Knowledge about long-term consequences of breast cancer treatment on shoulder and arm function and volume in stages I-II breast cancer survivors is limited. The effects of shoulder-arm function shortly after surgery on long-term function are unknown. One hundred and ninety-four women were examined pre-surgery (T0) and 6 weeks after surgery (T1). Of those, 110 were re-examined 7 years later (T2). Thirty-four women underwent sentinel lymph node biopsy (SLNB) and 76 underwent axillary lymph node dissection (ALND). Differences between affected and unaffected side were calculated for four ranges of motion functions, three strength functions and arm volume. These were used to analyse time and group effects. Differences exceeding 20° in range of motion, 20 % in strength and 200 ml in arm volume were considered clinically relevant. Multivariate regression analyses examined the effect of shoulder-arm function at T1 on shoulder-arm function at T2. Additional predictor variables included were age, follow-up time, Body Mass Index, complications, chemotherapy, radiation, SLNB/ALND and type of breast surgery. At T2, range of motion (except external rotation), abduction strength and arm volume were impaired compared to T0. After ALND, women had significantly more forward flexion impairment, increased arm volume and clinically relevant impairments (70 %) than after SLNB (41 %). T1 external rotation, abduction-external rotation, grip strength and arm volume were the strongest predictors of these variables at T2. Age was the strongest predictor of the remaining four variables. ALND predicted arm volume only. Seven years after breast cancer surgery, two-fifth of the women after SLNB and seven out of ten women after ALND had impairments. Impairments were found in five of eight shoulder-arm functions. After SLNB, women have less forward flexion impairment and less arm volume increase than after ALND. Shoulder-arm function at 6 weeks after surgery and age are the strongest predictors of long-term shoulder-arm function.