• Open Access

Regression curves of brain metastases after gamma knife irradiation: Difference by tumor and patient characteristics


To whom correspondence should be addressed.

E-mail: katsuraherbst@aol.com


Regression curves and local control rates of brain metastases after gamma knife treatment were evaluated to investigate differences in tumor response to radiation. A total of 203 metastases were serially evaluated using contrast-enhanced MRI (or computed tomography) at 1, 2, 3, 4.5 and 6 months after a 20-Gy dose. Differences were evaluated in regression curves and control rates between tumors ≥10 mm and tumors <10 mm in mean diameter, among three major histological subtypes of lung cancer, among adenocarcinomas of the lung, breast and colorectum, and between tumors in patients with above and below median hemoglobin levels. Smaller tumors shrank faster and yielded better control rates than larger tumors. Metastases from small cell and squamous cell carcinomas of the lung shrank faster than those from lung adenocarcinoma, but 6-month control rates were not different. Breast adenocarcinomas tended to shrink faster than lung adenocarcinomas, but the control rates were not different among adenocarcinomas of the lung, breast and colorectum. Tumors in patients with higher hemoglobin levels tended to shrink faster but the control rates were not different. Small cell and squamous cell carcinomas of the lung regress more rapidly than adenocarcinomas, although local control rates might not differ significantly.

The radiosensitivity of tumors depends on various patient-related and tumor-related factors, including the histology, the primary organ, the size and the oxygenation status of the tumors. Small cell carcinomas of the lung respond more quickly to radiation than other histological subtypes of lung cancer. In uterine cervical cancer treated by radiation, squamous cell carcinoma is generally associated with better prognosis than adenocarcinoma;[1, 2] such observations have led many oncologists to assume that small cell carcinomas and squamous cell carcinomas are more radiosensitive than adenocarcinomas.[3, 4] In addition, it is generally considered that breast cancers are more radiosensitive than adenocarcinomas in other organs,[5] and smaller tumors are more radiosensitive than larger tumors.[6, 7] Furthermore, substantial amount of clinical data shows that patients with high hemoglobin levels have a better prognosis than those with low levels following definitive radiotherapy of various cancers,[8-12] indicating that tumors in patients with high hemoglobin levels might be more radiosensitive than tumors in patients with low hemoglobin levels. However, these observations mostly come from retrospective analyses of clinical data and have not been evaluated in a prospective study.

In Japan, gamma knife (GK) treatment is now the most frequently used treatment for single or oligo brain metastases.[13] GK treatment usually entails a single session of relatively high-dose irradiation.[13, 14] Therefore, if imaging studies are carried out serially after GK, shrinkage patterns and local control of irradiated tumors following single high-dose radiation can readily be evaluated. Namely, regression curves of human tumors can be obtained, as are often generated for murine tumors in laboratory studies using a growth delay assay. Therefore, we considered that GK should be a useful tool to investigate and evaluate the radiosensitivity of various tumors. The present study evaluates the velocity of shrinkage and local control of metastatic brain tumors prospectively after a uniform GK dose of 20 Gy to clarify the differences according to size, histology and primary site of tumors, as well as the hemoglobin levels of the patients.

Materials and Methods

Study design and eligibility criteria

The present paper is a prospective study evaluating tumor response using contrast-enhanced MRI (or computed tomography [CT] in cases where MRI is contraindicated) at 1, 2, 3, 4.5 and 6 months following GK with a marginal dose of 20 Gy. To estimate the number of patients that should be accrued, the relative tumor area on MRI or CT at 2 months after GK was assumed to be 60% + 15% (SD) of the pretreatment size, on the basis of institutional experience.[15, 16] To detect differences of 20% in mean relative tumor size by histology and primary site with a significance level of P < 0.05 (two-sided) and a statistical power of 0.8, at least nine tumors were considered necessary per group.[17] With patients divided into two groups by tumor size, we planned to accrue at least 18 patients with lung adenocarcinomas, squamous cell carcinomas or small cell carcinomas, or with breast or colorectal adenocarcinomas, to detect possible differences between pairs of groups. Therefore, the study was continued until the number of patients in every group reached this number. Patients with metastases from lung adenocarcinomas were frequently seen, and accrual of such patients was stopped earlier than for other groups. The study protocol was approved by institutional review boards.

