rate of protocol-specified dose-limiting toxicity; however there were 3 deaths associated with SBRT at the two highest dose-levels. Thus, a patient’s eligibility for SBRT and differences in dose and margin based on whether the tumor is peripherally or centrally located are important and relevant considerations.

8. RECOMMENDED DOSE/VOLUME OBJECTIVES

Based on the data collected and the modeling results generated, Table 2 summarizes the required physical doses at isocenter and at the periphery of the PTV, assuming a prescription isodose line of 80%, to achieve a maximum local control rate with 3, 4 and 5 fraction regimens for T1, T2, and T1+T2 lesions. The derived α/β (Gy) that led to the overall best fit in the entire data set was approximately 21 Gy. Of note, since most of the data analyzed used 3D conformal techniques and not IMRT or VMAT, relationship between the dose distribution within the PTV and the dose at the isocenter or periphery of the PTV may be dependent on the treatment technique utilized and this may influence tumor control probabilities. Another caveat is that because there are fewer data points for isocenter dose < 52 Gy, the recommendation of 42 Gy in 3 fractions should be interpreted with caution and as a minimum dose required. Of note, a recently published guideline from the European Society for Radiotherapy and Oncology Advisory Committee on Radiation Oncology Practice reached a consensus that “risk-adapted” SBRT fractionation was achieved with 3 × 15 Gy for peripherally located lesions. For patients free from severe comorbidities and with favorable long-term OS expectancy, use of the maximum tolerated dose of 3 × 18 Gy should be considered[25].

9. FUTURE STUDIES

Maturation of reported studies with updated outcomes would benefit the robustness of our tumor control models in early stage NSCLC treated with SBRT. Consistent reporting of 5-year outcomes is needed in order to compare to the surgical outcomes, which is often considered the standard of care. For example, the early reports from RTOG 0236 noted 3-year LC rate of 97.6%, and nodal control rate of 87.2%. A subsequent report, with more mature follow-up data noted a 5-year LC rate of 93% and a nodal control rate of 62%[91]. It is reassuring that only 3 additional local recurrences occurred with longer follow-up compared to the 3-year outcomes.

10. REPORTING STANDARDS FOR OUTCOMES

More rigorous and consistent reporting of clinical outcomes together with the doses delivered is needed in order to improve the accuracy of TCP models for early stage NSCLC treated with SBRT (see Table 3 for summary). For example, as mentioned previously, consistent reporting of long-term outcomes of at least 5 years differentiating local control, interlobar control, local regional control (hilar/mediastinal recurrences), as well as incidence of distant metastases and overall survival are needed. In addition, reporting of details regarding the histology of treated tumors, the medical operability status of these patients, and tumor location will help refine TCP models for specific circumstances. Finally, standard dose-reporting metrics are needed. For example, reports should describe how the GTV is defined, how respiratory motion is accounted for at simulation and treatment, and the extent of expansion for the PTV. Doses to the GTV can vary by up to 20–30% depending on whether the dose is prescribed to the isocenter, prescribed to cover a certain percentage of the PTV, or prescribed to a specific isodose line (e.g. 80%). Such dramatic differences in the actual dose delivered may obscure any real dose-response when they are incorrectly represented. With regard to dose calculations, we recommend that modern algorithms be used for future published studies, tissue heterogeneity corrections should be accounted for using modern calculation algorithms.

Table 3:

Recommendations for reporting standards.

	Reporting recommendations
Clinical	Up to 5 years of outcome, including local control, interlobar control, local regional control (hilar/mediastinal recurrences), rates of distant metastases, overall survival, histology, tumor molecular makeup, medical operability status, tumor location, tumor size, and treatment related toxicities
Treatment simulation	Type of images, treatment immobilization, motion management strategies (breath-hold, gating, tracking, ITV, etc.)
Target volume definitions	Definition of GTV, ITV, CTV (if added), PTV margins
Dose calculation	Dose calculation algorithm used, heterogeneity correction, prescription parameters (isocenter, % of PTV, or specific isodose line), minimal GTV, ITV, or PTV coverage, mean doses, maximum doses, and equivalent uniform dose. % of GTV receiving > 110% of prescribed dose, the GTV D95%, Total dose, dose per fraction, total number of fractions, number of elapsed days during treatment
Dose delivery	Dose delivery machine, type of image guidance (4D, 3D, 2D, gated), frequency of image guidance, motion management strategies (breath-hold, gating, tracking,compression), motion monitoring

Perhaps new dose metrics are needed, as current prescription methods were designed for conventionally fractionated 3D conformal radiation therapy designed to deliver more homogenous dose distributions within the PTV. Due to increase heterogeneity of dose delivered with IMRT, the observed dose response (Figure 1) may be related to the GTV minimal dose, mean dose, maximal dose, or possibly to an equivalent uniform dose (with the model parameter ‘a’ to be determined). Other important metrics to report may include the percent of GTV receiving > 110% of the prescribed dose, the GTV D95%, or the percentage of microscopic disease coverage achieved by a particular treatment technique and dose fractionation schedule. In fact, treatment technique and dose fractionation schedule cannot be decoupled from each other but must be considered in combination. For instance, in a planning and modeling study, Arvidson et al., found that dose fractionation schedules and treatment techniques that varied by up to 174 Gy in LQ model 2 Gy fraction equivalents at the edge of the PTV yielding a similar predicted LC did cover more than 80% of possible microscopic disease extensions to a dose of 55 Gy or higher in 2Gy fraction equivalents[2]. They concluded that the high dose per fraction is necessary for some SBRT treatment regimens to obtain adequate microscopic extension coverage, while in other regimens yielding similar local control rates, lower prescribed doses per fraction can be employed since adequate microscopic extension coverage is obtained through added treatment margins. Therefore, a detailed description of target volume definition, treatment margins, and treatment technique is necessary in addition to dose parameters. Further support to reporting the percentage of microscopic disease coverage as a metric for a given dose fractionation schedule and treatment technique is provided by the systematic review of van Baardwijk et al., who found no significant relationship between the dose at the edge of the PTV and freedom of local progression[96]. More recently, Shaverdian et al. reported a clinical series of 120 consecutive stage I NSCLC patients treated with a physical dose regimen of 54 Gy in 3 fractions prescribed to either the PTV (ITV + 3–6 mm) or to the ITV alone[74]. The local control was 100% at 3 years in both clinical scenarios, confirming the hypothesis that doses to adjacent non-target tissue are likely sufficient to sterilize the microscopic disease near the GTV.

These metrics should be consistently reported in all future publications. Moreover, it is essential when different dose schemes are used in a single publication that a breakdown of the patients, tumor characteristics and LC outcomes be reported for each prescription scheme used. Ideally, a prospective web-based dosimetric and outcomes registry that is updated and curated in real-time for the entire country would facilitate future endeavors in analyzing radiation outcomes as a function of treatment parameters.