Use of three pins in Gamma Knife stereotactic radiosurgery for brain metastases

Abstract

Purpose

We present our institutional experience in treating brain metastases with GK-SRS and a headframe fixed to the skull with only 3 pins to avoid collisions between the headframe and the Gamma Knife (GK) machine.

Methods and materials

Among 3500 consecutive patients who received GK-SRS in 2011–2017, 50 had 1 of the 2 anterior pins removed immediately before treatment of ≥1 brain lesion. Endpoints were local control, dosimetric parameters, and toxicity.

Results

Median follow-up time for the 49 patients with follow-up was 7.0 months (range 0.2–57.0). Median number of lesions treated per session was 6 (range 1–18); a median 1 lesion was treated with 3-pin fixation (range 1–2) and a median 5 lesions treated with 4-pin fixation (range 0–17) during the same session. Lesions treated with 3-pin fixation were in the occipital lobe (n=41), cerebellum (n=9), or temporal lobe (n=1). No local failures were noted. The sole grade 2 toxicity (partial seizure) was attributed to treatment of a 4-pin-fixed lesion. Except for gradient index, dosimetry did not vary for lesions treated with 3-pin versus 4-pin fixation.

Conclusions

Treating brain metastases with 3-pin fixation did not compromise treatment outcome and is a good option for posterior brain metastases that cannot otherwise be treated with 4-pin GK-SRS.

Introduction

Stereotactic radiosurgery (SRS) is a preferred treatment option over whole-brain irradiation for limited brain metastases.[1,2,3] Gamma Knife SRS (GK-SRS) involves the use of multiple Co-60 sources to create 192 or 201 beams that converge on a single isocenter, with an accuracy between 0.1 and 1 mm.[4]

For patients undergoing GK-SRS, the skull is fixed to the stereotactic headframe by using 4 pins, at the left and right posterior and anterior posts. After the frame is placed, the potential for collision of a post or screw with the collimator is occasionally identified; although the frame’s gamma angle can be rotated from the neutral position (90°) to either 70° or 110° to allow the frame to move out of collision range, this is not always sufficient. Collisions are most common for extremely anterior or posterior lesions, particularly for patients with large heads.

After the various options for gamma angle and shot placement are exhausted, the patient can either undergo reframing or have a pin removed. Theoretically, reframing should include repeat imaging and generation of a new coordinate system; however, reframing is stressful for patients and increases treatment time. Another option is to remove 1 of the 4 pins while the patient is on the treatment couch and use the original 4-pin plan. However, concern has been expressed that target accuracy could be affected owing to displacement of the skull toward the removed pin site.

We previously evaluated outcomes after GK-SRS for brain metastases after removal of 1 of 4 headframe pins for skull immobilization [5] and found excellent local control and no additional toxicity. However, that study was limited to 20 patients, and 20 lesions. Here we sought to validate our previous findings by evaluating an additional 30 patients, for a total of 50 patients (51 lesions).

Methods and Materials

Patient Selection and Treatment

We retrospectively reviewed the records of approximately 3500 patients with brain metastases treated with GK-SRS at a single institution from September 2011 through 2017 by searching our de-identified, institutional review board–approved GK database. Due to extreme measure taken to protect patients’ confidentiality and retrospective nature of this study involving no more than minimal risk to the subjects, the review board approved the request for waiver of informed consent. All patients were treated with an Elekta Gamma Knife Perfexion system (Elekta, Stockholm, Sweden). Magnetic resonance imaging (MRI) scans for treating planning were acquired with a GE 1.5T MRI scanner (GE Healthcare, Chicago, Illinois, USA) with a gadolinium-enhanced GK-SRS protocol (3D, Axial, T1+C, EDR, FSPGR, TR/TE ~6.9/2.6 msec, 2 NEX, 220 mm FOV, 256×256 pixels, 1-mm slice thickness, no gap/overlap). MRI scans were obtained with a Leksell Coordinate Frame G attached to the patient’s skull and secured by 4 fixation posts/screws, placed under local anesthesia. The attending radiation oncologist chose the radiation dose on the basis of tumor size and the maximum tolerated doses established in the Radiation Therapy Oncology Group study 90-05.[6]

