Effectiveness of posterior decompression techniques compared with conventional laminectomy for lumbar stenosis

We included all types of prospective, controlled studies (randomised controlled trials (RCTs) and cohort studies).

The population consists of adult patients with symptomatic degenerative lumbar stenosis. We excluded studies that included cases of congenital lumbar stenosis (e.g. achondroplasia) or acquired lumbar stenosis due to trauma, infection or abnormal bone metabolism (e.g. Paget’s disease). We made no exclusions for age or gender of the population, or type, location or duration of symptoms.

We included all prospective studies comparing a posterior decompressive technique that avoids removal of posterior midline structures (spinous processes, vertebral arches, interspinous and supraspinous ligaments) or a technique involving only partial resection of the vertebral arch with conventional facet-preserving laminectomy. We also included studies that describe cases requiring decompression of more than one stenotic level or a concomitant discectomy or foraminotomy. We excluded studies involving cases of decompression through interspinous process devices or concomitant (instrumented) fusion procedures.

Eligible studies evaluate at least one of the main clinically relevant outcome measures using a valid instrument. Furthermore, predefined secondary outcome measures are compared, as mentioned below. The minimal duration of follow-up for studies considered for inclusion is 6 months. Long-term follow-up is defined as a minimum of 2 years of follow-up.

Primary outcomes

Secondary outcomes

An experienced librarian conducted a comprehensive electronic search for relevant literature in the Cochrane Central Register of Controlled trials, Medline, Embase and Web of Science up to June 2014. The trials registries ClinicalTrials.gov and World Health Organization International Clinical Trials Registry Platform (WHO ICTRP) were searched for ongoing and unpublished trials up to 5 June 2014. Additionally, we reviewed the reference sections and citation tracking results of all included studies for articles not found during the systematic search.

Two review authors (GO, WJ) independently examined titles and abstracts from the electronic search to identify eligible studies. We obtained full-text articles if necessary. We consulted a third review author (WP) if consensus was not reached. We listed unpublished, ongoing and excluded studies in the results and used no language restrictions to minimise publication bias.

Two independent review authors (CVL, CT) assessed risk of bias using the criteria advised by the Cochrane Back Review Group (CBRG) for evaluation of RCTs [18], listed in Table 2.

The list consists of 12 items, which are scored low, unclear or high. In accordance with the recommendations of the CBRG, review authors rated studies as having low risk of bias if at least six of the criteria were met and the study had no serious flaws (e.g. more than 20 % loss to follow-up).

The review authors discussed differences in scoring of the risk of bias assessment during a consensus meeting and consulted a third review author (RG) if necessary. Among other degenerative spinal conditions, all review authors involved in the risk of bias assessment are experts in the field of lumbar stenosis.

We recorded data describing study characteristics such as study design, characteristics and numbers of participants, surgical techniques and co-interventions, primary and secondary outcomes, follow-up time and study sponsorship. Data were entered into the statistical software of The Cochrane Collaboration, Review Manager 2011 [19].

We defined treatment effect as the differences between treatment groups on relevant outcome measures. We presented comparisons of continuous data as mean differences (MDs) with corresponding 95 % confidence intervals (CIs). For dichotomous outcomes (e.g. perceived recovery among treatment groups), we calculated odds ratios (ORs) and corresponding 95 % CIs. We compared continuous data directly or through calculation of standardised mean differences (SMDs) if outcome measures were not directly comparable (e.g. different measurement scales used). We analysed outcomes of included studies using a random-effects model.

We adopted the definition of a minimal clinically important difference [20]. We defined the minimal clinically important difference of the primary outcome measures (i.e. functional disability, perceived recovery, leg pain) as 30 % improvement from baseline. This corresponds to a mean difference of 1.5 for the visual analogue scale (VAS) (0–10), 2 for the numerical rating scale (0–10), 5 for the Roland Disability Questionnaire (0–24), 10 for the Oswestry Disability Questionnaire (0–100) and 20 for the Quebec Back Pain Disability Questionnaire (0–100). These thresholds represent a minimum important change from baseline for a patient. However, there is no generally accepted minimal clinically important difference between treatments.

We assessed the heterogeneity of RCTs first clinically, then statistically. We evaluated clinical heterogeneity for variability in participant characteristics (baseline symptom severity, duration of symptoms, number of stenotic levels involved, age and gender) and treatment characteristics (comparison of techniques, concomitant interventions and length of follow-up time). When we judged studies to be clinically homogeneous, we tested statistical homogeneity with an I ² test.

