Minimally invasive discectomy versus microdiscectomy/open discectomy for symptomatic lumbar disc herniation

Abstract

Background

Microdiscectomy or open discectomy (MD/OD) are the standard procedures for symptomatic lumbar disc herniation and they involve removal of the portion of the intervertebral disc compressing the nerve root or spinal cord (or both) with or without the aid of a headlight loupe or microscope magnification. Potential advantages of newer minimally invasive discectomy (MID) procedures over standard MD/OD include less blood loss, less postoperative pain, shorter hospitalisation and earlier return to work.

Objectives

To compare the benefits and harms of MID versus MD/OD for management of lumbar intervertebral discopathy.

Search methods

We searched the Cochrane Central Register of Controlled Trials (CENTRAL) (November 2013), MEDLINE (1946 to November 2013) and EMBASE (1974 to November 2013) and applied no language restrictions. We also contacted experts in the field for additional studies and reviewed reference lists of relevant studies.

Selection criteria

We selected randomised controlled trials (RCTs) and quasi‐randomised controlled trials (QRCTs) that compared MD/OD with a MID (percutaneous endoscopic interlaminar or transforaminal lumbar discectomy, transmuscular tubular microdiscectomy and automated percutaneous lumbar discectomy) for treatment of adults with lumbar radiculopathy secondary to discopathy. We evaluated the following primary outcomes: pain related to sciatica or low back pain (LBP) as measured by a visual analogue scale, sciatic specific outcomes such as neurological deficit of lower extremity or bowel/urinary incontinence and functional outcomes (including daily activity or return to work). We also evaluated the following secondary outcomes: complications of surgery, duration of hospital stay, postoperative opioid use, quality of life and overall participant satisfaction. Two authors checked data abstractions and articles for inclusion. We resolved discrepancies by consensus.

Data collection and analysis

We used standard methodological procedures expected by The Cochrane Collaboration. We used pre‐developed forms to extract data and two authors independently assessed risk of bias. For statistical analysis, we used risk ratio (RR) for dichotomous outcomes and mean difference (MD) for continuous outcomes with 95% confidence intervals (CI) for each outcome.

Main results

We identified 11 studies (1172 participants). We assessed seven out of 11 studies as having high overall risk of bias. There was low‐quality evidence that MID was associated with worse leg pain than MD/OD at follow‐up ranging from six months to two years (e.g. at one year: MD 0.13, 95% CI 0.09 to 0.16), but differences were small (less than 0.5 points on a 0 to 10 scale) and did not meet standard thresholds for clinically meaningful differences. There was low‐quality evidence that MID was associated with worse LBP than MD/OD at six‐month follow‐up (MD 0.35, 95% CI 0.19 to 0.51) and at two years (MD 0.54, 95% CI 0.29 to 0.79). There was no significant difference at one year (0 to 10 scale: MD 0.19, 95% CI ‐0.22 to 0.59). Statistical heterogeneity was small to high (I² statistic = 35% at six months, 90% at one year and 65% at two years). There were no clear differences between MID techniques and MD/OD on other primary outcomes related to functional disability (Oswestry Disability Index greater than six months postoperatively) and persistence of motor and sensory neurological deficits, though evidence on neurological deficits was limited by the small numbers of participants in the trials with neurological deficits at baseline. There was just one study for each of the sciatica‐specific outcomes including the Sciatica Bothersomeness Index and the Sciatica Frequency Index, which did not need further analysis. For secondary outcomes, MID was associated with lower risk of surgical site and other infections, but higher risk of re‐hospitalisation due to recurrent disc herniation. In addition, MID was associated with slightly lower quality of life (less than 5 points on a 100‐point scale) on some measures of quality of life, such as some physical subclasses of the 36‐item Short Form. Some trials found MID to be associated with shorter duration of hospitalisation than MD/OD, but results were inconsistent.

Authors' conclusions

MID may be inferior in terms of relief of leg pain, LBP and re‐hospitalisation; however, differences in pain relief appeared to be small and may not be clinically important. Potential advantages of MID are lower risk of surgical site and other infections. MID may be associated with shorter hospital stay but the evidence was inconsistent. Given these potential advantages, more research is needed to define appropriate indications for MID as an alternative to standard MD/OD.

Plain language summary

Surgery for leg and back pain caused by damage to spinal discs

Background

When the discs between vertebrae in the spine become damaged (herniated), the soft gel inside them pushes through the wall of the disc and presses against the nerves or the spinal cord, causing a burning pain in legs and pain in the back. When this happens in the lower back, it i s known as lumbar disc herniation.

Review question

The main treatment for this condition is lumbar discectomy, which involves removing the part of the disc pressing on the nerves. There are two main types of this surgery. The first type is standard microdiscectomy, which can be performed with the aid of microscope magnification or headlight loupe, or open discectomy where surgeons do not use a microscope or loupe (MD/OD). However, all steps of the operations are similar. The second type of operation is minimally invasive discectomy (MID) procedures. MID involves a smaller incision and less damage to the surrounding tissue. We reviewed the evidence to see if one type of surgery is more effective than the other type of surgery in terms of the results after surgery including pain in the legs, low back pain, problems with mobility or numbness and disability.

Study characteristics

We found 11 studies up to November 2013, examining 1172 people, with studies ranging from 22 to 325 participants, and people aged from 12 to 70 years. All had tried non‐surgical treatments and all had leg pain that was worse than their back pain. The follow‐up period after surgery ranged from five days to 56 months.

Key results

People who had a MD/OD had less pain in their legs, and less low back pain, but the difference was small. They were less likely to need a second operation because the first had been unsuccessful. They felt slightly better in some physical aspects of their quality of life, but again the difference was too small to be meaningful. In terms of complications, the two operations were similar, though people who had a MD/OD were more likely to have wound infections.

Quality of evidence

Many of the studies were carried out on a small number of people and had a high risk of bias, so the overall quality of the evidence for leg and low back pain was low.

Summary of findings

Summary of findings for the main comparison. Minimal invasive discectomy compared with micro/discectomy for lumbar disc herniation.

Minimal invasive discectomy compared with micro/discectomy for lumbar disc herniation
Participant or population: participants with lumbar disc herniation Settings: operated lumbar disc herniation Intervention: minimally invasive discectomy Comparison: micro/discectomy
Outcomes#	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No of participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Micro/discectomy	Minimal invasive discectomy
Mean leg pain intensity on a numerical scale, e.g. 0 (no pain) to 10 (maximum pain) ‐ 1‐year post operative	Mean leg pain score ranged across control groups from 0.1 to 1	Mean leg pain score in the intervention groups was 0.13 higher (0.09 to 0.16)	Not applicable	599 (4 studies)	⊕⊕⊝⊝ low1	There was a difference between the groups that was small in magnitude. However, the difference was not clinically important (< 1.5 out of 10 points)
Mean LBP intensity on a numerical scale, e.g. 0 (no pain) to 10 (maximum pain) ‐ 6‐month post operative	Mean LBP score ranged across control groups from 1 to 1.77	Mean LBP score in the intervention groups was 0.35 higher (0.19 to 0.51)	Not applicable	577 (3 studies)	⊕⊕⊝⊝ low2	There was a difference between the groups that was small in magnitude. However, the difference was not clinically important (<1.5 out of 10 points)
Mean LBP intensity on a numerical scale, e.g. 0 (no pain) to 10 (maximum pain) ‐ 1‐year post operative	Mean LBP score ranged across control groups from 0 to 1.75	Mean LBP score in the intervention groups was 0.19 higher (‐0.22 to 0.59)	Not applicable	577 (3 studies)	⊕⊝⊝⊝ very low3	There was no statistically significant difference
Mean LBP intensity on a numerical scale, e.g. 0 (no pain) to 10 (maximum pain) ‐ 2‐year post operative	Mean LBP score ranged across control groups from 0 to 1.94	Mean LBP score in the intervention groups was 0.54 higher (0.29 to 0.79)	Not applicable	577 (3 studies)	⊕⊕⊝⊝ low4	There was a difference between the groups that was small in magnitude. However, the difference was not clinically important (< 1.5 out of 10 points)
Persistent motor deficits post operative	Study population 343.7 per 1000	Study population 338.7 per 1000	Not applicable	126 (4 studies)	⊕⊕⊝⊝ low5	There was no statistically significant difference
Persistent sensory deficits post operative	Study population 550 per 1000	Study population 459 per 1000	Not applicable	165 (4 studies)	⊕⊕⊝⊝ low6	There was no statistically significant difference
Persistent reflex deficit post operative (12 months)	Study population 737 per 1000	Study population 500 per 1000	Not applicable	47 (2 studies)	⊕⊕⊝⊝ low7	There was a difference between the groups. However, the difference was not clinically important
Disability (higher ratings mean greater disability). Various instruments were used, e.g. in Oswestry Disability Index > 6 months' post operative 0% (no disability) to 100% (bedridden)	Mean disability score ranged across control groups from 10 to 13	Mean disability score in the intervention groups was 0.84 higher (‐0.21 to 1.88)	Not applicable	312 (3 studies)	⊕⊕⊝⊝ low8	There was no statistically significant difference
Side effects ‐ surgical site and other infections > 6 months' follow‐up	Study population 32 per 1000	Study population 2.3 per 1000	RR 0.23 (0.07 to 0.79)	931 (6 studies)	⊕⊕⊕⊝ moderate9	There was a difference between the groups that was small in magnitude. However, the difference was not clinically important (< 10%)
Side effects ‐ re‐hospitalisation due to recurrent disc herniation ≥ 6 months' follow‐up	Study population 43 per 1000	Study population 75 per 1000 (43 to 103)	RR 1.74 (1.03 to 2.94)	949 (6 studies)	⊕⊕⊕⊝ low10	The magnitude of this difference was small to moderate. However, the difference was not clinically important (< 10%)
SF‐36 Physical Functioning subclass > 6 months' follow‐up ‐ on a numerical scale (higher ratings mean higher quality of life), e.g. 0 (the worst) and 100 (the highest) quality	Mean HRQoL score ranged across control groups from 80.4 to 84	Mean HRQoL score in the intervention group was 4.7 lower (‐5.05 to ‐4.35)	Not applicable	385 (2 studies)	⊕⊕⊝⊝ low11	The magnitude of this difference was in the range of small to moderate. However, this difference was not clinically important (< 10%)
The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; HRQoL: health‐related quality of life; LBP: low back pain; RR: risk ratio; SF‐36: 36‐item Short Form. GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.

^# Persistent bladder dysfunction > six months' follow‐up, sciatica‐specific outcomes including the Sciatica Bothersomeness Index (SBI) and the Sciatica Frequency Index (SFI) have been written in the text but did not mention in the 'Summary of finding' table because there was only one study for each and no need for further evaluation and meta‐analysis.

¹ Two trials had unclear risk of bias due to non‐adequate randomisation, no allocation concealment and no blinding (Huang 2005; Mayer 1993). Meanwhile, Huang 2005 had no clear intention‐to‐treat analysis and Mayer 1993 had unclear co‐intervention. One other trial had high risk of bias for blinding of participants and personnel (performance bias) and unclear risk of bias for co‐interventions (Arts 2011). Another trial had unclear allocation concealment and high risk of bias for blinding (Teli 2010). We downgraded the level of evidence due to the high risk of bias and low precision due to the small number of trials for specific minimally invasive techniques. The analysis was consistent with no heterogeneity (I² = 0%).

² One trial had high risk of bias due to non‐adequate randomisation, no allocation concealment, no blinding, group differences at baseline and no clear intention‐to‐treat analysis (Righesso 2007). One other trial had high risk of bias for blinding of participants and personnel (performance bias) and unclear risk of bias for co‐interventions (Arts 2011). Another trial had unclear allocation concealment and high risk of bias for blinding (Teli 2010). We downgraded the level of evidence due to the high risk of bias and low precision because of the small number of trials for specific minimally invasive techniques. The analysis was almost consistent with small heterogeneity (I² = 35%).

³ One trial had high risk of bias due to non‐adequate randomisation, no allocation concealment, no blinding, group differences at baseline and no clear intention‐to‐treat analysis (Righesso 2007). One other trial had high risk of bias for blinding of participants and personnel (performance bias) and unclear risk of bias for co‐interventions (Arts 2011). Another trial had unclear allocation concealment and high risk of bias for blinding (Teli 2010). We downgraded the level of evidence due to the high risk of bias, low precision because of the small number of trials for specific minimally invasive techniques and high heterogeneity (I² = 90%).

⁴ One trial had high risk of bias due to non‐adequate randomisation, no allocation concealment, no blinding, group differences at baseline and no clear intention‐to‐treat analysis (Righesso 2007). One other trial had high risk of bias for blinding of participants and personnel (performance bias) and unclear risk of bias for co‐interventions (Arts 2011). Another trial had unclear allocation concealment and high risk of bias for blinding (Teli 2010). We downgraded the level of evidence due to the high risk of bias and low precision because of the small number of trials for specific minimally invasive techniques. The analysis was almost consistent with small heterogeneity (I² = 65%).

⁵ Two trials had high risk of bias due to non‐adequate randomisation, allocation concealment and no blinding (Ryang 2008; Righesso 2007). One trial had unclear risk of bias due to non‐adequate randomisation, no allocation concealment, no blinding and unclear co‐intervention (Mayer 1993). In another trial, there was no blinding of participants, care provider and outcome assessor; and authors did not mention whether there was any co‐intervention or not (Hermantin 1999). The analysis was consistent with small heterogeneity (I² = 15%).

⁶ Two trials had high risk of bias due to non‐adequate randomisation, allocation concealment and no blinding (Ryang 2008; Righesso 2007). One trial had unclear risk of bias due to non‐adequate randomisation, no allocation concealment, no blinding and unclear co‐intervention (Mayer 1993). In another trial, there was no blinding of participants, care provider and outcome assessor; and authors did not mention whether there was any co‐intervention or not (Hermantin 1999). The analysis was consistent with no heterogeneity (I² = 0%).

⁷ One trial had unclear risk of bias due to non‐adequate randomisation, no allocation concealment, no blinding and unclear co‐intervention (Mayer 1993). In another trial, there was no blinding of participants, care provider and outcome assessor; and authors did not mention whether there was any co‐intervention or not (Hermantin 1999). There was statistically significant difference between minimally invasive discectomy and microdiscectomy/open discectomy. However, we downgraded the evidence because of high risk of bias and low precision due to small sample size. The analysis was consistent with no heterogeneity (I² = 0%).

⁸ Only one trial had overall low risk of bias (Teli 2010). Both other trials had high risk of bias due to non‐adequate randomisation, allocation concealment and no blinding (Ryang 2008; Righesso 2007). There was no statistically significant difference between different types of minimally invasive discectomy (microendoscopic discectomy and minimal access trocar microdiscectomy) versus microdiscectomy. However, we downgraded the evidence because of high risk of bias and low precision due to small sample size. The analysis was consistent with no heterogeneity (I² = 0%).

⁹ Six trials evaluated side effects of postoperative surgical infection. One trial had high risk of bias for blinding of participants and personnel (performance bias) and unclear risk of bias for co‐interventions (Arts 2011). Another trial had unclear allocation concealment and high risk of bias for blinding (Teli 2010). One trial had high risk of bias due to no clear randomisation method, no clear allocation concealment, no clear blinding of participants and personnel, no clear selective reporting and no clear intention‐to‐treat analysis (Garg 2011). One trial had high risk of bias due to no clear randomisation, no clear allocation concealment, no clear blinding and no clear intention‐to‐treat analysis (Huang 2005). One trial had high risk of bias because of non‐randomisation, no allocation, no blinding, no clear group similarity at baseline and no clear intention‐to‐treat analysis (Ruetten 2008). However, there was no blinding of participants, care provider and outcome assessor. Authors did not mention whether there was any co‐intervention or not. Lack of blindness, and unclear co‐intervention were major problems of the one trial that had no other risk of bias (Hermantin 1999). We downgraded the evidence due to the high risk of bias. One participant out of 431 had infection following minimally invasive discectomy versus 16 out of 500 participants in microdiscectomy/open discectomy. The analysis was almost consistent with small heterogeneity (I² = 34%).

