Introduction

Stereotactic body radiotherapy (SBRT) is the standard treatment for inoperable early non-small cell lung cancer (NSCLC), with good local control rate and low incidence of side effects [1, 2]. SBRT has also achieved encouraging outcomes in patients with operable early NSCLC [3, 4]. The total physical dose varies greatly due to different fractionation [5, 6]. Even if the fractionation and the nominal prescription dose are the same, there is significant difference for near maximum dose or the average dose in the planning target volume due to the absence of the uniform standard for the selection of prescription dose isodose line [7]. In SBRT of early stage NSCLC, higher maximum dose in target volume can not only improve local control rate, but also reduce cancer specific death and promote overall survival (OS), without increasing toxicity [8]. Several studies have attempted to determine the most relevant dose parameters for tumor local control in SBRT of NSCLC, and tried to recommend the prescription dose [5, 6]. However, it is regrettable that the optimal dose prescription is still unclear.

In this study, published articles on SBRT in NSCLC were comprehensively searched, and the dose effect relationship between prescription dose and tumor local control rate was established through Probit model.

Materials and methods

Data sources and search strategy

A comprehensive literature search was conducted using the PubMed, the Web of Science and the Cochrane databases to determine the published articles of NSCLC patients treated with SBRT. We searched “stereotactic ablative radiotherapy”, “stereotactic body radiotherapy” “non-small cell lung cancer” or their synonym in the title or abstract, and the search was restricted to English-language (Additional File 1: Table S1). Our last literature search was on September 30, 2022.

The inclusion criteria

  1. 1.

    The enrolled patients were early-stage (T1-2 N0) NSCLC.

  2. 2.

    The treatment conforms to the SBRT technical specifications in the AAPM TG101 report.

  3. 3.

    The original study of complete prescription dose information and tumor local control rate was reported.

  4. 4.

    The median follow-up time should be at least 24 months and the sample size of patients should be at least 20.

The exclusion criteria

  1. 1.

    Review articles, articles about recommendations, consensus or guidelines.

  2. 2.

    Irrelevant articles or articles focusing on the research of radio diagnostics, medical physics and radiobiology.

  3. 3.

    The study included more than one dose fractionation, and did not report tumor local control for a cohort of single dose fractionation.

  4. 4.

    Proton beam stereotactic radiotherapy.

  5. 5.

    Articles focused on the combination of SBRT and other treatments, such as surgery, or the use of SBRT only as boost after conventional radiotherapy, or a salvage treatment for patients of recurrence.

  6. 6.

    Non-English articles.

Data extraction

After removing the duplicate literature, we screened the literature by title and abstract, and the remaining literature was screened by full text. When the patient data reported in the literature overlap, we selected the latest and most complete data. Literature screening and data extraction are conducted independently by two authors according to the inclusion and exclusion criteria, and objections are resolved through negotiation. The following data were extracted from all selected studies: 1. Article information: first author, publication year, country, patient treatment time and patient number; 2. Patient and tumor characteristics: median age, tumor stage, medically operable or inoperable, internal target volume (ITV) delineation and planning target volume (PTV) margin; 3. Treatment characteristics: prescription (nominal) dose, peripheral (near minimum) dose, central (near maximum) dose, number of minutes, interval time between fractions; 4. Three years local control rate. If 3 years local control rate is not available in the literature, obtained from Kaplan Meier curve is also recognized. If multiple prescription doses and corresponding local control are introduced in the article, subgroups data will be included in the analysis. The biologically effective doses (BED) were calculated using the linear quadratic (LQ) equation: BED10 = nd[1 + d/(α/β)], where d represents the fraction dose, n represents the fractions, and α/β equals to 10. Due to the high heterogeneity of the PTV dose in SBRT, in order to facilitate the dose effect analysis, this study defines several doses: nominal dose, which is the prescription dose; central dose, which is the maximum dose of PTV, isocenter dose or the minimum dose to the 2% volume of the maximally exposed region in PTV (D2); peripheral dose is the minimum dose of PTV or D95–98. When central dose BED10 and peripheral dose BED10 data are available, the average value is defined as average dose BED10.

Regression analysis

Probit model regression analysis was conducted by XLSTAT 2016 (Addinsoft, Paris, France). Select the mean or median dose BED10 as the quantitative dose. The number of patients is considered as observation weights in the dose effect analysis panel. The statistical significance was set at the level of P < 0.05. Subgroup analysis was conducted according to the country, type of study design, treatment era, operable or inoperable, proportion of male patients, median age, patient immobilization mode, CT scanning mode, PTV margin, number of fractions, interval time between fractions, treatment equipment, tissue heterogeneous correction or not, follow-up time.