We pooled results from individual studies if the studies were judged to be sufficiently homogeneous (clinically and statistically). We evaluated the quality of evidence for all primary outcome parameters, regardless of quantitative analysis, using the GRADE (Grading of Recommendations Assessment, Development and Evaluation) approach [21, 22]. In short, we judged the quality of evidence using the following criteria:

Depending on how many domains were met, we judged the quality of evidence as ‘high’, ‘moderate’, ‘low’ or ‘very low’ based on the following descriptions.

We considered important outcomes for which there are no trials to have ‘no evidence’. We automatically graded an outcome with only one trial as low quality, and if it also had high risk of bias, we graded the evidence as very low quality.

The literature search up to June 2014 yielded 5924 articles. We identified 10 studies that compared a posterior decompressive technique that avoids removal of posterior midline structures versus conventional laminectomy. All studies were randomised controlled trials. We found no eligible prospective cohort studies, and we identified no study that compared a technique involving only partial resection of the vertebral arch (with removal of posterior midline structures) or laminoplasty versus conventional laminectomy. Citation tracking and review of the reference sections of included articles yielded no additional articles eligible for inclusion. We identified one published study protocol of an ongoing study [23]. Review of trial registries revealed no other ongoing or unpublished trials. We subdivided studies evaluating posterior decompressive techniques into unilateral laminotomy for bilateral decompression [4, 24, 25], bilateral laminotomy [2‐4, 26] and split-spinous process laminotomy [27‐30] to ensure clinical homogeneity.

Included studies were published between 1993 and 2014 and included a total of 733 participants (see Table 1). All studies except one [3] performed analyses according to the intention-to-treat principle. The study by Postacchini et al. [3] compared three groups according to the treatment actually received, and thus compared 26 participants undergoing bilateral laminotomy, nine allocated to bilateral laminotomy but converted to conventional laminectomy and 32 participants allocated to and undergoing conventional laminectomy.

Two studies were excluded from the quantitative analysis as the result of clinical heterogeneity [3, 27]. Decisive arguments were the inclusion of concomitant intertransverse arthrodesis (in 12 out of 67 participants) and concomitant discectomy (in 7 out of 67 participants) by Postacchini et al. [3], and concomitant discectomy (in 44 out of 70 participants) by Cho et al. [27]. Concomitant discectomy was not among the predefined exclusion criteria of this review, but the exceptionally high rate of concomitant discectomy by Cho et al. compared with no discectomy in any of the other studies precludes a valid comparison of results.

We excluded eight comparative studies from this review because they employed a retrospective study design [7, 31‐37]. We excluded two studies that compared techniques of posterior decompression because of concomitant fusion procedures [38, 39]. Moreover, concomitant fusion procedures were unequally distributed among treatment groups, resulting in potential bias [39]. We excluded one randomised controlled trial because researchers reported follow-up of only three months [40]. This study compared unilateral laminotomy for bilateral decompression with conventional laminectomy. Perceived recovery favoured participants in the unilateral laminotomy group, but the difference was not clinically significant. Moreover, the duration of hospital stay was significantly shorter in the unilateral laminotomy group. We excluded eight studies because the control group did not meet the criteria of this review [6, 41‐47]. We excluded three studies because the design precluded a reliable comparison of decompression techniques [48‐50]. All studies report outcome measures of different decompression techniques that are the subject of this review, but the comparability of treatment groups cannot be ensured, and indications for the surgical treatment groups may vary. Moreover, two of these studies did not describe surgical techniques in sufficient detail [48, 49]. We excluded one study because study authors reported no outcome measures relevant to this review [51]. They compared only the extent of postoperative haematoma and the cross-sectional area of the spinal canal following unilateral laminotomy for bilateral decompression, bilateral laminotomy and conventional laminectomy.

Four out of 10 studies had low risk of bias [2, 4, 29, 30], having met at least six of the risk of bias criteria. Poor performance on the risk of bias assessment was the result of poor methodology or poor reporting. The risk of bias summary of trials is shown in Table 2.

