Effectiveness and toxicity of conventional radiotherapy treatment for painful spinal metastases: a detailed course of side effects after opposing fields versus a single posterior field technique

Patients with metastases in the cervical spine were excluded from the DBMS [1, 15]. In total, 348 patients were treated for painful metastases in the thoracic, lumbar, or sacral spine. Data on spinal location and treatment technique were available in 343 patients (99%). For spinal location, the treated level of the spine (i.e., thoracic, lumbar, or sacral) was registered, without specification of the specific vertebra or vertebrae irradiated. Treatment was performed with conventional treatment techniques, using either a PA field or APPA fields contributing each 50% of the total dose; no other techniques were used. The prescription depth for PA fields was typically at 4 to 6 cm, depending on the depth of the vertebra. The choice for treatment technique was left to the decision of the treating radiation oncologist and was mostly dependent on institutional policy.

At randomization and during follow-up, patients filled out 13 weekly questionnaires and monthly afterwards until 2 years of follow-up, death, or closure of the study in December 1998. The questionnaires consisted, among others, of the Rotterdam Symptom Checklist (RSCL) [16], a pain scale, pain medication intake, and questions about itching and painful skin. No data were available on the use of anti-emetics or anti-diarrhea medication. The following items were studied to determine toxicity: diarrhea, abdominal pain, nausea, and vomiting. These scores were grouped into the variable “abdominal complaints.” Itching and painful skin were grouped into the variable “skin complaints.” All items were rated on a 4-point Likert-type scale, ranging from 1 (no complaints) to 4 (severe complaints). To facilitate interpretation, all sum scores were standardized to the range of 0 (no complaints) to 10 (severe complaints). Besides sum scores, the individual item scores were also studied. As radiotherapy of the lower spine is more likely to affect the bowel, we studied the individual abdominal items for the treated thoraco-lumbar, lumbar, and lumbo-sacral vertebrae separately. Pain was measured using an 11-point numeric rating scale, ranging from 0 (no pain) to 10 (the worst pain imaginable). A pain score of at least 2 was required to enter the study [1].

Pain response was calculated according to international criteria, taking into account changes in pain medication and pain score [17]. No fixed time interval from the date of randomization was applied. A response was calculated if at least two successive follow-up pain scores were available, which was possible in 325 patients (95%).

To compare the categorical variables at baseline, chi-square tests were used. To visualize and compare the course of side effects over time, we used repeated measurement analyses (mixed procedure), a longitudinal data analysis technique. Analyses were also performed adjusted for treatment institute to take into account potential confounding by indication by institutional choice for treatment technique. p values are based on two-sided tests and considered significant if p < 0.05. Figures were created based on the least square means of the repeated measurements.

To assess which baseline variables were predictive for toxicity, the complaints variable was dichotomized into having or not having complaints. For that purpose, we compared the maximal complaint scores 1 to 4 weeks after treatment with the baseline scores. If a score was higher than the baseline score, the patient was considered as having side effects of radiotherapy. The time period of 4 weeks was chosen because by then most side effects would be present.

We applied multivariate logistic regression analyses to relate candidate predictors to toxicity. First, we started with a full model, including all preselected variables. Subsequently, we eliminated the variables by a backward selection process with a threshold p value of 0.20, based on likelihood-ratio test results. The chosen p value of 0.20 intends to limit the loss of information and to select also weaker predictors, although at the cost of including “noise” variables [18]. The preselected baseline variables, based on the literature and clinical experience, were primary tumor (breast, prostate, lung, or other cancer), age (≤ 65 years or > 65 years), gender (male or female), Karnofsky performance status (KPS) [19] (≤ 60, 70–80, or 90–100), pain score (2–4, 5–7, or 8–10), presence of visceral metastases (yes or no), concomitant systemic therapy (yes or no), treatment arm (1 × 8 or 6 × 4 Gy), opioids (yes or no), spinal localization (thoracic, thoraco-lumbar, lumbar, or lumbo-sacral spine), and treatment technique (PA or APPA).

The database was analyzed using IBM SPSS statistics for Windows version 20.0 (IBM Corp., Armonk, NY, USA) and SAS software (version 9.2, SAS Institute Inc., Cary, NC, USA).

