Adverse Effects of Preoperative Radiation Therapy for Rectal Cancer: Long-Term Follow-Up of the Swedish Rectal Cancer Trial

Irradiation given in addition to surgery to decrease local failure rates is now considered as standard treatment for many patients with rectal cancer.^1,2 In Sweden, as in most other European countries, preoperative radiation therapy is preferred to postoperative because of its more favorable dose-response relationship and a better tolerability.^1-4

The beneficial effects of preoperative irradiation, most importantly, reduced local recurrence and increased survival, are potentially shadowed by the adverse effects, which can be severe, causing the patients significant disability and potentially outweighing the benefits. This is of particular concern in patients with a low risk of local recurrence when treated with surgery alone. Adverse effects of irradiation depend on the radiation dose and the irradiated volumes.⁵ Acute adverse effects of irradiation (acute toxicity) in rectal cancer treatment have been well described and consist mainly of wound healing, gastrointestinal, genitourinary, and neurologic complaints.^2,4-7 Late adverse effects occur in the urinary tract and skin and in the gastrointestinal, vascular, and skeletal systems, but their magnitude has not been thoroughly documented, because few studies have analyzed the outcome after long-term follow-up.^2,4,8 Most evidence so far has come from the Swedish trials, in which a short course of preoperative irradiation, 5 × 5 Gy in 1 week, has been given.

The Swedish Rectal Cancer Trial, conducted in 1987 to 1990, has had a great impact on the use of radiation therapy for rectal cancer because it showed not only lower local failure rates but also improved overall survival among irradiated patients.⁶ Earlier studies of patients participating in the Swedish Rectal Cancer Trial have demonstrated, however, that irradiated patients have more perineal infections in the early postoperative period.⁹ Moreover, significantly impaired bowel function among irradiated patients was found after 5 years of follow-up.¹⁰ This has, together with impaired sexual function, also been reported from the Dutch Total Mesorectal Excision trial, using the same preoperative radiation dose of 5 × 5 Gy.¹¹ After 8 years of follow-up of the Swedish Rectal Cancer Trial, no difference in late morbidity was seen, but those results have only been published in a doctoral thesis.¹²

A 5-year follow-up concerning late adverse effects of preoperative irradiation (5 × 5 Gy) in the Stockholm I and II trials showed that the irradiated group had an increased risk of venous thromboembolism, femoral neck/pelvic fractures, and intestinal obstruction.⁸ In the Stockholm I trial, which ran from 1980 to 1987,¹³ a two-beam technique with the upper beam limits between L2 and L1 was used. The large radiation volume and insufficient shielding may have accounted for the increased postoperative mortality and higher incidence of late bowel obstruction seen in the Stockholm I trial.^8,13

The aim of this study was to analyze the occurrence of adverse effects possibly associated with radiation therapy for rectal cancer, with special focus on late adverse effects. The primary hypothesis was that irradiated patients are at increased risk of hospital admissions because of diseases in organs situated within or adjacent to the irradiated target.

Patients participating in the Swedish Rectal Cancer Trial, which ran from 1987 to 1990, constituted the population for the present analysis. In all, 1,147 patients were randomly assigned to preoperative radiation at a dose of 5 × 5 Gy with surgery the following week or to surgery alone. The target volume included the anal canal, the primary tumor, the mesorectal and presacral nodes, the lymph nodes along the internal iliac artery, the lumbar lymph nodes up to the level of the upper border of the fifth lumbar vertebra, and the lymph nodes at the obturator foramen. The treatment was, however, not individually planned for each patient; rather, the beam limits were defined by anatomic (bony) landmarks. It was delivered with three portals with the patient in a prone position or with a four-portal so-called box technique with the patient lying supine. The cranial beam limits of all beams were at the middle of the fourth lumbar vertebra, the caudal 1 cm below the anal verge, the ventral limits of the lateral beams approximately 3.5 cm in front of the promontory, and the dorsal limits behind the sacrum. The tumor dose was 25 Gy administered in five fractions over 1 week. The treatment was given with 5 to 16 megavoltage photons, except at one small hospital where Cobalt 60 was used.^9,14 Of the 572 patients randomly assigned to preoperative irradiation, 17 patients did not receive any irradiation, 175 patients were treated with the three-portal technique, 332 patients were treated with the four-portal technique, and 48 patients had anterior-posterior beams.⁹

