Table 3 Characteristics of included studies reporting associations between blood biomarkers and prognosis

 Berrino et al. (2005) [20]

107 postmenopausal breast cancer survivors. Mean age: 56.8 years. Patients with normal-, or overweight/obesity. Operated for breast cancer at least a year prior to enrolment. HT: about 38% of women were under tamoxifen treatment during the RCT. Subtypes: ER-, ER+, PR-, PR+, unknown. Stage not reported. Serum samples collected at baseline of the RCT. Follow-up: 5.5 years

Serum testosterone, estradiol, SHBG, glucose, insulin (Radioimmunoassay)

Cox proportional hazards models, adjusted for HR T (size of the primary) and N (axillary node status) with/without tertiles of baseline testosterone ^b

Recurrence (local relapse, distant metastasis, contralateral breast cancer)

Recurrent patients: ↑ testosterone, estradiol, glucose vs. patients without recurrence (0.52 vs. 0.38 ng/ml, p < 0.001; 8.06 vs. 5.52 pg/ml, p = 0.02; 96 vs. 91 mg/dl, p = 0.02, respectively). In tertiles: only ↑ testosterone (3rd tertile) was significantly associated with ↑ RR (HR: 7.19, 95% CI 2.42–21.35, p not shown). No association with RR were observed for estradiol, SHBG, fasting glucose and insulin

 Pasanisi et al. (2006) [21]

110 postmenopausal breast cancer survivors. Mean age: 56.8 years. Patients with normal-, or overweight/obesity, 16 women with MetS. Operated for breast cancer at least a year prior to enrolment. HT: about 38% of women were under tamoxifen treatment during the RCT. Subtypes: ER-, ER+, PR-, PR+. Stage not reported. Serum samples collected at baseline of the RCT. Follow-up: 5.5 years

Serum testosterone (Radioimmunoassay), glucose, insulin, triglycerides, HDL-c

Cox proportional hazards models, adjusted for age with/without pathological prognostic factors ^b

Recurrence (local relapse, distant metastasis, contralateral breast cancer)

Women with MetS and testosterone > 0.40 ng/mL: ↑ RR vs. women without MetS and testosterone ≤0.40 ng/mL (HR: 6.7, 95% CI 2.3–19.8, p not shown). No significant associations between each component of MetS and RR

 Pasanisi et al. (2008) [22]

110 postmenopausal breast cancer survivors. Mean age: 56.8 years. Patients with normal-, or overweight/obesity. Operated for breast cancer at least a year prior to enrolment. HT: about 38% of women were under tamoxifen treatment during the RCT. Subtypes: ER-, ER+, PR-, PR+. Stage not reported. Serum samples collected at baseline of the RCT. Follow-up: 5.5 years

IGF-1, PDGF (radioimmunoassay), fructosamine, CRP (Hitachi modular automatic analyzer)

Cox proportional hazards models, adjusted for pathologic prognostic factors with (FAM)/without tertiles of body weight and serum testosterone ^b

Recurrence (local relapse, distant metastasis, contralateral breast cancer)

Recurrent patients: ↑ PDGF vs. patients without recurrence (11.9 ng/mL vs. 9.4 ng/mL, respectively, p = 0.01). In FAM, PDGF and IGF-1 combined, women with ↑ PDGF and ↑ IGF-1 (> their median value) had ↑ RR vs. women with ↓ PDGF and ↓ IGF-1 (HR: 6.4, 95% CI, 1.5–26.7, p not shown). In quartiles: no biomarkers showed significant associations with RR

 Berrino et al. (2014) [23]

2092 breast cancer survivors (about 45% postmenopausal) at high RR because of metabolic or endocrine milieu. Mean age: 51.4 years. 897 women with WC: ≥ 85 cm, 419 women with MetS. Operated for breast cancer on average 1.7 years (0–5 years) before enrolment. HT: not reported. Subtypes: ER+, PR+, HER2+. Stage: I-III. Plasma samples collected at baseline. Follow-up: median of 2.8 years

Glucose, HOMA-IR, triglycerides, HDL-c (Routine laboratory techniques)

Logistic regression models, adjusted for age, education, stage at diagnosis, ER expression

Recurrence (loco-regional recurrences, distant metastasis and new primary breast cancer)

No significant associations between fasting glucose, HOMA-IR and recurrence. ↓ HDL-c (OR 1.83, 95% CI 1.24–2.70, p not shown) and ↑ TG (OR 1.58, 95% CI 1.01–2.46, p not shown) were associated with ↑ RR

 Vasson et al. (2020) [24]

113 breast cancer survivors (about 60% postmenopausal). Mean age: 52 years. Mean BMI: 27.9 kg/m², Patients with under-, normal-, overweight/obesity. Operated for breast cancer and completed chemotherapy and/or radiotherapy < 9 months before the randomization. HT: most women were under tamoxifen treatment during the RCT. Subtypes: HoR+, HER2+. Stage: not reported. Plasma samples collected at baseline of the RCT. Follow-up: 7 years