Patients who fulfilled the following criteria were considered eligible: (i) age >18 years; (ii) World Health Organization performance status score of 0–2; (iii) histologically proven adenocarcinoma, squamous cell carcinoma or small cell carcinoma of the lung, adenocarcinoma of the breast or adenocarcinoma of the colorectum as a primary tumor (histological subclassification according to tumor differentiation was not requested); (iv) measurable tumor(s) between 5 and 20 mm in mean diameter; (v) treatment by GK alone; (vi) no previous brain surgery and no previous whole-brain radiotherapy or GK; (vii) expectation of undergoing serial contrast-enhanced imaging studies for at least 3 months; and (viii) informed consent to undergo both GK and imaging studies. The diagnosis of brain metastases was made on the basis of imaging findings. In no patients was chemotherapy given within the 1.5 months prior to GK treatment. During follow up after GK (at 1.5–6 months), 24 patients with 29 evaluable lesions received chemotherapy with agents that do not freely cross the blood–brain barrier. Because tumor regression curves in patients undergoing chemotherapy were similar to those in patients not receiving chemotherapy, the contribution of chemotherapy was considered to be minimal and the two groups were analyzed together.

Patient characteristics

Between 2004 and 2009, 176 patients fulfilling the inclusion criteria entered the study. Unexpectedly, however, 15 patients could not undergo the imaging studies until 3 months after GK. They were, hence, deemed unevaluable and were excluded from analysis. As a result, 161 patients were evaluated: 91 men (57%) and 70 women (43%), with a median age of 64 years (age range 30–84 years). The patient and tumor characteristics are shown in Table 1.

Table 1. Primary cancer type, hemoglobin level and tumor size
Primary cancerHistologyNumber of patients Total Hb ≥ 12.3/<12.3 g/dLMean size ± SD (mm2) ≥78.5 mm2/<78.5 mm2



200 ± 84/34 ± 20



220 ± 91/28 ± 21
Small cell



230 ± 99/43 ± 24



210 ± 77/49 ± 21



233 ± 100/54 ± 25

Gamma knife treatment

The method for GK has been described previously.[15, 16] Briefly, contrast-enhanced MRI with a 1.5-T imager (Magnex150; Shimadzu, Kyoto, Japan and SIGNA EXCITE XL 1.5T; GE Healthcare, Milwaukee, WI, USA) or CT with a four-row multidetector scanner (LightSpeedPlus; GE Healthcare) was performed with a Leksell frame applied under local anesthesia for imaging of treatment planning. CT was used in two patients with contraindication for MRI or gadolinium agents. Treatment planning was performed using Leksell GammaPlan (Elekta AB, Stockholm, Sweden). The planning target volume was defined by the contrast-enhanced tumor region plus a 1-mm margin. GK was performed using a Leksell GK unit (Model C, Elekta AB). The peripheral dose at a 50% isodose line was uniformly set at 20 Gy.

In the present study, 186 tumors were treated with a single-isocenter plan, and only 17 lesions were treated with a multiple-isocenter plan. In both plans, tumors were covered with a 50% isodose line, so there appeared to be no differences in dose distribution.

Serial imaging studies

For imaging studies, MRI was absolutely the first choice, but in the two patients with contraindication for MRI or gadolinium agents, CT was used throughout because their tumors were clearly visible and measurable. All patients were scheduled to undergo contrast-enhanced MRI or CT before and at 1, 2, 3, 4.5 and 6 months after GK treatment. Because of disease progression, however, a proportion of patients became unevaluable before 6 months post-treatment. Among the 161 patients, those who could be evaluated until 3, 4.5 or 6 months were analyzed separately. In 42 patients with multiple tumors, one each of tumors with a mean diameter of 5–10 mm (preferably around 7.5 mm) and 10–20 mm (preferably around 15 mm) were selected and used for analysis. A maximum of two tumors were assessed per patient. Therefore, a total of 203 tumors were evaluated.

