Surgical versus non-surgical treatment for thoracolumbar burst fractures without neurological deficit

  • Review
  • Intervention

Authors

  • Minawaer Abudou,

    1. Xinjiang Medical University, The Eye Department of the First Affiliated Hospital, Xinjiang, China
    Search for more papers by this author
  • Xueyi Chen,

    1. First Affiliated Hospital of Xinjiang Medical University, Eye Department, Xinjiang, China, China
    Search for more papers by this author
  • Xiangyu Kong,

    1. Chengdu Medical College, Department of Epidemiology, Chengdu, China
    Search for more papers by this author
  • Taixiang Wu

    Corresponding author
    1. West China Hospital, Sichuan University, Chinese Clinical Trial Registry, Chinese Ethics Committee of Registering Clinical Trials, Chengdu, Sichuan, China
    • Taixiang Wu, Chinese Clinical Trial Registry, Chinese Ethics Committee of Registering Clinical Trials, West China Hospital, Sichuan University, No. 37, Guo Xue Xiang, Chengdu, Sichuan, 610041, China. txwutx@hotmail.com.

    Search for more papers by this author

Abstract

Background

Spinal burst fractures result from the failure of both the anterior and the middle columns of the spine under axial compression loads. Conservative management is through bed rest and immobilisation once the acute symptoms have settled. Surgical treatment involves either anterior or posterior stabilisation of the fracture, sometimes with decompression involving the removal of bone fragments that have intruded into the vertebral canal. This is an update of a review first published in 2006.

Objectives

To compare the outcomes of surgical with non-surgical treatment for thoracolumbar burst fractures without neurological deficit.

Search methods

We searched the Cochrane Bone, Joint and Muscle Trauma Group Specialised Register (October 2012), the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2012, Issue 8), MEDLINE (1946 to October 2012), EMBASE (1980 to October 2012) and the Chinese Biomedical Literature Database (1978 to October 2012). We also searched trial registers and reference lists of articles.

Selection criteria

Randomised or quasi-randomised controlled trials comparing surgical with non-surgical treatment of thoracolumbar burst fractures without neurological deficit.

Data collection and analysis

Two review authors independently assessed risk of bias and extracted data independently. Only limited pooling of data was done.

Main results

We included two trials that compared surgical with non-surgical treatment for patients with thoracolumbar burst fractures without neurological deficit. These recruited a total of 87 participants and reported outcomes for 79 participants at follow-up of two years or more. Both trials were judged at unclear risk of selection bias and at high risk of performance and detection biases, resulting from lack of blinding.

The two trials reported contrasting results for pain and function-related outcomes at final follow-up, and numbers returning to work. One trial found less pain (mean difference (MD) -15.09 mm, 95% CI -27.81 to -2.37; 100 mm visual analogue scale), and better function based on the Roland and Morris disability questionnaire results (MD -5.87, 95% CI -10.10 to -1.64; 24 points = maximum disability) in the surgical group. Based on the same outcome measures, the other trial found the surgical group had more pain (MD 13.60 mm, 95% CI -0.31 to 27.51) and worse function (MD 4.31, 95% CI 0.54 to 8.08). Neither trial reported a statistically significant difference in return to work. There were greater numbers of participants with complications in the surgical group of both trials (21/41 versus 6/38; RR 2.85, 95% CI 0.83 to 9.75; 2 trials), and only participants of this group had subsequent surgery, involving implant removal either for complications or as a matter of course. One trial reported that surgery was over four times more costly than non-surgical treatment.

Authors' conclusions

The contradictory evidence provided by two small and potentially biased randomised controlled trials is insufficient to conclude whether surgical or non-surgical treatment yields superior pain and functional outcomes for people with thoracolumbar burst fractures without neurological deficit. It is likely, however, that surgery is associated with more early complications and the need for subsequent surgery, as well as greater initial healthcare costs.

Résumé scientifique

Comparaison entre le traitement chirurgical et non chirurgical pour les fractures-éclatement thoracolombaires sans déficit neurologique

Contexte

Les fractures-éclatement de la colonne vertébrale résultent de lésions de la colonne vertébrale aussi bien antérieure que centrale sous l'effet de charges de compression axiale. La gestion conservatrice comporte un repos au lit et une immobilisation une fois que les symptômes aigus ont disparu. Le traitement chirurgical comporte une stabilisation antérieure ou bien postérieure de la fracture, parfois associée à une décompression comportant l'extraction de fragments osseux qui se sont introduits dans le canal vertébral. Ceci est une mise à jour d'une revue publiée pour la première fois en 2006.