Investigators used different disability questionnaires among the studies included in this review [2, 4, 24‐30]. Postacchini et al. did not report a validated disability score [3]. None of these studies, except the study by Fu et al. [26], demonstrated a significant difference between the techniques of posterior decompression. The mean difference between the bilateral laminotomy group [0.37 ± 0.96 standard deviation (SD)] and the conventional laminectomy group (3.37 ± 8.55 SD) as reported by Fu et al. does not seem clinically relevant when a minimal clinically important difference in the ODI of 10 is considered [26]. We provided pooled estimates of RCTs by calculating standardised mean differences in disability questionnaire scores. Between those who received unilateral laminotomy and those undergoing laminectomy, low-quality evidence shows no significant difference regarding disability scores [three RCTs, 166 participants, MD −1.11, 95 % CI −11.91 to 9.69). Between those who received bilateral laminotomy and those undergoing laminectomy, very low-quality evidence reveals a significant difference regarding disability scores in favour of bilateral laminotomy (three RCTs, 294 participants, MD −2.73, 95 % CI −4.59 to −0.87). The quality of evidence had to be decreased because of the relatively high impact of one low-quality study on the combined quantitative analysis [26]. The difference did not reach the criteria of a minimal clinically important difference. Between those who received split-spinous process laminotomy and those undergoing laminectomy, low-quality evidence suggests no significant difference regarding disability scores (three RCTs, 139 participants, MD −1.68, 95 % CI −8.50 to 5.13), see Table 3.

Thome et al. reported self-perceived overall recovery and found no significant difference between unilateral laminotomy and conventional laminectomy [4]. However, a significant difference in favour of the bilateral laminotomy group was assessed when compared with the conventional laminectomy group (36/37, 97.3 % vs 25/34, 73.5 %). Postacchini et al. used a scoring system based on the participant’s perceived recovery and the examiner’s evaluation, which included results of a neurological examination, the need for analgesics, the ability to work and carry out activities of daily living and walking ability. Investigators reported no significant differences between treatment groups: bilateral laminotomy 21/26 (81 %) with excellent/good recovery, allocated to the bilateral laminotomy group but converted to conventional laminectomy 7/9 (78 %) and conventional laminectomy 25/32 (78 %) [3]. Fu et al. assessed perceived recovery using a structural interview, which evaluated back and leg pain, walking ability and restriction from usual activities. In all, 68/76 (89 %) participants in the bilateral laminotomy group reported excellent/good results compared with 48/76 (63 %) in the conventional laminectomy group. The difference was statistically significant [26]. The recovery rate reported by Gurelik et al. is not included in this comparison because it is derived from the ODI [24]. Between those who received unilateral laminotomy and those undergoing laminectomy, low-quality evidence shows no significant difference (one high-quality RCT, 73 participants, OR 1.04, 95 % CI 0.37–2.98). Between those who received bilateral laminotomy and those undergoing laminectomy, low-quality evidence suggests a significant difference regarding self-perceived recovery in favour of bilateral laminotomy (two RCTs, 223 participants, OR 5.69, 95 % CI 2.55–12.71). The quality of evidence had to be decreased because of the relatively high impact of one low-quality study on the combined quantitative analysis [26], see Table 3.

Postacchini et al. and Thome et al. reported improvement in leg pain in 34/37 (92 %) participants and mean improvement of 71 out of 100 points among 35 participants undergoing bilateral laminotomy, respectively [3, 4]. Compared with 25/34 (74 %) participants reporting improvement and mean improvement of 84 out of 100 points among participants undergoing conventional laminectomy, Thome et al. concluded that a statistically significant difference favoured bilateral laminotomy, whilst Postacchini et al. found no statistically significant differences between treatment groups. Reporting of data in these studies does not allow for a quantitative comparison, nor a comparison with other studies. Celik et al. and Fu et al. compared leg pain VAS scores (0–10) of participants undergoing bilateral laminotomy and conventional laminectomy [2, 26]. Between those who received bilateral laminotomy and laminectomy, very low-quality evidence shows a significant difference regarding VAS leg pain in favour of bilateral laminotomy (two RCTs, 223 participants, MD −0.29, 95 % CI −0.48 to −0.11). The quality of evidence had to be decreased because of the relatively high impact of one low-quality study on the combined quantitative analysis and inconsistent results among studies [26]. The difference did not exceed the minimal clinically important difference of 1.5. Among participants who underwent unilateral laminotomy, Thome et al. reported an improvement in leg pain in 26/39 (68 %) compared with 25/34 (74 %) who underwent conventional laminectomy [4]. The difference was not significant. Liu et al. and Rajasekaran et al. compared leg pain VAS scores (0–10) of participants undergoing split-spinous process laminotomy and conventional laminectomy [28, 29]. Between those who received split-spinous process laminotomy and laminectomy, very low-quality evidence shows a significant difference regarding VAS leg pain in favour of split-spinous process laminotomy (two RCTs, 107 participants, MD –0.29, 95 % CI −0.41 to −0.17). The quality of evidence had to be decreased because of the relatively high impact of one low-quality study on the combined quantitative analysis [28]. Again, the difference did not exceed the minimal clinically important difference of 1.5, see Table 3.

