BACKGROUND: Although there is a strong biological rationale for early decompression of the injured spinal cord, the influence of the timing of surgical decompression for acute spinal cord injury (SCI) remains debated, with substantial variability in clinical practice. We aimed to objectively evaluate the effect of timing of decompressive surgery for acute SCI on long-term neurological outcomes. METHODS: We did a pooled analysis of individual patient data derived from four independent, prospective, multicentre data sources, including data from December, 1991, to March, 2017. Three of these studies had been published; of these, only one study previously specifically analysed the effect of the timing of surgical decompression. These four datasets were selected because they were among the highest quality acute SCI datasets available and contained highly granular data. Individual patient data were obtained by request from study authors. All patients who underwent decompressive surgery for acute SCI within these datasets were included. Patients were stratified into early (<24 h after spinal injury) and late (≥24 h after spinal injury) decompression groups. Neurological outcomes were assessed by American Spinal Injury Association (ASIA), or International Standards for Neurological Classification of Spinal Cord Injury (ISNCSCI), examination. The primary endpoint was change in total motor score from baseline to 1 year after spinal injury. Secondary endpoints were ASIA Impairment Scale (AIS) grade and change in upper-extremity motor, lower-extremity motor, light touch, and pin prick scores after 1 year. One-stage meta-analyses were done by hierarchical mixed-effects regression adjusting for baseline score, age, mechanism of injury, AIS grade, level of injury, and administration of methylprednisolone. Effect sizes were summarised by mean difference (MD) for sensorimotor scores and common odds ratio (cOR) for AIS grade, with corresponding 95% CIs. As a secondary analysis, change in total motor score was regressed against time to surgical decompression (h) as a continuous variable, using a restricted cubic spline with adjustment for the same covariates as in the primary analysis. FINDINGS: We identified 1548 eligible patients from the four datasets. Outcome data at 1 year after spinal injury were available for 1031 patients (66·6%). Patients who underwent early surgical decompression (n=528) experienced greater recovery than patients who had late decompression surgery (n=1020) at 1 year after spinal injury; total motor scores improved by 23·7 points (95% CI 19·2-28·2) in the early surgery group versus 19·7 points (15·3-24·0) in the late surgery group (MD 4·0 points [1·7-6·3]; p=0·0006), light touch scores improved by 19·0 points (15·1-23·0) vs 14·8 points (11·2-18·4; MD 4·3 [1·6-7·0]; p=0·0021), and pin prick scores improved by 18·3 points (13·7-22·9) versus 14·2 points (9·8-18·6; MD 4·0 [1·5-6·6]; p=0·0020). Patients who had early decompression also had better AIS grades at 1 year after surgery, indicating less severe impairment, compared with patients who had late surgery (cOR 1·48 [95% CI 1·16-1·89]; p=0·0019). When time to surgical decompression was modelled as a continuous variable, there was a steep decline in change in total motor score with increasing time during the first 24-36 h after injury (p<0·0001); and after 36 h, change in total motor score plateaued. INTERPRETATION: Surgical decompression within 24 h of acute SCI is associated with improved sensorimotor recovery. The first 24-36 h after injury appears to represent a crucial time window to achieve optimal neurological recovery with decompressive surgery following acute SCI. FUNDING: None.