Objectifs

Comparer les résultats du traitement chirurgical et non chirurgical pour les fractures-éclatement thoracolombaires sans déficit neurologique.

Stratégie de recherche documentaire

Nous avons effectué une recherche dans le registre spécialisé du groupe Cochrane sur les traumatismes ostéo-articulaires et musculaires (octobre 2012), le registre Cochrane des essais contrôlés - Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2012, numéro 8), MEDLINE (de 1946 à octobre 2012), EMBASE (de 1980 à octobre 2012) et la base de données de la littérature biomédicale chinoise (CBM, Chinese BioMedical Literature Database) (de 1978 à octobre 2012). Nous avons également recherché dans les registres d'essais et les listes bibliographiques d'articles.

Critères de sélection

Essais contrôlés randomisés ou quasi-randomisés comparant le traitement chirurgical au traitement non chirurgical pour les fractures-éclatement thoracolombaires sans déficit neurologique.

Recueil et analyse des données

Deux auteurs de la revue ont, de manière indépendante, évalué le risque de biais et extrait des données. Seul un regroupement limité des données a été réalisé.

Résultats principaux

Nous avons inclus deux essais qui comparaient le traitement chirurgical au traitement non chirurgical pour les patients présentant des fractures-éclatement thoracolombaires sans déficit neurologique. Ceux-ci portaient sur un total de 87 participants et indiquaient des résultats pour 79 participants après deux ans de suivi ou plus. Les deux essais ont été considérés comme présentant un risque peu clair de biais de sélection et un risque élevé de biais de performance et de détection, résultant de l'absence d’assignation secrète.

Les deux essais indiquaient des résultats contrastés pour les résultats liés à la douleur et à la fonction lors du suivi final, et le nombre de patients retournant au travail. Un essai indiquait une douleur moins importante (différence moyenne (DM) -15,09 mm, IC à 95 % -27,81 à -2,37 ; échelle visuelle analogique de 100 mm), et une meilleure fonction d'après les résultats au questionnaire sur l'incapacité de Roland et Morris (DM -5,87, IC à 95 % -10,10 à -1,64 ; 24 points = incapacité maximale) dans le groupe de patients subissant une intervention chirurgicale. Sur la base des mêmes mesures de résultat, l'autre essai indiquait une douleur plus importante (DM 13,60 mm, IC à 95 % -0,31 à 27,51) et une fonction plus mauvaise (DM 4,31, IC à 95 % 0,54 à 8,08) pour le groupe de patients subissant une intervention chirurgicale. Aucun essai n'indiquait de différence statistiquement significative pour le retour au travail. Il y avait un nombre plus important de participants souffrant de complications dans le groupe de patients subissant une intervention chirurgicale des deux essais (21/41 contre 6/38 ; RR 2,85, IC à 95 % 0,83 à 9,75 ; 2 essais), et seuls les participants de ce groupe subissaient une intervention chirurgicale subséquente, comportant le retrait de l'implant soit en raison de complications soit de manière systématique. Un essai indiquait que l'intervention chirurgicale était plus de quatre fois plus coûteuse que le traitement non chirurgical.

Conclusions des auteurs

Les données contradictoires fournies par deux petits essais contrôlés randomisés potentiellement biaisés sont insuffisantes pour conclure si le traitement chirurgical ou non chirurgical donne de meilleurs résultats en termes de douleur et de fonction chez les personnes présentant des fractures-éclatement thoracolombaires sans déficit neurologique. Cependant, l'intervention chirurgicale est susceptible d'être associée à davantage de complications précoces et à la nécessité d'une intervention chirurgicale subséquente, ainsi qu'à des dépenses initiales de soins de santé plus importantes.

Plain language summary

Surgical versus non-surgical treatment for thoracolumbar burst fractures without neurological deficit

The thoracolumbar region of the spine is composed of the thoracic (middle back) and lumbar (lower back) spine. One type of spinal injury is the burst fracture where a vertebra (one of several bones making up the spine) is fractured (broken) such that it loses height on both its back and front sides. This sort of fracture occurs most frequently in the bones situated at the junction of the thoracic and lumbar spine. These injuries are usually the result of a high-velocity accident such as a motor vehicle crash. These are serious injuries, particularly when the spinal cord is also damaged as this may result in the partial or complete loss of sensory and motor function in the legs, and bladder or bowel dysfunction. This review only included people whose nerve tissue was not damaged, although later damage could not be ruled out. People are treated in hospital either conservatively by being placed in a lying position that reduces strain on that part of the spine followed by fitting a cast or brace so that they can move around, or surgically by stabilising the affected part of the spine using various implants and procedures.