Celik et al. reported no significant difference regarding length of hospital stay after bilateral laminotomy (mean 2.2 days) compared with conventional laminectomy (2.3 days) [2]. The quality of evidence is low (only one high-quality RCT, 71 participants, MD −0.10, 95 % CI −0.89 to 0.69). Yagi et al. reported a significantly shorter duration of hospital stay after unilateral endoscopic laminotomy (mean 4 days) compared with conventional laminectomy (mean 13 days) [25]. Results of studies comparing split-spinous process laminotomy with laminectomy are conflicting. Cho et al. reported a significantly shorter duration of hospital stay after split-spinous process laminotomy (mean 4 days) compared with conventional laminectomy (mean seven days) [27], but Rajasekaran et al. reported equal duration of hospital stay after split-spinous process laminotomy (mean 4.5 days) compared with conventional laminectomy (mean 4.4 days) [29]. The quality of evidence is low (only one high-quality RCT, 51 participants, MD −0.10, 95 % CI −0.46 to 0.66).

All studies reported procedure-related complications. None of the studies included in this review reported procedure-related mortality. The most commonly reported complication of the surgical procedure was a dural tear. Celik et al. and Thome et al. reported a significantly lower incidence of incidental dural tear in the bilateral laminotomy group compared with the laminectomy group (1/37 vs 7/34 and 2/40 vs 8/40, respectively) [2, 4]. Other studies did not report a significantly different incidence of dural tears among treatment groups [3, 26, 28, 29]. None of the studies included in this review reported significant differences between treatment groups regarding iatrogenic neurological impairment, wound infection or epidural haematoma. Between those who received bilateral laminotomy and those undergoing conventional laminectomy, low-quality evidence shows no significant differences regarding cumulative incidence of complications (three RCTs, 303 participants, OR 0.33, 95 % CI 0.07–1.59). Between those who received unilateral laminotomy and those undergoing conventional laminectomy, low-quality evidence indicates no significant differences regarding cumulative incidence of complications (three RCTs, 173 participants, OR 0.73, 95 % CI 0.24–2.20). Finally, between those who received split-spinous process laminotomy and those undergoing conventional laminectomy, low-quality evidence shows no significant differences regarding cumulative incidence of complications (three RCTs, 141 participants, OR 1.21, 95 % CI 0.20–7.16), see Table 3.

All studies, except the studies by Rajasekaran et al. and Watanabe et al., reported surgically induced spinal instability. Investigators used flexion–extension radiographs to assess spinal instability [2‐4, 24‐30]. Among studies comparing unilateral laminotomy with conventional laminectomy, Thome et al. reported postoperative radiological instability in 2/39 participants in the unilateral laminotomy group compared with 3/34 participants in the conventional laminectomy group [4]. All participants with instability underwent instrumented fusion. Gurelik et al. reported no postoperative instability in the unilateral laminotomy group (26 participants) compared with 5 out of 26 patients in the conventional laminectomy group [24]. It must be noted though that investigators performed a unilateral total facetectomy in seven participants in the laminectomy group and in none of those in the laminotomy group. Yagi et al. reported no postoperative spondylolisthesis in the unilateral laminotomy group compared with 2/21 participants in the conventional laminectomy group [25]. Between those who received unilateral laminotomy and those undergoing laminectomy, low-quality evidence showed no significant differences regarding the incidence of instability (three RCTs, 166 participants, OR 0.28, 95 % CI 0.07–1.15). Among studies comparing bilateral laminotomy with conventional laminectomy, Celik et al. reported no radiological instability in 37 participants in the bilateral laminotomy group compared with 3 out of 34 participants in the conventional laminectomy group [2]. Two of these participants underwent instrumented fusion, and the other participant declined subsequent surgery. Fu et al. reported no postoperative instability in 76 participants in the bilateral laminotomy group compared with 6 out of 76 participants in the conventional laminectomy group [26]. Four of these participants underwent instrumented fusion. Postacchini et al. reported no postoperative instability in the bilateral laminotomy group compared with 3/41 participants treated with conventional laminectomy [3]. Thome et al. reported no postoperative instability in 37 participants in the bilateral laminotomy group compared with 3 out of 34 participants in the conventional laminectomy group [4]. All participants with instability underwent instrumented fusion. Between those who received bilateral laminotomy and those undergoing laminectomy, low-quality evidence suggests a significantly higher incidence of instability in the conventional laminectomy group (three RCTs, 294 participants, OR 0.10, 95 % CI 0.02–0.55). The quality of evidence had to be decreased because of high risk of bias due to a non-standardised assessment of spinal instability. Cho et al. reported no postoperative instability in 40 participants in the split-spinous process laminotomy group [27]. Two out of 30 participants in the conventional laminectomy group developed postoperative spondylolisthesis, and one underwent subsequent instrumented fusion. The difference was not significant. Liu et al. reported no difference among 27 participants treated with split-spinous process laminotomy compared with 29 participants treated with conventional laminectomy [28]. No participants in either group developed instability. The quality of evidence is very low (only one low-quality RCT), see Table 3.