In general, patients with spinal metastases did not differ from the entire population of 1157 patients with bone metastases. Table 1 shows baseline characteristics of the study population (n, 343). Primary tumors were mainly breast (42%), prostate (24%), and lung (20%) cancer. The mean age was 65 years (range 32–89 years) and 52% was male. The majority of patients was in good to moderate condition; 71% had a KPS of 70 or higher. The mean pain score at baseline was 6.4 (range 2–10). Visceral metastases were documented in 28% of patients; 55% of patients received concomitant systemic therapy at the time of randomization.

At baseline, the mean scores of the individual complaints items were low, varying from 1 (no complaints) to 2 (minor complaints), on a scale from 1 to 4. The mean sum score of abdominal complaints was 1.3 (range 0–7.5, on a scale from 0 to 10) and the mean sum score of skin complaints was 0.7 (range 0–8.3, on a scale from 0 to 10). No baseline differences in items or sum scores between the two treatment groups were observed.

A preference per treatment institute was noticed for treatment technique. Institute policy and preferences mainly determined the choice of technique, instead of individual patient characteristics.

In total, 241 (74%) of the 325 evaluable patients had a pain response to radiotherapy, with no significant difference between the two treatment techniques (74% each). The pain response rate is comparable to that of the entire DBMS population.

Side effects were minor. In general, patients experienced more abdominal complaints than skin complaints. Four and 8 weeks after treatment, respectively, 264 (77%) and 229 (67%) patients returned questionnaires. Figure 1 shows the course of complaints in the first weeks after treatment. Patients treated with an APPA technique experienced more abdominal complaints compared to patients treated with a PA field. This difference was temporary; abdominal complaints were comparable 5 weeks after treatment. Skin complaints increased minimally over time, irrespective of treatment technique (Fig. 1a). For both techniques, patients receiving multiple fractions experienced more abdominal complaints than patients receiving a single fraction (Fig. 1b). Differences in skin complaints were hardly noticed (Fig. 1c).

In patients treated with both techniques, the course of complaints was similar for all anatomical localizations, although most outspoken for the lumbar spine in patients treated with an APPA technique (Fig. 2). For skin complaints, only the two patients treated at the lumbo-sacral spine with an APPA technique showed a distinctive increase, to a maximum mean score of 3.4 (scale of 0–10).

Studying the separate side effects for all 343 patients, a trend was noticed towards more vomiting and abdominal pain in patients treated with the APPA technique (p = 0.054 and p = 0.053, respectively). Patients treated with the APPA technique had significantly more severe complaints of diarrhea (p = 0.044). No significant difference was noticed for nausea. Studying the abdominal side effects for the lower spine (all patients, excluding those treated on the thoracic spine only), there were statistically significant differences between treatment techniques. For all studied items (nausea, vomiting, abdominal pain, and diarrhea), patients treated with the APPA technique experienced more complaints than patients treated with the PA technique (p values all <0.009).

Table 2 shows the results of the multivariate analyses. Treatment schedule and location were independent predictors for abdominal complaints. Patients treated with a single fraction had a lower risk of abdominal complaints (OR 0.49 (95% CI 0.29–0.81)) compared to multiple fractions. Patients treated at the thoraco-lumbar and lumbar spine had a higher risk of abdominal complaints (OR 2.51 (0.93–6.80) and 2.29 (1.34–3.93), respectively), compared to radiotherapy of the thoracic spine. For skin complaints (Table 3), primary tumor and localization were predictive in multivariate analyses. Patients with lung cancer (OR 2.27 (1.20–4.30) compared to breast cancer) had a higher risk of skin complaints. Patients treated at the lumbo-sacral spine (OR 1.83 (0.52–6.49) compared to radiotherapy of the thoracic spine) had a higher risk, while patients treated at the lumbar spine had a lower risk of skin complaints (OR 0.54 (0.32–0.91)) compared to the thoracic spine.

Treatment technique did not predict for abdominal or skin toxicity after radiotherapy. When studying patients per treatment arm, treatment technique was not significantly associated with abdominal or skin toxicity.