We included data from two trials, which included 87 participants. The trials compared surgical with non-surgical treatment for these fractures in the thoracolumbar region of the spine. Both trials had limitations in their methods that could reduce the reliability of their results. They reported contrasting results for patient pain and function at a minimum of two years after treatment. One study found patients had less pain and better function after surgery compared with patients who did not have surgery. The other trial found the opposite. Both trials found there were more early complications in the surgical group and only participants of this group had subsequent additional surgery. This involved the removal of the implant either to resolve a complication or routinely. One trial reported that surgery was over four times more costly than non-surgical treatment.

Our review concluded that the weak evidence from these two trials was insufficient to say whether surgery or non-surgical treatment was better for these fractures. However, surgery is likely to be associated with more early complications and the need for subsequent surgery, as well as greater initial healthcare costs.

Résumé simplifié

Comparaison entre le traitement chirurgical et non chirurgical pour les fractures-éclatement thoracolombaires sans déficit neurologique

La région thoracolombaire de la colonne vertébrale se compose de la colonne vertébrale thoracique (partie centrale du dos) et lombaire (bas du dos). Un type de lésion médullaire est la fracture-éclatement dans laquelle une vertèbre (un des quelques os qui constituent la colonne vertébrale) est fracturée (cassée) de manière telle que sa hauteur est réduite aussi bien à l'arrière qu'à l'avant. Cette sorte de fracture affecte le plus fréquemment les os situés à la jonction de la colonne vertébrale thoracique et lombaire. Ces lésions sont généralement le résultat d'un accident survenant à vitesse élevée tel qu'un accident de véhicule à moteur. Il s'agit de lésions graves, en particulier lorsque la moelle épinière est également endommagée car il peut s'ensuivre une perte partielle ou complète de la fonction sensorielle et motrice des jambes, et un dysfonctionnement de la vessie et de l'intestin. La présente revue ne comprenait que des personnes dont le tissu nerveux n'était pas endommagé, bien que des dommages ultérieurs ne puissent pas être exclus. Les personnes sont traitées à l'hôpital soit de manière conservatrice en étant placées dans une position allongée qui réduit les contraintes sur cette partie de la colonne vertébrale puis en portant un plâtre ou une attelle de manière à pouvoir se déplacer, soit chirurgicalement par stabilisation de la partie affectée de la colonne vertébrale à l'aide de divers implants et procédures.

Nous avons inclus les données de deux essais, qui comportaient 87 participants. Les essais comparaient le traitement chirurgical au traitement non chirurgical pour ces fractures dans la région thoracolombaire de la colonne vertébrale. Les méthodes des deux essais présentaient des limitations pouvant réduire la fiabilité de leurs résultats. Ils présentaient des résultats contrastés pour la douleur ressentie par les patients et la fonction à un minimum de deux ans après le traitement. Une étude a révélé que les patients ressentaient une douleur moins importante et présentaient une meilleure fonction après une opération chirurgicale en comparaison avec les patients qui n'avaient pas subi d'opération chirurgicale. L'autre essai indiquait le contraire. Les deux essais ont mis en évidence qu'il y avait davantage de complications précoces dans le groupe de patients subissant une opération chirurgicale et que seuls les participants de ce groupe subissaient une intervention chirurgicale subséquente additionnelle. Celle-ci consistait à retirer l'implant soit pour résoudre une complication soit de manière systématique. Un essai indiquait que l'intervention chirurgicale était plus de quatre fois plus coûteuse que le traitement non chirurgical.

Notre revue a conclu que les données peu probantes de ces deux essais étaient insuffisantes pour déterminer si le traitement chirurgical ou non chirurgical était plus approprié pour ces fractures. Cependant, l'intervention chirurgicale est susceptible d'être associée à davantage de complications précoces et à la nécessité d'une intervention chirurgicale subséquente, ainsi qu'à des dépenses initiales de soins de santé plus importantes.

Notes de traduction

Traduit par: French Cochrane Centre 16th July, 2013
Traduction financée par: Pour la France : Minist�re de la Sant�. Pour le Canada : Instituts de recherche en sant� du Canada, minist�re de la Sant� du Qu�bec, Fonds de recherche de Qu�bec-Sant� et Institut national d'excellence en sant� et en services sociaux.