Three studies reported paraspinal muscle denervation/atrophy [25, 28, 30]. After 1 year, Yagi et al. compared muscle atrophy ratios of multifidus and erector spinae muscles following unilateral microendoscopic laminotomy of 13 vs 35 % following conventional laminectomy [25]. The difference was statistically significant, and the quality of evidence was very low (only one low-quality RCT). Liu et al. and Watanabe et al. compared muscle atrophy ratios of multifidus and erector spinae muscles following split-spinous process laminotomy and conventional laminectomy after 3 months and 1 month, respectively [28, 30]. Between those who received split-spinous process laminotomy and laminectomy, low-quality evidence suggests a significant difference regarding postoperative back muscle atrophy ratios in favour of split-spinous laminotomy (two RCTs, 90 participants, MD −12.07, 95 % CI −20.01 to −4.13). The quality of evidence had to be decreased because of the relatively high impact of one low-quality study on the combined quantitative analysis [28].

Five studies reported muscle cell injury (creatine phosphokinase levels) [25, 27‐30]. All studies but one (Rajasekaran et al.) reported statistically significant differences, with higher creatine phosphokinase levels (CPK-MM) in the conventional laminectomy groups. Following unilateral microendoscopic laminotomy, CPK-MM after 24 h was 270 IU/L, and following conventional laminectomy, CPK-MM was 620 IU/L (Yagi et al.) [25]. As only one low-quality RCT compared muscle cell injury after unilateral laminotomy versus conventional laminectomy, the quality of evidence is very low. Following a split-spinous process laminotomy, the CPK-MM was 161 IU/L, and it was 276 IU/L following conventional laminectomy (Cho et al.) [27]. Between those who received split-spinous process laminotomy and laminectomy, low-quality evidence shows no significant differences regarding postoperative creatine kinase levels (three RCTs, 141 participants, MD −194.87, 95 % CI −456.95 to 67.20). The quality of evidence had to be decreased because of severely variable standard deviations between studies, possibly as the result of different methods of measuring CPK-MM (Liu et al.: measurement on postoperative day three; Rajasekaran et al.: difference between preoperative measurement and measurement on postoperative day one; Watanabe et al.: measurement on postoperative day three) [28‐30].

Only Gurelik et al. reported actual walking distance, as assessed by walking distance on a treadmill [24]. Walking distance was not significantly different among patients who underwent unilateral laminotomy (288.7 m) versus conventional laminectomy (203.7 m). Celik et al. and Thome et al. compared participants’ self-reported walking distance [2, 4]. They reported no significant differences regarding walking distance between participants treated with bilateral laminotomy (3663 m), unilateral laminotomy (2958 m) and conventional laminectomy (2318 m) (Thome et al.). Celik et al. reported no significant differences regarding pain-free walking distance between participants treated with bilateral laminotomy (97 m) and those undergoing conventional laminectomy (94 m). Fu et al. reported walking tolerance [26]. After a mean of 40 months, 100 % of participants treated with bilateral laminotomy were able to walk longer than 30 min, 97 % were able to walk longer than 60 min and 89 % could walk an unlimited distance. Compared with participants treated with conventional laminectomy, of whom 100 % were able to walk longer than 30 min, 86 % were able to walk longer than 60 min and 51 % were able to walk an unlimited distance, only the percentage of participants who reported an unlimited walking distance increased significantly. In summary, low-quality evidence suggests that walking distance after bilateral laminotomy and conventional laminectomy did not differ (three RCTs, 294 participants), and very low-quality evidence shows that walking distance after unilateral laminotomy and conventional laminectomy does not differ (two RCTs, 125 participants). A quantitative comparison of data was not possible.