Patients for whom framing and treatment planning were initially done with 3 pins were excluded from this study. A total of 50 patients were identified who had at least 1 lesion treated with 3 pins (1 pin having been removed while the frame was docked to the treatment couch) and without verification imaging. Treatment plans were developed from the planning data obtained when all 4 pins were secured. All of the removed pins were anterior, because our practice involves acquiring a new planning computed tomography (CT) or MR scan after removal of a posterior pin. For these 50 patients, unsuccessful attempts had been made to resolve the clearance issues. If moving the violating shot to a treatable position resulted in less coverage, consideration was given to changing the gamma angle, shot pattern, and weighting to cover the lesion, possibly resulting in greater dose to normal tissue. These changes were judged by the treating radiation oncologist as acceptable or not. If treating at a treatable position with less selectivity was acceptable, then treatment could go forward with all 4 pins. If less selectivity was felt not to be acceptable, but removing a pin allowed for treating at the violating position with better selectivity, then a pin was removed and the patient was treated at the desired position. Before the pin was removed, all patients had completed the portion of treatment delivered with all 4 posts intact and were given a break if needed. Patients were then “docked” onto the treatment couch, still with 4 pins and posts, and checked for clearance to re-verify the site of pin removal and to confirm that the plan could be delivered if the pin were removed. Once the pin and post were removed, patients remained in the treatment position, with the 3-pin configuration, throughout the entire treatment. If adequate lesion coverage could not be obtained with pin removal, then patients were reframed and not included in this study. Afterward, patients typically returned to the clinic, where they underwent neurologic examination and MRI of the brain with and without contrast, every 2 to 3 months. All medical records and images available up until March 2018 were reviewed for all 50 patients.

Dosimetric Parameters

The prescription dose (D_P), prescription isodose line (IDL), number of shots, tumor volume (V_T), tumor volume covered by IDL (V_p), prescription isodose volume (V_d), volume covered by 50% of the prescription isodose surface (V_0.5*d), and maximum dose (D_max) were collected from each patient’s treatment plan (stored in the GammaPlan system). The conformity index, a measure of conformity of the plan, was calculated as (V_d/V_T). The gradient index (GI), a measure of dose fall-off, was calculated as (V_0.5*d/V_d). Tumor coverage was calculated as (V_p/V_T), and in all cases was at least 99.8%.

Statistical Analyses

All time intervals were measured from the date of GK-SRS. Fisher’s exact tests were used to compare lesion characteristics and Wilcoxon rank sum tests to compare dosimetric data. Actuarial survival was estimated by Kaplan-Meier analysis. For progression-free survival, local failure was scored as an event, and patients were censored at each follow-up MRI session up to March 2018. For overall survival, patients were censored at the last clinic follow-up visit. Statistical results were analyzed with IBM SPSS Statistics Version 24 (International Business Machines Corporation, Armonk, New York, USA)

Treatment-related toxicity was assessed and scored with the Common Terminology Criteria for Adverse Events (CTCTAE) version 4.03.

Local failure was defined as a function of increased lesion size (at least a 50% increase in the product of the 2 largest perpendicular diameters for lesions >5 mm measured at the baseline scan acquired for GK-SRS or a 100% increase in the product of the 2 largest perpendicular diameters for lesions ≤5 mm). All images available from the time of treatment until March 2018 were imported into GammaPlan to evaluate changes in size. If initial follow-up of a lesion met the size-increase criteria for “local failure” but later imaging revealed a decrease in size to below those criteria, then the lesion was not classified as a local failure. Lesions that met the size-change criteria were also classified qualitatively. In addition to the size-change criteria, later pathologic findings were also evaluated for local failure.

To account for differences in lesion size and location, we also compared the location and size of the 3-pin lesions with those of the 4-pin control lesions in the same patient. Location-matching involved matching a 3-pin lesion in the patient’s occipital, cerebellar, or temporal lobe with 4-pin lesions located in the corresponding occipital, cerebellar, or temporal lobes. Size-matching involved matching a 3-pin lesion with 4-pin lesions that were located in occipital, cerebellar, or temporal lobes (regardless of which lobe the 3-pin lesion was in) with a volume between 0.5 to 2 times that of the 3-pin lesion treated in the same patient. Of the 51 3-pin lesions, 17 could not be matched to a 4-pin lesion (4 of those 17 were the only lesion in that treatment session), and 34 had at least one 4-pin match. Among the 277 lesions treated with 4 pins, 71 matches were found: 16 for both location and size, 25 for location only, and 30 for size only.