Background

Description of the condition

Spinal burst fractures result from failure of both the anterior and the middle columns of the spine under axial compression loads, frequently associated with flexion loading. In the three column theory of spinal stability (Denis 1983), the anterior column is composed of the anterior part of the vertebral body and its adjacent soft tissue complex; the middle column consists of the posterior half of the vertebral body and the soft tissue complex; and the posterior column is formed by the osseous and ligamentous structures posterior to the vertebral body.

Thoracolumbar burst fractures account for up to 17% of major spinal fractures (Denis 1983). Males are at four times higher risk than females. Motor vehicle accidents are the commonest cause of injury, followed by falls and sports-related injuries (Price 1994). Other organ system injuries occur in up to 50% of thoracolumbar trauma patients (Purcell 1981; Weinstein 1988). High-energy injuries associated with paraplegia have a 7% first-year mortality rate (Shikata 1990). The incidence of neurological deficit resulting from thoracolumbar burst fractures is estimated to be 50% to 60% (Denis 1983; McEvoy 1985). Most of these fractures occur in the thoracolumbar junction section of the spine, which comprises one or more thoracic vertebrae (10th, 11th and 12th thoracic: T10, T11 and T12) and one or more lumbar vertebrae (first and second lumbar: L1 and L2) according to the various definitions in the literature. Common definitions of this section are T10 to L2, T11 to L2 and T12 to L2 (Denis 1983; McEvoy 1985). The junction forms a transition zone between the relatively fixed and kyphotic thoracic spine above and the relatively mobile and lordotic lumbar spine below, which causes stress forces to concentrate upon the thoracolumbar vertebral column and predisposes this area to injury (Gertzbein 1992).

The clinical features of thoracolumbar fracture include acute back pain, limited motion, and swelling at the fracture site. Damage to the nerve root or spinal cord, or both, from protruding fragments of vertebral body may result in partial loss of sensory and motor function in the lower extremity or, in cases of severe disruption, may produce full paralysis, urinary and fecal incontinence, and permanent neurological injury. Characteristic radiological findings of the fractures include moderate to marked anterior wedging of the vertebral body, an increased interpedicular distance and narrowing of the spinal canal as a result of displacement of vertebral body fragments. Although many thoracolumbar spinal injuries do not result in paralysis, they may leave an unstable spinal segment due to disruption of bony elements and soft tissues. Late paralysis may result from the residual instability (Denis 1983).

The classification of thoracolumbar injury remains controversial. Denis 1983 proposed a classification of thoracolumbar fractures into four types, based on the three column theory, and defined three patterns of instability. The Arbeitsgemeinschaft für osteosynthesefragen (AO) classification system, developed by Müller and colleagues (OTA 1996) and subsequently updated (Marsh 2007), has also been commonly applied in clinical practice. In 2005 a new classification system, the Thoracolumbar Injury Classification and Severity Score (TLICS), was suggested by Vaccaro et al (Vaccaro 2005) in which the classification of the thoracolumbar injury is based on three injury characteristics: 1) morphology of injury determined by radiographic appearance, 2) integrity of the posterior ligamentous complex, and 3) the neurological status of the patient. The calculation of thoracolumbar injury severity score is based on these characteristics, with a higher score indicating a more severe injury.

Description of the intervention

In clinical practice, thoracolumbar burst fractures without neurological deficit are managed either non-surgically (conservative treatment) or surgically.

Non-surgical treatments include recumbent bed rest and postural reduction, followed by casting or bracing with early ambulation. In the traditional conservative management of unstable thoracolumbar injuries, the patient was confined to bed, with great attention paid to lordotic posture, for as long as six to eight weeks followed by gradual mobilisation (Bedbrook 1975). However, this regimen resulted in mortality rates as high as 90% (Bedbrook 1975). Holdsworth 1970 recommended postural reduction of the deformity that resulted from the burst fracture, followed by hyperextension casting and patient mobilisation. This resulted in significantly improved morbidity and mortality rates. Postural reduction can produce an indirect decompression of the spinal canal, and a well-fitted cast or orthosis may also maintain reduction and lessen deformity while late neurologic deterioration is uncommon (Cantor 1993; Chow 1996).