Low back pain after surgery, assessed with a VAS, was reported by six studies [2, 25‐29]. Among participants treated with bilateral laminotomy, Fu et al. reported a significantly lower VAS 0–10 (0.05), compared with conventional laminectomy (0.63) [26]. However, another study comparing bilateral laminotomy (VAS 4.2) versus conventional laminectomy (VAS 4.4) yielded no significant differences between treatment groups [2]. Among participants treated with unilateral microendoscopic laminotomy (VAS 0.8) and those treated with conventional laminectomy (VAS 3.4 cm), Yagi et al. reported a statistically significant difference in favour of unilateral microendoscopic laminotomy [25]. Among participants treated with bilateral laminotomy, unilateral laminotomy and conventional laminectomy, Thome et al. reported no statistically significant differences regarding improvement in back pain at rest, but described significantly improved back pain during walking among participants treated with bilateral laminotomy versus those treated with conventional laminectomy [4]. According to the as-treated analysis provided by Postacchini et al., the mean improvement on the VAS scale for back pain was significant among participants treated with bilateral laminotomy compared with those who crossed-over or were allocated to conventional laminectomy [3]. Two studies comparing split-spinous process laminotomy versus conventional laminectomy reported significantly decreased postoperative back pain VAS 0–10 [27, 28]. Cho et al. reported one-year postoperative VAS of 2.4 in the split-spinous process laminotomy group compared with 4.0 in the conventional laminectomy group [27]. Liu et al. reported one-year postoperative VAS back pain of 1.0 and 2.6, respectively [28]. In comparison, Rajasekaran et al. did not find a significant difference between split-spinous process laminotomy (2.5) and conventional laminectomy (3.0) regarding postoperative back pain after 1 year [29]. In summary, low-quality evidence showed that back pain is greater after conventional laminectomy than after bilateral laminotomy, but the mean difference does not reach the criteria of a minimal clinically important difference (two RCTs, 223 participants, MD −0.51, 95 % CI −0.80 to −0.23). The quality of evidence had to be decreased because of the relatively high impact of one low-quality study on the combined quantitative analysis [26]. A quantitative comparison of postoperative back pain after unilateral laminotomy (two RCTs, 114 participants) was not possible because of different reporting of outcome measures. Between those who receive split-spinous process laminotomy and those undergoing laminectomy, low-quality evidence shows a significant difference regarding back pain in favour of split-spinous process laminotomy, but the mean difference does not reach the criteria of a minimal clinically important difference (two RCTs, 97 participants, MD −1.07, 95 % CI −2.15 to −0.00), see Table 3.

Eight studies reported length of the surgical procedure [2‐4, 25, 27‐30]. Thome et al. reported a significantly increased duration of bilateral laminotomy (90 min) compared with unilateral laminotomy (77 min) or conventional laminectomy (73 min) [4]. Postacchini et al. reported a significantly increased duration of bilateral laminotomy in cases of multiple-level decompression, but not when comparing single-level decompression [3]. By contrast, Celik et al. reported a shorter duration of bilateral laminotomy (83 min) compared with conventional laminectomy (107 min) [2]. Yagi et al. reported a longer duration of unilateral laminotomy (71 min) compared with conventional laminectomy (64 min), but the difference was not statistically significant [25]. Comparatively, Cho et al. reported a long duration of surgical procedures of 259 min among participants treated with a split-spinous process laminotomy and 193 min among those treated with conventional laminectomy, but they performed a concomitant discectomy in most participants and decompressed 2.6 levels on average per participant (see Table 1) [27]. Two studies comparing a split-spinous process laminotomy with conventional laminectomy reported a non-significantly shorter duration of conventional laminectomy [28, 29], and one study reported a non-significantly longer duration of conventional laminectomy [30]. Between those who receive bilateral laminotomy and those undergoing conventional laminectomy, low-quality evidence suggests no significant difference regarding length of the procedure (two RCTs, 142 participants, MD 0.3, 95 % CI −39.2 to 39.8). Between those who receive unilateral laminotomy and those undergoing conventional laminectomy, low-quality evidence shows no significant difference regarding length of the procedure (two RCTs, 114 participants, MD 6.3, 95 % CI −0.7 to 13.2). Between those who receive split-spinous process laminotomy and those treated with conventional laminectomy, low-quality evidence indicates no significant difference regarding length of the procedure (three RCTs, 141 participants, MD 4.6, 95 % CI −5.1 to 14.3).

