Impact of body-mass factors on setup displacement during pelvic irradiation in patients with lower abdominal cancer

Cancers in the lower abdomen, such as prostate, rectal, and gynecological cancers, are common malignancies worldwide.1 Pelvic irradiation is frequently used in the treatment of these patients. However, acute or chronic gastrointestinal or genitourinary toxicities might jeopardize the treatment compliance and quality of life in some patients. As a modern technique such as intensity-modulated radiation therapy (IMRT) is capable of dose painting and has been implemented to deliver tumoricidal doses to the target volume while sparing the adjacent normal tissues2, setup accuracy is more critical to minimize deviation from the planned target. Currently, treatment alignment is carried out by lining up skin markers with an equipped laser system. In some circumstances however, the effectiveness of skin alignment might be offset because the exact external position does not always match the internal anatomy accurately. The uncertainties, leading to inadequate dosage to the tumors or untoward toxicities, can be attributed to setup errors or organ motion.

Image-guided radiation therapy (IGRT) using kilo-voltage imaging, and cone beam computed tomography (CBCT) have been widely applied to quantify geometrical uncertainties for daily treatment setup3, 4; however, they are not feasible for widespread use due to the increasing treatment time, cost, and daily dose to the patients5, the technique and frequency of using IGRT should be adjusted based on the clinical conditions. In some developing countries, not all cancer patients requiring radiotherapy are able to receive adequate treatment.6 Particularly, patients who can undergo weekly or daily IGRT were limited even in some institutes where patient load was huge.7 In Europe, IGRT was available in only 49% of all linear accelerators.8 Therefore, tailored use of IGRT for patients with a high risk of setup displacement is an important issue, particularly in countries or institutions where IGRT resources are limited.

Many studies have reported that greater margins are required for obese patients due to higher setup uncertainties.3, 4, 9, 10, 11 However, most studies investigated only the relationship between body mass index (BMI) and the magnitude of setup errors. The impact of patient-related parameters or body-mass factors (BMF) on setup displacement in patients receiving pelvic irradiation remains to be clarified. We hypothesized that the uncertainties can be scored according to the BMFs. Therefore, this study investigated the effect of BMFs on the magnitude of setup displacement during pelvic radiotherapy. As a result, patients with high-risk features or those who requiring large margins between the planning target volume (PTV) and clinical target volume (CTV) can be determined.

In the training cohort, a total of 1976 setup images including the CBCT or OBI were analyzed. As listed in Table 2, the population SE / REs were 1.1 / 2.6 mm, 1.1 /2.0 mm, and 1.9 /5.0 mm in the SI, AP and ML directions, respectively. The SEs and RE of CR were 0.23 and 0.44 degrees. According to Van Herk’s formula19, 20, the suggested CTV- PTV margins for minimizing setup uncertainties were 4.5, 4.0 and 8.1 mm in the AP, ML and SI directions, respectively.

As shown in Figure 2, a linear relationship existed between the individual setup errors and certain BMFs, especially between ML-SE and umbilical AP diameter and between ML-RE and umbilical AP diameter (Coefficient: 0.536 and 0.604, respectively). Table 3 shows the results of univariate and multivariate analyses of the binary logistic regression in the training cohort. Female gender was associated with increasing uncertainties of ML-SE

Figure 2

only in univariate analysis. We found that a large HLD correlated with a greater SI-SE and AP-SE (P = 0.036 and 0.044), whereas a large UAPD correlated with a greater ML-SE and ML-RE (P = 0.021 and 0.001). In addition, a higher SI-RE was associated with a large HC (P = 0.008). Furthermore, patients without previous surgery were vulnerable to a greater CR-RE (P = 0.003).

To differentiate the risk groups, the ROC curves showed the optimal cutoffs of the BMFs in predicting the setup errors as illustrated in Figure 3. The values were 36.5 cm for HLD and 102.3 cm for HC in the SI direction, 34.5 cm for HLD in the AP direction, and 22.1cm for UAPD in the ML direction. A scoring system to stratify the risk groups was proposed according to the scores of these predictors. In the SI direction, the two BMFs (HLD and HC) were utilized to score the risk of setup errors. Each positive predictor scored one point and accordingly patients were dichotomized into groups at high risk and low risk (0 versus 1-2 points) for setup errors. In the AP and ML direction, patients were grouped according to the HLD and UAPD, respectively. Based on the scores, the required

Figure 3

PTV-CTV margin for the SI direction in the high-and low-risk groups were 5.4 mm and 3.8 mm, whereas those for the ML direction were 8.2 mm and 4.2 mm, respectively (Table 4).

In the validation cohort, a total of 959 setup images were retrieved. There was no difference between the training and validation cohorts regarding gender or BMI (gender 1:1, median BMI 25.3). The population SE / REs were 1.0 /1.6 mm, 1.2 /2.4 mm, and 1.6 / 2.8 mm in the SI, AP, and ML directions, respectively. As listed in Table 5, a similar trend of a greater population RE and required PTV-CTV margins could be found when using the same scoring criteria to classify the low- and high-risk groups.