¹⁰ One trial had high risk of bias for blinding of participants and personnel (performance bias) and unclear risk of bias for co‐interventions (Arts 2011). Another trial had unclear allocation concealment and high risk of bias for blinding (Teli 2010). However, four other trials had high risk of bias because of non‐randomisation, no allocation and no blinding of participants and personnel (Garg 2011; Mayer 1993; Ruetten 2008; Ryang 2008). We downgraded the evidence to the low level due to high risk of bias and imprecision (wide 95% confidence intervals). The analysis was consistent with no heterogeneity (I² = 0%).

¹¹ One trial had high risk of bias for blinding of participants and personnel (performance bias) and unclear risk of bias for co‐interventions (Arts 2011). Another trial had high overall risk of bias because of non‐randomisation, no allocation concealment, no blinding, unclear group similarity at baseline and no clear intention‐to‐treat analysis (Ryang 2008). We downgraded the level of evidence due to the high risk of bias and low precision because of small number of trials. The analysis was consistent with no heterogeneity (I² = 0%).

Background

Lumbar discectomy involves removal of all or part of one or more intervertebral discs. Intervertebral discs separate and cushion the spinal vertebral bodies. In people with a protruded or herniated disc, the soft gel inside pushes through the wall of the disc. Pathophysiology of lumbar discopathy suggests that probably inflammation is more important than pressure. By removing part of the intervertebral disc, lumbar discectomy can relieve symptoms.

The lumbar discectomy procedure remained basically unchanged until the operating microscope enhanced the visualisation of the operative field in 1978. This new operation was described as lumbar microdiscectomy because it was performed through a smaller incision, with less dissection than open lumbar discectomy (Arts 2008). Microdiscectomy is generally regarded as a technical modification of standard discectomy, rather than a distinct procedure (Koebbe 2002). In one systematic review by Gibson and Waddell, results of microdiscectomy for treatment of lumbar disc prolapse were "broadly comparable" to results of open lumbar discectomy (Gibson 2007). Microdiscectomy is now the most common surgical procedure for lumbar disc herniation (Koebbe 2002).

More recently, several minimally invasive surgical approaches have been introduced for the surgical management of symptomatic lumbar disc herniation, utilising technologies to reduce incision size and the area of dissection further (Deen 2003). Systematic reviews have compared specific types of minimally invasive lumbar surgery for the management of lumbar disc herniation and radiculopathy (Dasenbrock 2012; Jacobs 2012; Nellensteijn 2010; Singh 2009a; Singh 2009b), but did not yield conclusive results, in part due to limited evidence. Given the availability of newer evidence, we performed a systematic review of the literature to evaluate the benefits and harms of minimally invasive discectomy (MID) versus microdiscectomy/open discectomy (MD/OD).

Description of the condition

Lumbar discopathy (disc herniation) often presents clinically as radiculopathy (sciatica), which results from compression of one or more spinal nerve roots, and manifests as radiating leg pain and paraesthesias (a sensation of tingling, burning and numbness), with or without a neurological deficit. Approximately 3% to 4% of people presenting with low back pain (LBP) have lumbar radiculopathy. Lumbar disc herniation, defined as displacement of central disc material (nucleus) beyond the margins of the intervertebral disc space, is considered to be the most common cause (90%) of radiculopathy (Hahne 2010; Koes 2007). In people who have severe symptoms refractory to conservative management for six to eight weeks, imaging studies are often indicated. If imaging demonstrates correlative disc pathology, the person may be a surgical candidate (Koes 2007). Progressive or severe neurological deficits is an indication for more urgent surgery. Neurological symptoms can manifest as progressive lower extremity muscle weakness or urinary or bowel (or both) incontinence. Surgical intervention for radiculopathy (discectomy) is based on removal of herniated disc materials to relieve nerve root irritation or compression or both (Gibson 2007).

Description of the intervention

Although the first lumbar disc surgery was performed in 1934 by Mixter and Barr (Mixter 1934), few technical changes occurred in this field until the 1970s, when the operating microscope and subsequent microdiscectomy were introduced. The use of an operating microscope allowed a smaller incision, with outcomes generally comparable to open discectomy (Gibson 2007). Microdiscectomy, or discectomy that is performed under microscope, is now a common procedure for the management of lumbar radiculopathy resulting from lumbar disc herniation (Arts 2008; Gibson 2007; Haines 2002; Thomé 2005; Tullberg 1993). Since the mid‐1970s, several minimally invasive procedures have been developed as alternatives to standard MD/OD. Minimally invasive spine surgeries use technologies to reduce incision size and the area of tissue dissection further (Krappel 2001), potentially resulting in reduced recovery time and better cosmetic results (Jaikumar 2002).

The first generation of minimally invasive procedures involved blind percutaneous techniques including chemonucleolysis, percutaneous nucleotomy (Kahanovitz 1990; Schreiber 1986), automated percutaneous nucleotomy (Stevenson 1995), and laser disc decompression (Mathews 2002). Newer techniques such as biportal arthroscopic intradiscal discectomy utilise rigid discoscopes and flexible endoscopes to provide additional visualisation of the operative field (Mathews 2002; Sharif‐Alhoseini 2011). Transmuscular tubular discectomy replaces the subperiosteal muscle dissection required in standard MD/OD with a muscle‐splitting transmuscular approach.

See Table 2 for a description of minimally invasive procedures included in this review.

1. Brief description of the minimally invasive procedures.

Reference	Minimal invasive procedure	Description of the procedure
Kahanovitz 1990	Percutaneous nucleotomy	Under fluoroscopy in the posterior‐lateral position, a K‐wire was advanced into the intervertebral space and a dilator and working cannula were introduced into the disc space step by step. Discectomy was performed through the cannula using pituitary forceps
Onik 1985	Automated percutaneous discectomy	Through a lateral oblique percutaneous approach, and insertion of 2‐mm disk‐aspiration probe, nucleus pulposus was mechanically removed
Ditsworth 1998	Percutaneous endoscopic lumbar discectomy (PELD)	The identified symptomatic disc can be dissected by interlaminar or transforaminal approach using endoscope
Arts 2009	Transmuscular tubular microdiscectomy	Tubular discectomy utilises a transmuscular approach rather than a subperiosteal dissection. In this method, a guidewire is inserted percutaneously into the inferior part of the lamina, and its location is confirmed using fluoroscopy. Then, dilators of increasing diameter are inserted sequentially over the guidewire. The tubular retractor is then inserted over the final dilator

How the intervention might work

MD/OD removes the intervertebral disc portion that has compressed the nerve root or spinal cord, or both. MID procedures are performed using techniques that allow for smaller incisions or less dissection (or both) than microdiscectomy, potentially resulting in lower blood loss, less postoperative pain, shorter hospitalisation and earlier return to work compared with MD/OD (Deen 2003; Mathews 2002). Potential disadvantages of minimally invasive lumbar surgery include a sharp learning curve for the surgeon. In other words, the duration of surgical time of any new technique such as MID will decrease over the course of the learning curve, and evaluation of the efficacy of the MID against MD/OD may be affected by when the comparison occurred on the learning curve (Mathews 2002; Rahimi‐Movaghar 2009; Rahimi‐Movaghar 2010; Rahimi‐Movaghar 2011). In addition, studies have raised the question of whether minimally invasive procedures are as safe and efficient as expected (Arts 2011; Chatterjee 1995; Chung 1999; Fourney 2010; Franke 2009; Hermantin 1999; Krappel 2001; Lew 2001; Mayer 1993; Ruetten 2008).

Why it is important to do this review

Previous studies examining the efficacy and safety of different types of MID (percutaneous nucleotomy, automated percutaneous discectomy, percutaneous endoscopic lumbar discectomy (PELD), transmuscular tubular microdiscectomy) have yielded inconsistent findings. Given the availability of new evidence, the objective of this systematic review was to clarify whether these procedures provide an advantage compared with MD/OD.

Objectives

To compare the benefits and harms of MID versus MD/OD for management of lumbar intervertebral discopathy.

Methods

Criteria for considering studies for this review

Types of studies

For this review, we primarily considered randomised controlled trials (RCTs) and quasi‐randomised controlled trials (QRCTs) comparing MD/OD with a minimally invasive procedure. In RCTs, computer‐generated sequences or randomisation tables are typically used to allocate people to different groups. QRCTs use a method of allocation, such as date of birth or day of the week, which is not truly randomised, resulting in a greater risk of selection bias. We adapted our methodology to follow the recommendations of the Cochrane Back Review Group (Furlan 2009).

Types of participants

We selected studies of adults (aged greater than 12 years and less than 70 years) undergoing surgical treatment for lumbar radiculopathy secondary to herniated discs. Number of levels could be one or multiple. We did not exclude people based on gender or duration of symptoms. We excluded trials of people with lumbar radiculopathy due to causes other than prolapsed disc.

Types of interventions

We included trials of MID compared with MD/OD. We included MID procedures in which the treatment mirrored that of MD/OD in terms of removal of the problematic disc material. These MID procedures include percutaneous endoscopic interlaminar or transforaminal lumbar discectomy (Ditsworth 1998; Mathews 1996), transmuscular tubular microdiscectomy and automated percutaneous lumbar discectomy. We excluded trials of chemical nucleolysis, intradiscal electrothermal annuloplasty (Pauza 2004; Saal 2002), laser discectomy, the Dekompressor and Coblation nucleoplasty (radiofrequency radio waves) in which the mechanism of action involved destruction or disruption of the disc causing mechanical compression using energy or chemicals, rather than removal of disc materials.

Types of outcome measures

We evaluated clinical and functional outcome measures. Where outcome measures were composite, we tried to use item sub‐scores, but we also performed analyses at the composite level. We categorised follow‐up periods as immediately postoperative (first six weeks post operative, which is more a reflection of 'surgical' pain), short term (from six week's post‐operative to one year), medium term (one to five years) or long term (longer than five years).

Primary outcomes

We considered the following variables as primary outcomes:

1. pain measure by visual analogue scale (VAS) for each of sciatica and LBP;

2. sciatica‐specific outcomes: the Sciatica Bothersomeness Index (SBI) and the Sciatica Frequency Index (SFI) (Grøvle 2008);

3. neurological deficit of lower extremity or bowel/urinary incontinence;

4. functional outcome including daily activity and return to work, measured by scales such as the Oswestry Disability Index (ODI) or the Roland‐Morris Disability score.

Secondary outcomes

1. Complications of surgery including mortality and common adverse events including:

   1. thromboembolic complications;

   2. surgical site and other infections;

   3. procedure‐related complications;

   4. re‐hospitalisation due to recurrent disc herniation;

   5. re‐hospitalisation due to other causes;

   6. surgical re‐intervention;

   7. dural tear.

2. Duration of hospital stay.

3. Opioid use.

4. Quality of life measured by 36‐item Short Form (SF‐36) or 12‐item Short Form (SF‐12), and overall satisfaction of participants, which is usually reported using a Likert scale.

Search methods for identification of studies

Search strategies for the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE and EMBASE are in Appendix 1. We used specific search terms to identify studies of MID to exclude non‐randomised studies. We applied no language or date restrictions. We did not consider unpublished records in this review.

Electronic searches

We searched the following databases:

• CENTRAL (October 2013) (Ovid SP);

• MEDLINE (1946 to 22 November 2013) (Ovid SP);

• EMBASE (1974 to 22 November 2013) (Ovid SP) (Figure 1).

1.

Study flow diagram.

Searching other resources

We contacted experts in the field and reviewed reference lists of relevant articles.

Data collection and analysis

Selection of studies

One review author (MRR) independently reviewed each title/abstract and list potentially relevant references. A second review author (VRM) checked data abstractions and articles for inclusion. We used a consensus process resolved discrepancies (MRR, VRM, MML, RC, FS).

Data extraction and management

We extracted data into pre‐developed forms. From each study, we collected basic information concerning authors (affiliation, sponsoring), methods (study design, sample size), participants (selection criteria and diagnoses, pain location, age, gender, neurological deficit, quality of life, functional disability), treatments (surgical technique, materials used, levels involved), control treatments (MD/OD) and outcome variables with results. Two review authors extracted data independently. A second review author re‐checked entered data; there were 0.85% minor errors, which were corrected.

Assessment of risk of bias in included studies

We assessed risk of bias using the 12 criteria recommended by the Cochrane Back Review Group (Furlan 2009) (Appendix 3). We scored the items as 'low risk', 'high risk' or 'unclear'. We considered studies to have a 'low risk of bias' overall when they met at least six of the 12 criteria and the study had no serious methodological flaws.

Authors of trials did not assess their own trial. We used the 'Risk of bias' assessment in sensitivity analysis to compare the results after excluding trials with high risk of bias compared with the results including all studies.

We pilot tested our assessment tool to ensure that a similar approach was used across the review team. Two review authors (MRR, VRM) independently assessed the study design and the risk of bias. We calculated the inter‐observer reliability for risk of bias assessment. There were 11 studies with 132 items for risk of bias. Both observers agreed on 84 low risk of bias and agreed on 34 high or unclear risk of bias. Two observers had disagreements in 14 (10+4) items. Therefore, observed agreement was 89%. Expected agreement was 57%. The kappa statistic (or kappa coefficient) was 0.74, which we considered as substantial agreement. Two observers worked together and solved 14 out of 132 items. We contacted authors of trials published since the mid‐2000s to obtain missing information about study methods (randomisation, allocation concealment, blinding).

Measures of treatment effect

We extracted both dichotomous outcomes expressed or calculated as risk ratios (RR) and continuous outcomes reported or calculated as mean difference (MD). For dichotomous outcomes, we reported risk differences in addition to RR estimates. However, for outcomes such as functional status or quality of life, which are measured using different scales, we reported the standardised mean difference (SMD). When calculating an MD, we converted scales to similar measures (e.g. all measures converted to a 100‐ or 10‐point scale). We extracted or calculated a 95% confidence interval (CI) for each outcome. We evaluated clinical relevance with the five questions listed in Appendix 2 (Furlan 2009). We used the guideline suggested by Ostelo 2008 to define clinically important changes.

We summarised effects according to the timing of the outcomes: immediately postoperative (first six weeks' post operative), short term (from six weeks' postoperative to one year), medium term (one to five years) or long term (longer than five years).

Unit of analysis issues

For cluster randomised trials, we conducted the analysis at the same level as the allocation, using a summary measurement from each cluster. If data were available at the individual level, we analysed these data while accounting for the cluster design.

Dealing with missing data

We contacted authors of included trials to request missing information about methodological properties (randomisation, allocation concealment, blinding) and missing data including missing standard deviations (SD) of the trials. We only included trials that were missing less than 20% of clinical data for immediate, short‐ and medium‐term follow‐ups in the analysis. We estimated missing information about parameter variability from ranges if provided, or estimated from comparable trials.

Assessment of heterogeneity

We examined whether the included studies were sufficiently clinically similar by examining the populations, interventions, controls and outcomes (PICO) and other study characteristics (e.g. study design: RCT or QRCT). We did not pool studies with large clinical or statistical differences.