Results

Patient and Tumor Characteristics

Patient and tumor characteristics are summarized in Table 1.

Table 1.

Patient Characteristics.

Characteristic	Value or No. (%)
Total no. of patients	50
Age, y, median (range)	61 (26–83)
Sex, no. (%)
Male	29 (58)
Female	21 (42)
Race, no. (%)
White	42 (84)
Hispanic	4 (8)
Black	1 (2)
Asian	3 (6)
Primary malignancy, no. (%)
Lung	18 (36)
Melanoma	14 (28)
Renal	6 (12)
Breast	3 (6)
Salivary gland	1 (2)
Thyroid	1 (2)
Testicular	1 (2)
Esophagus	1 (2)
Neuroendocrine carcinoma	4 (8)
Sarcoma	1 (2)
KPS score, median (range)	90 (70–100)
GPA, median (range)	1.5 (0.5–3.5)
No. of lesions treated in 1 GK session, median (range)	6 (1–18)
Follow-up time for 49 patients, mo., median (range)	7.0 (0.2–57.0)
Imaging follow-up time for 44 patients, mo., median (range)	6.7 (0.3–55.6)
Follow-up time for the 17 living patients, mo., median (range)	12.1 (6.1–57.0)
Survival time for the 33 dead patients, mo., median (range)	4.6 (0.7–37.2)

Abbreviations: GPA, graded prognostic assessment; KPS, Karnofsky performance status

Among the 50 patients in this analysis, 51 lesions were treated with 3 pins and 277 lesions with 4 pins. Four of the 51 lesions treated with 3 pins were the only lesion treated for that patient that day. All of the 3-pin treatment sessions had one anterior pin and post removed (none of the removed pins was on the posterior side) after the patient had been “docked” to the couch. Of the 51 lesions treated with 3 pins, 30 were on the left (and thus the anterior-right pins were removed) and 21 were on the right (anterior-left pins were removed). Eleven of the 51 lesions were treated at a gamma angle of 110°, and 40 at a gamma angle of 90°. Median treatment time for the 50 3-pin sessions was 19.9 minutes (range 10.7–108.4 minutes, mean 26.9 minutes). Median treatment time for the 4-pin sessions was 119.5 minutes (range 33.3–275.1 minutes, mean 133.8 minutes).

Survival

Of the 50 patients, 17 were alive at the time of this analysis, with a median follow-up time of 12.1 months (range 6.1–57.0 months, mean 18.9 months). The median survival time for the other 33 patients was 4.6 months (range 0.7–37.2 months, mean 8.0 months). Six patients had no follow-up imaging; one patient who died 1.2 months after GK-SRS had no follow-up visit at all. The maximum survival time for these 6 patients was 2.2 months. The median follow-up time for the 49 patients who had follow-up visits was 7.0 months (range 0.2–57.0 months, mean 11.5 months). The median imaging follow-up time for the 44 patients who had follow-up imaging was 6.7 months (range 0.3–55.6 months, mean 11.5 months).

Lesion and Dosimetry Characteristics

Lesion and dosimetry characteristics for the 3-pin and 4-pin lesions are summarized in Table 2. Only one 3-pin lesion was not located in either the occipital or cerebellar lobe, as compared to 193 4-pin lesions (2.0% vs. 70.0%, P<0.001). The gamma angle was 110 ^o for 11 3-pin lesions and 23 4-pin lesions (21.6% vs. 8.3%, P<0.01). The difference in gamma angle for 3-pin or 4-pin lesions was no longer significant when comparing only the location-matched (p=0.6) or size-matched lesions (P=0.3).

Table 2.

Characteristics of the 3-Pin and 4-Pin Lesions.