Surgery generally involves open reduction and stabilisation, using either a posterior or anterior approach. Sometimes surgery also involves decompression, where bone fragments that have intruded into the vertebral canal are removed. One now dated technique using the posterior approach involved posterior reduction and stabilisation using the Harrington distraction rod system, a kind of long-segment fixation. Dickson 1978 demonstrated that this technique produced restoration of vertebral height at the injured level, reduction of the kyphosis, and spinal canal clearance to some degree. However, the instrumentation had some drawbacks. It required fixation of an extensive area of the spine, at least two levels caudal (lower) and three levels cephalad (higher) than the fracture. In addition, it offered only semirigid fixation, thus supplemental external support was often necessary. A more modern and more popular technique for internal fixation is posterior pedicle screw fixation. As summarised in the Cochrane review on pedicle screw fixation for traumatic thoracolumbar fractures, "pedicle screw fixation can be divided into several categories based on the lengths of the fixed segments and of pedicle screws in the injured vertebra" (Cheng 2013). Cheng 2013 refer to three main categories: short-segment fixation pedicle screw instrumentation where two pedicle screws are inserted above and below the fractured vertebra, respectively (Dick 1985); long-segment pedicle screw instrumentation where four pedicle screws are inserted into four vertebrae, two above and two below the fractured vertebra, respectively (Tezeren 2005); and monosegmental pedicle screw instrumentation, where pedicle screws are inserted at the level of the fracture and one level adjacent depending on the location of the intact endplate (Liu 2009). Additional procedures include the use of autogenous bone grafts to fuse together adjacent segments (Wang 2006).

How the intervention might work

The goals of treatment for thoracolumbar burst fractures are to obtain early patient mobilisation and a painless, balanced, stable vertebral column with maximum spine mobility and optimal neurological function (Eskenazi 2000).

Surgical management offers the prospect of immediate stability, correction of deformity, early walking, reduced reliance on casts and braces, and theoretical protection against future neurological deterioration. However, this comes at the cost of surgery-related morbidity and risk of surgical complications such as infection, hardware-related problems and fixation failure (McLain 1993), and iatrogenic injury. Advocates of conservative treatment cite cohort studies demonstrating good outcomes, low progression of deformity, less back pain than surgical treatment, satisfactory work status, low incidence of neurological deterioration, progressive bony remodeling and diminution of canal compromise (Cantor 1993; Chow 1996; Weinstein 1988). Although some studies argue that the degree of kyphosis does not correlate to clinical outcomes, others have demonstrated a relationship between significant deformity (> 30 degrees) and increased pain (Gertzbein 1992; Weinstein 1988). Although in one study of non-surgical treatment, with poor follow-up, the incidence of late neurological deterioration was 17% (Denis 1984) others have reported much lower rates of between 0% and 3% (Cantor 1993; Chow 1996; Weinstein 1988).

Why it is important to do this review

The decision to treat thoracolumbar burst fractures without neurological deficit surgically or non-surgically remains controversial (Gnanenthiran 2012; Okuyama 1996). Thus, an update of this systematic review that was first published in 2006 (Yi 2006) assessing the evidence from randomised controlled trials comparing different methods of treatment for patients with thoracolumbar burst fractures without neurological deficit is justified.

Objectives

To compare the outcomes of surgical with non-surgical treatment for thoracolumbar burst fractures without neurological deficit.

Methods

Criteria for considering studies for this review

Types of studies

Randomised controlled trials and quasi-randomised trials comparing surgical with non-surgical treatment of thoracolumbar burst fractures without neurological deficit.

Types of participants

Adults, aged 18 years or over, with a radiologically confirmed, recent (less than three weeks) thoracolumbar burst fracture without neurological deficit.

Types of interventions

Any method of surgical treatment (for example isolated posterior stabilisation with or without decompression, or isolated anterior decompression and stabilisation) versus any method of conservative treatment (including lying down or bed rest, postural reduction followed by casting or bracing, and early ambulation).

Types of outcome measures

Primary outcomes
  • Proportion of patients who recovered according to self or clinician assessment, or both (e.g. pain assessed by visual analogue scale (VAS))

  • Proportion of patients who had an improvement in function measured on a disability or quality of life scale (e.g. Oswestry Disability Scale or Roland-Morris Disability Questionnaire for low back pain and leg pain (RMDQ-24))

  • Return to work

  • Rate of subsequent surgery

  • The rate of complications: 1) fatal, 2) life threatening, 3) neurological deterioration, 4) decubitus ulcers, 5) infection, 6) implant failure

Secondary outcomes
  • Sagittal plane kyphosis

  • Sagittal balance of the patient

  • Degree of spinal canal compromise

  • Mean duration of hospitalisation

  • Economic evaluation

We also collected data on the correlation between the final amount of kyphosis or canal compromise and the pain or disability reported.