We evaluated statistical heterogeneity by assessing for non‐overlapping CIs, and defined significant statistical heterogeneity as a Q‐test with a P value lower than 0.1 or an I² statistic greater than 75. If we determined that the results were too heterogeneous to pool, we tried to identify explanations for the heterogeneity through subgroup and sensitivity analyses (see Sensitivity analysis). As background factors such as mean age of the participants and sex ratio may affect treatment outcome, we also tried to adjust these factors using meta‐regression by STATA.

Assessment of reporting biases

We did not formally assess for publication bias because there were too few studies to perform visual or statistical methods for markers of publication bias reliably (e.g. small sample effects).

Data synthesis

Clinical homogeneity is a prerequisite for pooling studies in a meta‐analysis. In the case of statistical heterogeneity, we tried to find an explanation using subgroup and sensitivity analyses. We described individual studies that could not be pooled narratively.

Regardless of whether there were sufficient data available to perform quantitative analyses to summarise the data, we assessed the overall quality of the evidence for each outcome. To accomplish this, we used the GRADE approach, as recommended in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), and adapted in the updated Cochrane Back Review Group guidelines (Furlan 2009). Factors that may decrease the quality of the evidence are: study design and risk of bias, inconsistency of results, indirectness (not generalisable), imprecision (sparse data) and other factors (e.g. reporting bias). The quality of the evidence for a specific outcome was reduced by one level, according to the performance of the studies against these five factors.

• High quality evidence: there were consistent findings among at least 75% of RCTs with no limitations of the study design, consistent, direct and precise data and no known or suspected publication biases. Further research is unlikely to change either the estimate or our confidence in the results.

• Moderate quality evidence: one of the domains was not met. Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.

• Low quality evidence: two of the domains were not met. Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.

• Very low quality evidence: three of the domains were not met. We are very uncertain about the results.

• No evidence: no RCTs were identified that addressed this outcome.

We used five questions listed in Appendix 2 to evaluate the clinical relevance of the review results. The results of this assessment informed the final results and conclusions.

Subgroup analysis and investigation of heterogeneity

To investigate heterogeneity, we analysed subgroups according to participant characteristics, including neurological deficit, radicular pain and LBP. We also analysed subgroups according to the different surgical procedures in specific minimally invasive techniques and in two subgroups of microdiscectomy and discectomy.

Sensitivity analysis

If we had a sufficient number of studies and comparisons, we used sensitivity analyses to determine the robustness of the review findings. In particular, we performed sensitivity analysis to determine whether the results of the review change when trials are excluded for the following reasons: 1. high risk of bias, 2. unclear methods of randomisation and 3. missing data were estimated. We also included the source of funding in the sensitivity analysis comparing studies that were funded by the industry with studies that were funded by other sources (government, non‐profit organisations and institutions).

Results

Description of studies

We identified 11 studies (sample sizes ranged from 22 to 325 participants; total number of participants across studies was 1172) of MID versus microdiscectomy/discectomy (Arts 2011; Chatterjee 1995; Garg 2011; Hermantin 1999; Huang 2005; Mayer 1993; Righesso 2007; Ruetten 2008; Ryang 2008; Shin 2008; Teli 2010). Eight studies evaluated PELD, one of these eight studies simultaneously evaluated transforaminal and intralaminar endoscopic discectomy, two studies evaluated transmuscular tubular microdiscectomy and one study evaluated automated percutaneous discectomy. We assessed four out of 11 studies as having low overall risk of bias with clear methods of random sequence generations (Arts 2011; Hermantin 1999; Shin 2008; Teli 2010). We assessed the remaining seven studies as having high overall risk of bias or unclear/non‐random sequence generations. With regards to the comparison groups, eight studies evaluated microdiscectomy and three studies evaluated open discectomy. Microdiscectomy was performed by microscope magnification in six studies; headlight loupe in one study and with both microscope magnification and headlight loupe in one study. Duration of follow‐up in the studies ranged from six to 56 months. One study only evaluated participants in the first five days post operation (Shin 2008).

Results of the search

Searches identified 841 references. After exclusion of 370 duplicates, 471 references remained: 74 from CENTRAL, 261 from MEDLINE and 136 from EMBASE. We added three additional references from other resources. Eleven studies met inclusion criteria.

For some of the studies there were missing data or missing information (or both) relevant to the risk of bias assessment. We requested this information from study authors but did not receive any additional information.

Included studies

We included 11 studies (Arts 2011; Chatterjee 1995; Garg 2011; Hermantin 1999; Huang 2005; Mayer 1993; Righesso 2007; Ruetten 2008; Ryang 2008; Shin 2008; Teli 2010).

Excluded studies

We excluded five studies because they compared MD versus OD or discectomy versus sequestrectomy or different types of MID (Henriksen 1996; Lagarrigue 1994; Thomé 2005; Tullberg 1993; Türeyen 2003). We excluded a retrospective observational comparison of MID and discectomy (Harrington 2008). Finally, we excluded three RCTs because they did not meet our inclusion criteria (Franke 2009; Karasek 2000; van den Akker 2011). See Characteristics of excluded studies tables for more details.

Risk of bias in included studies

We assessed seven out of 11 studies as having high overall risk of bias (Chatterjee 1995; Garg 2011; Huang 2005; Mayer 1993; Righesso 2007; Ruetten 2008; Ryang 2008). We described risk of bias for each study in the 'Risk of bias' tables, and the ratings across studies are summarised in Figure 2.

2.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Allocation

We assessed four studies as having low risk for random sequence generation (Arts 2011; Hermantin 1999; Shin 2008; Teli 2010). However, only two studies reported allocation concealment and we assessed these as having low risk for selection bias (Arts 2011; Hermantin 1999). We assessed eight studies as having unclear risk of selection bias (Chatterjee 1995; Garg 2011; Huang 2005; Mayer 1993; Righesso 2007; Ryang 2008; Shin 2008; Teli 2010). We assessed one study as having high risk of selection bias (Ruetten 2008), due to inadequate generation of a randomised sequence and inadequate concealment of allocations prior to assignment.

Blinding

No study was blinded to both participants and outcome assessors. Six studies did not clearly report use of both types of blinding (Chatterjee 1995; Garg 2011; Huang 2005; Mayer 1993; Ryang 2008; Shin 2008), and five were not blinded (Arts 2011; Hermantin 1999; Righesso 2007; Ruetten 2008; Teli 2010).

Incomplete outcome data

We assessed all studies as having low risk of attrition bias except two studies that did not mention it clearly (Ryang 2008; Shin 2008).

Selective reporting

We assessed all studies as having low risk of selective reporting bias except one study that did not mention it clearly (Garg 2011).

Other potential sources of bias

We identified no other potential sources of biases in four studies (Arts 2011; Hermantin 1999; Mayer 1993; Teli 2010). We assessed six studies as having unclear risk of bias related to use of intention‐to‐treat analysis (Garg 2011; Huang 2005; Righesso 2007; Ruetten 2008; Ryang 2008; Shin 2008). We assessed two studies as having high risk of bias related to group differences at baseline (Chatterjee 1995; Righesso 2007). Two other studies did not mention anything about group difference at baseline. Therefore, we rated them as having unclear risk of bias (Ruetten 2008; Shin 2008). Authors did not send us any additional information for missing data, including unclear risk of bias and on SDs).

Effects of interventions

See: Table 1

Results for comparisons between MID versus MD/OD are shown in Table 3. We were unable to perform meta‐regression due to the small number of trials.

2. Summary of 11 studies and their interventions and evaluated outcomes.

	S1	S2	S3	S4	S5	S6	S7	S8	S9	S10	S11
Q1	c	b	a	a	a	a	a	a1/a2	c	a	a
Q2	1	1	2	2	2	1	1b	1	1	1	1a/1b
PO1	+	‐	‐	‐	‐	‐	+	‐	‐	+	+
PO2	+	‐	‐	+	+	+	+	+	+	+	+
PO2a	+	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐
PO2b	+	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐
PO3a	‐	‐	‐	+	‐	+	+	‐	+	‐	+
PO3b	‐	‐	‐	+	‐	+	+	‐	+	‐	‐
PO3c	‐	‐	‐	+	‐	+	+	‐	‐	‐	‐
PO3d	‐	‐	‐	‐	‐	‐	‐	‐	+	‐	‐
PO3e	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐
PO4a1	‐	‐	+	‐	‐	‐	+	+	+	‐	+
PO4a2	+	‐	‐	+	‐	‐	‐	‐	‐	‐	‐
PO4b	+	‐	‐	+	‐	‐	‐	+	‐	‐	‐
SO1	‐	‐	+	‐	‐	‐	‐	‐	‐	‐	+
SO1a	+	‐	+	‐	+	‐	‐	+	+	‐	‐
SO1b	+	‐	+	+	+	‐	‐	+	‐	‐	+
SO1c	+	‐	+	‐	‐	‐	‐	+	‐	‐	‐
SO1d	+	‐	+	‐	‐	+	‐	+	+	‐	+
SO1e	+	‐	+	‐	‐	‐	‐	+	‐	‐	‐
SO1f	+	‐	+	‐	‐	+	‐	+	+	‐	‐
SO1g	+	‐	+	‐	‐	‐	‐	+	+	‐	+
SO2	+	+	+	‐	+	‐	+	‐	+	‐	‐
SO3	‐	‐	+	+	‐	‐	‐	+	‐	‐	‐
SO4a	+	‐	‐	‐	‐	‐	‐	‐	+	‐	+
SO4b	‐	‐	+	+	+	‐	‐	+	‐	‐	‐
No.	325	71	112	60	22	40	40	200	60	30	212
Severity	++	++	++	++	++	++	++	++	++	++	++
Age (years)	18‐70	NA	26‐57	15‐67	39 ± 11	12‐55	< 60	20‐68	21‐69	43 ± 18/ 48 ± 11	18‐65
M : F MID	84 : 82	NA	36 : 19	22 : 8	6 : 4	12 : 8	10 : 11	84 : 116	13 : 17	7 : 8	45 : 25
M : F M/D	88 : 71	NA	44 : 13	17 : 13	9 : 3	14 : 6	13 : 6		19 : 11	5 : 10	94 : 48
F/O	≥ 24	≥ 6	12‐18	19‐42	10‐25	≥ 24	24‐56	≥ 24	6‐26	5 days	24‐29

Q1. Minimally invasive discectomy

a. percutaneous endoscopic discectomy (1. transforaminal = lateral approach; 2. interlaminar = posterior approach),

b. automated percutaneous discectomy,

c. transmuscular tubular microdiscectomy (guidewire, sequential dilators, tubular retractor with microscopic magnification)

Q2. Microdiscectomy/discectomy

1. Microdiscectomy (a. microscope magnification, b. headlight loupe)
2. Discectomy (without microscope or loupe)
3. Microdiscectomy/discectomy (with or without magnification by microscope or headlight loupe)

++. failure to respond to non‐operative measures ‐ 4‐8 weeks of conservative treatment with rest, analgesia, non‐steroidal anti‐inflammatory drugs and physiotherapy

M : F MID. Male : female ratio in the minimal invasive discectomy group

M : F M/D. Male : female ratio in the micro/discectomy

F/O: duration of follow‐up (range in months; exception is S10 with 5 days' follow‐up)

Primary outcomes (PO)

PO1a. Leg pain assessed by visual analogue scale (VAS)

PO1b. Low back pain assessed by VAS

PO2a. Sciatica Bothersomeness Index (SBI)

PO2b. Sciatica Frequency Index (SFI)

PO3a. Persistent motor deficit

PO3b. Persistent sensory deficit

PO3c. Persistent reflex changes

PO3d. Persistent urinary incontinence

PO3e. Persistent bowel incontinence

PO4a. Functional outcome: daily activity ‐ 1. Oswestry Disability Index; 2. Roland‐Morris Disability score

PO4b. Functional outcome: return to work/duration postoperative disability

Secondary outcomes (SO)

SO1. Complications ‐ mortality

SO1a. Complications ‐ thromboemboli ‐ deep vein thrombosis (DVT)

SO1b. Complications ‐ surgical site and other infections ‐ urinary tract infection (UTI)

SO1c. Complications ‐ procedure related

SO1d. Complications ‐ re‐hospitalisation due to recurrent disc herniation

SO1e. Complications ‐ re‐hospitalisation due to other causes

SO1f. Complications ‐ surgical re‐intervention

SO1g. Complications ‐ dural tear

SO2. Duration of hospital stay

SO3. Postoperative opioid use

SO4a. Quality of life measured by 36‐item Short Form (SF‐36) or 12‐item Short Form (SF‐12)

SO4b. Overall satisfaction of participants

No. = total number of participants in each study

Studies:

S1. Arts 2011

S2. Chatterjee 1995

S3. Garg 2011

S4. Hermantin 1999

S5. Huang 2005

S6. Mayer 1993

S7. Righesso 2007

S8. Ruetten 2008

S9. Ryang 2008

S10. Shin 2008

S11. Teli 2010

Outcome assessment was performed in the study: positive (+)

Outcome assessment was not performed in the study: negative (‐)

Outcome summary

Primary outcomes (PO)

1. Pain measure by VAS for each of:

• sciatica (PO1a) ‐ Analysis 1.1; Analysis 1.2; Analysis 1.3; Analysis 1.4;

• LBP (PO1b) ‐ Analysis 2.1; Analysis 2.2.

1.1. Analysis.

Comparison 1 Leg pain in two groups of minimally invasive discectomy (MID) and micro/discectomy, Outcome 1 Medium term (1‐5 years).

1.2. Analysis.

Comparison 1 Leg pain in two groups of minimally invasive discectomy (MID) and micro/discectomy, Outcome 2 Leg pain in 2 groups of MID and micro/discectomy ‐ short term (at 1 day).

1.3. Analysis.

Comparison 1 Leg pain in two groups of minimally invasive discectomy (MID) and micro/discectomy, Outcome 3 Leg pain in 2 groups of MID and micro/discectomy ‐ short term (at 3 days).

1.4. Analysis.

Comparison 1 Leg pain in two groups of minimally invasive discectomy (MID) and micro/discectomy, Outcome 4 Leg pain in 2 groups of MID and micro/discectomy ‐ short term (at 5 days).

2.1. Analysis.

Comparison 2 Low back pain ‐ minimally invasive discectomy (MID) versus micro/discectomy, Outcome 1 Sensitivity analysis for low back pain in 1 year.

2.2. Analysis.

Comparison 2 Low back pain ‐ minimally invasive discectomy (MID) versus micro/discectomy, Outcome 2 Endoscopic discectomy vs. micro/discectomy.

2. Sciatica‐specific outcomes:

• SBI (PO2a) ‐ data and analysis: not available;

• SFI (Grøvle 2008) (PO2b) ‐ data and analysis: not available.

3. Neurological deficit of lower extremity:

• motor (PO3a) ‐ Analysis 3.1;

• sensory (PO3b) ‐ Analysis 3.2;

• reflex change (PO3c) ‐ Analysis 3.3;

• bowel (PO3d) ‐ Analysis 3.4;

• urinary incontinence (PO3e).

3.1. Analysis.

Comparison 3 Neurological deficit of lower extremity or bowel/urinary incontinency, Outcome 1 Persistent motor deficits post operative.

3.2. Analysis.

Comparison 3 Neurological deficit of lower extremity or bowel/urinary incontinency, Outcome 2 Persistent sensory deficits post operative.

3.3. Analysis.

Comparison 3 Neurological deficit of lower extremity or bowel/urinary incontinency, Outcome 3 Persistent reflex deficit postoperative (12 months).

3.4. Analysis.

Comparison 3 Neurological deficit of lower extremity or bowel/urinary incontinency, Outcome 4 Persistent bladder dysfunction > 6 months' follow‐up.

