Assessing Axis-specific Set-up Errors in Rectal Cancer Radiotherapy: A Prospective Cone-beam Computed Tomography-Based Study with Body Mass Index Correlation

Abstract

Purpose:

The purpose of the study was to quantify set-up errors and derive optimal clinical target volume to Planning Target Volume (PTV) margins for rectal cancer patients undergoing radiotherapy, while also evaluating the influence of body mass index (BMI) on set-up accuracy.

Materials and Methods:

Data from 41 patients and 1102 daily cone-beam computed tomography (CBCT) scans were analyzed. For each patient and fraction, the mediolateral (X), craniocaudal (Y), and antero-posterior (Z) translational shifts were measured. Population systematic (Σ) and random (σ) errors were calculated from per-patient summary statistics (see Methods), and PTV margins were derived using the van Herk formula (margin = 2.5 Σ +0.7 σ).

Results:

The mean systematic set-up error was small: 0.04 mm (X), 0.57 mm (Y), and 0.13 mm (Z), reflecting high reproducibility of daily image-guided positioning. Using Σ and σ, the derived PTV margins were 7.4 mm (X), 9.1 mm (Y), and 9.7 mm (Z). A moderate positive correlation between BMI and the Z-axis set-up error was observed (r = 0.55, P = 0.002). Overall, 23.1% of fractions required corrections >5 mm, underlining the value of daily CBCT.

Conclusion:

Nonuniform, axis-specific margins are essential to accommodate anatomical and physiological variability in rectal cancer radiotherapy. The use of daily CBCT significantly enhanced set-up precision. Findings align with ICRU 62/83 and QUANTEC recommendations and support individualized planning approaches, especially in diverse patient populations where BMI and pelvic anatomy may affect positioning accuracy.

INTRODUCTION

Colorectal cancer is a major global health burden, ranking as the second leading cause of cancer-related mortality worldwide, with an estimated 1.9 million new cases in 2022.[1] In India, it is the fourth most common malignancy, significantly contributing to the national cancer burden.[2,3] Radiotherapy (RT) remains a cornerstone in the multidisciplinary treatment of rectal cancer, especially in the neoadjuvant setting. However, the success of RT hinges on precise dose delivery, which can be compromised by various geometric uncertainties.[4]

Set-up errors, resulting from patient positioning inaccuracies, mechanical factors, or interoperator variability, represent systematic deviations that persist throughout treatment.[5] In contrast, random errors arise from organ motion or anatomical changes during therapy and vary from fraction to fraction.[6] These deviations, if uncorrected, may lead to underdosing the target or overdosing adjacent healthy tissues.[7,8] To mitigate such risks, the International Commission on Radiation Units and Measurements (ICRU) recommends defining a Planning Target Volume (PTV) by expanding the Clinical Target Volume (CTV) to account for these uncertainties.[8,9,10]

Van Herk et al. proposed a widely accepted formula for margin calculation, ensuring that 90% of patients receive at least 95% of the prescribed dose to the CTV.[11] Adopting nonuniform, axis-specific margins can enhance dose conformity while minimizing exposure to surrounding organs at risk (OAR). Daily image guidance using cone-beam computed tomography (CBCT) further improves set-up reproducibility, particularly for pelvic malignancies like rectal cancer, which are prone to anatomical variability.[12,13,14]

We conducted this study at a regional oncology center to provide population-specific data on set-up reproducibility. Regional differences in body habitus, pelvic morphology, and patterns of disease presentation can affect immobilization and set-up reproducibility; therefore, local evidence may better inform institution-specific margins and image-guidance policies.[15,16,17,18] Therefore, investigating population-specific data can provide insight into whether anthropometric differences, such as body mass index (BMI), influence set-up reproducibility and required planning margins. This study was motivated by the need for localized, evidence-based PTV margin protocols that reflect patient variability in the Kashmiri population.

This study aims to quantify set-up errors and determine optimal CTV-to-PTV margins in rectal cancer radiotherapy at a regional oncology center in Kashmir. The influence of BMI on set-up variation is also analyzed, with the goal of informing individualized margin protocols for enhanced treatment precision.