Three hundred sixteen patients who were included in the Swedish Rectal Cancer Trial were also included in the Stockholm II trial. The Stockholm II trial started as a part of the Swedish Rectal Cancer Trial in 1987, but the Stockholm group continued to randomly assign patients until 1993, whereas the Swedish Rectal Cancer Trial was closed in 1990.^8,15

The Stockholm area, together with one other hospital, did not use appropriate shielding of tissues considered not at risk of containing tumor cells as recommended in the Swedish Rectal Cancer Trial protocol. As a result, 181 patients were irradiated with insufficient shielding.¹⁶ These shields covered tissues behind the lumbar vertebrae and parts of the sacrum, tissues (mainly small bowel) lateral to the lumbar vertebrae, and tissues anterior to the anus and lower rectum (mainly urethra and part of vagina), unless the cancer grew anteriorly in the lower rectum.

Through their personal identification numbers, the patients were matched against the Swedish Hospital Discharge Register. This Register includes all hospital admissions in Sweden and records the main and the secondary diagnoses, the date of admission and discharge, and the hospital and department.¹⁷ The Ninth International Classification of Diseases (ICD-9) was used, analyzing diagnoses grouped according to ICD chapter and as specific diagnoses registered as a four-digit code. Admissions for the primary treatment and up to 30 days postoperatively were excluded, as the postoperative morbidity has been evaluated earlier.⁹ Eighteen patients who died postoperatively, of whom 12 patients were in the irradiated group, were not available for analyses.⁹

Early and late hospital admissions were defined as admissions occurring before and after 6 months from the resection of the primary rectal cancer, respectively. The first admission, with a diagnosis according to ICD chapter or a specific diagnosis, was recorded in each patient, both for early and late admissions.

Both the main and the secondary diagnoses were taken into account, and if a patient was given two or more diagnoses at the same admission, all diagnoses were analyzed. The inclusion of the secondary diagnosis was done to minimize the underreporting of an adverse event not registered as a main diagnosis. The reason is that patients with rectal cancer admitted to a hospital are often registered with the cancer diagnosis as the main diagnosis, even though some other disease registered as a secondary diagnosis caused the admission.

Admissions with rectal cancer as the only diagnosis and admissions with no diagnosis registered in the Hospital Discharge Register were included in the overall analysis. The specific causes of admissions in these patients could not be ascertained from the data in the register, so these admissions were not possible to include in the analysis of diagnosis according to ICD chapters or specific diagnosis.

To avoid confounding from diagnoses related to the presence of cancer, admissions for noncuratively treated patients and admissions during the 3 months before the diagnosis of a local recurrence or metastasis were excluded. Admissions owing to secondary malignancies were also excluded, because they have been analyzed separately.¹⁸ The ethics committee of the Uppsala University approved the study.

Data are presented as medians with 5 and 95 percentiles. The nonparametric Mann-Whitney U test was used to calculate the significance of differences in continuous variables, whereas the χ² test was applied in cases of dichotomous response parameters and to test differences in proportions between groups. Differences in the time to admission between preoperatively irradiated and surgery-alone patients were tested for significance by the log-rank method. Relative risk (RR) analysis, with 95% CIs, was performed with intention-to-treat analysis on time to admission using Cox proportional hazards models, where the calculated hazard ratio was used as the RR estimate. The duration of time to admission was measured from the date of the primary treatment to the first admission or to the censoring times, which were 3 months before a local/distal recurrence, dates of death, or end of the study period (December 31, 2001).

For the 1,147 patients matched to the Swedish Hospital Discharge Register 6,496 admissions were registered. Of these, 999 admissions were excluded as they belonged to the 239 noncuratively treated patients, 1,712 admissions were excluded because of cancer recurrence, and 343 were excluded because of secondary malignancies. In all, more admissions were excluded from the nonirradiated group (Table 1). For admissions excluded because of noncurative surgery or cancer recurrence, a rectal cancer or a metastasis was the only diagnosis registered in 56% of the admissions.

Age and sex did not differ between the treatment groups. Because a larger proportion of the irradiated than the nonirradiated patients survived for more than 5 and 10 years, the number of person-years at risk for hospital admissions was higher in the irradiated group (Table 1). Of the 908 patients analyzed, 661 (73%) were admitted at least once to a hospital after the treatment of the primary rectal cancer. More irradiated patients were admitted to a hospital, both in the early and late postoperative periods (Table 1).

No difference in RR for all admissions was found between the groups (RR = 1.07; 95% CI, 0.91 to 1.26). However, there was an increase in RR for early admissions in irradiated patients (RR = 1.64; 95% CI 1.21 to 2.22), but no difference in the rate of all late admissions was seen between the groups (RR = 0.95; 95% CI, 0.8 to 1.12; Fig 1). The groups did not differ in the median number of admissions per patient or in the duration of hospitalization for each admission (Table 1).