HDL-c (colorimetry methods), plasma testosterone (ELISA), CA 15–3 (Clermont Ferrand) Categorized in quartiles

Survival curves using Kaplan-Meier’s method, comparison of curves using the Log-rank test. Cox proportional hazard model, no adjustments

Disease-free survival (absence of local or distant (nodes, metastasis, and/or contralateral breast cancer))

In quartiles: ↑ HDL-c was associated with the best survival without recurrence (p = 0.047). ↓ testosterone and CA 15–3 were associated with longer disease-free survival (p = 0.001 and 0.03, respectively). Based on survival curves, testosterone was relevant for disease-free survival only in patients treated with HT (p = 0.012 vs. p = 0.69 for patients with and without HT). With the Cox model, only ↑ testosterone was associated with ↑ RR (HR 5.06, 95% CI 1.66–15.41, p = 0.004)

 Emond et al. (2011) [25]

447 postmenopausal breast cancer survivors without baseline hot flash symptoms. Age: 18–70 years. Metabolic diseases and BMI not reported. Operated for breast cancer and had completed chemo- and/or radiotherapy. HT: about 63% of women were under tamoxifen treatment at baseline of the RCT. Subtypes: not reported. Stage: I II, III. Serum samples collected at baseline of the RCT. Follow-up: mean of 7.3 years

Bioavailable and total estradiol, bioavailable and total testosterone (radioimmunoassay), SHBG (two-site chemiluminometric sandwich assay)

Cox proportional hazards models, adjusted for intervention arm, baseline hormone concentration, site, antiestrogen use, # of positive nodes, tumor size, oophorectomy status, and previous hormone replacement therapy use ^b

Recurrence (local, regional, or distant, or new primary events)

↑ Bioavailable testosterone associated with ↑ RR (HR for one unit increase in ln-transformed values: 1.69, 95% CI 1.00–2.84, p = 0.049). No significant associations with estradiol or SHBG were observed

 Al-Delaimy et al. (2011) [26]

510 Recurrent breast cancer patients (about 75% postmenopausal) vs. 510 non-recurrent breast cancer patients (80% postmenopausal). Age: 18–70 years. Metabolic diseases and BMI not reported. Operated for breast cancer and had completed chemo- and/or radiotherapy. HT: about 58% of women were under tamoxifen treatment at baseline of the RCT. Subtypes: ER-, ER+. Stage: I II, III. Serum samples collected within 6 months after the enrollment. Follow-up: mean of 7.3 years

Insulin, leptin (Luminex technology), IGF-1, IGFBP-1, IGFBP-3, adiponectin (immunoassay) ^c

Cox proportional hazards models, adjusted for tamoxifen use and menopausal status ^b

Breast cancer event (recurrent/new primary breast cancer event)

Insulin, IGFBP-1, IGFBP-3, leptin, and adiponectin did not predict a breast cancer event in the 188 case-control pairs. No significant associations were found in the larger sample of 510 case-control pairs analyzed for IGF-1 (in quartiles and as a continuous measurement)

Matched on randomization

 Villaseñor et al. (2013) [27]

2919 breast cancer survivors (about 80% postmenopausal). Age: 27–74 years. Patients with under-, normal-, or overweight/obesity. Operated for breast cancer and had completed chemo- and/or radiotherapy HT: about 69% of women were under tamoxifen treatment at baseline of the RCT. Subtypes: ER+/PR+, ER+/PR-, ER−/PR+, ER−/PR-, HER2+, HER2-, unknown. Stage: I II, IIIA. Serum samples collected at a mean of 23.6 months post-diagnosis. Follow-up: mean of 7.4 years

hsCRP (High-sensitivity electrochemiluminescence assay)

Cox proportional hazards models, adjusted for age at diagnosis, time since diagnosis, race/ethnicity, and stage and grade (model 1), model 1 plus BMI (model 2), model 2 plus anti-estrogen medication use and ER/PR status (model 3) ^b

Additional breast cancer events (recurrence or new primary breast cancer) and breast cancer mortality

↑ InCRP was associated with ↑ additional breast cancer events (HR 1.13, 95% CI 1.03–1.24, p = 0.03, model 3) and ↑ breast cancer mortality (HR 1.16 1.01–1.31, p t = 0.03, model 3). In cut-off: ↑ hsCRP levels (> = 10 mg/L) vs. no inflammation levels (<  1,0 mg/L) were associated with ↑ additional breast cancer events (HR 1.65, 95% CI 1.15–2.38, p = 0.03, model 3), and ↑ breast cancer mortality (HR 1.88, 95% CI 1.11–3.18, p = 0.03, model 3)

All-cause mortality

↑ InCRP was associated with ↑ all-cause mortality (HR 1.19 1.05–1.34, p = 0.006, model 3). In cut-off: higher hsCRP levels (≥ 10 mg/L) vs. no inflammation levels (<  1 mg/L) were associated with ↑ all-cause mortality (HR 1.92, 95% CI 1.20–3.08, p = 0.006, model 3)