The present study found a non-negligible incidence of SA in 13 of 119 patients (10.9%) at a mean time interval of 7 weeks (50.5 days) after surgery. Recovery was observed in 8 of the 13 patients 16 weeks (108.6 days) later, in keeping with other observations. Lotze et al. reported that serratus anterior palsy occurred in 30% of patients undergoing axillary dissection immediately after surgery, but returned to normal in all patients up to 6 months after the intervention . de Oliveira et al. reported that the post-operative incidence of SA was 73.3% immediately after axillary lymphadenectomy, 65.6% after 90 days and 27.7% at the end of follow-up (416 days) . Meininger et al. reported that most cases of SA resolved within six to nine months . In Martin and Fish’s review, most cases of isolated serratus anterior palsy resolved with conservative treatment within one to twenty-four months .
In line with other authors who compared the SA incidence according to ALND or SNB , ALND was one of the most significant risk factor of SA, with SA observed in 10 of 47 (21.3%) ALND patients, as compared with 3 of 72 (4.2%) SNB patients (Table 1). Even though the long thoracic nerve is identified and preserved during axillary dissection [33–36], a higher risk of damage than with sentinel nodes biopsy can be expected to occur. We note that using a logistic regression model that includes age, body mass index, and ALND (Table 2), the expected risk of SA would range from 0.4% in the lowest risk group (older overweight patients treated with SNB), to 63% in the highest risk group (younger leaner patients treated with ALND). This suggests that the variability of SA incidence might be explained, at least in part, by the heterogeneity of populations.
The relationship of lean body weight with increased risk of SA, or conversely the apparent decreased risk of SA with large body weight, is intriguing. We searched the literature on scapular winging of all causes, but found no direct mention of any relationship between SA and weight or BMI. However, in 25 papers that we found reporting pictures of patients, counting multiple photographs of the same patient as only one to avoid duplicated counting, we identified 47 distinct cases: all were lean or average body frame patients, there was no photograph of any overweight case [1, 2, 10, 32, 37–57]. The published cases lend support to our observation that weight is inversely related with SA. A tentative explanation is that lean patients might be more at risk of nerve and muscle injury than overweight patients, as there would be less axillary room and fat to move around to spare the long thoracic nerve, and a higher risk of indirect damage by vascular disruption, scarring, or compression against the chest wall. An alternative plausible explanation is that SA is more readily overlooked in overweight patients, in whom positional changes of the scapula would be masked by the overlying adipose tissues. If that is the case, then the true incidence of SA might have been underestimated. We note that in our one patient who had onset of post-RT SA, we found no hint to attribute SA to surgery or to radiotherapy. She was 45 years old, had breast-conserving surgery, sentinel nodes biopsy without ALND, irradiation to the breast without regional node irradiation. But, between the pre-RT assessment and the post-RT assessment, she experienced a weight loss of 10 kg, from a pre-RT weight of 67 kg, her BMI dropped from 24.9 kg/m2 before radiotherapy, to 21.2 kg/m2 thereafter. Incidentally, we found only one case report of SA occurring early after radiotherapy . Our patient would represent the second case so reported to the literature.
Our analyses found that younger age was a significant risk factor for SA. The literature provides scarce and contradictory data regarding age and the incidence of SA after breast cancer surgery. In Pereira et al.’s series of patients, the mean age was 60.3 years, but the relationship of SA with age was not investigated . Contrarily to our observation, Ribeiro et al. reported in an abstract that age >60 years by logistic regression was associated with an increased SA relative risk of 3.14 . Crude figures were not provided, hence the consistency of Ribeiro et al.’s logistic regression with data could not be ascertained, whereas our logistic regression was concordant with our raw data. de Oliveira et al. found no significant association of SA with age or any other characteristic . However, in de Oliveira et al.’s report, at mean follow-up of 416 days, the relative risk of SA for age >65 vs. age <65 years was 0.53 (95% CI 0.26-1.07), P = 0.06, concordantly with our results. There is no obvious explanation why young age would be a risk factor. We can only remark that outside the context of breast cancer, reports of conditions related to scapular winging appear with regard to young and active patients [49, 51]. The largest case series of serratus anterior paralysis reported for 197 patients with a mean age of 31.6 years . The literature that we browsed in the discussion about weight was also striking by the preponderance of young patients close to that age.
Regarding the relationship of SA with shoulder/arm morbidity, we encountered two particular issues. One issue is contralateral shoulder/arm morbidity, which we recently found was correlated with ipsilateral morbidity . The other issue was the different measurement scales using different units. We implemented the percentage change of measurement that occurred over time on the ipsilateral limb, therefore avoiding the need to rely on measurements of the contralateral limb, providing the same scale to the measurements, and further allowing links with the common terminology criteria for adverse events .
Some discrepancies could be noted in the relationship between SA and shoulder/arm morbidities, such as improved anteflexion and improved endorotation, albeit non-significant (Table 3). Yet, the overall results indicate that SA might be an important early indicator of higher risk of shoulder/arm morbidity. As shown in Table 3 and Figure 3, patients with SA prior to RT (that is, on average seven weeks after surgery) presented more frequently with altered shoulder/arm assessment. Interestingly, the Brown’s combined test which takes into account the correlation between outcomes was significant. This matches the clinical interpretation of shoulder/arm assessments: while each measurement considered separately might show only small alterations, taken all together the measurements might indicate more substantial risk of morbidity, notably lymphedema or loss of motion (Table 4 and Figure 4).
We are aware of the limitations of the present study. No physical assessment was done prior to surgery, precluding the possibility of analyzing the impact of pre-existing morbidities. The number of patients was small, which did not allow comprehensive analyses, modeling gave results difficult to interpret (Appendix 6 in Additional file 1), therefore limiting the scope of the present study to a descriptive stance. The follow-up was short. Though SA appeared as a predictor of early shoulder/arm toxicities, its value as a predictor of long-term toxicities remains unknown. We did not assess compliance of patients with preventive physical therapy. We did not assess the reproducibility of measurements. It has been argued there is no consistent evidence that any examination procedure used in shoulder assessments has acceptable levels of reliability . Contrariwise, assessment of scapular positioning and winging has been reported to be reliable [62, 63]. In order to evaluate inter-observer variability, the present study could have benefited from repeated assessment by different observers. This was not built into the trial’s design in view of the trial’s time constraints and examinations that patients underwent. Until the present study, we had no a priori reason to give precedence to SA assessment. The study could also have benefited from advanced scapular motion tracking and from electromyographic confirmation of serratus palsy . But, for the same reason that multi-observer assessments were not done, there was no a priori indication to perform motion tracking or electromyography.
The strengths of the study are its prospective nature, the patients were consistently evaluated clinically by the same team within the same institution throughout the study duration. Good internal consistency of measurements done by the same observer could be expected [13, 62]. The physical therapy assessment was blinded to patients’ randomization allocation. Furthermore, the assessors were not involved in the physical therapeutic management of the patients. We have mentioned as a limitation that compliance was not assessed. Yet, this concurred to strengthen the study against bias that could have resulted from knowing patients’ treatments. We believe that the results are robust and warrant further investigations.