Association of Radiation and Procarbazine Dose With Risk of Colorectal Cancer Among Survivors of Hodgkin Lymphoma

Key Points Question Is there a dose-response association of radiation dose to the large bowel and/or procarbazine dose with subsequent colorectal cancer risk in Hodgkin lymphoma survivors? Findings This nested case-control study of 316 patients who underwent treatment for Hodgkin lymphoma at 5 hospitals in the Netherlands found a linear dose-response association between radiation dose to the large bowel and colorectal cancer risk; the dose-response association became steeper with higher doses of procarbazine. Meaning The findings of this case-control study support the use of colorectal cancer screening for survivors of Hodgkin lymphoma who have been treated with subdiaphragmatic radiation therapy and/or procarbazine; this dose-response association can be used to estimate the risk of colorectal cancer among these patients.

Total number of unique individuals 78 (100.0) 151 (63.4) Patients were eligible for inclusion if they had first been treated for Hodgkin lymphoma between 1965 and 2000, when they were between 15 and 50 years of age, and had survived for at least 5 years after receiving treatment. If radiotherapy to the abdomen or the pelvis was given for reasons other than Hodgkin lymphoma, the patient was ineligible for inclusion in the case-control study. No., number; RT, radiotherapy eTable 2. Distribution of the factors used for matching (sex, year of HL diagnosis, age at HL diagnosis), follow-up time and age at end of follow-up of HL survivors who developed colorectal cancer and matched controls

Matching and selection of controls
Cases and controls were ineligible for inclusion if information from their medical record was incomplete or if radiotherapy to the abdomen or the pelvis was given for reasons other than Hodgkin lymphoma (HL, eTable 1). Controls also had to stay alive and remain free of colorectal cancer for a time interval equal to the interval from date of HL diagnosis to the date of colorectal cancer for the matched case (cut-off). Controls were selected by incidence density sampling: all survivors remained eligible as controls until they either had a case-defining event or were censored. 1 Survivors could be a control for more than one case.
For each case with colorectal cancer (CRC), up to five controls were selected from the cohort. Controls were individually matched to cases on sex, age at HL diagnosis (within one year) and date of HL diagnosis (within three years). Matching criteria were relaxed up to a maximum difference of three years for age at HL diagnosis and four years for date of HL diagnosis to enable selection of controls. Controls were ranked based on the matching criteria from closest (most optimal) control to the furthest (least optimal) control. Insufficient follow-up time was not a reason to exclude a control a priori, as updated information on this variable could be collected from the medical record. For the five most optimal controls for each case, we checked whether they met the eligibility criteria, especially the interval criterion, using data from the medical files. For many cases, there were less than five eligible controls available. To increase power, we kept controls ranked fourth or fifth when they fitted the eligibility criteria and used a variable matching ratio in analysis.

Treatment
All included HL survivors had been treated with radiotherapy and/or chemotherapy, the majority according to European Organization for Research and Treatment of Cancer Lymphoma Group protocols for primary treatment. 2 Briefly, in the 1960s, RT was usually delivered with cobalt-60 and sometimes with orthovoltage; from the 1970s onwards, linear accelerators were used. Individual blocks were used to shield normal tissues where possible. Patients usually received 40 Gray (Gy) when they were treated with RT alone, and 30-36 Gy when they also received chemotherapy, in both cases with 1.5-2.0 Gy fractions. Extended-field RT including mantle, para-aortic (±spleen/splenic hilum) and iliac and inguinal fields, was commonly given until the late 1980s.
Following this, involved-field RT was gradually introduced.

Development of a representative computed tomography library
The Eclipse treatment planning system (TPS) (Version 13.0.28, Varian Medical Systems, Palo Alto, CA) was used to retrospectively reconstruct the radiotherapy fields delivered to cases and controls.
Twelve representative computed tomography (CT) data sets (six female and six male) were chosen from a cohort of 66 recently treated Hodgkin lymphoma (HL) patients (30 female and 36 male), based on body surface separation. These CTs were taken after two or three cycles of chemotherapy, and were therefore closest in time sequence to the measurements taken for radiotherapy planning in the historic cohort. Separation was measured as the distance between the anterior and posterior body surfaces on axial CT slices at the following points: upper neck, lower neck, mid-mediastinum, lower mediastinum, para-aortic and ileocecal. These points match the anatomical landmarks used in radiotherapy dose calculations for the cohort.
Separation at the same anatomical points was extracted from the radiotherapy prescription charts of patients in the cohort where available. The cohort was then split into tertiles, separate for each sex, based on their mid-mediastinal separation and again based on their para-aortic separation.
Representative CTs were chosen from the modern cohort such that they most closely matched the median separation for each tertile of the historic cohort (eTable 3). The chosen CT was reviewed to ensure acceptable image quality, and absence of major anatomical variations or persistent largevolume lymphoma. The final representative CT library contains 12 full-body CT scans: three each for supradiaphragmatic dose-reconstruction for females and males, and three each for subdiaphragmatic dose-reconstruction for females and males.
All the abdominal-pelvic organs were contoured as per published guidelines 3,4 by a clinical oncologist (RS or DC) and counter-checked by the other. Where there was any doubt or disagreement, the contours were checked by a radiologist. The bowel was outlined to include the external bowel wall plus contents.
For each individual in the cohort who was treated with radiotherapy, a representative CT was chosen based on his or her separation at either the para-aortic or mid-mediastinal level.
Individuals treated with both supra-and subdiaphragmatic fields had the representative CT chosen that most closely matched the separation at the para-aortic level, as dose to abdominal-pelvic organs was the primary focus of this study. Individuals treated with only supra-or subdiaphragmatic radiotherapy had the representative CT chosen based on only the mid-mediastinal or para-aortic separation, respectively.  20.9 -25.5 23.20 23.8 a Chosen to ensure that the 'small' and 'medium' representative CTs were not too close in separation. cm, centimeter; CT, computed tomography; n, number