MATERIALS AND METHODS

Patient selection

This prospective observational study included 41 patients with histologically confirmed rectal cancer treated between October 2020 and January 2022 in the Department of Radiation Oncology. The study received Institutional Ethics Committee approval, and informed consent was obtained from all participants. Treatment modalities included volumetric-modulated arc therapy (VMAT) (n = 26), intensity-modulated radiation therapy (IMRT) (n = 3), and three-dimensional conformal radiation therapy (3DCRT) (n = 1).

Radiotherapy planning and set-up evaluation

CT simulation was performed using a 64-slice Siemens CT scanner. Patients were positioned supine with arms placed on the chest and immobilized using a knee block and foot wedge. A contrast-enhanced planning CT scan (2.5 mm slice thickness) was acquired from the L3–L4 vertebral level to 1.5 cm below the lesser trochanter. Patients followed a standardized bladder and rectum preparation protocol, which included voiding followed by ingestion of 300–400 mL of water 30–45 min before scanning and each treatment session, aiming for a consistently full bladder and empty rectum.

The planning CT datasets were transferred to the ECLIPSE treatment planning system (Varian Medical Systems, Palo Alto, CA). The CTV and OARs were contoured per institutional protocols. A PTV was generated by uniformly expanding the CTV by 7 mm. A uniform 7 mm expansion was initially applied during planning to ensure adequate target coverage. This uniform isotropic expansion served as the baseline institutional practice against which the need for axis-specific margins was evaluated. However, the goal of this study was to evaluate whether such an isotropic margin accurately reflects the true, direction-specific set-up variations observed in our patient population. Radiotherapy was delivered on a Varian True Beam STx linear accelerator. Daily CBCT was used for image guidance through the on-board imaging system. The alignment of CBCT images with planning digitally reconstructed radiographs (DRRs) was assessed in three orthogonal directions, mediolateral (ML, X-axis), craniocaudal (CC, Y-axis), and anteroposterior (AP, Z-axis), using bony landmarks [Figure 1].

Figure 1.

Bony landmarks defining the radiotherapy coordinate axes. Illustration of the three orthogonal translational axes used for set-up evaluation: medio-lateral (X), cranio-caudal (Y), and antero-posterior (Z), defined relative to pelvic bony anatomy and digitally reconstructed radiographs

Cone-beam computed tomography workflow and use in analysis

At each treatment fraction, a pretreatment CBCT was acquired and registered to the planning DRRs using bony landmarks in the ML (X), CC (Y), and AP (Z) directions. The set-up errors analyzed in this study were taken from these initial pre-correction CBCT shifts, as they represent the raw set-up uncertainty prior to any therapist or oncologist adjustment. When a displacement exceeded the institutional action level (>5 mm or >0.5 cm), the patient was repositioned, and a repeat CBCT was acquired for verification. Only the initial, pre-correction CBCT scans were used in the error and margin analysis, as they represent the true set-up uncertainty prior to intervention. The repeat verification CBCTs were excluded from systematic and random error computation. These repeat CBCTs were used only to confirm accurate repositioning and were not included in the calculation of systematic (Σ) and random (σ) errors. In total, 1102 initial CBCT scans were analyzed across all fractions of the 41 patients.

Body mass index assessment

Each patient’s height and weight were recorded at the time of simulation. BMI was calculated using the standard formula, BMI (kg/m²) = weight (kg)/(height [m]) ². Patients were stratified into BMI categories (underweight, normal, overweight, and obese) based on World Health Organization guidelines. These BMI values were used to evaluate their correlation with set-up deviations along each anatomical axis to explore whether body habitus influenced positional reproducibility.

