Scapula alata in early breast cancer patients enrolled in a randomized clinical trial of post-surgery short-course image-guided radiotherapy

The study population consisted of women who participated in the TomoBreast clinical trial (NCT00459628, ISRCTN21164902) approved by the University Hospital of Brussels’ ethics board. The trial recruited women aged 18 years or older, presenting with a primary breast carcinoma completely removed by mastectomy or by breast-conserving surgery, pathological stage pT1-3N0M0 or pT1-2N1M0 with pathological nodal status assessed by axillary lymph node dissection (ALND) or by sentinel nodes biopsy (SNB), who were to receive post-surgery radiotherapy. Women who gave written informed consent were allocated to either a control group or to an experimental group by computer randomization. In order to reduce the risk of imbalance due to the small size of the trial, randomization used Efron’s biased coin design: instead of a fixed 1/2 probability, the probability of a new patient being allocated to the control or to the experimental group was assigned as 1/3, as 1/2, or as 2/3, depending on how many preceding patients, stratified by nodal status, type of surgery, and chemotherapy sequence, had been previously allocated in one or the other group [11]. In the control group, a dose of 50 Gy was delivered in 25 fractions over five weeks to the chest wall using tangential photon fields, and in cases of pN1 status, to the supraclavicular, infraclavicular and axillary nodes using an anterior field matched to the tangential fields. Breast-conserved patients received, in addition, a sequential boost of 16 Gy delivered in 8 fractions over two weeks to the initial tumor bed using a direct electron field. In the experimental group, a dose of 42 Gy was delivered in 15 fractions over three weeks to the chest wall in cases of mastectomy, or to the whole breast in cases of breast-conserving surgery, and to the supraclavicular, infraclavicular and axillary nodes in cases of pN1 status, using the image-guided radiotherapy system TomoTherapy (TomoTherapy Inc., Madison, WI, USA). Breast-conserved patients received a simultaneous integrated boost of 9 Gy delivered in 15 fractions over the three weeks.

Per protocol, radiotherapy had to start within six weeks of breast surgery, or, in case of adjuvant chemotherapy, within six weeks after completion of the adjuvant chemotherapy. Quality of life, arm mobility and volume, pulmonary function and heart function tests were scheduled prior to radiotherapy, at one to three months after completion of radiotherapy, then yearly. The primary endpoint of the trial was the combined pulmonary and cardiac toxicities as determined by medical imaging and functional tests during follow-up versus pre-treatment evaluation. The secondary endpoint was locoregional recurrence. Formal comparisons of the endpoints and quality of life between treatment groups are ongoing but are not the purpose of the present study. The focus of the study is scapula alata and the physical therapy assessment made prior to radiotherapy and at the first follow-up one to three months after radiotherapy.

Written informed consent was obtained from the patients for publication of this report and any accompanying images.

Patient’s subjective arm symptoms, physical shoulder-arm evaluation, and presence or not of scapula alata were assessed by a physical therapist after the patient’s consent to participate in the trial but before radiotherapy (pre-RT evaluation) and at one to three months after the last radiotherapy session.

Subjective arm symptoms were recorded as present or absent. Arm symptoms were considered present when the patient reported for the operated arm/hand any symptom of dysesthesia, heaviness, swelling, fatigue, more effort needed, warmth, burning or pain. Arm symptoms were considered absent when the patient reported none of these symptoms.

Shoulder-arm evaluation recorded the following measurements (Figure 1):

Arm volume, computed from circumferential measurements using the mean of the frustum sign and the cylinder model method as detailed in Appendix 1 of Additional file 1 [12].

Maximum range of active lateral elevation of the arm (abduction).

Maximum range of active forward elevation of the arm (anteflexion).

Maximum range of active backward elevation of the arm (retroflexion).

Maximal functional endorotation measured by counting the vertebrae between C7 and the most cranial vertebra the patient could reach with her thumb on her back.

Scapular distance (the lateral scapular slide test), measured as the distance between the spine and the angulus inferior of the scapula, with the arms elevated 90° in the scapular plane [13].