MRI was performed using 1.5-Tesla systems (Magnex150; Shimadzu and Gyroscan Intera; Philips Medical Systems, Best, the Netherlands). Our standard imaging procedures have been described previously.[15, 16, 18] The allowance for MRI slice thickness was 5 + 1 mm. CT was performed using a four-row multidetector scanner (LightSpeedPlus, GE). The slice thickness was 5 mm. A single shot of gadopentetate meglumine (0.1 mmol/kg body weight) or 100 mL iopamidol (300 mgI/mL) was used for contrast enhancement.

Data evaluation

The size of brain metastases was measured at an axial plane of maximal tumor size. In all cases, the longest diameter of an enhancing lesion and the diameter perpendicular to it were measured after appropriate magnification on picture archiving and communicating system monitors, and tumor areas were estimated by π(a/2·b/2). The relative tumor size was obtained by the ratio of the tumor size at each follow up to the pretreatment size. Note that a circular lesion of 10 mm in diameter corresponds to an area of approximately 78.5 mm2. Retreatment for local failure or new brain metastases was at the discretion of an attending physician; when whole-brain radiation or second GK was performed for the tumors under evaluation during the 6-month follow-up period, evaluation of the tumors was terminated at that time. In contrast, even when new brain metastases were treated with GK, evaluation of the GK-treated tumors was continued if the dose of the second GK to the tumors under evaluation was considered negligible. Differences in tumor size at each follow up were examined by t-test. Differences between the whole regression curves were examined by repeated-measure analysis of covariance. Local failure was defined by >25% increase of the tumor area from the smallest area after GK. Tumor control rates at 6 months after GK were calculated using the Kaplan–Meier method, and a log-rank test was used to compare the rates. All statistical analyses were carried out using StatView 5.0 (SAS Institute, Cary, NC, USA). A P-value of <0.05 was defined as statistically significant.


Responses by tumor size

Figure 1 shows regression curves of brain metastases with tumor areas greater than or less than 78.5 mm2 in patients who could be evaluated for 3 months, in those who were evaluable for 4.5 months and in those evaluable for 6 months. Smaller tumors shrank faster than larger tumors; relative tumor sizes at 1, 2, 3 and 4.5 months after GK were significantly larger in tumors >78.5 mm2. The local control rates at 6 months were 70% for tumor areas >78.5 mm2 and 86% for tumor areas <78.5 mm2; the difference was significant (P = 0.0073). Therefore, the tumors were divided into two groups in the following analyses: greater than or less than 78.5 mm2.

Figure 1.

Changes of relative tumor sizes (areas) after gamma knife treatment of tumors of greater than (◆) and less than (○) 78.5 mm2 in patients who were evaluable for 3 months (upper panel), for 4.5 months (middle panel) and for 6 months (lower panel). Bars represent SE.

Responses by histology among lung cancers

Figure 2 shows regression curves of metastases from three subtypes of lung cancer. Small cell carcinomas of the lung shrank faster than adenocarcinomas of the lung. Regression curves between adenocarcinomas and small cell carcinomas were significantly different in all analyses except for tumors <78.5 mm2 followed for 6 months. Squamous cell carcinomas also tended to decrease in size faster than adenocarcinomas, and there were significant differences in overall regression curves in smaller tumors with an area <78.5 mm2. The local control rates at 6 months, however, did not differ significantly among the three histological subtypes of lung cancer (Table 2).

Figure 2.

Changes of relative tumor sizes after gamma knife treatment for three subtypes of lung cancer in patients who were evaluable for 3 months (upper panel), for 4.5 months (middle panel) and for 6 months (lower panel). ◆, adenocarcinoma (Ad); ○, squamous cell carcinoma (Sq); and Δ, small cell carcinoma (Sm). Bars represent SE.