Statistical analysis

For each patient (i = 1…N) with nᵢ treatment fractions, we first calculated the patient mean set-up error (μᵢ) as the average of all shifts across fractions: μᵢ = (1/nᵢ) Σⱼ shiftᵢⱼ. The within-patient random variation (sᵢ) was then calculated as the standard deviation (SD) of shifts around this mean: sᵢ = √([1/(nᵢ − 1)] Σⱼ [shiftᵢⱼ − μᵢ]²). Using these patient-level statistics, the population systematic error (Σ) was defined as the SD of the patient means (μᵢ) across the cohort, that is, Σ = SD (μᵢ) = √([1/(N − 1)] Σᵢ [μᵢ − μ]²), where μ is the grand mean of all patient means. The population random error (σ) was defined as the root mean square of the individual patient SD s, σ = √([1/N] Σᵢ sᵢ²). Margins for CTV to PTV expansion were calculated using the van Herk formula: Margin = 2.5 Σ + 0.7 σ. For example, for the X-axis where Σ = 2.6 mm and σ = 1.3 mm, the calculated margin was (2.5 × 2.6) + (0.7 × 1.3) = 7.41 mm, reported as 7.4 mm in Table 1. To examine the influence of patient body habitus, a correlation between BMI and set-up deviations was performed using Pearson’s correlation between BMI and each patient’s mean error (μᵢ). Spearman’s correlation was used as a nonparametric sensitivity check, and a linear mixed-effects model with patient as a random effect was additionally fitted to account for repeated measures. A two-sided P < 0.05 was considered statistically significant. All analyses were performed using IBM SPSS Statistics version 23 (IBM Corp., Armonk, NY, USA), and R Version X. X for mixed-effects modeling.

Table 1.

Configuration and geometric errors for rectum (mm)

Parameter	Horizontal (x)	Vertical (y)	Longitudinal (z)
Mean error±SD	0.04±0.83	0.57±1.76	0.13±0.18
Systematic error (Σ) (95% CI)	2.6 mm (2.41–2.79)	3.2 mm (2.96–3.44)	3.4 mm (3.15–3.65)
Random error (σ) (95% CI)	1.3 mm (1.20–1.40)	1.6 mm (1.48–1.72)	1.7 mm (1.57–1.83)
CTV to PTV margin	7.4	9.1	9.7

CI: Confidence interval, PTV: Planning target volume, CTV: Clinical target volume, SD: Standard deviation

RESULTS

Configuration errors

The study included 41 patients with rectal cancer (13 men and 28 women; mean age 51.7 ± 14.5 years), as summarized in Table 2. Among them, 27% belonged to the 50–60 year age group. All patients were treated using VMAT, IMRT, or 3DCRT techniques. A total of 1102 precorrection CBCT images were acquired and analyzed to evaluate translational displacements along the ML (x), CC (y), and AP (z) axes. The mean positional deviations across all treatment fractions were minimal, 0.04 mm in the X-direction, 0.57 mm in the Y-direction, and 0.13 mm in the z-direction [Table 1 and Figure 2], indicating high accuracy in daily image-guided set-up corrections. The systematic (Σ) and random (σ) errors derived from patient-level statistics were subsequently used to determine appropriate PTV margins.

Table 2.

Demographic characteristics of patients enrolled for study

Characteristic	Value, n (%)
Age (years), mean±SD	51.7±14.5
Age range (years)	20–80
Age group distribution
20–30	4 (9.8)
30–40	6 (14.6)
40–50	7 (17.0)
50–60	12 (29.3)
60–70	9 (22.0)
70-80	3 (7.3)
Sex (male/female)	13/28
BMI (kg/m2), mean±SD	25.6±4.3
BMI categories
Underweight	3 (7.3)
Normal weight	17 (41.5)
Overweight	13 (31.7)
Obese	8 (19.5)
Treatment technique
VMAT	26 (63)
IMRT	3 (7)
3DCRT	1 (2)

IMRT: Intensity modulated radiation therapy, SD: Standard deviation, BMI: Body mass index, 3DCRT: Three-dimensional conformal radiation therapy, VMAT: Volumetric-modulated arc therapy

Figure 2.

Displacement trends across radiotherapy sessions. Distribution of precorrection set-up displacements for 1102 cone-beam computed tomography fractions along the medio-lateral (x), cranio-caudal (y), and antero-posterior (z) axes. Data are shown as mean ± standard deviation across patients and fractions, reflecting daily set-up reproducibility prior to image-guided correction

Set-up errors and calculated margins

Table 1 summarizes the population mean shift (Mean ± SD), the population systematic error (Σ), random error (σ), and the derived CTV to PTV margins for each axis. The “Mean Error ± SD” values represent the population mean shift and its SD (i.e., the mean of all fraction shifts across the cohort) and are not directly substituted into the van Herk formula. These values reflect the accuracy of daily imaging corrections, whereas the systematic (Σ) and random (σ) errors are the true determinants of margin estimation using the van Herk formula. The van Herk margins were computed from Σ and σ (see Methods). For example, for the x-axis Σ (X) = 2.6 mm and σ (X) = 1.3 mm give Margin (X) = 2.5 × 2.6 + 0.7 × 1.3 = 7.41 mm [reported as 7.4 mm in Table 1]. When we compared our results with previous studies, we concluded that they were in line with them [Figure 3].[19,20,21]

Figure 3.