This is the first study to report the impact of image-derived BMFs and other patient-related parameters to score the magnitude of setup displacement during pelvic radiotherapy in patients with lower abdominal cancers. Our results disclosed that certain BMFs have a significant effect on setup errors in specific translational directions. The displacement in the SI direction was greater in patients with higher HC and HLD. A higher HLD and UAPD were associated with greater shifts in the AP and ML directions, respectively. Furthermore, a scoring system for the high-risk group was proposed and validated.

Wong et al. investigated the correlation between BMI and daily setup deviation in 117 patients who received IGRT for prostate cancer.9 They reported that setup shifts greater than 10 mm in the ML direction increase significantly as the BMI increases, with a 1.3% shift for those with normal body weight to a 21.2% shift for those with severe obesity. Strong correlations were found between some BMFs such as subcutaneous adipose-tissue thickness, BMI, body weight, and the SDs of daily displacements in the ML direction. Lin et al. demonstrated that mean absolute shifts in three translational directions positively correlated with the BMI in 30 patients with endometrial cancer treated with adjuvant pelvic IMRT10. Kim et al. revealed that the mean shifts in the ML direction were 0.9 mm for those with a BMI ≥ 30 and 0.1 mm for those with a BMI <30 (P = 0.02)3. In addition, Bray et al. revealed that obese patients had larger mean displacements and REs in the ML direction.4 Undoubtedly, some BMFs have a great impact on setup uncertainties. However, a scoring system is required to identify high-risk patients for daily IGRT or to employ a large PVT-CTV margin.

In this study, the required CTV-PTV margins for all populations in the SI, AP, and ML directions were 4.5, 4.0 and 8.1 mm, respectively. The greatest setup uncertainties were present in the ML direction, similar to previous studies.3, 4 Although daily IGRT could reduce setup variations in patients receiving pelvic irradiation3, 4, it is not always accessible due to limited facilities in some institutions as well as concerns about the increased daily dose to patients.8 Based on our scores, we could adapt the required PTV-CTV margins (5.4 mm for SI and 8.2 mm for ML) for patients with high risk features. Certainly, the clinical validity of the scoring system needs to be verified by external validation.

Laaksomaa et al.21 investigated the influence of gender on setup uncertainties in patients with pelvic cancers and found larger SEs and REs in women. As a result, women required greater PTV-CTV margins in the three translational directions. They also suggested that the difference in the amount of subcutaneous fat between sexes might contribute to this difference. In multivariate analysis in our study however, female gender did not impact the setup uncertainty. The discrepancy could be attributed to the fact that various distributions of accumulated adipose had been included in the BMF analyses, and consequently the impact of gender was diluted.

In several studies, the setup uncertainties were larger in obese patients despite the use of immobilization devices.3, 4, 9, 10, 11 Particularly, obesity has a negative influence on toxicity for prostate cancer patients treated with 3-dimensional radiotherapy without IGRT.22 Therefore, for prostate cancer patients who cannot be managed with IGRT or surgical treatment, a sophisticated guidance for PTV-CTV margin to reduce setup uncertainty during radiotherapy is required. Currently, obesity is usually determined by BMI alone. However, there are two kinds of obesity, the central and peripheral types, depending on the area of fat accumulation. The BMI is not able to distinguish entirely central obesity from the peripheral type.23 Based on the external surface markers on the belly, the type of obesity might influence the setup errors because the skin folds would be more movable in central obesity. To overcome this limitation, this study retrieved the UC, HC, and diameters of in the AP and lateral directions from the simulation CT, which could include the effects of different types of obesity. Thus, our data evidenced that the abdominal or hip circumferences and diameters are more effective in predicting greater setup uncertainties compared with the BMI.

This study was subject to several limitations. First, the circumferences and diameters of the patients were collected retrospectively from CT images instead of direct measurement of the girdle of the bodies. Although the mean deviation between the two methods was less than 5% according to a previous comparison test, the concordance of the two approaches should be assessed further. Second, the strength of the validation test was limited because of the small sample size. However, a trend of a greater RE in the high-risk group could be found among the three translational directions. Finally, organ motion or tumor regression may affect daily treatment accuracy, and the values across various cancers might be different. Our study did not explore the impact of these two factors through daily CBCT, as well as weekly dosimetric changes. Future studies should enroll patients prospectively and evaluate subsequent dosimetric changes according to evolution of the BMFs. Furthermore, external validation is needed to facilitate widespread utility of the scoring system.

Several BMFs including the HLD, HC, and UAPD are associated with greater setup uncertainties in patients receiving pelvic irradiation for lower abdominal cancers. Based on the scores, IGRT can be suggested for patients with high risk features, or required PTV margins could be adapted for patients who cannot be managed with IGRT.