Search methods for identification of studies

Electronic searches

We searched the Cochrane Bone, Joint and Muscle Trauma Group Specialised Register (September 2012), the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2012, Issue 8), MEDLINE (1946 to August week 5 2012), EMBASE (1980 to Week 36 2012) and the Chinese Biomedical Literature Database (CBM) (January 1978 to October 2012). No language restrictions were applied.

In MEDLINE (Ovid Online), a subject-specific search was combined with all three phases of the Cochrane optimal trial search strategy (Higgins 2005) and modified for use in The Cochrane Library, EMBASE and the Chinese Biomedical Literature database. These search strategies can be found in Appendix 1.

We searched the WHO International Clinical Trials Registry Platform for ongoing and unpublished trials (October 2012).

Searching other resources

We searched references lists of articles and contacted authors of relevant trials.

Data collection and analysis

Selection of studies

One review author (MA) scanned the results of the search, and retrieved the full articles for all potentially relevant trials. Two review authors (MA and TXW) independently assessed each of these trials for inclusion and scrutinised each trial report for multiple publications from the same data set. We resolved any disagreement through discussion.

Data extraction and management

Two review authors (MA and TXW) independently extracted data using a piloted data extraction form. We extracted data on study characteristics including methods, participants, interventions, and outcomes. We resolved any disagreements by referring to the trial report and through discussion. Where individual patient data were available, we calculated means and standard deviations for continuous outcomes.

Assessment of risk of bias in included studies

Three review authors (MA, XYK and TXW) independently assessed the risk of bias of each trial in terms of generation of allocation sequence, allocation concealment, blinding (of participants, treatment providers and outcome assessors), completeness of outcome data, selective reporting and other bias using the criteria of the Cochrane Collaboration's 'Risk of bias' tool (Higgins 2011). Based on these criteria, risk of bias for the individual domains was assessed as: high, low or unclear. All inter-rater differences were resolved by discussion.

Measures of treatment effect

Risk ratios and 95% confidence intervals were calculated for dichotomous outcomes, and mean differences and 95% confidence intervals were calculated for continuous outcomes.

Dealing with missing data

We tried to contact the original trial authors for missing data and information. We performed intention-to-treat analyses where possible. We did not impute missing standard deviations.

Assessment of heterogeneity

We assessed heterogeneity using the Cochrane Q statistic with significance set at a P value of less than 0.10. We used the I² statistic to estimate the percentage of heterogeneity between trials (Higgins 2011; Chapter 9):

  • 0% to 40%, might not be important;

  • 30% to 60%, may represent moderate heterogeneity;

  • 50% to 90%, may represent substantial heterogeneity;

  • 75% to 100%, considerable heterogeneity.

Assessment of reporting biases

When a sufficient number of trials (10 or more) become available, we plan to assess publication bias by using funnel plots.

Data synthesis

We did not pool data where there was statistically significant heterogeneity. Where it was appropriate to pool data, we initially used the fixed-effect model. However, if substantial and unexplained heterogeneity was detected, we used the random-effects model. We used 95% confidence intervals throughout. Should pooling of continuous outcome data based on different scales be undertaken in the future, we plan to calculate standardised mean differences.

Subgroup analysis and investigation of heterogeneity

When there are sufficient numbers of trials, we intend to explore the effects of different study characteristics to determine possible causes of heterogeneity. Subgroup analyses will be performed based on different age groups, different genders, and different types of surgery. We will investigate whether the results of subgroups are significantly different by inspecting the overlap of confidence intervals and performing the test for subgroup differences available in RevMan.

Sensitivity analysis

We planned to test the robustness of the evidence by sensitivity analysis of pooled data in three ways:

  • repeating the analysis excluding unpublished studies (if any);

  • repeating the analysis excluding studies at high or unknown risk of selection bias from inadequate methods of sequence generation and allocation concealment;

  • comparing the results of the fixed-effect model to the random-effects model (robust evidence should not be reversed by changing the model).

Results

Description of studies

Results of the search

Overall, for this update (search completed October 2012) and for the previous version (search completed June 2005) of the review, a total of 597 records were screened from the following databases: Cochrane Bone, Joint and Muscle Trauma Group Specialised Register (24 records); CENTRAL (39); MEDLINE (348); EMBASE (185); and the Chinese Biomedical Literature Database (1). We did not identify any potentially eligible studies from other sources.

