The reported complication rates after various surgical techniques used to create a lumbar fusion vary within wide ranges. In a previous paper, the Swedish Lumbar Spine Study Group have reported on the clinical outcome of lumbar spine fusion for chronic low back pain in a comparably homogeneous patient population where there were no significant differences between baseline sociodemographic, clinical and paraclinical characteristics. In this report we compared the complication rates of the surgical procedures used in that study and analyzed the association between complications and baseline variables, and between outcome results and complications. A multicenter randomized study was conducted where 211 patients aged 25-65 were treated with lumbar fusion according to three different surgical techniques: noninstrumented posterolateral fusion (PLF, n=71), instrumented posterolateral fusion (VSP, n=68), and in the third procedure we added an interbody fusion with solid autogenous bone grafts ("360", n=72). We categorized complications as: early/late, major/minor. The association between complications and sociodemographic characteristics (age, gender, comorbidity, previous surgery, smoking), and technical variables (surgical technique, levels fused, hospital category) was analyzed. The association between outcome variables (patient global assessment, pain, disability, depressive symptoms) and complications was analyzed. A literature review was conducted. There was no mortality. There was no significant difference in clinical outcome between the surgical groups after 2 years, although the power to detect such a difference was low. The total complication rate after 2 years in the PLF group was 12%, compared with 22% in the VSP group, and 40% in the "360" group (P=0.0003). After exclusion of complications, there was still no difference in outcome between the groups. The odds ratio (confidence intervals) of having a complication was 5.3 (2.2-12.7) when "360" was used compared with PLF, and 2.4 (1.1-5.3) for "360" compared with VSP. There was no association between clinical outcome and complications on a group level. The reintervention rate was 6% in the PLF group, 22% in the VSP, and 17% in the "360" group (P=0.020). The odds ratio (confidence intervals) of having a reintervention was 4.0 (1.3-11.9) when instrumentation was used compared with non-instrumented fusion. In this prospective randomized study comparing three lumbar fusion techniques in a comparably homogeneous patient population, complications increased significantly with increasing technicality of the surgical procedure. Even though we did not find a significant association between clinical outcome and complications after 2 years, the increased morbidity inflicted on an individual patient was not negligible. In this light, and as no fusion technique produced superior clinical outcome irrespective of whether complications were included or excluded in the analyses, the patient and the treating physician should carefully discuss the possible advantages and drawbacks of the different surgical options before making a decision. In order to make valid comparisons of both complication and reintervention rates after lumbar fusion, there is a need for a consensus in the spinal society regarding the definition of these entities.