Background The functional and clinical basis on which to choose whether or not to retain the posterior cruciate ligament during total knee arthroplasty surgery remained unclear after a Cochrane systematic review and meta-analysis in 2005, which contained eight clinical trials. Several new trials have been conducted since then. Hence, an update of the review was performed. Objectives Our aim was to assess the benefits and harms of retention compared to sacrifice of the posterior cruciate ligament in total knee arthroplasty in patients with osteoarthritis of the knee. Search methods An extensive search was conducted in CENTRAL, MEDLINE (PubMed), EMBASE, Web of Science, CINAHL, Academic Search Premier, Current Contents Connect and Science Direct. All databases were searched, without any limitations, up to 6 December 2012. References of the articles were checked and citation tracking was performed. Selection criteria Randomised and quasi-randomised controlled trials comparing retention with sacrifice of the posterior cruciate ligament in primary total knee arthroplasty in patients with osteoarthritis of the knee. Data collection and analysis Data were collected with a pre-developed form. Risk of bias was assessed independently by two authors (WV, LB). The level of evidence was graded using the GRADE approach. Meta-analysis was performed by pooling the results of the selected studies, when possible. Subgroup analyses were performed for posterior cruciate ligament retention versus sacrifice using the same total knee arthroplasty design, and for studies using a posterior cruciate ligament retaining or posterior stabilised design, and when sufficient studies were available subgroup analyses were performed for the same brand. Main results Seventeen randomised controlled trials (with 1810 patients and 2206 knees) were found, described in 18 articles. Ten of these were new studies compared to the previous Cochrane Review. One study from the original Cochrane review was excluded. Most new studies compared a posterior cruciate ligament retaining design with a posterior stabilised design, in which the posterior cruciate ligament is sacrificed (a posterior stabilised design has an insert with a central post which can engage on a femoral cam during flexion). The quality of evidence (graded with the GRADE approach) and the risk of bias were highly variable, ranging from moderate to low quality evidence and with unclear or low risk of bias for most domains, respectively. The performance outcome 'range of motion' was 2.4 ° higher in favour of posterior cruciate ligament sacrifice (118.3 ° versus 115.9 °; 95% confidence interval (CI) of the difference 0.13 to 4.67; P = 0.04), however the results were heterogeneous. On the item 'knee pain' as experienced by patients, meta-analysis could be performed on the Knee Society knee pain score; this score was 48.3 in both groups, yielding no difference between the groups. Implant survival rate could not be meta-analysed adequately since randomised controlled trials lack the longer term follow-up in order to evaluate implant survival. A total of four revisions in the cruciate-retention and four revisions in the cruciate-sacrifice group were found. The well-validated Western Ontario and McMaster Universities osteoarthritis index (WOMAC) total score was not statistically significantly different between the groups (16.6 points for cruciate-retention versus 15.0 points for cruciate-sacrifice). One study reported a patient satisfaction grade (7.7 points for cruciate-retention versus 7.9 points for cruciate-sacrifice on a scale from 0 to 10, 10 being completely satisfied) which did not differ statistically significantly. Complications were distributed equally between both groups. Only one study reported several re-operations other than revision surgery; that is patella luxations, surgical manipulation because of impaired flexion. The mean functional Knee Society Score was 2.3 points higher (81.2 versus 79.0 points; 95% CI of the difference 0.37 to 4.26; P = 0.02) in the posterior cruciate ligament sacrificing group. Results from the outcome Knee Society functional score were homogeneous. All other outcome measures (extension angle, knee pain, adverse effects, clinical questionnaire scores, Knee Society clinical scores, radiological rollback, radiolucencies, femorotibial angle and tibial slope) showed no statistically significant differences between the groups. In the subgroup analyses that allowed pooling of the results of the different studies, no homogeneous statistically significant differences were identified. Authors' conclusions The methodological quality and the quality of reporting of the studies were highly variable. With respect to range of motion, pain, clinical, and radiological outcomes, no clinically relevant differences were found between total knee arthroplasty with retention or sacrifice of the posterior cruciate ligament. Two statistically significant differences were found; range of motion was 2.4 ° higher in the posterior cruciate ligament sacrificing group, however results were heterogeneous; and the mean functional Knee Society Score was 2.3 points higher in the posterior cruciate ligament sacrificing group. These differences are clinically not relevant.