The search update resulted in the identification of one new included study (Siebenga 2006) and one new excluded study (Wan 2005).

In total, there are two included trials (Siebenga 2006; Wood 2003), two excluded studies (Shen 2001; Wan 2005) and no studies awaiting classification. We did not identify any ongoing trials.

Included studies

See the Characteristics of included studies for details of the two included studies.

Design

Both studies (Siebenga 2006; Wood 2003) were randomised using a parallel design. Siebenga 2006 was a multicentre study and Wood 2003 was conducted at one centre with patients recruited from three associated trauma facilities.

Sample sizes

Siebenga 2006 recruited 34 participants and reported results for 32 participants. Wood 2003 recruited 53 participants, reporting results for 47 participants.

Setting

Siebenga 2006 was conducted in Gemany and the Netherlands, and Wood 2003 was conducted in Minneapolis, USA.

Participants

Both trials recruited patients with thoracolumbar fractures without neurological deficit. Where stated, the distribution of fracture sites was T12 (8 patients), L1 (18), L2 (2), L3 (3) and L4 (1) in Siebenga 2006; and T11 (1), T12 (8), L1 (28) and L2 (10) in Wood 2003. Patient characteristics were provided only for those followed-up in both trials. The mean age of these participants was 41.8 years in Siebenga 2006 and 41.4 years in Wood 2003. There were more males than females in both trials: 20 males and 12 females in Siebenga 2006; and 32 males and 15 females in Wood 2003.

Interventions

Siebenga 2006 compared surgical treatment with short-segment posterior stabilisation followed by a Jewett hyperextension orthosis for three months versus non-surgical therapy involving horizontal bed rest for a minimum of five days and a Jewett hyperextension orthosis for three months.

In Wood 2003, the participants in the surgical group were treated with posterior or anterior arthrodesis and instrumentation, while the non-surgical intervention involved the application of a body cast or orthosis.

Outcomes

The mean length of follow-up in Siebenga 2006 was 4.3 years (range 2.0 to 6.6 years). In Wood 2003, the mean duration of follow-up was 44 months (range 24 to 118 months).

Siebenga 2006 reported on pain, disability, return to work, complications and subsequent surgery, kyphosis, and length of hospital stay. Individual patient data were available for visual analogue scale (VAS) pain, the VAS spine score (Knop 2001), the Roland and Morris disability questionnaire, and kyphosis.

Wood 2003 reported on pain, disability, quality of life, return to work, complications and subsequent surgery, kyphosis, canal compromise, length of hospital stay, and costs. Individual patient data were available for VAS pain, the Roland and Morris disability questionnaire, the Oswestry back-pain questionnaire, kyphosis, and degree of canal compromise.

Excluded studies

Two studies were excluded for reasons detailed in the Characteristics of excluded studies.

Risk of bias in included studies

See Figure 1 and Figure 2 for summaries of the results of the risk of bias assessment.

Figure 1.

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Figure 2.

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

Allocation

Siebenga 2006 gave no details of their randomisation procedure and was judged at unclear risk of bias. Wood 2003 described a blind, computer-generated randomisation procedure, but gave insufficient details of allocation concealment and thus was judged at unclear risk of selection bias.

Blinding

Performance bias relating to lack of blinding was inevitable in these two trials. Neither study used blinding and thus potentially each was at high risk of detection bias.

Incomplete outcome data

Two patients were lost to follow-up in Siebenga 2006. In Wood 2003, two participants from each group were lost to follow-up and two additional patients of the non-surgical group died from other causes unrelated to their injuries before the final follow-up could be performed. Both trials were judged at low risk of attrition bias.

Selective reporting

We were unable to check the registered protocol or ethics committee approval to compare the difference according to Chan's measure (Chan 2004), therefore the potential of reporting bias was unclear.

Other potential sources of bias

This was judged as unclear.

Effects of interventions

Although the two included trials had many outcomes in common, their results were significantly heterogeneous for several of these and only limited pooling was performed. As none of the primary outcomes were reported as stipulated in the Types of outcome measures, we presented alternatives from the same categories below. Final follow-up results (minimum two years in both trials) are presented below unless stated otherwise.