4. Functional outcome:

• daily activity ‐ ODI (PO4a1) ‐ Analysis 4.1;

• Roland‐Morris Disability score (PO4a2);

• return to work (PO4b1) ‐ Analysis 4.2;

• postoperative work disability days (PO4b2) ‐ Analysis 4.3.

4.1. Analysis.

Comparison 4 Functional outcomes including daily activity and return to work, Outcome 1 Oswestry Disability Index > 6 months post operative.

4.2. Analysis.

Comparison 4 Functional outcomes including daily activity and return to work, Outcome 2 Number of participants returned to work.

4.3. Analysis.

Comparison 4 Functional outcomes including daily activity and return to work, Outcome 3 Postoperative work disability days ‐ return to work.

Secondary outcomes (SO)

1. Complications of surgery (SO1):

• mortality (SO1);

• common adverse events:

  • thromboembolic complications (SO1a);

  • surgical site and other infections including urinary tract infections (UTI) (SO1b) ‐ Analysis 5.1;

  • procedure‐related complications (SO1c) ‐ Analysis 5.2;

  • re‐hospitalisation due to recurrent disc herniation (SO1d) ‐ Analysis 5.3;

  • re‐hospitalisation due to other causes (SO1e);

  • surgical re‐intervention (SO1f) ‐ Analysis 5.4;

  • dural tear (SO1g) ‐ Analysis 5.5.

5.1. Analysis.

Comparison 5 Secondary outcomes ‐ complications of surgery, Outcome 1 Surgical site and other infections.

5.2. Analysis.

Comparison 5 Secondary outcomes ‐ complications of surgery, Outcome 2 Procedure‐related complications.

5.3. Analysis.

Comparison 5 Secondary outcomes ‐ complications of surgery, Outcome 3 Re‐hospitalisation due to recurrent disc herniation ‐ ≥6 months.

5.4. Analysis.

Comparison 5 Secondary outcomes ‐ complications of surgery, Outcome 4 Surgical re‐intervention.

5.5. Analysis.

Comparison 5 Secondary outcomes ‐ complications of surgery, Outcome 5 Dural tear.

2. Duration of hospital stay (SO2)

3. Postoperative opioid use (SO3)

4. Quality of life:

• measured by SF‐36 or SF‐12 (SO4a) ‐ Analysis 6.1; Analysis 6.2; Analysis 6.3; Analysis 6.4; Analysis 6.5;

• overall success/satisfaction of participants (SO4b) ‐ Analysis 6.6; Analysis 7.1;

• Preoperative pain versus postoperative pain at 12 months ‐ Analysis 8.1.

6.1. Analysis.

Comparison 6 Secondary outcomes ‐ quality of life measured by SF‐36 or SF‐12, and overall satisfaction of participants, which is usually reported by a Likert scale, Outcome 1 SF‐36 Physical Functioning subclass > 6 months.

6.2. Analysis.

Comparison 6 Secondary outcomes ‐ quality of life measured by SF‐36 or SF‐12, and overall satisfaction of participants, which is usually reported by a Likert scale, Outcome 2 SF‐36 Bodily Pain subclass > 6 months.

6.3. Analysis.

Comparison 6 Secondary outcomes ‐ quality of life measured by SF‐36 or SF‐12, and overall satisfaction of participants, which is usually reported by a Likert scale, Outcome 3 SF‐36 General Health subclass > 6 months.

6.4. Analysis.

Comparison 6 Secondary outcomes ‐ quality of life measured by SF‐36 or SF‐12, and overall satisfaction of participants, which is usually reported by a Likert scale, Outcome 4 SF‐36 Physical Health component summary (6 months).

6.5. Analysis.

Comparison 6 Secondary outcomes ‐ quality of life measured by SF‐36 or SF‐12, and overall satisfaction of participants, which is usually reported by a Likert scale, Outcome 5 SF‐36 Mental Health component summary (6 months).

6.6. Analysis.

Comparison 6 Secondary outcomes ‐ quality of life measured by SF‐36 or SF‐12, and overall satisfaction of participants, which is usually reported by a Likert scale, Outcome 6 Overall success (number of participants).

7.1. Analysis.

Comparison 7 Automated percutaneous discectomy versus microdiscectomy, Outcome 1 SO4b. Overall satisfaction of participants.

8.1. Analysis.

Comparison 8 Preoperative pain versus postoperative pain at 12 months, Outcome 1 Low back pain at 12 months in discectomy/microdiscectomy.

Subgroup analyses

A. MID versus microdiscectomy (not open discectomy)

B. Microendoscopy versus MD/OD for leg pain ‐ Analysis 9.1; LBP ‐ Analysis 9.2; ODI ‐ Analysis 9.3 and re‐operations due to recurrent of discopathy ‐ Analysis 9.4.

9.1. Analysis.

Comparison 9 Microendoscopy versus microdiscectomy/open discectomy (MD/OD), Outcome 1 Leg pain ‐ medium‐term follow‐up.

9.2. Analysis.

Comparison 9 Microendoscopy versus microdiscectomy/open discectomy (MD/OD), Outcome 2 Low back pain ‐ 6 months' follow‐up.

9.3. Analysis.

Comparison 9 Microendoscopy versus microdiscectomy/open discectomy (MD/OD), Outcome 3 Oswestry Disability Index (ODI) > 6 months' follow‐up.

9.4. Analysis.

Comparison 9 Microendoscopy versus microdiscectomy/open discectomy (MD/OD), Outcome 4 Re‐operations due to recurrence of discopathy.

Primary outcomes

PO1a. Leg pain assessed by visual analogue scale

Four trials (599 participants) examined leg pain from six months' to two years' follow‐up (Arts 2011; Huang 2005; Mayer 1993; Teli 2010) (Figure 3). Two trials had unclear risk of bias due to non‐adequate randomisation, no allocation concealment and no blinding (Huang 2005; Mayer 1993). Meanwhile, Huang 2005 had no clear intention‐to‐treat analysis and Huang 2005 and Mayer 1993 had unclear co‐interventions. One trial had high risk of bias for blinding of participants and personnel (performance bias) and unclear risk of bias for co‐interventions (Arts 2011). Another trial had unclear allocation concealment and high risk of bias for blinding (Teli 2010). We downgraded the level of evidence to low due to the high risk of bias and the small number of participants for minimally invasive techniques. The analysis was consistent with no heterogeneity (I² = 0%). Based upon four studies, MID was associated with slightly greater leg pain versus MD/OD at 12 months (MD 0.13 on a 0 to 10 scale, 95% CI 0.09 to 0.16) (Figure 3) (Table 1 ‐ Leg pain). In both the MID and MD/OD groups, leg pain scores improved significantly from baseline to one year. Following MID, pain decreased a mean of 5.8 points on a 0 to 10 scale (95% CI 5.57 to 6.03), and following MD/OD, pain decreased 6.45 points (95% CI 6.25 to 6.64). One trial found tubular discectomy associated with slightly more leg pain from one to 104 weeks after MID than microdiscectomy (MD 3.3, 95% CI 0.2 to 6.2) (Arts 2011), but three other trials found no difference between endoscopic discectomy and MD/OD (MD 0.08, 95% CI ‐0.04 to 0.20). One trial found no difference in postoperative leg pain between microendoscopic discectomy (MED) and microdiscectomy from one to five days (Shin 2008).

3.

Forest plot of comparison: Outcome 1. Leg pain in two groups of minimally invasive discectomy (MID) and micro/discectomy, medium term (one to five years).

PO1b. Low back pain assessed by visual analogue scale

Three trial (577 participants) examined LBP at 6‐month follow‐up (Arts 2011; Righesso 2007; Teli 2010) (Figure 4). One trial had high risk of bias due to non‐adequate randomisation, no allocation concealment, no blinding, group differences at baseline and no clear intention‐to‐treat analysis (Righesso 2007). One trial had high risk of bias for blinding of participants and personnel (performance bias) and unclear risk of bias for co‐interventions (Arts 2011). Another trial had unclear allocation concealment and high risk of bias for blinding (Teli 2010). We downgraded the level of evidence to low due to the high risk of bias and the small number of participants for minimally invasive techniques. The analysis was almost consistent with low heterogeneity (I² = 35%). Therefore, there was low‐quality evidence that MID (tubular discectomy or endoscopic discectomy) was associated with slightly greater LBP versus MD/OD at six months (MD 0.35 on a 0 to 10 scale, 95% CI 0.19 to 0.51) and at two years (MD 0.54 on a 0 to 10 scale, 95% CI 0.29 to 0.79). There was no significant difference at one year (MD 0.19 on an 0 to 10 scale, 95% CI ‐0.22 to 0.59). Statistical heterogeneity was small to high (I² = 35% at six months, 90% at one year and 65% at two years) (Figure 4) (Table 1 ‐ LBP). Pooled results were similar to results from the largest trial (325 participants), which compared tubular discectomy versus microdiscectomy. In both the MID and MD/OD groups, LBP scores improved significantly from baseline to one year. Following MID, pain decreased a mean of 1.31 points on an 0 to 10 scale (95% CI 1.25 to 1.38), and following micro/discectomy, pain decreased 1.49 points (95% CI 1.39 to 1.58). One study found MED associated with lower LBP intensity at one to five days' follow‐up versus microdiscectomy (Shin 2008).

4.

Forest plot of comparison: Outcome 2. Low back pain in two groups of minimally invasive discectomy (MID) and micro/discectomy at six months, one year and two years.

PO2a. Sciatica Bothersomeness Index

Based on one trial (325 participants), there was low‐quality evidence of no statistically significant difference between tubular discectomy versus microdiscectomy on the SBI from one to 104 weeks (MD 0.5 on a 0 to 24 scale, 95% CI ‐0.3 to 1.3) (Arts 2011). We rated the quality of evidence as low due to availability of only one trial reporting SBI and imprecision/sparse data.

PO2b. Sciatica Frequency Index

Based on one trial (325 participants), there was low‐quality evidence of no statistically significant difference between tubular discectomy versus microdiscectomy on the SFI at 12 months (MD 0.5 on a 0 to 24 scale, 95% CI ‐0.5 to 1.4) (Arts 2011). We rated the quality of evidence as low due to availability of only one trial reporting SFI, and imprecision/sparse data.

PO3a. Persistent motor deficit

Based on four studies (126 participants who had preoperative motor deficit), there was low‐quality evidence of no statistically significant difference between MID and MD/OD for persistent motor deficit at at least six months' follow‐up (RR 0.96, 95% CI 0.56 to 1.63) (Hermantin 1999; Mayer 1993; Righesso 2007; Ryang 2008). Statistical heterogeneity was low (I² = 15%), despite variability across studies in the proportion of participants with motor deficits that resolved and in the duration of follow‐up (range six to 56 months). We rated the quality of evidence as low due to the high risk of bias and small number of participants with motor deficits in the studies, resulting in imprecise estimates.

PO3b. Persistent sensory deficit

Based on four studies (165 participants who had preoperative sensory deficit), there was low‐quality evidence of no statistically significant difference between MID and MD/OD for persistent sensory deficit at at least six months' follow‐up (RR 0.86, 95% CI 0.65 to 1.15) (Hermantin 1999; Mayer 1993; Righesso 2007; Ryang 2008). Statistical heterogeneity was very low (I² = 0%), despite variability in the proportion of participants with sensory deficits that resolved and duration of follow‐up. One trial showed most participants improved completely following either endoscopic discectomy or microdiscectomy 24 months after operation (persistent deficit in 1/13 participants following endoscopic discectomy and 5/16 participants following microdiscectomy) (Mayer 1993). In one trial, persistent sensory neurological deficits were reported in 16/26 participants following endoscopic discectomy and 18/28 participants following discectomy after at least 19 months (Hermantin 1999). We rated the quality of evidence as very low due the high risk of bias, small number of participants with baseline sensory deficits and the use of subjective evaluations to assess this outcome.

PO3c. Persistent reflex changes

Based on two studies (47 participants who had preoperative reflex changes), there was low‐quality evidence of statistically significant difference between percutaneous endoscopic discectomy and microdiscectomy or open discectomy for persistent reflex changes at 12 months' follow‐up (RR 0.68, 95% CI 0.49 to 0.96) (Hermantin 1999; Mayer 1993). Statistical heterogeneity was low (I² = 0%), despite variability in the proportion of participants with reflex changes that resolved and duration of follow‐up. One trial showed most participants improved completely in both groups of endoscopic discectomy (persistent deficit in 2/10 participants) and microdiscectomy (persistent deficit in 2/7 participants) 24 months after operation (Mayer 1993). In the other study, persistent reflex changes were reported with 12/18 participants that underwent endoscopic discectomy and 12/12 participants who underwent standard discectomy (Hermantin 1999). We rated the quality of evidence as low due the high risk of bias and small number of participants with baseline reflex changes. We did not consider this a significant change because of the role of just one study with small number of participants in the analysis (RR 0.68, 95% CI 0.49 to 0.96) (Hermantin 1999).

PO3d. Persistent bowel incontinence

One study reported persistent bowel incontinence, but included no participants with bowel incontinence at baseline (Righesso 2007). Therefore, there was no evidence for the role of MID and MD/OD on postoperative persistent bowel incontinence.

PO3e. Persistent urinary incontinence

Two studies evaluated persistent urinary incontinence, but only included three participants with urinary incontinence at baseline, resulting in unreliable estimates and no evidence for the role of MID and MD/OD on postoperative persistent urinary incontinence (Righesso 2007; Ryang 2008).

PO4a1 and 4a2. Functional outcome: daily activity ‐ Oswestry Disability Index and Roland‐Morris Disability score

Based on five studies (624 participants), there was moderate‐quality evidence due to the high risk of bias of no statistically significant difference between MID and MD/OD on the ODI at more than six months (MD 0.84 on a 0 to 100 scale, 95% CI ‐0.21 to 1.88; I² = 0%) (Table 1 ‐ ODI) (Garg 2011; Righesso 2007; Ruetten 2008; Ryang 2008; Teli 2010). All trials found no significant difference between different types of MID (four studies of MED and one study of minimal access trocar microdiscectomy) versus MD/OD.

Based on one trial (325 participants), there was low‐quality evidence of no statistically significant difference between MID and microdiscectomy on the Roland‐Morris Disability score at most follow‐up time points between four and 104 weeks after surgery (Arts 2011). A statistically significant difference in favour of microdiscectomy over transmuscular tubular microdiscectomy was reported at 52 weeks' follow‐up (MD 1.3, 95% CI 0.03 to 2.6). We rated the quality of evidence as low due to availability of only one trial reporting this outcome, and imprecision/sparse data.

PO4b1. Functional outcome: return to work

Based on one study (60 participants), there was low‐quality evidence of no statistically significant difference between endoscopic discectomy and discectomy for return to work at 19 to 42 months' follow‐up (OR 2.07, 95% CI 0.18 to 24.15) (Hermantin 1999). Twenty‐nine out of 30 participants in endoscopic discectomy and 28/30 participants in discectomy returned to their work. We rated the quality of evidence as low due to only one trial reporting return to work, and imprecision/sparse data.

PO4b2. Functional outcome: postoperative work disability days

Two trials (503 participants) reported inconsistent results for the outcome postoperative work disability days at 24 months' follow‐up (Arts 2011; Ruetten 2008). One study found no clear difference in length of postoperative work disability between transmuscular tubular microdiscectomy (2 weeks, 95% CI 1.6 to 2.4) versus microdiscectomy (2.1 weeks, 95% CI 1.8 to 2.5) (Arts 2011). The other study reported fewer mean postoperative work disability days following percutaneous endoscopic discectomy (24 days) versus microdiscectomy (49 days), but did not report the SD (Ruetten 2008). We rated the quality of evidence as very low due to the high risk of bias, small number of trials and presence of inconsistency.