In the analysis of specific diagnoses, no diagnosis was registered for 27 admissions (1%), and the diagnosis of rectal cancer was the only diagnosis registered for 314 admissions (9%). Of these, 215 admissions were seen in the irradiated group, and 126 admissions were seen in the surgery-alone group, but the difference was nonsignificant (P = .09).

An increase in RR among irradiated patients during the first 6 months was observed for infections and gastrointestinal diagnoses, and a trend for an increased RR was seen for endocrine and cardiovascular diagnoses (Table 2). Further analyses of the specific diagnoses did not reveal any differences.

Regarding late admissions, no diagnoses according to ICD chapters showed a significant increase in RR in the irradiated group, but trends were seen for infectious and gastrointestinal diagnoses and for fractures (Table 3; Fig 2). Diagnoses such as endocrine, lung, and psychiatric diseases, with no known data from earlier studies or not reasonably affected by pelvic radiotherapy, were not analyzed further.

Admissions resulting from specific skin, gynecologic, urologic, neurologic, and orthopedic diagnoses were not correlated with radiation therapy. Concerning infectious diagnoses, an increase in RR among irradiated patients was noted for nonspecific infections (n = 10; RR = 8.06; 95% CI, 1.02 to 63.69). In cardiovascular diagnoses, which was the one with most events, the risk of arrhythmia was reduced in the irradiated group (n = 74; RR = 0.57; 95% CI, 0.36 to 0.91).

In the group of gastrointestinal diagnoses, increased RRs in irradiated patients were observed for bowel obstruction (after approximately 8 years), nausea, and unspecific abdominal pain (Table 4; Fig 3), whereas the risk for inguinal hernia was lower among irradiated patients (Table 4).

A tendency toward an increased risk was found for fractures overall (n = 94; RR = 1.45; 95% CI, 0.95 to 2.19; Fig 4). There was, however, no significant differences between the two groups regarding different types of fractures or fractures of bones within or adjacent to the irradiated volume (n = 54; RR = 1.01; 95% CI, 0.59 to 1.72; Fig 4).

When the patients treated with insufficient shielding of radiation beams were compared with those treated with optimal shielding, no statistically significant differences were seen in the occurrence of late complications. However, when the patient groups treated with different radiation techniques were compared, a trend for more bowel obstruction was seen for those treated with anterior-posterior beams (Fig 5).

All analyses were also done including only the main diagnosis registered and also including patients undergoing noncurative surgery; these additional analyses did not change the results.

The present study shows that patients with rectal cancer receiving radiation therapy in addition to surgery have an increased risk to be admitted to hospital during the early postoperative period compared with patients treated with surgery alone. There was, however, no overall increase in late admissions in the irradiated group. These results may thus be reassuring and would seem to rule out substantially increased risks of late toxicity from preoperative radiotherapy of such a magnitude as to result in a hospital admission. Analyses of specific diagnoses, however, showed increased risks of late admissions for gastrointestinal disorders, with small bowel obstruction as the disease likely of most importance. This is in concordance with observations by others,^3,4,8,19 and it is important to analyze this issue in greater detail before the full merits of preoperative radiotherapy can be judged.

Small bowel–related complications are thought to be proportional to the radiation dose and the volume of small bowel within the irradiated volume.^5,20,21 For a given total dose, the use of higher fraction doses results in more damage, both acute and late. If the total dose is lower (eg, 25 Gy using the 5 × 5 Gy schedule as compared with 40 to 50 Gy using daily fractions of 1.8 to 2 Gy for 4 to 6 weeks), the damage may be less. However, the precise relationship between total dose, fraction size, and small bowel toxicity is not known with any great certainty. The timing of the radiation therapy in relation to the surgery may also be of relevance, with higher risks of developing small bowel obstruction,^3,4 diarrhea, and anastomotic stricture³ in postoperatively irradiated patients.

Besides having a proper definition of the tumor target, there are several methods to prevent small bowel irradiation, including multiple-beam techniques, small bowel contrast for planning, a prone position, bladder distension, and an open table-top device.²² These preventive methods may be more applicable for patients irradiated preoperatively, because the small bowel is then generally not adhered in the pelvis, as it may be postoperatively. The efficacy of these methods has not been studied sufficiently, except for the multiple-beam technique, which is the most important of the preventive methods, because it substantially decreases the small bowel volume that receives the same radiation dose as the tumor target.⁵ Patients treated with anterior-posterior beams, actually violating the protocol, have higher postoperative mortality,⁹ and the present study shows a clear trend for an increase in late bowel obstruction compared with patients treated with the multiple-beam technique.

