Background: Guidelines recommend primary prophylaxis (PP) with granulocyte colony stimulating factors (G-CSF) for patients above a febrile neutropenia (FN) risk threshold of 20%. When considering the role of PP with G-CSF, practitioners often use FN rates of regimens based on data from randomized controlled trials (RCTs), which are often comprised of highly selected patients. Patients who received these regimens in the real world may be at higher risk of FN. Methods: A systematic literature search was conducted using MEDLINE, EMBASE, and CENTRAL databases for full-length articles reporting FN rates for breast cancer-related chemotherapies between Jan/96 and Feb/14. A regimen was included if there was at least 1 RCT and 1 observational study (ObS). Any pilot, dose-finding, feasibility, phase I, or phase II studies were excluded. Meta-regression using logistic regression with fixed-effects and random-effects was used to model the odds ratio (OR) of FN with 95% confidence intervals (CI). Results: 128 studies involving 29 regimens and 50,069 patients were identified. 65 ObS (n = 7,812) and 110 RCT (n = 42,257) cohorts were included. The unadjusted FN rate was 11.7% in ObS and 7.9% in RCT cohorts. The univariable fixed effects OR for FN in the ObS compared to RCT cohorts was 1.58 (CI: 1.09-2.28; p= 0.017). The FN rates remained significantly higher in the ObS compared to RCT cohorts (OR = 1.74; CI: 1.15-2.62; p =0.012) in the multivariable mixed effects model adjusted for age, chemotherapy intent, chemotherapy regimen and accounted for random effects of study and chemotherapy regimen. This meant that a 13% FN rate in RCT would translate into 20% FN rate in ObS study. The unadjusted FN rates were higher in taxane regimens than non-taxane regimens in both the ObS (16.5% vs. 6.4%) and RCT (10.0% vs. 4.7%) cohorts. Conclusions: FN rates in ObS are significantly higher than suggested by RCTs in many breast cancer-related chemotherapy regimens. A 13% FN rate in RCT appeared to correspond to the 20% FN rates in the real world. Large population-based studies are needed to confirm FN rates of different regimens in the real world to ensure optimal utilization of G-CSF.