Secondary outcomes

SO1. Complications ‐ mortality

One study (212 participants) reported mortality but reported no deaths after at least six months of follow‐up (Teli 2010).

SO1a. Complications ‐ thromboemboli ‐ deep vein thrombosis

Based on one study (325 participants), there was low‐quality evidence due to the high risk of bias, one single trial and the small number of participants of no statistically significant difference between MID and MD/OD for thromboemboli and deep vein thrombosis after at least six months' follow‐up (Arts 2011). Statistically heterogeneity was not applicable.

SO1b. Complications ‐ surgical site and other infections including urinary tract infection

Six trials (931 participants) evaluated side effects of postoperative surgical infection (Arts 2011; Garg 2011; Hermantin 1999; Huang 2005; Ruetten 2008; Teli 2010) (Figure 5). One trial had high risk of bias for blinding of participants and personnel (performance bias) and unclear risk of bias for co‐interventions (Arts 2011). Another trial had unclear allocation concealment and high risk of bias for blinding (Teli 2010). One trial had high risk of bias due to no clear randomisation, no clear allocation concealment, no clear blinding of participants and personnel, no clear selective reporting and no clear intention‐to‐treat analysis (Garg 2011). One trial had high risk of bias due to no clear randomisation, no clear allocation concealment, no clear blinding and no clear intention‐to‐treat analysis (Huang 2005). One trial had high risk of bias because of non‐randomisation, no allocation, no blinding, no clear group similarity at baseline and no clear intention‐to‐treat analysis (Ruetten 2008). However, no blinding of participants, care provider and outcome assessor and no clear co‐intervention were major problems of the one trial that had no other risk of bias (Hermantin 1999). We downgraded the level of evidence to the moderate level due to the high risk of bias. One participant (1/431) had infection following MID versus 16/500 participants following MD/OD. The analysis was almost consistent with small heterogeneity (I² = 34%). Therefore, based upon six studies, MID was associated with lower risk of surgical site and other infections including UTI versus MD/OD after at least six months' follow‐up (RR 0.23, 95% CI 0.07 to 0.79).

5.

Forest plot of comparison: Secondary outcome ‐ complication ‐ surgical sites and other infections

SO1c. Complications ‐ procedure related

Based on seven studies (991 participants), there was low‐quality evidence of no statistically significant difference between MID and MD/OD for procedure‐related complications after at least six months' follow‐up (RR 1.01, 95% CI 0.61 to 1.66) (Arts 2011; Garg 2011; Hermantin 1999; Huang 2005; Ruetten 2008; Ryang 2008; Teli 2010). Statistical heterogeneity was low (I² = 33%). We rated the quality of evidence as low due to high risk of bias and some imprecision in estimates.

SO1d. Complications ‐ re‐hospitalisation due to recurrent disc herniation

Six studies (949 participants) evaluated re‐hospitalisation due to recurrent disc herniation (Arts 2011; Garg 2011; Mayer 1993; Ruetten 2008; Ryang 2008Teli 2010) (Figure 6). One trial had high risk of bias for blinding of participants and personnel (performance bias) and unclear risk of bias for co‐interventions (Arts 2011). Another trial had unclear allocation concealment and high risk of bias for blinding (Teli 2010). However, four other trials had high risk of bias because of non‐randomisation, no allocation concealment, and no blinding of participants and personnel (Garg 2011; Mayer 1993; Ruetten 2008; Ryang 2008). We downgraded the evidence to the low level due to high risk of bias and imprecision (high 95% CI). The analysis was consistent with no heterogeneity (I² = 0%). Based on six studies, there was low‐quality evidence that MID was associated with increased risk of re‐hospitalisation due to recurrent disc herniation versus MD/OD after at least 12 months' follow‐up (RR 1.74, 95% CI 1.03 to 2.94) (Arts 2011; Garg 2011; Mayer 1993; Ruetten 2008; Ryang 2008Teli 2010).

6.

Forest plot of comparison: Secondary outcome 1e ‐ complications ‐ re‐hospitalisation due to recurrent disc herniation six months or greater.

SO1e. Complications ‐ re‐hospitalisation due to other causes

No study evaluated re‐hospitalisation die to other causes.

SO1f. Complications ‐ surgical re‐intervention

Based on three studies (425 participants), there was low‐quality evidence of no statistically significant difference between MID and MD/OD for surgical re‐intervention after at least six months' follow‐up (RR 1.46, 95% CI 0.68 to 3.14) (Arts 2011; Mayer 1993; Ryang 2008). Statistically heterogeneity was low (I² = 15%). We rated the quality of evidence as low due to the high risk of bias and the small number of events, resulting in imprecise estimates.

SO1g. Complications ‐ dural tear

Based on five studies (887 participants), there was low‐quality evidence of no statistically significant difference between MID and MD/OD for dural tear (RR 1.63, 95% CI 0.82 to 3.22, I² = 16%) (Arts 2011; Garg 2011; Ruetten 2008; Ryang 2008; Teli 2010). We rated the quality of evidence as low due to high risk of bias and some imprecision in estimates.

SO2. Duration of hospital stay

Five studies (731 participants) examined duration of hospital stay (Arts 2011; Garg 2011; Huang 2005; Ryang 2008; Teli 2010). One study was an outlier with large effect size and unusually high duration of hospital stay for MD/OD (mean 12 days, range 5 to 21) (Garg 2011). Inclusion of this study in the meta‐analysis produced high heterogeneity (I² = 99%) and significant difference between MID and MD/OD regarding duration of hospital stay (MD ‐2.29, 95% CI ‐4.15 to ‐0.43). In stratified analyses based on the specific minimally invasive technique used, two trials found no difference between tubular microdiscectomy and microdiscectomy (MD ‐0.01, 95% CI ‐0.26 to 0.23, I² = 0%) (Arts 2011; Ryang 2008). Three trials of endoscopic discectomy reported inconsistent results, with differences versus micro/discectomy ranging from ‐9.0 to +0.21 days (Garg 2011; Huang 2005; Teli 2010). We rated the level of evidence as very low due to the inconsistency, high risk of bias and the small number of participants for specific minimally invasive techniques.

SO3. Postoperative opioid use

Based on only one study (60 participants), there was low‐quality evidence (due to the small number of trials (imprecision)) of no statistically significant difference between endoscopic discectomy and discectomy for postoperative opioid use at 19 to 42 months' follow‐up (OR 0.06, 95% CI 0.00 to 1.15) (Hermantin 1999). Zero out of 30 participants who underwent endoscopic discectomy and 6/30 participants who underwent open discectomy used postoperative opioids.

SO4a. Health‐related quality of life measured by 36‐item Short Form or 12‐item Short Form

Two trials evaluated HRQoL (Arts 2011; Ryang 2008). Ryang 2008 had high overall risk of bias. We downgraded the level of evidence to low due to the high risk of bias and small number of participants. The analysis was consistent with no heterogeneity (I² = 0%). Based on two studies (385 participants), there was no statistically significant difference between MID and MD/OD in the Physical Health component summary of the SF‐36 at six months' follow‐up (MD 0.96, 95% CI ‐0.12 to 2.03) (Arts 2011; Ryang 2008). However, MID was associated with worse quality on life on three Physical Health component subclasses: Physical Functioning (MD ‐4.70, 95% CI ‐5.05 to ‐4.35), Bodily Pain (MD ‐3.70, 95% CI ‐4.11 to ‐3.28) and General Health (MD ‐2.52, 95% CI ‐2.92 to ‐2.11). There was no significant difference on the Mental Health component summary score at six months' follow‐up (MD ‐4.31, 95% CI ‐9.96 to 1.33).

SO4b. Overall success/satisfaction of participants

Based on five studies (443 participants), there was moderate‐quality evidence due to the high risk of bias of no statistically significant difference between MID and MD/OD for participant‐rated overall satisfaction after at least six months' follow‐up (OR 1.04, 95% CI 0.99 to 1.10) (Chatterjee 1995; Garg 2011; Hermantin 1999; Huang 2005; Ruetten 2008). There was no heterogeneity (I² = 0%).

Summary for quality of evidence

There was moderate‐quality evidence for lower side effects of surgical site and other infections after at least six months' follow‐up for MID versus MD/OD. However, there was low‐quality evidence for higher leg pain, LBP, side effects of re‐hospitalisation due to recurrent disc herniation after at least six months' follow‐up and lower SF‐36 Physical Functioning subclass after more than six months' follow‐up.

Subgroup analyses

A. Subgroup analysis of minimally invasive discectomy versus microdiscectomy (not open discectomy)

We compared MID with microdiscectomy. Eight of the 11 included studies examined microdiscectomy. Our primary outcome measures showed similar results for the primary analysis. There was low‐quality evidence of a statistically significant difference for medium‐term follow‐up for leg pain (MD 0.13, 95% CI 0.09 to 0.16; I² = 0%). There was low‐quality evidence of a statistically significant difference at six months' follow‐up for LBP (MD 0.36, 95% CI 0.30 to 0.41; I² = 3%); among secondary outcome measures, there was no statistically significant difference for surgical site and other infections (RR 0.23, 95% CI 0.05 to 1.02). The control groups for both studies that evaluated HRQoL used microdiscectomy (385 participants) (Arts 2011; Ryang 2008). There was no statistically significant difference between MID and microdiscectomy in the Physical Health component summary of the SF‐36 at six months' follow‐up (MD 0.96, 95% CI ‐0.12 to 2.03). However, MID was associated with worse quality of life on three Physical Health component subclasses: Physical Functioning (MD ‐4.70, 95% CI ‐5.05 to ‐4.35), Bodily Pain (MD ‐3.70, 95% CI ‐4.11 to ‐3.28) and General Health (MD ‐2.52, 95% CI ‐2.92 to ‐2.11). There was no statistically significant difference on the Mental Health component summary score at six months' follow‐up (MD ‐4.31, 95% CI ‐9.96 to 1.33). Therefore, regarding HRQoL, analysis of MID versus microdiscectomy was similar results for the primary analysis.

B. Subgroup analysis of microendoscopy versus microdiscectomy/open discectomy

We compared MED with MD/OD. Eight of the 11 included studies examined MED. There was no statistically significant difference for medium‐term follow‐up for leg pain (MD 0.09, 95% CI ‐0.03 to 0.21) or for LBP at 6 months' follow‐up (MD 0.29, 95% CI ‐0.19 to 0.77) and ODI after at least six months' follow‐up (MD 0.85, 95% CI ‐0.21 to 1.90). Among secondary outcome measures, a comparison of MED versus MD/OD showed that re‐operations due to recurrences of discopathy was statistically significant (RR 2.13, 95% CI 1.01 to 4.49). In particular, there was a statistically significant difference for surgical site and other infections (RR 0.22, 95% CI 0.06 to 0.82), such that participants in the MED group were associated with lower risk of surgical site and other infections including UTI compared with participants in the MD/OD group after at least six months' follow‐up. There was no statistically significant difference for duration of hospital stay (MD ‐3.71, 95% CI ‐10.24 to 2.8). No study evaluated HRQoL between the MED and MD/OD groups.

Sensitivity analysis

Sensitivity analysis of minimally invasive discectomy versus microdiscectomy

We excluded studies with high risk of bias and studies with unclear methods of randomisation. Four studies remained (Arts 2011; Hermantin 1999; Shin 2008; Teli 2010). We performed sensitivity analyses for outcomes that had shown significant results in the primary analyses. We found a statistically significant difference for MID versus microdiscectomy at one year' follow‐up for leg pain (MD 0.13, 95% CI 0.09 to 0.17), but differences were small (less than 0.5 points on a 0 to 10 scale) and did not meet standard thresholds for clinically meaningful differences. There was low‐quality evidence that MID was associated with worse LBP than microdiscectomy at six months' follow‐up (MD 0.36, 95% CI 0.30 to 0.41). For secondary outcomes, MID was not associated with lower risk of surgical site and other infections (RR 0.30, 95% CI 0.05 to 1.70), but there was higher risk of re‐hospitalisation due to recurrent disc herniation (RR 2.13, 95% CI 1.13 to 4.02). In addition, MID was associated with slightly lower quality of life (RR ‐4.70, 95% CI ‐5.05 to ‐4.35) (less than 5 points on a 100‐point scale) on Physical Functioning subclass after at least six months. However, MID was not associated with shorter duration of hospitalisation than microdiscectomy (MD 0.13, 95% CI ‐0.07 to 0.33).

Discussion

Summary of main results

The results of this review showed that leg pain and LBP were worse with MID compared with MD/OD. The differences were small and did not meet standard thresholds for clinically meaningful differences. There were no statistically significant differences on other primary outcomes such as measures of function or persistent neurological deficits, but conclusions were limited by the small number of participants in the trials with neurological deficits at baseline. With secondary outcomes, MID was associated with lower risk of surgical site and other infections and UTI, but higher risk of re‐hospitalisation due to recurrent disc herniation. In addition, MID was associated with slightly lower HRQoL after at least six months' follow‐up (less than 5 points on a 100‐point scale) on some measures, such as some Physical subclasses. Once more, the differences were small and did not meet standard thresholds for clinically meaningful differences.

When we performed subgroup analyses of MID versus microdiscectomy (not open discectomy), which was used in eight of the 11 included studies, there was low‐quality evidence of a statistically significant difference in leg pain at medium‐term follow‐up and LBP at six‐month and two‐year follow‐up, but not at one‐year follow‐up. Meanwhile, MID was associated with increased risk of re‐hospitalisation due to recurrent disc herniation versus microdiscectomy after at least 12 months' follow‐up. Moreover, MID was associated with worse HRQoL after six‐month follow‐up on three Physical Health component subclasses: Physical Functioning, Bodily Pain and General Health. However, there was no statistically significant difference for surgical site and other infections. Finally, subgroup analysis showed primary outcome measures for MID versus microdiscectomy was the same as MID versus MD/OD. In other words, better outcomes for pain improvement are related to using a microscope in the standard MD/OD.

It was difficult to compare specific MID procedures with MD/OD because of the few studies available. Since eight of the 11 MID procedures were percutaneous MED, comparison of MED versus MD/OD showed that there was a low level of evidence of no statistically significant difference for pain and function. In addition, MED was associated with lower risk of surgical site and other infections and UTI, but higher risk of re‐hospitalisation due to recurrent disc herniation.

Overall completeness and applicability of evidence

Although all of the included studies evaluated leg pain following surgery, more than half of the studies reported no other primary outcome, potentially resulting in biased or less precise estimates. Some older studies may have evaluated surgical techniques that would now be considered outdated, potential limiting their applicability (Chatterjee 1995;Hermantin 1999;Mayer 1993). The studied interventions in more recent trials generally appear applicable to current practice. No study was funded by a device company. We were unable to assess for publication bias formally using statistical or graphical methods due to the small number of studies.

Concerning clinical relevance, the positive effects of MD/OD did not exceed the threshold for minimum clinically important differences (MCID), which is 1.5 point (on a 0 to 10 scale) improvement for leg pain and LBP (Ostelo 2008). Moreover, although MID was associated with more re‐operations due to recurrence of discopathy, the difference between the two groups of MID and MD/OD was less than an MCID of 10%. For measures related to quality of life, differences were small (less than 5 points on a 0‐ to 100‐point scale), which is lower than the threshold typically considered to meet MCID (Copay 2008). In contrast, potential advantages of MID were lower risk of surgical site and other infections, and shorter hospital stay. However, similar to the findings described above, these two outcome measures did not meet standard thresholds for clinically meaningful differences.