The target volume used in the Swedish Rectal Cancer Trial was larger than that generally used today (eg, used in the Dutch TME trial). If the increase in bowel obstruction seen after approximately 8 years in this study is true, this can thus be anticipated to be smaller in the Dutch TME study,²³ where the upper beam limits were slightly above the promontory, which would mean exposure of a lower volume of small bowel to the irradiation compared with the Swedish Rectal Cancer Trial, where the upper limit was at mid L4.

The increased risk of infectious complications directly after the operation in patients irradiated preoperatively is well known from previous trials and mainly concerns perineal wound infections.^7,9,13 The reasons for this are not clear and are likely to be multifactorial, with contributions by immunologic and wound healing factors.

Cardiovascular diagnoses constituted the individual disease category with most events, reflecting the commonness of these diseases in an older population. Increased risk of deep venous thrombosis or other thromboembolic disorders have been reported from other trials,^8,24 but were not seen in this study. The lower risk of arrhythmias in the irradiated group was the only significant finding, but conclusions from this result should be drawn with caution, as it has not been reported from other studies and likely is an effect of mass significance.

In a study analyzing the risks of adverse effects based on matching between registries, there is always a risk of underestimation or misclassifications of the risks. The unique personal identification numbers and the high completeness of the Swedish registers tell that these risks are limited. The dropout from the Swedish Hospital Discharge Register is only approximately 2%, and the main diagnosis is missing in less than 1% of the hospital admissions,¹⁷ as confirmed by the present study. A wrong ICD coding has been seen in 7% of the ICD chapter and in 14% of the specific diagnoses registered.²⁵ It is likely, however, that this misclassification will be the same in both the radiation and nonradiation groups. In the present study, approximately 9% of the admissions had rectal cancer as the only diagnosis registered; the reason for this could be that a patient was admitted for endoscopy or for an operation for stoma reversal. However, it is also possible that some of those patients were admitted because of complications of the rectal cancer treatment.

Of greater relevance is that this study only concerned diseases or symptoms that led to admission to a hospital. Thus disabling symptoms treated in outpatient clinics such as sexual and urinary tract dysfunction, enteritis, proctitis, and incontinence would not have been detected.^10,19,26,27 Further, the study had a follow-up of 14 years from radiation treatment, so that adverse effects occurring even later would not be discovered. In a previous analysis of the Swedish Rectal Cancer Trial after 8 years of follow-up,¹² no relevant increase in late adverse effects was detected. Since then there have been a few findings, particularly the late increase in bowel obstruction reported here, that requires close observation. In a population with a median age at diagnosis of approximately 70 years, the number of patients at risk beyond 15 years will be limited, diminishing the ability to detect important late effects.

Because the presence of cancer frequently is associated with many symptoms and is a common reason for hospital admissions, patients with residual cancer after the primary treatment were excluded, and patients with recurrent cancer were censored 3 months before the diagnosis of a recurrence. A calculation including patients with residual cancer after the surgery did not alter the relative risks between irradiated and nonirradiated patients, neither during the first 6 months nor later. The most likely reason for this is that few individuals with nonradical surgery became long-term survivors, with less than 20% surviving for more than 5 years, compared with 60% of the curatively treated patients. Not censoring patients with a recurrence would have increased hospital admissions proportionally more in nonirradiated patients, because more recurrences, particularly local ones, 26% versus 9%, were seen in this group.²⁸ Another factor that indicates that the admissions in these patients were related to the cancer disease is that admissions of noncuratively treated patients and patients with cancer recurrence generated in more than half of the cases no other diagnosis than rectal cancer or metastasis, compared with less than 10% of the admissions in patients with no residual disease or recurrence.

In conclusion, gastrointestinal diagnoses seem to be the most common disease category affected by irradiation leading to both early and late admissions, with bowel obstruction as the most important diagnosis. There is a need to further improve the radiation delivery and to evaluate the preventive methods currently in use, and, above all, to identify the patient groups at low risk for local failure after surgery alone to prevent unnecessary radiation.

The authors indicated no potential conflicts of interest.

We thank Niclas Eriksson, statistician, and his coworkers at the Uppsala Clinical Research Centre for their counseling and calculations.