Table 2. Local control rates at 6 months according to histology, primary tumor site and hemoglobin level
SubjectHistologySize (mm2)P-valueSize (mm2)
  1. Ad, adenocarcinoma; Br, breast cancer; Co, colorectal cancer; Sm, small cell carcinoma; Sq, squamous cell carcinoma.

Lung cancerAdeno71%82%Ad versus Sm0.420.48
Squamous67%83%Ad versus Sq0.660.98
Small83%91%Sq versus Sm0.260.58
Primary cancer
AdenocarcinomaLung71%82%Br versus Co0.340.46
Breast68%89%Br versus Lung0.910.49
Colorectum58%100%Co versus Lung0.310.32
Hemoglobin (mg/dL)
Patients with adenocarcinoma≥12.366%60%≥12.3 vs <12.3 mg/dL1.00.28
< 12.381%91%

Responses by primary tumor

Figure 3 shows regression curves of brain metastases from adenocarcinomas of the lung, breast and colorectum. In tumors >78.5 mm2, breast cancer metastases shrank faster than lung adenocarcinoma metastases at 1 and 2 months, but otherwise, there were no differences in the regression curves among the tumor types. The local control rates also did not differ significantly among the three types of adenocarcinomas.

Figure 3.

Changes of relative tumor sizes after gamma knife treatment for adenocarcinomas of the lung (Lu, ◆), breast (Br, ○) and colorectum (Co, Δ) in patients who were evaluable for 3 months (upper panel), for 4.5 months (middle panel) and for 6 months (lower panel). Bars represent SE.

Responses by hemoglobin level

The median hemoglobin (Hb) level before GK was 12.3 g/dL for all patients (range, 7.7–20.1) as well as for all adenocarcinoma patients (range 7.9–16.6), so the patients were classified into two groups: Hb ≥ 12.3 and Hb < 12.3 g/dL. Figure 4 shows regression curves according to the hemoglobin level. For adenocarcinoma metastases with an area >78.5 mm2, relative tumor areas at 1 month were smaller in patients with Hb ≥ 12.3 g/dL than in patients with Hb < 12.3 g/dL (P = 0.04), but there were no differences in tumor sizes and control rates at 6 months (Table 2).

Figure 4.

Changes of relative tumor sizes after gamma knife treatment in adenocarcinoma patients with greater than (◆) and less than (○○) median hemoglobin level (12.3 g/dL) who were evaluable for 3 months (upper panel), for 4.5 months (middle panel) and for 6 months (lower panel). Bars represent SE.

When all tumors were analyzed, the curves for tumors in patients with Hb ≥ 12.3 g/dL also tended to lie below the curves for tumors in patients with Hb < 12.3 g/dL, especially in tumors >78.5 mm2, but the differences were not significant (data not shown). The influence of hemoglobin levels was also examined using the cutoff levels 11 and 10 g/dL; similar trends were observed, but the tumor sizes at 1–6 months and the local control rates at 6 months did not differ significantly between the two groups (data not shown).


Various factors are considered to be associated with the radiosensitivity of tumors. This study investigated the issue by evaluating the rate of tumor regression and local control rate at 6 months. In many studies, smaller tumor volume has been shown to be a significant prognostic factor in patients with brain metastases from lung cancer.[19-21] Smaller tumors not only have fewer clonogenic cells but also tend to have a lower hypoxic fraction,[6] so it appears quite reasonable that smaller tumors are more easily controlled. In a study analyzing 191 patients with brain metastases from lung cancer, the shrinkage rate was significantly different between tumor volumes of 2 cm3 or greater and those <2 cm3.[22] However, in another study analyzing 51 patients with brain metastases treated by GK excluding patients with small cell lung cancer, neither the histological type nor the tumor volume was a significant prognostic factor.[23] The absence of difference might be due to the small patient number. Our prospective study including 203 tumors clearly indicated that tumor regression rates as well as local control rates were poorer in larger tumors.