Quality of the evidence

The low level of evidence across outcomes was due to the high risk of bias and the small number of trials for specific minimally invasive techniques. The analysis was consistent with no/small heterogeneity in these outcome measures: leg pain at one year' follow‐up (I² = 0%); LBP at six months' follow‐up (I² = 35%); LBP at two years' follow‐up (I² = 65%); disability after six months' follow‐up (I² = 0%); surgical site and other infections after six months' follow‐up (I² = 34%); re‐ operations due to recurrence of discopathy after six months' follow‐up (I² = 0%); and HRQoL after six months' follow‐up (I² = 0%). There was high heterogeneity for LBP at one year' follow‐up (I² = 90%). We wrote these data just to avoid selective reporting results. In one case, we found medium‐quality evidence for lower side effects of surgical site and other infections after six months' follow‐up for MID versus MD/OD. This was due to high risk of bias in only a few studies.

Potential biases in the review process

We were unable to assess for publication bias formally due to the small number of studies. Although we did not detect signs of potential selective reporting bias, we did not have the original trial protocols to review and several primary outcomes for this review were not reported in included studies. We did not apply language restrictions and utilised methods to reduce potential effects of bias.

Agreements and disagreements with other studies or reviews

Our findings are generally consistent with other systematic reviews. One review found no difference between MED versus MD/OD in the ODI, based on four RCTs of Garg 2011; Huang 2005; Righesso 2007 and Teli 2010 (Smith 2013). One trial in this review reported an increased number of severe complications in the MED group. Another systematic review found no clear differences in benefits or harms between various minimally invasive techniques versus MD/OD, based on six RCTs (Dasenbrock 2012). This review did not include several trials included in our review (Arts 2011; Chatterjee 1995; Garg 2011; Hermantin 1999; Mayer 1993). Another systematic review compared the effectiveness of transforaminal endoscopic surgery and open microdiscectomy in people with symptomatic lumbar disc herniations, but it only included one RCT of Hermantin 1999 (Nellensteijn 2010).

Two systematic reviews evaluated percutaneous lumbar mechanical disc decompression utilising the Dekompressor and laser, but we excluded these interventions from this review (Singh 2009a; Singh 2009b).

Finally, Jacobs et al. performed a systematic review for surgical techniques for sciatica due to herniated disc. However, there were some differences in our inclusion criteria and risk of bias assessment (Jacobs 2012). Moreover, in our review, we focused on outcomes that were highly relevant to participants. Therefore, these differences contributed to differences in overall findings.

Authors' conclusions

Implications for practice.

It could be argued that leg pain is the main reason for performing surgery in people with discopathy. Our findings show both arms of microdiscectomy or open discectomy (MD/OD) and minimally invasive discectomy (MID) achieved the minimum clinically important difference (MCID) pre‐post intervention of more than 1.5 points out of 10 (or 15 out of 100) according to Ostelo et al. (Ostelo 2008). However, there was more reduced postoperative leg pain and LBP following MD/OD compared with MID. This group difference is almost always less than 0.5 points out of 10 (or 5 out of 100). Our results for leg pain, LBP and other outcome measures were limited to short‐ and medium‐term follow‐ups and there was a lack of information on long‐term outcomes. MD/OD are the standard of care because of their long record of efficacy and safety. We found no evidence that MID is superior with regards to key participant‐focused outcomes such as pain, function and re‐operation. However, our results show that potential advantages of MID include lower risk of surgical site infections and urinary tract infections, as well as shorter hospital stay.

Implications for research.

More trials are needed to define what role, if any, there is for MID. Future trials should address all important outcomes related to discectomy surgeries; in particular, more research is needed to understand benefits and harms associated with specific minimally invasive techniques, including their impact on potential harms, such as bowel and bladder incontinence, and to clarify further effects on quality of life and other outcomes in the long term. In addition, future research should focus on identifying participants and procedures that are associated with better or similar outcomes along with lower/similar costs/risks.

Appendices

Appendix 1. Search Strategy (Medline via OVID SP)

EBM Reviews ‐ Cochrane Central Register of Controlled Trials (CENTRAL) <October 2013>

1 intervertebral disk degeneration/

2 intervertebral disk displacement/

3 dis?opath$.tw,ot.

4 spondylodiscitis.tw,ot.

5 (spondylochondrosis or chondrosis).tw,ot.

6 (hernia$ or perfora$ or ruptur$ or degenerat$ or degradat$ or displac$ or prolaps$ or protru$ or avuls$ or compress$ or extru$).tw,ot.

7 or/1‐6

8 lumbar vertebrae/

9 lumbosacral region/

10 intervertebral disk/ or (intervertebral or intradiscal or intradiscal).tw,ot.

11 (or/8‐9) and 10

12 (lumb$ adj (disc$ or disk$)).tw,ot.

13 or/11‐12

14 exp surgical procedures, minimally invasive/

15 (microdis?ectom$ or nucleotom$ or nucleoplast$ or annuloplasty or (microscop$ adj dis?otom$)).tw,ot.

16 ((mini$ adj3 invas$) or mini?invas$).tw,ot.

17 automated percutaneous discectomy.tw,ot.

18 laser.tw,ot.

19 ((percutaneous or transforaminal) adj (microendoscop$ or endoscop$ or dis?oscop$ or arthroscopy$)).tw,ot.

20 transmuscular tubular.tw,ot.

21 or/14‐20

22 7 and 13 and 21

Ovid MEDLINE(R) In‐Process & Other Non‐Indexed Citations and Ovid MEDLINE(R) <1946 ‐ 22 November 2013>

1 intervertebral disk degeneration/

2 intervertebral disk displacement/

3 dis?opath$.tw,ot.

4 spondylodiscitis.tw,ot.

5 (spondylochondrosis or chondrosis).tw,ot.

6 (hernia$ or perfora$ or ruptur$ or degenerat$ or degradat$ or displac$ or prolaps$ or protru$ or avuls$ or compress$ or extru$).tw,ot.

7 or/1‐6

8 lumbar vertebrae/

9 lumbosacral region/

10 intervertebral disk/ or (intervertebral or intradiscal or intradiskal).tw,ot.

11 (or/8‐9) and 10

12 (lumb$ adj (disc$ or disk$)).tw,ot.

13 or/11‐12

14 exp surgical procedures, minimally invasive/

15 (microdis?ectom$ or nucleotom$ or nucleoplast$ or annuloplasty or (microscop$ adj dis?otom$)).tw,ot.

16 ((mini$ adj3 invas$) or mini?invas$).tw,ot.

17 automated percutaneous discectomy.tw,ot.

18 laser.tw,ot.

19 ((percutaneous or transforaminal) adj (microendoscop$ or endoscop$ or dis?oscop$ or arthroscopy$)).tw,ot.

20 transmuscular tubular.tw,ot.

21 or/14‐20

22 randomized controlled trial.pt.

23 controlled clinical trial.pt.

24 randomized.ab.

25 placebo.ab.

26 drug therapy.fs.

27 randomly.ab.

28 trial.ab.

29 groups.ab.

30 or/22‐29

31 (animals not (humans and animals)).sh.

32 30 not 31

33 7 and 13 and 21 and 32

 

EMBASE via Ovid SP <1974 to 22 November 2013>

1 exp intervertebral disk hernia/

2 intervertebral disk degeneration/

3 lumbar disk hernia/

4 dis?opath$.tw,ot.

5 spondylodiscitis.tw,ot.

6 (spondylochondrosis or chondrosis).tw,ot.

7 (hernia$ or perfora$ or ruptur$ or degenerat$ or degradat$ or displac$ or prolaps$ or protru$ or avuls$ or compress$ or extru$).tw,ot.

8 or/1‐7

9 lumbar vertebra/

10 intervertebral disk/ or (intervertebral or intradiscal or intradiscal).tw,ot.

11 9 and 10

12 (lumb$ adj (disc$ or disk$ or region or vertebra?)).tw,ot.

13 or/11‐12

14 minimally invasive surgery/

15 (microdis?ectom$ or nucleotom$ or nucleoplast$ or annuloplasty or (microscop$ adj dis?otom$)).tw,ot.

16 ((mini$ adj3 invas$) or mini?invas$).tw,ot.

17 automated percutaneous discectomy.tw,ot.

18 laser.tw,ot.

19 ((percutaneous or transforaminal) adj (microendoscop$ or endoscop$ or dis?oscop$ or arthroscopy$)).tw,ot.

20 Transmuscular tubular.tw,ot.

21 or/14‐20

22 crossover‐procedure/

23 double‐blind procedure/

24 randomized controlled trial/

25 single‐blind procedure/

26 random$.tw,ot.

27 factorial$.tw,ot.

28 cross?over$.tw,ot.

29 placebo$.tw,ot.

30 ((doubl$ or singl$) adj blind$).tw,ot.

31 assign$.tw,ot.

32 allocat$.tw,ot.

33 volunteer$.tw,ot.

34 or/22‐33

35 8 and 13 and 21 and 34

36 limit 35 to animals

37 limit 35 to human

38 36 not 37

39 35 not 38

 

Appendix 2. Assessment for clinical relevance

1. Were the participants described in detail so that you could decide whether they were comparable to those that you see in your practice?

2. Were the interventions and treatment settings described well enough so that you could provide the same for your patients?

3. Were all clinically relevant outcomes measured and reported?

4. Is the size of the effect clinically important?

5. Are the likely treatment benefits worth the potential harms?

Appendix 3. Criteria for assessing risk of bias for internal validity

Random sequence generation (selection bias)

Selection bias (biased allocation to interventions) due to inadequate generation of a randomised sequence

There was a low risk of selection bias if the investigators described a random component in the sequence generation process such as: referring to a random number table, using a computer random number generator, coin tossing, shuffling cards or envelopes, throwing dice, drawing of lots, minimisation (minimisation may be implemented without a random element, and this is considered to be equivalent to being random).

There was a high risk of selection bias if the investigators described a non‐random component in the sequence generation process, such as: sequence generated by odd or even date of birth, date (or day) of admission, hospital or clinic record number; or allocation by judgement of the clinician, preference of the participant, results of a laboratory test or a series of tests, or availability of the intervention.

Allocation concealment (selection bias)

Selection bias (biased allocation to interventions) due to inadequate concealment of allocations prior to assignment

There was a low risk of selection bias if the participants and investigators enrolling participants could not foresee assignment because one of the following, or an equivalent method, was used to conceal allocation: central allocation (including telephone, web‐based and pharmacy‐controlled randomisation); sequentially numbered drug containers of identical appearance or sequentially numbered, opaque, sealed envelopes.

There was a high risk of bias if participants or investigators enrolling participants could possibly foresee assignments and thus introduce selection bias, such as allocation based on: using an open random allocation schedule (e.g. a list of random numbers); assignment envelopes were used without appropriate safeguards (e.g. if envelopes were unsealed or non‐opaque or not sequentially numbered); alternation or rotation; date of birth; case record number or other explicitly unconcealed procedures.

Blinding of participants

Performance bias due to knowledge of the allocated interventions by participants during the study

There was a low risk of performance bias if blinding of participants was ensured and it was unlikely that the blinding could have been broken; or if there was no blinding or incomplete blinding, but the review authors judged that the outcome is not likely to be influenced by lack of blinding.

Blinding of personnel/care providers (performance bias)

Performance bias due to knowledge of the allocated interventions by personnel/care providers during the study

There was a low risk of performance bias if blinding of personnel was ensured and it was unlikely that the blinding could have been broken; or if there was no blinding or incomplete blinding, but the review authors judged that the outcome was not likely to be influenced by lack of blinding.

Blinding of outcome assessor (detection bias)

Detection bias due to knowledge of the allocated interventions by outcome assessors

There was low risk of detection bias if the blinding of the outcome assessment was ensured and it was unlikely that the blinding could have been broken; or if there was no blinding or incomplete blinding, but the review authors judged that the outcome was not likely to be influenced by lack of blinding, or:

• for participant‐reported outcomes in which the participant was the outcome assessor (e.g. pain, disability): there was a low risk of bias for outcome assessors if there was a low risk of bias for participant blinding (Boutron 2005);

• for outcome criteria that were clinical or therapeutic events that will be determined by the interaction between participants and care providers (e.g. co‐interventions, length of hospitalisation, treatment failure), in which the care provider was the outcome assessor: there was a low risk of bias for outcome assessors if there was a low risk of bias for care providers (Boutron 2005);

• for outcome criteria that were assessed from data from medical forms: there was a low risk of bias if the treatment or adverse effects of the treatment could not be noticed in the extracted data (Boutron 2005).

Incomplete outcome data (attrition bias)

Attrition bias due to amount, nature or handling of incomplete outcome data

There was a low risk of attrition bias if there were no missing outcome data; reasons for missing outcome data were unlikely to be related to the true outcome (for survival data, censoring unlikely to be introducing bias); missing outcome data were balanced in numbers, with similar reasons for missing data across groups; for dichotomous outcome data, the proportion of missing outcomes compared with the observed event risk was not enough to have a clinically relevant impact on the intervention effect estimate; for continuous outcome data, the plausible effect size (difference in means or standardised difference in means) among missing outcomes was not enough to have a clinically relevant impact on observed effect size, or missing data were imputed using appropriate methods (if drop‐outs are very large, imputation using even 'acceptable' methods may still suggest a high risk of bias) (van Tulder 2003). The percentage of withdrawals and drop‐outs should not exceed 20% for short‐term follow‐up and 30% for long‐term follow‐up and should not lead to substantial bias (these percentages are commonly used but arbitrary, not supported by literature) (van Tulder 2003).

Selective reporting (reporting bias)

Reporting bias due to selective outcome reporting

There was low risk of reporting bias if the study protocol was available and all of the study's pre‐specified (primary and secondary) outcomes that were of interest in the review were reported in the pre‐specified way, or if the study protocol was not available but it was clear that the published reports included all expected outcomes, including those that were pre‐specified (convincing text of this nature may be uncommon).

There was a high risk of reporting bias if not all of the study's pre‐specified primary outcomes were reported; one or more primary outcomes was reported using measurements, analysis methods or subsets of the data (e.g. subscales) that were not pre‐specified; one or more reported primary outcomes were not pre‐specified (unless clear justification for their reporting was provided, such as an unexpected adverse effect); one or more outcomes of interest in the review were reported incompletely so that they could not be entered in a meta‐analysis; the study report did not include results for a key outcome that would be expected to have been reported for such a study.

Group similarity at baseline (selection bias)

Bias due to dissimilarity at baseline for the most important prognostic indicators 

There was low risk of bias if groups were similar at baseline for demographic factors, value of main outcome measure(s), and important prognostic factors (examples in the field of back and neck pain are duration and severity of complaints, vocational status, percentage of participants with neurological symptoms) (van Tulder 2003).

Co‐interventions (performance bias)

Bias because co‐interventions were different across groups

There was low risk of bias if there were no co‐interventions or they were similar between the index and control groups (van Tulder 2003).

Compliance (performance bias)

Bias due to inappropriate compliance with interventions across groups

There was low risk of bias if compliance with the interventions was acceptable, based on the reported intensity/dosage, duration, number and frequency for both the index and control intervention(s). For single‐session interventions (e.g. surgery), this item is irrelevant (van Tulder 2003).

Intention‐to‐treat analysis

There was low risk of bias if all randomised participants were reported/analysed in the group to which they were allocated by randomisation.

Timing of outcome assessments (detection bias)

Bias because important outcomes were not measured at the same time across groups

There was low risk of bias if all important outcome assessments for all intervention groups were measured at the same time (van Tulder 2003).

Other bias

Bias due to problems not covered elsewhere in the table

There was a low risk of bias if the study appeared to be free of other sources of bias not addressed elsewhere (e.g. study funding).