No. of Lesions (%) or Median (Range)
Characteristic	Lesions Treated with 3 Pins (n = 51)	Lesions Treated with 4 Pins (n = 277)	P Value	4-Pin Size-Matched or Location-Matched Lesions (n=71)	P Value (vs. 3-pin)	4-Pin Location-Matched Lesions (n=41)	P Value (vs. 3-pin)
Side of brain, no. (%)			0.49		0.58		0.68
Right	21 (41)	120 (43)		33 (46)		19 (46)
Left	30 (59)	148 (53)		38 (54)		22 (54)
Central	0 (0)	9 (3)		0 (0)		0 (0)
Location, no. (%)
Occipital	41 (80)	33 (12)		27 (38)		24 (59)
Cerebellum	9 (18)	51 (18)		28 (39)		16 (39)
Frontal	0 (0)	95 (34)		0 (0)		0 (0)
Parietal	0 (0)	40 (14)		0 (0)		0 (0)
Temporal	1 (2)	38 (14)		16 (23)		1 (2)
Thalamus	0 (0)	4 (1)		0 (0)		0 (0)
Brainstem	0 (0)	7 (3)		0 (0)		0 (0)
Other	0 (0)	9 (3)		0 (0)		0 (0)
Location			0.00		0.00		1.00
Occipital/cerebellum	50	84		55		40
Other	1	193		16		1
Gamma Angle			0.01		0.65		0.63
90o	40	254		58		30
110o	11	23		13		11
Prescription dose, Gy	20 (15–20)	20 (12–24)	0.03	20 (15−22)	0.34	20 (15−22)	0.80
Prescription isodose line	50 (40–80)	50 (42–85)	0.00	50 (42−75)	0.12	50 (42−75)	0.36
Dmax, Gy	40 (22.5–45.45)	40 (17.33–44.44)	0.02	40 (28.57−44.44)	0.44	40 (28.57−44.44)	0.21
Dmin, Gy	22.5 (15.4–30)	23 (12.75–33.2)	0.05	22.75 (16−28.8)	0.14	22.4 (16−28.8)	0.74
No. of shots (range)	2 (1–26)	2 (1–34)	0.29	2 (1−34)	0.99	2 (1−34)	0.58
Tumor volume, cm3	0.13 (0.013–8.84)	0.06 (0.002–22.03)	0.00	0.08 (0.004−16.24)	0.11	0.2 (0.004−16.24)	0.54
Treatment time, min	19.9 (10.7–108.4)*	17.5 (6.8–97.1)/lesion**	0.05	20.5 (8.2−90.99)	0.90	23.56 (8.2−90.99)	0.41
Prescription isodose volume, cm3	0.44 (0.06–13.84)	0.14 (0.02–46.92)	0.00	0.24 (0.05−25.1)	0.03	0.3 (0.05−25.1)	0.22
50% isodose volume, cm3	1.2 (0.2–39.48)	0.57 (0.13–131.87)	0.00	0.9 (0.2−70.9)	0.24	1.75 (0.2−70.9)	0.98
Conformity index	2.52 (1.27–11.76)	2.53 (1.22–42.1)	0.97	2.32 (1.31−16.32)	0.34	2.15 (1.31−12.92)	0.06
Gradient index	2.91 (2.48–4.08)	3.21 (2.24–6.95)	0.00	3.20 (2.24−5.96)	0.00	3.17 (2.58−5.96)	0.01

Abbreviations: Dmax, maximum dose; Dmin, minimum dose; Gy, gray

Statistical comparisons were made with Fisher’s exact test for gamma angle, lesion side, and lesion location and with Wilcoxon rank-sum test for all other characteristics.

^*The mean treatment time for 51 lesions was 26.4 min; for the 50 lesions treated in 3-pin sessions, 26.9 min.

^**The median (range) of total treatment time for the 4-pin lesions treated in a single session was 119.5 (33.3–275.1) min.

The 3-pin and 4-pin lesions were significantly different (P<0.05) in terms of D_P, IDL, D_max, minimum dose (D_min), V_T, prescription isodose volume, V_0.5*d (volume covered by the 50% isodose surface), and GI. When matched for location, the only difference between the 3-pin lesions and the 4-pin lesions was in the GI (P=0.005) (Table 2).

Tumor Control

Thirty-three of the 50 patients (66%) had distant failures, some of which were subsequently treated with more Gamma Knife sessions. The median time to distant failure was 3.0 months (range 1.0−14.2 months, mean 4.3 months). Of all 328 lesions, 7 were considered to have failed locally (according to size criteria), in 5 patients (one patient had 3 local failures), all of which had been treated as 4-pin lesions (Table 3). Radiologically, these lesions were not considered treatment failures but rather were interpreted as stable (n=1), leptomeningeal involvement (n=2), hemorrhage (n=1), or hematoma (n=3, all in 1 patient). No treated lesions examined pathologically after resection or biopsy were found to be local failures.