Tumor regression occurred faster in metastases from small cell lung cancer and squamous cell carcinomas of the lung than in metastases from lung adenocarcinomas. However, there were no differences in local control rates at 6 months among the three tumors. A GK study also indicated no difference in 1-year control between small cell carcinoma and nonsmall cell carcinoma metastases.[20] It is well known that small cell carcinoma responds to radiation quickly and this is considered to be due to apoptosis.[24] In addition, squamous cell carcinomas have been thought to be more radiosensitive than adenocarcinomas by many oncologists other than radiation oncologists, especially in Japan.[25, 26] In contrast, it is well known that there is no difference in the prognosis of lung cancer patients with these two histological subtypes after radiation therapy.[27, 28] Our results indicate that small cell carcinomas and squamous cell carcinomas regress more rapidly than adenocarcinomas, so the former two could be more radiosensitive than the latter. However, radiosensitivity in terms of local control probability might not differ significantly among the three histological subtypes, as has also been suggested by radiosensitivity testing studies.[29-31] Alternatively, however, it might also be interpreted that a 20-Gy marginal dose of GK is generally sufficient to control lung cancer metastases of 2 cm or less for 6 months. The difference in radiosensitivity in terms of local control between small cell and nonsmall cell carcinomas did not become apparent in the present study.

Among adenocarcinomas, breast cancer is often regarded as more radiosensitive than colorectal cancer.[5] Although brain metastases from colorectal cancer are relatively rare, a GK study suggests this trend.[32] In addition, a substantial amount of data on stereotactic body radiotherapy (SBRT) for the lung indicates a worse local control rate for metastases from colorectal cancer compared with those of other lung metastases.[33, 34] In the present study, although breast cancer metastases of >10 mm in mean diameter shrank faster than lung adenocarcinoma metastases, metastases from adenocarcinomas of the lung, breast and colorectal cancer showed similar control rates at 6 months. The local control rates for metastases from lung adenocarcinoma and those from colorectal cancer were 71% and 58%, respectively, in tumors >78.5 mm2, but the local control rates were 82% vs 100% in tumors <78.5 mm2. Although neither of these differences were significant, the trend might become clearer if more numbers of colorectal cancer metastases are analyzed (in the present study, n = 15 for >78.5 mm2 and n = 7 for <78.5 mm2). In addition, the absence of differences in local control rates might be attributable to the relatively short period of evaluation. Because brain metastasis patients generally have a poor prognosis and most patients die of extracranial disease progression, it is rather difficult to evaluate the control rate over a longer period after GK. This can be considered a limitation of this study, and the issue of local control by primary organ might better be investigated in SBRT studies for lung metastases.

Anemia has been reported to be a factor associated with poor outcome after radiation therapy.[8-10] It has been hypothesized that low hemoglobin levels lead to decreased tumor oxygenation and resistance to radiotherapy.[35] In the present study, hemoglobin levels above the median tended to be associated with early response to GK, especially in larger tumors. This observation concurs with the findings of previous studies,[8-12] but the entire regression curves and 6-month local control rates were not different. Again, evaluating the 6-month local control might be a reason for not detecting differences due to the hemoglobin level, and the influence of hemoglobin level should also be further evaluated in future studies.

In conclusion, the radiosensitivity of tumors should not be estimated by early responses alone. Tumors, like adenocarcinomas, respond to radiation rather slowly, but they might possess reasonable radiosensitivity in terms of local control. Small cell carcinomas and squamous cell carcinomas of the lung shrank faster than adenocarcinomas, but local control rates at 6 months were not different. This observation should contribute to reevaluating the assumption that these two subtypes of lung cancer are more radiosensitive than adenocarcinomas. Further studies on tumor radiosensitivity using solitary lung or liver tumors that can be followed for longer periods are warranted.


The authors would like to thank Drs Naoki Hayashi and Kyota Oda, Mrs Nobuhiko Muramatsu, Kimio Fukazawa, Makoto Tsuchiya, Masahiro Hagiwara, Hisato Nakazawa, Takashi Kimura, Ms Yoko Morishita, Ryoko Ohsaka and Hiromi Teramoto for their valuable support in data collection and technical assistance.

Disclosure Statement

The authors have no conflicts of interest to declare.