Appendix 4. Definitions

Reports of symptoms in the past week were assessed at baseline and follow‐up, including the frequency (from 0 'not at all' to 6 'always') and bothersomeness (from 0 'not bothersome' to 6 'extremely bothersome') of low back pain, leg pain, leg or foot weakness, leg numbness, and pain in the back or leg while sitting. Sciatica frequency and bothersome indexes, each with scores ranging from 0 to 24, were created by summing the four leg‐related questions.

Roland‐Morris Low Back Pain and Disability Questionnaire (RMQ): RMQ is a self administered disability measure in which greater levels of disability are reflected by higher numbers on a 24‐point scale. For example, at the beginning of treatment, a participant's score was 12 and, at the end of treatment, their score was 2 (10 points of improvement), we would calculate an 83% (10/12 x 100) improvement (Stratford 1996).

Oswestry Disability Index (ODI): an index derived from a low back pain disability questionnaire used to measure a participant's permanent functional disability. The questionnaire contains topics concerning intensity of pain, lifting, ability to care for oneself, ability to walk, ability to sit, sexual function, ability to stand, social life, sleep quality and ability to travel. The index is scored from 0 (no disability) to 100 (maximum disability).

Data and analyses

Comparison 1. Leg pain in two groups of minimally invasive discectomy (MID) and micro/discectomy.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Medium term (1‐5 years)	4	599	Mean Difference (IV, Random, 95% CI)	0.13 [0.09, 0.16]
2 Leg pain in 2 groups of MID and micro/discectomy ‐ short term (at 1 day)	1	30	Mean Difference (IV, Random, 95% CI)	‐0.40 [‐2.33, 1.53]
3 Leg pain in 2 groups of MID and micro/discectomy ‐ short term (at 3 days)	1	30	Mean Difference (IV, Random, 95% CI)	‐0.5 [‐1.95, 0.95]
4 Leg pain in 2 groups of MID and micro/discectomy ‐ short term (at 5 days)	1	30	Mean Difference (IV, Random, 95% CI)	0.10 [‐1.24, 1.44]

Comparison 2. Low back pain ‐ minimally invasive discectomy (MID) versus micro/discectomy.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Sensitivity analysis for low back pain in 1 year	3		Mean Difference (IV, Random, 95% CI)	Subtotals only
1.1 6 months post operative	3	577	Mean Difference (IV, Random, 95% CI)	0.35 [0.19, 0.51]
1.2 1 year post operative	3	577	Mean Difference (IV, Random, 95% CI)	0.19 [‐0.22, 0.59]
1.3 2 years post operative	3	577	Mean Difference (IV, Random, 95% CI)	0.54 [0.29, 0.79]
2 Endoscopic discectomy vs. micro/discectomy	1		Mean Difference (IV, Random, 95% CI)	Subtotals only
2.1 Early post operative	1	212	Mean Difference (IV, Random, 95% CI)	‐0.5 [‐0.79, ‐0.21]
2.2 6 months post operative	1	212	Mean Difference (IV, Random, 95% CI)	0.5 [0.21, 0.79]
2.3 1 year post operative	1	212	Mean Difference (IV, Random, 95% CI)	0.0 [‐0.29, 0.29]
2.4 2 years post operative	1	212	Mean Difference (IV, Random, 95% CI)	0.5 [0.21, 0.79]

Comparison 3. Neurological deficit of lower extremity or bowel/urinary incontinency.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Persistent motor deficits post operative	4	126	Risk Ratio (M‐H, Random, 95% CI)	0.96 [0.56, 1.63]
2 Persistent sensory deficits post operative	4	165	Risk Ratio (M‐H, Random, 95% CI)	0.86 [0.65, 1.15]
3 Persistent reflex deficit postoperative (12 months)	2	47	Risk Ratio (M‐H, Random, 95% CI)	0.68 [0.49, 0.96]
4 Persistent bladder dysfunction > 6 months' follow‐up	1	3	Risk Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]

Comparison 4. Functional outcomes including daily activity and return to work.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Oswestry Disability Index > 6 months post operative	3	312	Mean Difference (IV, Random, 95% CI)	0.84 [‐0.21, 1.88]
2 Number of participants returned to work	1	60	Odds Ratio (M‐H, Random, 95% CI)	2.07 [0.18, 24.15]
3 Postoperative work disability days ‐ return to work	1	178	Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]

Comparison 5. Secondary outcomes ‐ complications of surgery.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Surgical site and other infections	6	931	Risk Ratio (M‐H, Random, 95% CI)	0.23 [0.07, 0.79]
2 Procedure‐related complications	7	991	Risk Ratio (M‐H, Random, 95% CI)	1.01 [0.61, 1.66]
3 Re‐hospitalisation due to recurrent disc herniation ‐ ≥6 months	6	949	Risk Ratio (M‐H, Random, 95% CI)	1.74 [1.03, 2.94]
4 Surgical re‐intervention	4	637	Risk Ratio (M‐H, Random, 95% CI)	1.46 [0.68, 3.14]
5 Dural tear	5	887	Risk Ratio (M‐H, Random, 95% CI)	1.63 [0.82, 3.22]
6 Re‐hospitalisation due to recurrent disc herniation ‐ 2 years' follow‐up	4	777	Risk Ratio (M‐H, Random, 95% CI)	1.89 [1.09, 3.27]
7 Subgroup analysis for duration of hospital stay	2		Mean Difference (IV, Random, 95% CI)	Subtotals only
7.1 Sensitivity analysis based on randomisation for duration of hospital stay	2	537	Mean Difference (IV, Random, 95% CI)	0.13 [‐0.07, 0.33]

5.6. Analysis.

Comparison 5 Secondary outcomes ‐ complications of surgery, Outcome 6 Re‐hospitalisation due to recurrent disc herniation ‐ 2 years' follow‐up.

5.7. Analysis.

Comparison 5 Secondary outcomes ‐ complications of surgery, Outcome 7 Subgroup analysis for duration of hospital stay.

Comparison 6. Secondary outcomes ‐ quality of life measured by SF‐36 or SF‐12, and overall satisfaction of participants, which is usually reported by a Likert scale.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 SF‐36 Physical Functioning subclass > 6 months	2	385	Mean Difference (IV, Random, 95% CI)	‐4.70 [‐5.05, ‐4.35]
2 SF‐36 Bodily Pain subclass > 6 months	2	385	Mean Difference (IV, Random, 95% CI)	‐3.70 [‐4.11, ‐3.28]
3 SF‐36 General Health subclass > 6 months	2	378	Mean Difference (IV, Random, 95% CI)	‐2.52 [‐2.92, ‐2.11]
4 SF‐36 Physical Health component summary (6 months)	2	272	Mean Difference (IV, Random, 95% CI)	0.96 [‐0.12, 2.03]
5 SF‐36 Mental Health component summary (6 months)	2	272	Mean Difference (IV, Random, 95% CI)	‐4.31 [‐9.96, 1.33]
6 Overall success (number of participants)	5	443	Risk Ratio (M‐H, Random, 95% CI)	1.04 [0.99, 1.10]

Comparison 7. Automated percutaneous discectomy versus microdiscectomy.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 SO4b. Overall satisfaction of participants	1	71	Risk Ratio (M‐H, Random, 95% CI)	0.89 [0.67, 1.17]

Comparison 8. Preoperative pain versus postoperative pain at 12 months.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Low back pain at 12 months in discectomy/microdiscectomy	2	602	Mean Difference (IV, Random, 95% CI)	2.33 [1.93, 2.73]

Comparison 9. Microendoscopy versus microdiscectomy/open discectomy (MD/OD).

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Leg pain ‐ medium‐term follow‐up	3	274	Mean Difference (IV, Random, 95% CI)	0.09 [‐0.03, 0.21]
2 Low back pain ‐ 6 months' follow‐up	2	252	Mean Difference (IV, Random, 95% CI)	0.29 [‐0.19, 0.77]
3 Oswestry Disability Index (ODI) > 6 months' follow‐up	2	252	Mean Difference (IV, Random, 95% CI)	0.85 [‐0.21, 1.90]
4 Re‐operations due to recurrence of discopathy	4	564	Risk Ratio (M‐H, Random, 95% CI)	2.13 [1.01, 4.49]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Arts 2011.

Methods	RCT
Participants	325 participants Severity of condition: herniation lasting 8 weeks and refractory to conservative treatment Age: 18‐70 years Gender (M : F): tubular discectomy group 84 : 82; microdiscectomy group 88 : 71 Follow‐up: 24 months
Interventions	Tubular discectomy vs. microdiscectomy
Outcomes
Dichotomous outcomes	Primary: urination deficit, sensory deficit, muscle weakness, asymmetrical deep‐tendon reflexes in knees, asymmetrical deep‐tendon reflexes in ankles, pain on straight‐leg raising test, pain on crossed straight‐leg raising test, pain on slump test Secondary: participant's preference for tubular discectomy, blood loss < 50 mL, total intraoperative complications, dural tear, nerve root injury, exploration started at wrong level, total postoperative complications, wound haematoma, wound infection, UTI, cerebrospinal fluid leakage, miction disturbances, deep venous thrombosis in leg, increase of sensory deficit, increase of motor deficit, timing of mobilisation ‐ same day as surgery or day 1, repeated surgery within 1 or 2 years, recurrent disk herniation, recurrent disk herniation same level at 2 years, stenosis (at 2 years), fibrosis 1 and 2 years, cerebrospinal fluid leakage, cauda equina syndrome at 2 years, instrumented fusion at 1 and 2 years
Continuous outcomes	Primary: Roland‐Morris Disability Questionnaire score, pain in leg (VAS ‐ mm), LBP (VAS ‐ mm), SFI score, SBI score, time to recovery (weeks), Prolo scale ‐ Functional score, Prolo scale ‐ Economic score Secondary: general health, SF‐36 score ‐ Bodily Pain, SF‐36 score ‐ Physical Functioning, operation time (minutes), weight of disk removal (mean and SD in mg), number of days in hospital
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	No biased allocation to interventions due to adequate generation of a randomised sequence
Allocation concealment (selection bias)	Low risk	No biased allocation to interventions due to adequate concealment of allocations prior to assignment
Blinding of participants and personnel (performance bias) All outcomes	High risk	No performance bias due to no knowledge of the allocated interventions by participants during the study. Surgeon was not blind
Blinding of outcome assessment (detection bias) All outcomes	Low risk	No detection bias due to no knowledge of the allocated interventions by outcome assessors
Incomplete outcome data (attrition bias) All outcomes	Low risk	No attrition bias due to low amount, nature or handling of incomplete outcome data
Selective reporting (reporting bias)	Low risk	No reporting bias due to no selective outcome reporting
Group similarity at baseline	Low risk	No selection bias due to similarity at baseline for the most important prognostic indicators
Co‐interventions	Unclear risk	Authors did not mention whether they performed any different non‐surgical management between 2 groups
Compliance	Low risk	No performance bias due to appropriate compliance with interventions across groups
Intention‐to‐treat analysis	Low risk	No risk of bias because all randomised participants are reported and analysed in the group to which they were allocated by randomisation
Timing of outcome assessments	Low risk	No detection bias because important outcomes were measured at the same time across groups
Other bias	Low risk	No bias due to other problems

Chatterjee 1995.

Methods	RCT
Participants	71 participants Severity of condition: all participants had radicular pain and the sciatica was more severe than LBP. Conservative treatment was performed for at least 6 weeks Age: not reported Gender: not reported Follow‐up: ≥ 6 months, but evaluations at 3 weeks and 2 and 6 months
Interventions	Automated percutaneous lumbar discectomy vs. microdiscectomy
Outcomes	MacNab criteria: success (excellent or good); failure (fair or poor)
Dichotomous outcomes	Primary: success rate
Continuous outcomes	Secondary: duration of hospital stay
Notes	We included Chatterjee (1995) because the study was an RCT to compare automated percutaneous lumbar discectomy and microdiscectomy. Chatterjee was stopped early, which is an issue, but we do not consider that stopping a trial early is necessarily a reason to exclude a trial a priori; in some cases, early stopping is warranted and the trial results can provide useful information The Chatterjee study's outcome (MacNab criteria) was not reported in other trials, and, therefore, it was not included in any meta‐analysis for the results of surgery (success vs. failure). MD/OD group had 14/17 (82%) excellent/good results. However, the automated percutaneous lumbar discectomy group even with those participants who failed and treated with MD/OD had 9/12 (75%) excellent/good results. We presented the Chatterjee results qualitatively
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Not reported
Blinding of outcome assessment (detection bias) All outcomes	Low risk	No detection bias due to no knowledge of the allocated interventions by outcome assessors
Incomplete outcome data (attrition bias) All outcomes	Low risk	No attrition bias due to low amount, nature or handling of incomplete outcome data
Selective reporting (reporting bias)	Low risk	No reporting bias due to no selective outcome reporting
Group similarity at baseline	High risk	Selection bias due to dissimilarity at baseline for the duration of symptoms. We did not exclude participants based on duration of symptoms. However, preoperative duration of LBP was 18 (Range: 2‐44) months in the automated percutaneous group but 33 (Range: 2‐60) months in microdiscectomy group
Co‐interventions	High risk	Authors simultaneously performed other co‐interventions not only on small contained disc herniation, but also all types of disc herniation
Compliance	Low risk	No performance bias due to appropriate compliance with interventions across groups
Intention‐to‐treat analysis	Low risk	No risk of bias because all randomised participants were reported and analysed in the group to which they were allocated by randomisation. Authors wrote both data for intention‐to‐treat and not intention‐to‐treat analyses
Timing of outcome assessments	Low risk	No detection bias because important outcomes were measured at the same time across groups
Other bias	High risk	Trial was stopped after an Interim analysis

Garg 2011.

Methods	RCT
Participants	112 participants Severity of condition: a single‐level disc herniation ‐ persistent radiculopathy (despite at least 6 weeks of conservative therapy) and positive tension signs in both straight and crossed leg raising tests without neurological deficit. Conservative treatment included bed rest for a short period (4‐7 days) with restriction of lifting, bending, climbing and heavy exertional activity. No steroids. All had non‐steroidal anti‐inflammatory drugs and physiotherapy Age: 26‐57 years Gender (M : F): MED group 36 : 19; OD group 44 : 13 Follow‐up: 12‐18 months, but evaluations at 6 weeks, 6 months, 1 year
Interventions	MED vs. OD
Outcomes
Dichotomous outcomes	Primary: overall satisfaction at week 6, overall satisfaction at 6 months, overall satisfaction at 1 year; clinical, neurological and electrophysiological improvement was similar in both groups Secondary: complication (dural leak), temporary urinary retention, urinary tract infection, nerve root damage, wound infection, recurrence of discopathy after 7 months treated by OD
Continuous outcomes	Primary: mean ODI score: preoperative, postoperative at week 1, week 6, 6 months, 1 year Secondary: surgical time, anaesthesia time, blood loss, weight of removed disc, hospital stay (days)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Not reported
Blinding of outcome assessment (detection bias) All outcomes	Low risk	No detection bias due to no knowledge of the allocated interventions by outcome assessors
Incomplete outcome data (attrition bias) All outcomes	Low risk	No attrition bias due to low amount, nature or handling of incomplete outcome data
Selective reporting (reporting bias)	Unclear risk	Not reported
Group similarity at baseline	Low risk	No selection bias due to similarity at baseline for the most important prognostic indicators
Co‐interventions	Low risk	No performance bias because no co‐interventions were different across groups
Compliance	Low risk	No performance bias due to appropriate compliance with interventions across groups
Intention‐to‐treat analysis	Unclear risk	Not reported
Timing of outcome assessments	Low risk	No detection bias because important outcomes were measured at the same time across groups
Other bias	Low risk	No bias due to other problems

Hermantin 1999.