Results

Description of the included studies

After comprehensive search, no article about regression analysis on dose effect relationship between dose and tumor local control rate based on published data was found. A total of 3699 potential related studies were identified using a systematic literature retrieval strategy. After duplicates removed, 22 eligible studies were obtained through title, abstract and full text screening and included in probit model regression analysis, involving 1861 patients in total, as shown in Additional File 1: Fig. S1. These included studies were from 6 countries, with the most published studies coming from the United States (nine), followed by Japan (seven), Italy and Sweden (two each), Spain and Canada (one each). Among them, there are 13 prospective studies, 9 retrospective studies. The main characteristics of the included studies are presented in Table 1.

Table 1 The main characteristics of the included studies
Full size table

Probit analyses

The range of prescription dose is 30–66 Gy, the fractions range is 1–5, and the actuarial or rough 3-years local control rate is 73.5–100.0%. There is no significant dose effect relationship between nominal BED10 and peripheral BED10 versus 3-years local control probability. There were significant dose effect relationships between the center BED10 and the average BED10 versus the 3 years local control probability, with P values are 0.001 and < 0.0001, respectively, as shown in Figs. 1 and 2. According to this model, the central BED10 and the average BED10 corresponding to probabilities of 90% 3 years local control were 167.5 GyBED10 (135.0–219.6 GyBED10) and 145.7 GyBED10 (126.7–172.2 GyBED10) respectively. The 3 years local control rate of 90.5% (87.5–92.1%) and 89.5% (86.7–91.0%) can be expected at the center BED10 of 180 Gy and the average BED10 of 140 Gy, prospectively. The Probit analyses based on subgroups show that the central BED10 and peripheral BED10 have most cumulative number of significance (Ten each), followed by the average BED10 (Eight), and nominal BED10 (Six), as shown in Table 2.

Fig. 1
figure 1

The probit model for the relationship between central BED10 and 3-years local control

Full size image
Fig. 2
figure 2

The probit model for the relationship between average BED10 and 3-years local control

Full size image
Table 2 The Probit analyses based on subgroups between BED10 and 3-years local control
Full size table

Discussion

SBRT is an effective treatment strategy for early-stage NSCLC [31, 32]. Because of the volume effect of normal tissue tolerance and the small tumor volume in SBRT, the PTV of SBRT can be delivered a higher BED than conventional radiotherapy, and the BED range is relatively large. Therefore, it is necessary to establish the dose effect relationship between dose and local control rate, so as to obtain an optimal BED to guide clinical practice.

Through systematic review and meta regression analysis based on Probit model, this study found that there was no significant dose effect relationship between nominal BED10 or peripheral BED10 and 3 years local control rate, but there was a significant dose effect relationship between PTV central BED10 or PTV average BED10 and local control rate. This result is similar to that of Eriguchi et al. [5]. Through systematic review and meta-analysis, they determined that the correlation between central BED10 of PTV and tumor local control is stronger than that between peripheral BED10 and local control.

Although peripheral BED10 and nominal BED10 were significantly correlated with local control in subgroup analysis, the nominal BED10 or peripheral BED10 do not have the predictability of local control for whole included studies in our study. It seems to be in great contradiction with the previous radiobiological theory and clinical practice. However, after careful consideration, it is found that this result is reasonable. In conventional fractionated intensity modulated radiotherapy, the prescription dose is given with a unified normative standard, which encourages absorbed-dose homogeneity in the PTV. However, in SBRT, radiation oncologists seem to accept the extreme dose inhomogeneity within the PTV due to the characteristics of small field dosimetry and the advantage of dose inhomogeneity in PTV. In Radiation Therapy Oncology Group (RTOG) study 0813 and RTOG 0915, the prescription dose isodose line is selected at 60–90% of the maximum dose, that is, the maximum dose in the PTV is about 110–166% [23, 27, 33]. In terms of radiation physics, sufficiently high tumor dose inhomogeneity is also conducive to a sharp drop in the dose of normal tissues outside PTV [7]. At present, there is no uniform standard for the prescription dose method of SBRT. Some prescription doses are defined as isocenter dose [23, 24, 34], and some are defined as isodose lines surrounding 95% PTV volume [32, 35, 36]. At the same time, the degrees of heterogeneity of dose in PTV for different centers may be diverse. The maximum dose in PTV at some centers is about 120% of the prescribed dose [35], while some is more than 160% [31]. Therefore, for the same nominal dose, it may occur that the actual dose given to PTV is not comparable [37].

For the cohort study of the same prescription dose rule and PTV dose inhomogeneity, the correlation results between prescription dose BED10 or peripheral BED10 and tumor local or regional control have been obtained. Jeon et al. [38] found that PTV prescription dose BED10 > 100 Gy can obtain better 3 years regional control, 69.4% versus 100%, P = 0.004. Suzuki et al. [39] conducted a retrospective analysis on SBRT of 383 patients with early-stage NSCLC. The patients were divided into four groups based on GTV volume (11 cc) and PTV peripheral BED10 (85 Gy). The results demonstrated that among patients with GTV > 11 cc, the 5-years local control rate of patients with BED10 ≥ 85 Gy was better than that of patients with BED10 < 85 Gy, 65% versus 94%, P = 0.002.