Table 3.

Seven local failures according to size criteria*

Patient ID	Sex	Time to Local Failure, months	Radiologist Interpretation	KPS	GPA	No. of Lesions	Primary	Age, years	Location	Dose, Gy	Volume, mm3	Size-Matched
1	M	4.6	Stable	100	2.0	18	Lung	44.2	Left Frontal	20	26.0	—
2	F	17.4	Leptomeningeal	100	1.0	6	Melanoma	62.4	Right Temporal	20	1390.0	Yes
3	F	18.2	Leptomeningeal	80	1.5	8	Neuroendocrine	78.2	Brainstem	16	29.6	—
4	M	14.9	Hemorrhage	90	2.5	2	Renal	82.9	Left Cerebellum	20	84.2	Yes
5	M	3.4	Hematomas†	80	1.5	11	Melanoma	47.9	†	20	†	—

Abbreviations: GPA, graded prognostic assessment; Gy, gray; KPS, Karnofsky performance status

^*No lesions had been location-matched; all had been treated with 4 pins.

^†Patient 5 had 3 hematomas, 2 in the right parietal lobe (one 53.0 mm³ and one 457.0 mm³) and 1 in the right frontal lobe (476.2 mm³).

Toxicity

Of the 49 patients with follow-up visits in this study, 6 experienced 6 different toxic effects: 1 patient had 2 toxic effects (nausea and headache), and the other 5 had 1 toxic effect each (weakness, nausea, paresthesia/numbness, dysarthria, intracranial hemorrhage). All toxic effects were CTCAE grade 1, except for 1 patient who had grade 2 weakness at 1 month after GK-SRS that resolved with steroids. Four of these 6 patients had been reported in our previous paper²; those toxic effects were not considered to be related to GK-SRS. The other 2 patients were from the new cohort: 1 had intracranial hemorrhage at 1 month after GK-SRS that occurred at two sites treated with 4 pins. The hemorrhages at both sites resolved within 6 months. The other patient developed left-sided numbness in the lower face at 1 month after GK-SRS; this patient had had 3 lesions treated, 2 with 4 pins (1 [0.1 cm³] in the left frontal lobe and 1 [0.6 cm³] in the left ventricle) and 1 with 3 pins (in the right occipital lobe [0.05 cm³]). The numbness disappeared within 1 month.

In addition to these 6 patients, 1 patient had a grade 2 partial seizure, which was reported in our previous paper,[5] that was attributed to a 4-pin lesion. No other patients had seizures related to GK-SRS. Another 21 patients suffered grade ≤ 2 toxic effects that were deemed unrelated to GK-SRS: nausea (n=3), vomiting (n=2), loss of appetite (n=4), hearing loss (n=1), dizziness (n=6), neuralgia (n=1), paresthesia (n=8), headache (n=7), weakness (n=1), fatigue (12), depression (n=3), intracranial hemorrhage (n=2), and memory loss (n=2). For these patients, symptoms had been present before GK-SRS and either did not worsen or were attributed to disease progression or other treatments (e.g., whole-brain irradiation, surgery, chemotherapy). No patients had hypersomnia or falls related to GK-SRS.

Discussion

Our results strengthen our previous findings [5] that receiving GK-SRS to lesions that have had 1 of the 4 pins removed does not affect local control or toxicity. Only 1 instance of grade 2 toxicity was attributable to GK-SRS, and this resulted from treatment of a control (4-pin) lesion. The 7 instances initially considered to be local failures (according to size criteria) were all redefined upon further radiologic analysis. We conclude that removing 1 anterior pin is a safe and reasonable alternative to frame repositioning or cancellation of treatment, both of which may require additional reimaging, re-planning, time, and discomfort for the patient. In most cases, accurate target positioning can be achieved after removal of a single pin or post. However, extra care should be taken to minimize torque, as high torque after removal of a posterior pin has been found to result in up to 1.2 mm error.[7] Anterior pin removal has been shown to cause less deviation from control, especially under low and moderate torque, ranging from 0.07 to 0.34 mm.[7] To mitigate torque forces and deviation of target coordinates, pin removal in our study was done after the patient’s headframe had already been docked to the treatment couch. Notably, our study did not find any anteriorly located peripheral lesions with (posterior) pin removal.