Methods	RCT
Participants	60 participants Severity of condition: a single disc herniation at the lumbar level with associated radiculopathy that matches clinic and radiology; not more than half of the anteroposterior diameter of the spinal canal; an absence of osseous or ligamentous stenosis; availability of the disc for both arthroscopic microdiscectomy and laminotomy; failure to respond to non‐operative measures; more pain in the lower extremities than in the back; the presence of positive straight leg raising test ± an associated neurological deficit; no previous operation on the low back and the absence of any Workers' Compensation claim Age: 15‐67 years Gender (M : F): arthroscopic discectomy group 22 : 8; discectomy group 17 : 13 Follow‐up: 19‐42 months; arthroscopic discectomy 32 months; discectomy 31 months
Interventions	Arthroscopic discectomy vs. discectomy
Outcomes
Dichotomous outcomes	Primary: participants were very satisfied with the results of operation (overall success), motor deficit (neurological deficit), sensory deficit, reflexes, functional outcome (return to work) Secondary: neurovascular complications, postoperative infections, occasional narcotic use (at latest follow‐up)
Continuous outcomes	Primary: pain score (radiculopathy), duration postoperative, disability (return to work) Secondary: frequency of postoperative use of narcotics per day, duration of use of narcotics (days)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk
Allocation concealment (selection bias)	Low risk
Blinding of participants and personnel (performance bias) All outcomes	High risk
Blinding of outcome assessment (detection bias) All outcomes	High risk
Incomplete outcome data (attrition bias) All outcomes	Low risk
Selective reporting (reporting bias)	Low risk
Group similarity at baseline	Low risk
Compliance	Low risk
Intention‐to‐treat analysis	Low risk
Timing of outcome assessments	Low risk
Other bias	Low risk

Huang 2005.

Methods	RCT
Participants	22 participants, MED group 10 participants, OD group 12 participants Severity of condition: symptomatic herniated lumbar discs not responded to conservative treatment after 3 months, and 6 participants with acute severe back and leg pain that had no improvement after 1‐2 weeks of absolute bed rest Age (mean ± SD): MED group 39.2 ± 10.8 years; OD group 39.8 ± 11.0 years Gender (M : F): MED group 6 : 4; OD group: 9 : 3: Follow‐up: 10‐25 months, mean 18.9 months
Interventions	MED vs. OD
Outcomes
Dichotomous outcomes
Continuous outcomes	Primary: pain severity of leg (VAS) preoperative, pain severity of leg (VAS) after mean postoperative follow‐up of 18.9 months; overall success (clinical outcome) ‐ MacNab Secondary: postoperative hospital stay, intraoperative blood loss, operation length (including the set‐up time) (minutes), mean size of skin incision, complications, nerve root sleeve tear, superficial wound infection
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Incomplete outcome data (attrition bias) All outcomes	Low risk
Selective reporting (reporting bias)	Low risk
Group similarity at baseline	Low risk
Co‐interventions	Low risk
Compliance	Low risk
Timing of outcome assessments	Low risk
Other bias	Low risk

Mayer 1993.

Methods	RCT
Participants	40 participants Severity of condition: all participants had radicular symptoms such as a positive straight‐leg raising test, sciatica, sensory disturbances, mild motor weakness. Conservative therapy had been unsuccessful in all participants Age: 12‐55 years Gender (M : F): percutaneous endoscopic discectomy group 12 : 8; microdiscectomy group 14 : 6 Follow‐up: ≥ 48 months
Interventions	Percutaneous endoscopic discectomy vs. microdiscectomy
Outcomes
Dichotomous outcomes	Primary: LBP; sciatica; sensory deficits; motor deficits; reflex deficits; qualitative assessments (all at 2 years postoperative): excellent, good, moderate, bad, subjective assessments (all at 2 years postoperative): excellent, good, moderate, bad Secondary: re‐operation, recurrence, spondylolisthesis, other complications
Continuous outcomes	Primary: modified Suezawa and Schreiber (LBP, sciatica, motor, sensory, reflex): 0 = very bad, 10 = excellent, preoperative and postoperative score at 2 years Secondary: operation time (minutes)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Not reported
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	Attrition bias due to amount, nature or handling of incomplete outcome data
Selective reporting (reporting bias)	Low risk	Reporting bias due to selective outcome reporting
Group similarity at baseline	Low risk	Selection bias due to dissimilarity at baseline for the most important prognostic indicators
Co‐interventions	Unclear risk	Not reported
Compliance	Low risk	Performance bias due to inappropriate compliance with interventions across groups
Intention‐to‐treat analysis	Low risk	Risk of bias if all randomised participants are not reported and analysed in the group to which they were allocated by randomisation
Timing of outcome assessments	Low risk	Detection bias if important outcomes were not measured at the same time across groups
Other bias	Low risk	Bias due to problems not covered elsewhere in the table

Righesso 2007.

Methods	RCT
Participants	40 participants Severity of condition: sciatica caused by herniated lumbar discs who did not respond to 4‐8 weeks of conservative treatment with rest, analgesia, non‐steroidal anti‐inflammatory drugs and physiotherapy Age: < 60 years Gender (M : F): OD group 10 : 11; MED group 13 : 6 Follow‐up: 24‐56 months
Interventions	OD vs. MED
Outcomes
Dichotomous outcomes	Primary: neurological deficits, motor deficits, sensory deficits, altered reflexes
Continuous outcomes	Primary: pain surgical wound pain (VAS), disability (ODI)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding of participants and personnel (performance bias) All outcomes	High risk	Performance bias due to knowledge of the allocated interventions by participants during the study
Blinding of outcome assessment (detection bias) All outcomes	High risk	Detection bias due to knowledge of the allocated interventions by outcome assessors
Incomplete outcome data (attrition bias) All outcomes	Low risk	No attrition bias due to low amount, nature or handling of incomplete outcome data
Selective reporting (reporting bias)	Low risk	No reporting bias due to selective outcome reporting
Group similarity at baseline	High risk	Selection bias due to dissimilarity at baseline for the most important prognostic indicators, because MED group had preoperative pain (VAS) less than OD group (P value = 0.03)
Co‐interventions	Low risk	No performance bias because no co‐interventions were different across groups
Compliance	Low risk	No performance bias due to appropriate compliance with interventions across groups
Intention‐to‐treat analysis	Unclear risk	Not reported
Timing of outcome assessments	Low risk	No detection bias because important outcomes were measured at the same time across groups
Other bias	Low risk	No bias due to other problems

Ruetten 2008.

Methods	RCT
Participants	200 participants Severity of condition: clinically symptomatic disc herniation ‐ radicular pain symptoms and presence of neurological deficits Age: range 20‐68 years (mean 43 years) Gender (M : F): 84 : 116 Pain duration ranged from 1 day to 16 months (mean 82 days) 162 participants had received a mean of 9 weeks' conservative treatment The indication for surgery was defined according to present‐day standards Follow‐up: ≥ 24 months
Interventions	Full‐endoscopic interlaminar and transforaminal vs. conventional microdiscectomy
Outcomes
Dichotomous outcomes	Primary: leg pain, LBP, overall success (satisfaction and would undergo operation again), functional outcome including daily activity and return to work (postoperative work disability) days Secondary: serious complications, dural injury, nerve injury, cauda equine syndrome, transient postoperative, dysaesthesia, postoperative bleeding, delayed wound healing, soft tissue infection, transient urinary retention, spondylodiscitis, thrombosis, recurrence, mean operation time (minutes), mean operative time in revisions (minutes), revision (spinal canal expansion and fusion), postoperative pain and pain medication
Continuous outcomes	LBP, leg pain, disability, North American Spine Society Instrument scores (NASS pain), NASS neurology
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Biased allocation to interventions due to the alternate assignment of the participants to 2 groups of intervention and control
Allocation concealment (selection bias)	High risk	Biased allocation to interventions due to open randomisation and no concealment of allocations prior to assignment
Blinding of participants and personnel (performance bias) All outcomes	High risk	Performance bias due to knowledge of the allocated interventions by participants during the study
Blinding of outcome assessment (detection bias) All outcomes	High risk	Detection bias due to knowledge of the allocated interventions by outcome assessors.
Incomplete outcome data (attrition bias) All outcomes	Low risk	No attrition bias due to low amount, nature or handling of incomplete outcome data
Selective reporting (reporting bias)	Low risk	No reporting bias due to no selective outcome reporting
Group similarity at baseline	Unclear risk	No mention of similarity of age, sex and duration of symptoms in 2 groups
Co‐interventions	Low risk	No performance bias because no co‐interventions were different across groups
Compliance	Low risk	No performance bias due to appropriate compliance with interventions across groups
Intention‐to‐treat analysis	Unclear risk	Not reported
Timing of outcome assessments	Low risk	No detection bias because important outcomes were measured at the same time across groups
Other bias	Low risk	No bias due to other problems

Ryang 2008.

Methods	RCT
Participants	60 participants Severity of condition: a single level virgin lumbar disc herniation, typical monoradicular symptoms associated with the similar lumbar segment, sciatica more severe than back pain, failure of 8‐12 weeks of conservative treatment Age: 21‐69 years Gender (M : F): tubular discectomy group 13 : 17; microdiscectomy group 19 : 11 Follow‐up: 6‐26 months (mean 16 months)
Interventions	Tubular discectomy vs. microdiscectomy
Outcomes
Dichotomous outcomes	Complications
Continuous outcomes	Leg pain
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Selective reporting (reporting bias)	Low risk
Group similarity at baseline	Low risk
Co‐interventions	Low risk
Compliance	Low risk
Timing of outcome assessments	High risk
Other bias	Low risk

Shin 2008.

Methods	RCT
Participants	30 participants Severity of condition: single‐level unilateral herniated nucleus pulposus ‐ refractory to > 6 weeks of conservative treatment Age: MED group 43 ± 18 years; microdiscectomy group 48 ± 11 years Gender (M : F): MED group 7 : 8; microdiscectomy group 5 : 10 Follow‐up: 1‐5 days
Interventions	MED vs. microscopic discectomy
Outcomes	LBP and leg pain from day 1 to 5
Dichotomous outcomes
Continuous outcomes	LBP (VAS in mm); leg pain (VAS in mm)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk
Blinding of outcome assessment (detection bias) All outcomes	Low risk
Selective reporting (reporting bias)	Low risk
Compliance	Low risk
Timing of outcome assessments	Low risk
Other bias	Low risk

Teli 2010.

Methods	RCT
Participants	212 participants Severity of condition: symptomatic, single level lumbar disc herniation ‐ pain or neurological signs (or both) in the same distribution lasting > 6 weeks of conservative treatment consisting of systemic drugs for pain relief or epidural steroid (or both) Age: 18‐65 years Gender (M : F): MED group 45 : 25; discectomy/microdiscectomy 94 : 48 Follow‐up: 24‐29 months
Interventions	MED vs. discectomy/microdiscectomy
Outcomes
Dichotomous outcomes	Secondary: death, dural tear, root injury, recurrent herniation (operated), wound infection, spondylodiscitis, worsening motor deficit, hospital stay (hour)
Continuous outcomes	Primary: leg pain (mean VAS), back pain (mean VAS), disability (ODI) Secondary: SF‐36 (Physical Health), SF‐36 (Mental Health)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	No biased allocation to interventions due to adequate generation of a randomised sequence
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding of participants and personnel (performance bias) All outcomes	High risk	Performance bias due to knowledge of the allocated interventions by participants during the study. Authors wrote that the local ethical committee expressed approval of the study with the caveat that blindness had to be omitted "in order to protect the possibility of a free choice for patients"
Blinding of outcome assessment (detection bias) All outcomes	High risk	Detection bias due to knowledge of the allocated interventions by outcome assessors. Authors wrote that the local ethical committee expressed approval of the study with the caveat that blindness had to be omitted "in order to protect the possibility of a free choice for patients"
Incomplete outcome data (attrition bias) All outcomes	Low risk	No attrition bias due to low amount, nature or handling of incomplete outcome data
Selective reporting (reporting bias)	Low risk	No reporting bias due to no selective outcome reporting
Group similarity at baseline	Low risk	No selection bias due to similarity at baseline for the most important prognostic indicators
Co‐interventions	Low risk	No performance bias because no co‐interventions were different across groups
Compliance	Low risk	No performance bias due to appropriate compliance with interventions across groups
Intention‐to‐treat analysis	Low risk	No risk of bias because all randomised participants are reported and analysed in the group to which they were allocated by randomisation
Timing of outcome assessments	Low risk	No detection bias because important outcomes were measured at the same time across groups
Other bias	Low risk	No bias due to other problems

LBP: low back pain; MED: microendoscopic discectomy; OD: open discectomy; ODI: Oswestry Disability Index; RCT: randomised controlled trial; SD: standard deviation; SF‐36: 36‐item Short Form; VAS: visual analogue scale.

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Brock 2008	Compared different minimally invasive techniques to each other. We only included studies of minimally invasive vs. standard/microdiscectomy
Franke 2009	Compared MID (Dilator inserted within muscular microscopic discectomy) to MD. However, it did not report any of our included outcome measures. Rather, it reported the "sum VAS" score (leg plus back pain). The main outcome measure of the study was to compare operation time between 2 centres (index and transfer)
Harrington 2008	Retrospective comparison of MID and discectomy
Henriksen 1996	Compared standard MD and OD without microscope. Authors did not compare MD/OD with MID
Karasek 2000	In our protocol, we agreed to exclude intradiscal thermal annuloplasty from our list of relevant interventions. We excluded trials of intradiscal electrothermal annuloplasty in which the mechanism of action involves destruction or disruption of the disc causing mechanical compression using energy or chemicals, rather than removal of disc materials
Katayama 2006	Compared standard MD and OD without microscope. Authors did not compare MD/OD with MID
Lagarrigue 1994	Compared standard MD and OD without microscope. Authors did not compare MD/OD with MID
Thomé 2005	Compared standard lumbar microdiscectomy versus sequestrectomy
Tullberg 1993	Compared standard MD and OD without microscope. Authors did not compare MD/OD with MID
Türeyen 2003	Compared standard MD and OD without microscope. Authors did not compare MD/OD with MID
van den Akker 2011	Although the study was original and published in Neurosurgery, the focus was on cost utility, which is unrelated to our primary and secondary outcomes

MD: microdiscectomy; MID: minimally invasive discectomy; OD: open discectomy; VAS: visual analogue scale.

Differences between protocol and review

We included all pre‐planned primary or secondary (or both) outcomes in the review; however, there is no stool dysfunction evaluation reported in the identified trials.

However, we found two minor additional secondary outcome measures for minimally invasive discectomy; 1. intraoperative blood loss and 2. time (minutes) of surgical operation. We did not add these two new outcomes to the review because we had not pre‐specified them.

Contributions of authors

Mohammad R. Rasouli: protocol, clinical interpretation, selection of studies, methodological evaluation, draft review, quality control, discussion.

Vafa Rahimi‐Movaghar: protocol, clinical interpretation, selection of studies, methodological evaluation, writing and editing the draft, data extraction, data entry in Review Manager 5, data analysis, quality control, discussion.

Maziar Moradi‐Lakeh: protocol, selection of studies, methodological evaluation, draft review.

Farhad Shokraneh: protocol, search strategy, selection of studies.

Roger Chou: protocol review, selection of studies, methodological evaluation, draft review, quality control, discussion.

Sources of support

Internal sources

• Tehran University of Medical Sciences, Iran.

  The protocol was supported by a grant for corresponding author confirmed 10 May 2011.

External sources

• No sources of support supplied

Declarations of interest

The authors performed no RCTs to compare these two methods. We declare that we have no interest in the results.

New