Comparison of clinical target volume-to-planning target volume margins for rectal cancer with previous studies. Axis-specific margins derived in the present study (X = 7.4 mm, Y = 9.1 mm, Z = 9.7 mm) compared with published values from Yamaguchi et al.,[19] Yang et al.,[20] and Peters et al.[21] The figure illustrates consistency of our anisotropic margins with prior literature, while emphasizing population-specific variability

Out of the 1102 CBCT scans analyzed, 255 scans (23.2%) revealed translational set-up errors exceeding the institutional correction threshold of 0.5 cm in any direction. Among these, 157 instances (14.3% of total scans) required full patient repositioning under oncologist supervision due to displacements <5 mm. The remaining 98 cases (8.9%) were corrected using couch-based translational shifts by radiation therapists. These findings highlight the practical frequency of significant set-up deviations and emphasize the importance of daily image guidance to ensure accurate treatment. Notably, the majority of errors requiring repositioning were observed along the CC (Y) and AP (Z) axes, consistent with the higher systematic and random errors calculated in these directions.

To evaluate the potential influence of body habitus on set-up reproducibility, a correlation analysis was conducted between patient BMI and mean set-up errors across the three orthogonal axes. A moderate positive correlation was observed between BMI and set-up error along the AP (Z) axis (r = 0.55, P = 0.002), suggesting that patients with higher BMI exhibited greater variability in this direction. A weaker correlation was noted along the CC (Y) axis (r = 0.32, P = 0.082), whereas an unexpected moderate inverse correlation was found along the ML (X) axis (r = −0.40, P = 0.028), indicating potential anatomical or positioning compensations in that plane. These results are summarized in Table 3. To visually represent these findings, a scatter plot was generated [Figure 4], illustrating the relationship between BMI and set-up error in the z-direction. The red dashed line denotes the line of best fit, with a corresponding legend box in the upper-right corner of the plot. This graphical representation reinforces the observed trend that higher BMI may contribute to increased set-up uncertainty in pelvic radiotherapy, particularly along the AP axis.

Table 3.

Correlation between body mass index and set-up errors across axes

Axis	Correlation coefficient (r)	P	Interpretation
Medio-lateral (x)	−0.40	0.028	Moderate inverse correlation
Cranio-caudal (y)	0.32	0.082	Weak positive correlation
Antero-posterior (z)	0.55	0.002	Moderate positive correlation

Figure 4.

Correlation between body mass index (BMI) and z-axis set-up error. Scatter plot of patient BMI versus mean antero-posterior (Z-axis) set-up error (per patient, averaged across fractions). Pearson’s correlation analysis showed a moderate positive relationship (r = 0.55, P = 0.002). The red dashed line represents the line of best fit, with shaded area indicating 95% confidence interval. BMI: Body mass index

DISCUSSION

This study quantified axis-specific set-up errors and derived corresponding CTV to PTV margins for rectal cancer patients treated with image-guided conformal radiotherapy at a regional center. The margins, 7.4 mm (X), 9.1 mm (Y), and 9.7 mm (Z) reflect anisotropic variability in patient positioning, particularly in the CC and AP directions. These findings are consistent with those reported in similar,[19,20,21] affirming the reliability and reproducibility of our measurements within an Indian clinical context.

Systematic errors in our cohort ranged from 2.6 mm to 3.4 mm, and random errors from 1.3 mm to 1.7 mm. These values closely align with prior research,[19,20,21] reinforcing the accuracy of our institutional set-up and imaging workflow. We employed the Van Herk formula, a validated model in radiotherapy planning, to convert positional uncertainties into clinically applicable margin recommendations.[11] Notably, 23.1% of treatment fractions required correction for translational deviations exceeding 5 mm, highlighting the practical significance of implementing daily image guidance protocols to maintain treatment accuracy.