Before lesions were matched for size or location, the dosimetric characteristics of the 3-pin and 4-pin lesions were significantly different in all parameters studied except for CI and number of shots. However, with only one exception, the 3-pin lesions were located in the occipital or cerebellar lobes. Previous analyses have shown the incidence of potential collision points to be 32% in the frontal lobe and 57% in the occipital lobes. With optimized calculation, single lesions located laterally or inferiorly can be treated with a single placement of the headframe.[8]

When 3-pin and 4-pin lesions were matched for location, the difference in gamma angle was no longer significant, indicating that a gamma angle of 110° is used more often for lesions in the occipital and cerebellar lobes. Figures 1 and 2 show that many of the 4-pin lesions were quite small, requiring a higher prescribed IDL to obtain reasonable conformity. Under normal circumstances, smaller tumors would be associated with lower prescription isodose volumes and lower 50% isodose volumes, and a higher IDL would result in lower Dmax and higher GI.[9] Collectively, this partially explains the significant differences in dosimetric characteristics between the 3-pin and 4-pin lesions. When using the lesions matched in terms of location or size, only the GI remained different, probably because nearly all the 3-pin lesions were close to the skull, thereby limiting the volume of the 50% isodose line, and resulting in a lower GI.

Figure 1.

Prescription isodose line as a function of tumor volume for 3-pin and 4-pin lesions. Many 4-pin lesions were so small that they required higher isodose lines, thus resulting in a higher gradient index.

Figure 2.

Comparison of tumor volumes between the lesions treated with 3 pins and those treated with 4 pins.

This study had several inherent limitations owing to its retrospective nature. Both the patient population and the tumor characteristics were heterogeneous, and the brief median follow-up time (7.0 months) limits our ability to assess long-term tumor control. Although the treatments were all delivered with an Elekta Gamma Knife Perfexion system, differences between frames and stereotactic modalities should be considered. Our results are strengthened by our matching of the 3-pin lesions with 4-pin lesions treated during the same session and in the same patient, along with separate analyses in which lesions were matched for location and size.

Use of Elekta’s Leksell Gamma Knife Icon system, which features on-board imaging capability, allows the use of cone-beam CT to verify the geometric integrity of treatment after pin removal, which was not an option with the Perfexion system. Stieler et al. recently validated frame-based positional accuracy with cone-beam CT when neither patients nor phantoms had been docked before the pin was removed.[10] In that study, 4 patients were treated with the Icon system with 3 posts; 2 of those patients required new planning CTs and treatment plans with 3-pin geometry because of unacceptable deviation between planning CT and pretreatment cone-beam CT. In this workflow, with these deviations, these investigators recommended either that cone-beam CT be used for pretreatment position verification after pin removal or that the planning CT be repeated using 3 posts. However, with one patient, a planning CT obtained with 3 posts did not lead to reproducible patient positioning, and thus a new planning cone-beam CT was still required with the patient in the treatment position. In that study, no verification cone-beam CT was acquired immediately before the pin removal, and therefore the source of the deviation could not be further isolated. If the deviation between the planning CT and the 4-post cone-beam CT is already large (which is possible if the frame is not sufficiently secured even with 4 pins), then the difference between the planning CT and the 3-post cone-beam CT cannot be attributed entirely to post removal. In our study, we sought to minimize deviations by docking the patient before pin removal. How well this approach reduces deviations is a topic requiring further investigation.

It should be emphasized that repositioning the headframe to avoid collisions is still the most precise and accurate option, although it requires additional imaging, planning, and treatment time. Treated lesions in this study were cortical and thus were not near structures critically sensitive to radiation. The acceptable level of accuracy for brain lesions should be determined by the treating physician, considering the aforementioned factors and patient tolerance.

Conclusions

In some situations, removal of a pin from the stereotactic headframe during GK-SRS is preferable to reframing to avoid collisions between the post and the collimator. Our results strengthen previous findings that treatment of brain metastases using only 3 pins does not affect toxicity or local control.