Treatment reconstruction
Treatment planning variables, including prescribed dose, fractionation regime, beam arrangement, beam energy, source-to-skin distance, field size, and field shielding were extracted from the patients' original radiotherapy prescription cards. This information was used in combination with the original simulation films to individually reconstruct the treatments for 72 cases and 131 controls, blinded to whether the individual was a case or control. The relationship of block borders to anatomic landmarks, shown on the simulation films, was used as a guide to reconstruct the field shielding in the treatment planning system. An example of a reconstruction is given in eFigure 1. eFigure 1. An example of (a) a radiotherapy simulation film and (b) reconstructed radiotherapy plan The original simulation films (a) were used to reconstruct the radiotherapy fields on a digitally reconstructed radiograph from the representative CT data set (b), and block borders (black in a, orange in b) matched to anatomic landmarks.
For patients who were treated with a combination of mediastinal and subdiaphragmatic fields, common practice was to include a slip zone between the inferior border of the mediastinal field and the superior border of the subdiaphragmatic field (typically an inverted Y or para-aortic field).
These borders were moved up or down during treatment; the aim of this was to minimize the effect of hot or cold spots at the junction of the two fields. Where this technique was used, this was incorporated into the individual reconstruction.
When treating patients with bulky mediastinal disease, common practice was to start radiotherapy with a wide mediastinal field and then narrow it once the disease started to respond and shrink. Where this was the case, and when the simulation films indicated the borders of the wide and narrow fields, this was also incorporated into the individual reconstruction.
Where simulation films were not available (26 cases and 34 controls, total 232 fields), fields were reconstructed using multiple other sources of information. The field type and size were extracted from the radiotherapy prescription, along with information from any diagrams or dose calculations contained within the prescription. This information was used in combination with the fields used in other patients treated at the same treatment center at a similar time to the individual in question.
For 25 patients, there was no information on fractionation or beam energy for primary and/or relapse treatment, and this was imputed from other individuals of similar size (separation) treated at the same center with the same field type in the same period. For five cases and six controls, there was no information on prescribed dose for primary and/or relapse treatment, this was again imputed from other individuals treated at the same center with the same field types in the same period. For one case and one control, information on radiation dose, field and energy was not sufficient to impute a radiation dose, these two patients were included in the unknown radiation dose category. A sensitivity analysis was done to determine the effect of imputing missing subdiaphragmatic RT doses (three cases and three controls); results are presented in the main article.
Seven patients were treated with orthovoltage fields (30 fields, 3.4% of patients, 2.9% of fields) which we were unable to reconstruct using the Eclipse TPS. However, these fields were typically used as boost to the groins or axilla, for example following megavoltage or cobalt radiotherapy, and as such, the dose to the colon or rectum would have been negligible. These fields were therefore omitted. Patients were treated with between one and 20 fields each (median four); 1027 fields were reconstructed for 203 unique patients (73 cases, 131 controls).

Dose calculation
The anisotropic analytical algorithm (AAA) photon dose calculation model in the Eclipse TPS was used to calculate the dose distributions for patients who received radiotherapy using a linear accelerator (170 patients [83.3%], 919 fields [89.5%]). The pencil beam convolution (PBC) algorithm was used for patients who received radiotherapy using cobalt-60 (23 patients [11.3%], 63 fields [6.1%]). AAA results in more accurate dose distributions compared to PBC, because it accounts for lateral electron transfer, which is especially relevant in heterogeneous and low density tissue (such as lung) 5 . However, as cobalt-60 is a historic treatment, there is no AAA model available to use in combination with a cobalt-60 beam model in the Eclipse TPS. Given that our focus is on doses to the large bowel, which is a more homogeneous and higher density tissue than lung, it is likely that the impact of using PBC rather than AAA is extremely small. All radiotherapy TPS's are unreliable at evaluating out-of-field dose, and therefore low-dose estimates (less than 1Gy) are unlikely to be accurate. 6 This is unlikely to influence our dose-response relationship as the lowest dose category starts at 1.0 Gy, and the reference category was those not treated with radiotherapy or who received less than 1.0 Gy to the affected colon segment. Once the dose from each field had been calculated, the dose distributions from all fields were summed for each individual, and the mean doses to each segment of colon and rectum were extracted: cecum, ascending colon, hepatic flexure, transverse colon, splenic flexure, descending colon and sigmoid colon. The location of these structures in relation to the five main types of radiotherapy fields is shown in eFigure 2. For three cases exact information on the location of the subsequent colon tumor was not available; for these cases (and their corresponding controls) doses were estimated for the right colon (one case, CRC in the right part of colon) or the whole colon (two cases, unknown CRC location).