Studies comparing perioperative blood loss among participants treated with bilateral laminotomy and those treated with conventional laminectomy did not report a statistically significant difference [2‐4]. However, Thome et al. and Yagi et al. did find a statistically significant difference in favour of unilateral laminotomy when compared with conventional laminectomy (blood loss 177 vs 227 mL and 37 vs 71 mL, respectively). One study that compared a split-spinous process laminotomy versus conventional laminectomy reported a significant decrease in perioperative blood loss [28], whilst the other studies reported no significant difference [27, 29, 30]. Between those who receive bilateral laminotomy and those undergoing conventional laminectomy, moderate-quality evidence suggests no difference regarding perioperative blood loss (two RCTs, 142 participants, MD −20.2, 95 % CI −89.5 to 49.2). Between those who receive unilateral laminotomy and those treated with conventional laminectomy, low-quality evidence shows less perioperative blood loss with unilateral laminotomy (two RCTs, 114 participants, MD −34.1, 95 % CI −37.7 to −30.4). The quality of evidence had to be decreased because of the relatively high impact of one low-quality study on the combined quantitative analysis [25]. Between those who receive split-spinous process laminotomy and those undergoing conventional laminectomy, moderate-quality evidence shows no difference regarding perioperative blood loss (three RCTs, 141 participants, MD −3.8, 95 % CI −36.4 to 28.8).

Celik et al. reported no statistically significant difference regarding postoperative use of pethidine (mg) among participants treated with bilateral laminotomy compared with those treated with conventional laminectomy [2]. The quality of evidence was low (only one high-quality RCT, 71 participants, MD −53.0, 95 % CI −215.8 to 109.8). Yagi et al. reported significantly less use of diclofenac following unilateral (microendoscopic) laminotomy compared with conventional laminectomy, but study authors did not provide a quantitative comparison [25]. Watanabe et al. compared the use of non-steroidal anti-inflammatory drugs among participants treated with split-spinous process laminotomy versus conventional laminectomy during the first three days of follow-up [30]. The difference was not significant (mean 1.7 tablets in the split-spinous process laminotomy group and 2.3 tablets in the conventional laminectomy group).

Evidence provided by this systematic review for the effects of unilateral laminotomy for bilateral decompression, bilateral laminotomy and split-spinous process laminotomy compared with conventional laminectomy on functional disability, perceived recovery and leg pain is of low or very low quality. Therefore, further research is necessary to establish whether these techniques offer a safe and effective alternative to conventional laminectomy. Proposed advantages of these techniques regarding the incidence of iatrogenic instability and postoperative back pain are plausible, but definitive conclusions are limited by poor methodology and poor reporting of outcome measures among included studies. Future research is necessary to establish the incidence of iatrogenic instability using standardised definitions of radiological and clinical instability at comparable follow-up intervals. Additionally, long-term results of these techniques are currently lacking.

More methodologically rigorous studies are needed to compare techniques of decompression for lumbar stenosis before high-quality evidence-based recommendations can be made. The methodological quality of studies can be vastly improved with the use of adequate randomisation methods and blinding of participants and outcome assessors. Comparability of studies can be improved by standardising the outcome measures and the follow-up time points. In addition, more long-term outcome data (i.e. 5 years) are needed. More specifically, future research is necessary to allow for the distinction of subgroups based on anatomical characteristics of stenosis (e.g. single- vs multiple-level stenosis, bony vs soft stenosis). Further, the clinical relevance of radiological instability regarding symptom severity and reoperation rate with instrumented fusion should be addressed. Finally, researchers should conduct studies that compare a technique involving only partial resection of the vertebral arch with removal of the posterior midline structures or laminoplasty versus conventional laminectomy.