Background: Early neurological deterioration (END) following acute stroke is associated with poorer long-term outcomes. Identification of patients at risk could assist early monitoring and treatment decisions. This review summarised the evidence describing non-radiological biomarkers for END. Summary: Electronic searches from January 1990 to March 2017 identified studies reporting a blood/cerebrospinal fluid (CSF)/urine biomarker measurement within 24 h of acute stroke and at least 2 serial assessments of clinical neurological status (< 24 h and < 7 days). Out of 12,895 citations, 82 studies were included, mostly focusing on ischaemic stroke. Using higher neurological thresholds, the n-weighted END incidence for ischaemic stroke was 11.9% (95% CI 11.4–12.4%) and 18.6% (17.9–19.2%) for lower thresholds. Incidence decreased with advancing study publication year (Pearson r-squared 0.23 and 0.15 for higher and lower threshold studies). After classification into 3 broad categories, meta-analysis showed that biomarkers associated with increased END risk (n; fixed-effects mean difference; 95% CI) were “metabolic” (glucose [n = 9,481; 0.90 mmol/L; 0.74–1.06], glycosylated haemoglobin [n = 3,146; 0.33%; 0.19–0.46], low-density lipoprotein [n = 4,839; 0.13 mmol/L; 0.06–0.21], total cholesterol [n = 4,762; 0.21 mmol/L; 0.11–0.31], triglycerides [n = 4,820; 0.11 mmol/L; 0.06–0.17], urea [n = 1,351; 0.55 mmol/L; 0.14–0.96], decreasing albumin [n = 513; 0.33 g/dL; 0.05–0.61]); “inflammatory and excitotoxic” (plasma glutamate [n = 688; 60.13 µmol/L; 50.04–70.22], CSF glutamate [n = 369; 7.50 µmol/L; 6.76–8.23], homocysteine [n = 824; 2.15 µmol/L; 0.68–3.61], leucocytes [n = 3,766; 0.54 × 10<sup>9</sup>/L; 0.34–0.74], high-sensitivity C-reactive protein [n = 1,707; 3.79 mg/L; 1.23–6.35]); and “coagulation/haematological” (fibrinogen [n = 3,132; 0.32 g/L; 0.25–0.40]; decreasing haemoglobin [n = 3,586; 2.38 g/L; 0.15–4.60]). Key Messages: Declining incidence of END may represent improving care standards; however, it remains a frequent occurrence. Although statistical associations exist between biomarkers and an increased risk of END, the most promising still need prospective evaluation to determine their additional value relative to baseline radiological and clinical characteristics.