For the central BED10, several studies have yielded correlations with local control. Onishi et al. [40] reviewed 245 patients from 13 Japanese institutions, and found that when BED10 < 100 Gy at the isocenter, the local recurrence rate was 26.4%, significantly higher than 8.1% in the group with BED10 ≥ 100 Gy, P < 0.05. For operable patients, the 3-years OS of patients with BED10 ≥ 100 Gy at isocenter was 88.4%, which was significantly higher than that of patients with < 100 Gy (69.4%, P < 0.05). Guckenberger et al. [41] performed a retrospective logistic regression analysis on 399 patients with NSCLC, and found that there was a significant dose effect relationship between PTV maximum BED10 and tumor local control: when PTV maximum BED10 reached 176 Gy, 90% of tumor control could be expected. This result is consistent with that of the Probit model in this study.

For the average BED10, several studies have suggested that there are significantly relationships between that and the local recurrence rate, progression free survival rate or OS. Kestin et al. [42] found that in the SBRT of NSCLC, the average dose BED10 of PTV was significantly correlated with the local recurrence rate. The 2-years local recurrence rate of patients with average BED10 < 125 Gy was 17%, while that of patients with average BED10 > 125 Gy was only 4%, P = 0.01. Zhao et al. [43] conducted univariate analysis on SBRT data of 1092 NSCLC patients, indicated that the average BED10 > 130 Gy was related to 97% local/lobar control, and suggested that it should be one of the aims of plan optimization. The univariate analysis results of Shamp et al. [44] showed that BED10 > 150 Gy in the center of the PTV was related to the improvement of disease-free survival (DFS) and OS, with P values of 0.014 and 0.0132 respectively. These results are similar to that of this study. This study reiterated the importance of the average BED10 of PTV. When the average BED10 dose reaches 140 Gy, a 3-years local control rate of 89.5% (86.7–91.0%) can be expected.

Based on the results of the Probit regression analysis in this study, combined with the definitions of central BED10 and average BED10, the ED90 of peripheral BED10 can be calculated is 123.9 GyBED10. This result is slightly higher than that of Lee et al. [45]. Lee et al. tried to determine the optimal dose prescription of NSCLC SBRT through linear quadratic, universal survival curve, and regrowth models fitting. The results show that of the 3 models, the regrowth model provides the best fit to the clinical data. To achieve the maximal local control rate, the predicted physical dose schemes when prescribed at the periphery of the PTV are 43 ± 1 Gy in 3 fractions, 47 ± 1 Gy in 4 fractions, and 50 ± 1 Gy in 5 fractions. According to linear quadratic model calculation, the BED10 of the above dose schemes is 100.0–104.6 Gy. Our results pay more attention to the dose heterogeneity within the PTV, and suggest that a set of standard prescription dose specifications, including prescription dose definition and dose heterogeneity within the PTV, should be established.

From the results of our Probit model, they seem to give a false impression, that is, when the central BED10 or the average BED10 reaches 80 Gy or 70 Gy respectively, about 85% of the 3-years local control rate can be obtained. This is because, driven by previous dose effect conclusions [40], most BEDs are high, and data on lower BEDs is absence. In addition, there is also the possibility of publication bias, that is, it is difficult to publish the results with low local control rate at low BEDs or the authors are unwilling to publish.

This study has some limitations. There is a large heterogeneity in the patient cohort among the selected studies. These heterogeneities, including tumor location, size, number, and general status, are important confounding factors affecting local control of patients. In addition, the definition of local recurrence, the intervals and imaging modalities of follow-ups and recurrent diagnosis also affected the results. Due to the different ways of reporting in various literatures, the values of peripheral dose and central dose also have some uncertainties. Based on the Probit model in XLSTAT, it is not possible to conduct multifactor analysis for other factors besides the dose. The BED included in the Probit model analysis is calculated based on the LQ model. Therefore, the correctness and universality of the LQ model and the uncertainty brought by the value of α/β also affect the model results.

In conclusion, for NSCLC treated with SBRT, more attention should be paid to the central dose and average dose of PTV. The 3-years local control rate of 90.5% (87.5–92.1%) can be expected to be obtained at the central BED10 dose of 180 Gy; the 3-years local control rate of 89.5% (86.7–91.0%) can be expected to be obtained at the average BED10 dose of 140 Gy. A set of clear definition in the dose prescription should be established to ensure the effectiveness and comparability of treatment.