Note that impairment of arm mobility is indicated by decreased abduction, anteflexion, and/or retroflexion. However, impairment of endorotation would be marked by the inability of the hand on the back to reach closer to the neck, with consequently an increased count of vertebrae. Likewise, shoulder injury might entail decreased ability of the scapula to slide toward the spine, with consequently an increased scapular distance [14, 15]. All measurements were made on both arms, ipsilaterally and contralaterally to the operated side. For the present study, we took into consideration only measurements pertaining to the ipsilateral arm.

Scapula alata was assessed through visual observation of tilting and winging of the scapulae (Figure 2). The observation was performed with the subject instructed to stand relaxed and perform active elevation of the arms in the scapular plane until shoulder height. No differentiation in the amount of elevation was specified. The patient was observed from dorsal (frontal plane) and lateral (sagittal plane). Normally, the inferior angle should be flat against the chest wall [16] and the scapulae should be 30° internally rotated with respect to the frontal plane [17]. Scapular positioning was deemed impaired when:

the inferior angle of the scapula became prominent dorsally (rotating about the horizontal axis - tilting;

the entire medial border of the scapula became prominent dorsally (rotation about the vertical axis - winging.

If one or both criteria listed above were fulfilled, we scored scapula alata as 1 (SA present), only if there was a clear observation of the positioning fault. If none of the criteria were met, we judged scapula alata as 0 (no SA). Each position was observed and evaluated once.

In order to evaluate shoulder/arm morbidity on a common scale and to avoid reliance on the contralateral arm measurements, we computed the outcome of a shoulder/arm measurement as the percentage change of the measurement that occurred over time, between pre-RT assessment (= time T0) and post-RT assessment (= time T1) of the ipsilateral arm. That is, for volume, the percent change of volume was computed as: 100 x (volume of arm at T1 – volume of arm at T0)/volume of arm at T0. Likewise, for abduction, the percentage change of abduction was computed as 100 x (abduction at T1 – abduction at T0)/abduction at T0, and so on for retroflexion, anteflexion, endorotation, and scapular distance. The percentage changes of the measurements were analyzed as continuous variables, and were also analyzed as categorized variables. Categorization used cutoffs for limb edema and for motion impairment derived from the Common Terminology Criteria for Adverse Events version 4.0 (CTCAE 4.03) [18]. For arm swelling, the cutoffs applied were Grade 0 = less than 5%, Grade 1 = 5% to less than 10%, Grade 2 = 10% to less than 30%, and Grade 3 = 30% or more increase of arm volume, where percentage increases are computed as defined above, in order to avoid reliance on the contralateral limb [19]. For loss of range of motion, the cutoffs applied were Grade 0 = 5% or less, Grade 1 = more than 5% to 25%, Grade 2 = more than 25% to 50%, and Grade 3 = more than 50% loss of motion. CTCAE 4.03 does not specify a lower bound in the definition of Grade 1 toxicity. We implemented a lower bound of 5% in order to take into account the normal variability of range of motion [20‐25].

Fisher’s exact test was used for the analysis of data categorized in contingency tables [26]. Cochran-Armitage’s trend test was used for ordinal tables [27] (pp 504-509). Odds ratio relating SA with patients’ characteristics were computed by conditional maximum likelihood. Logistic regression was used to evaluate the multivariate association of patients’ characteristics with SA. Significance testing of continuous measurements used Student’s t-test. P-values from one-sided or two-sided tests are indicated as 1P or 2P, respectively. The overall assessment of multiple outcomes used Brown’s method to combine non-independent tests of significance [28].

All statistical computations used R version 2.14.1 [29]. Missing data were imputed using the method of multivariate imputation by chained equations from package ‘mice’ [30]. Variables used for imputation are listed in Appendix 2 of Additional file 1. Fisher’s exact test and odds ratios were computed using the function ‘fisher.test’. Ordinal test of proportions used the function ‘prop.trend.test’. Logistic regression used the function ‘glm’ [31]. Brown’s method for combining non-independent tests of significance [28] was computed using an in-house R script (Additional file 2).