A key component of this study was the investigation of body habitus as a contributing factor to set-up variability. Correlation analysis revealed a moderate positive association between BMI and set-up error along the AP (Z) axis (r = 0.55, P = 0.002). This suggests that patients with higher BMI may experience increased soft-tissue motion or displacement in the pelvic region, potentially compromising positional reproducibility. Lesser correlations were found in the CC (Y) axis (r = 0.32), and an unexpected inverse relationship in the ML (X) axis (r = −0.40), which may reflect compensatory alignment mechanisms or anatomical differences. These findings underscore the potential value of BMI-stratified margin planning in pelvic radiotherapy, particularly in populations with wide variability in body composition.

When comparing our underlying error components with prior reports, our systematic errors (2.6–3.4 mm) and random errors (1.3–1.7 mm) are highly consistent with the ranges reported by previous studies. This concordance at the Σ and σ level reinforces the validity of our institutional measurements and supports the generalizability of our results.[19,20,21]

The moderate inverse correlation between BMI and lateral (X) error is unexpected. Possible explanations include greater lateral stability in patients with larger pelvic girth, differences in skin mark alignment, or easier reproduction of bony landmarks in broader pelvises. Future studies should incorporate pelvic width or body composition indices. Clinically, the positive correlation between BMI and AP (Z) variability suggests that patients with higher BMI may benefit from daily CBCT and possibly modestly larger AP margins. Each center should test BMI thresholds (e.g. ≥30 kg/m²) in their own data before implementation.[22]

The use of daily CBCT was central to our methodology. CBCT enabled high-resolution, volumetric verification of patient positioning, allowing detection of subtle translational and rotational deviations throughout the treatment course. Compared to two-dimensional portal imaging, CBCT provides superior soft-tissue visualization, making it particularly valuable in pelvic radiotherapy where interfraction motion due to bladder filling, rectal contents, or diaphragm motion can significantly impact dose delivery.[23,24] The integration of CBCT into our workflow supports the use of tighter margins, which can enhance dose conformity to the target while sparing adjacent healthy structures.

Among the three axes, the Z-axis exhibited the highest variability, likely due to challenges in maintaining consistent anterior reference marks and the influence of variable abdominal wall movement. Factors such as table height, external skin markings, and organ motion (including diaphragm shifts) may contribute to this directional sensitivity. These observations are clinically significant, as underestimating longitudinal set-up error may result in geographic miss of the tumor or overdosing of critical adjacent structures.

While the magnitude of our margins is consistent with ICRU principles that expansions should reflect measured institutional uncertainties, the key novel finding is that anisotropic (axis-specific) margins, rather than isotropic ones, are justified for rectal cancer in our population.[25,26]

Our study also contributes to the limited literature on population-specific radiotherapy margins in the Indian subcontinent, particularly from the Kashmir region, where anatomical, dietary, and genetic factors may influence pelvic anatomy and treatment reproducibility. Given the moderate correlation between BMI and Z-axis error, future studies should consider incorporating pelvic girth or body composition indices to refine margin estimation further.

CONCLUSION

In this study, we determined the CTV to PTV set-up margins for rectal cancer to be 7.4 mm, 9.1 mm, and 9.7 mm along the x, y, and z axes, respectively. These nonuniform margins reflect the anatomical and physiological variability of the rectum and surrounding pelvic structures, ensuring adequate target coverage and consistent dose distribution. The observed correlation between BMI and set-up error, particularly along the Z-axis, suggests that body habitus may further influence positioning accuracy. Our findings align with ICRU Reports 62 and 83 and are consistent with QUANTEC recommendations, reinforcing the reliability and reproducibility of our margin estimates. The use of daily CBCT imaging in our protocol contributed significantly to set-up accuracy, supporting its continued role in image-guided radiotherapy. Overall, these results validate our margin recommendations and underscore the importance of individualized, evidence-based planning strategies for rectal cancer patients in similar clinical settings.

Conflicts of interest

There are no conflicts of interest.

Funding Statement

Nil.