Primary outcomes

Pain

Both trials used a VAS to assess the degree of pain. However, the direction of pain intensity was in opposite directions: in Siebenga 2006, pain was measured using a VAS (0 mm = worst pain imaginable; 100 mm = no pain at all); and in Wood 2003, pain was reported using a 10 cm VAS (0 = no pain; 10 cm = worst pain imaginable). The data from both trials were adjusted for presentation in the analyses so that they had the same direction of effect and scale. We calculated means and standard deviations based on the raw data given in the original articles (Table 1; Table 2).

Table 1. Siebenga 2006: individual patient data for patient-reported outcomes of pain and disability
  1. VAS = visual analogue scale

VAS pain (0 to 100 mm: no pain)

at final follow-up

VAS spine score (0 to 100 mm: best outcome)

at final follow-up

RMDQ-24 at final follow-up
SurgicalNon-surgicalSurgicalNon-surgicalSurgicalNon-surgical
55804580144
1006010029015
95659955010
761558111324
9065945429
10098869400
1009010082013
100656553214
951006210000
79100748610
100658634015
70658956610
8280768050
99559246015
8680906225
80 74 3 
77 88 4 
Table 2. Wood 2003: individual patient data for patient-reported outcomes of pain and disability
  1. ** = missing value
    VAS = visual analogue scale

VAS pain (0 to 10 cm: worst)
at final follow-up
Roland and Morris Questionnaire
at final follow-up
Oswestry Questionnaire
at final follow-up
SurgicalNon-surgicalSurgicalNon-surgicalSurgicalNon-surgical
125**4**
15312450
1.5061140
52160202
111264
551032412
32130302
001020
5140200
21.5422412
2060200
7.50191400
601423010
110002
197241652
402080
4191346
386201820
741994022
611524822
1061444
6019**48**
115104
5 6 4 

The two trials presented opposite and significantly heterogeneous results (Chi² = 8.90, df = 1 (P = 0.003); I² = 89%) for pain and thus we did not pool their data (see Analysis 1.1). In Siebenga 2006, the surgically-treated group had significantly less pain at final follow-up (MD -15.09 mm, 95% CI -27.81 to -2.37). Wood 2003 found the converse, with the surgical group having more pain at final follow-up, although the effect did not reach statistical significance (MD 13.60 mm, 95% CI -0.31 to 27.51).

Function measured on a disability or quality of life scale

Both trials used the Roland and Morris disability questionnaire (RMDQ-24) to estimate the functional disability at the time of pre-injury and at the last follow-up. We calculated the mean and standard deviation from the raw data given in the original articles (Table 1; Table 2). The two trials presented opposite and significantly heterogeneous results (Chi² = 12.41, df = 1 (P = 0.0004); I² = 92%) and thus we did not pool their data (see Analysis 1.2).

In Siebenga 2006, the surgical group had a statistically significantly lower mean RMDQ-24 score and thus less disability than the non-surgical group at follow-up (MD -5.87, 95% CI -10.10 to -1.64). In contrast, the surgical group in Wood 2003 had a statistically significantly higher mean RMDQ-24 at follow-up (MD 4.31, 95% CI 0.54 to 8.08).

Siebenga 2006 also reported more favourable results in the surgical group for the VAS spine score at the final follow-up (MD 20.10 mm, 95% CI 5.13 to 35.07) (see Analysis 1.3).

Wood 2003 found statistically significantly higher (thus worse function) Oswestry questionnaire scores (0 to 100: worst function) at the final follow-up evaluation in the surgical group (mean score 20.8 versus 10.7; MD 10.08; 95% CI 1.02 to 19.14) (see Analysis 1.4). Wood 2003 also reported significant differences in favour of the non-surgical group for two out of eight Short Form (SF)-36 domains at follow-up: physical function (mean score 63 versus 86; reported P = 0.002) and role, physical (mean score 51 versus 85; reported P = 0.003). Table 3 presents the results for all eight domains.

Table 3. Wood 2003: SF-36 scores for 8 domains at final follow-up
DomainSurgeryNon-surgeryP Value
Pain59720.07
Health Perception69720.4
Physical Function63860.002
Social Function84830.8
Role; Physical51850.003
Role; Emotional80780.92
Mental Health81750.3
Energy/Fatigue84560.8
Return to work

The two trials presented contrasting and heterogeneous results (Chi² = 5.51, df = 1 (P = 0.02); I² = 82%) for return to work and thus we did not pool these data (see Analysis 1.5). In Siebenga 2006, 11 of 13 working patients in the surgical group returned to their previous job whereas only five of 13 participants with an employment record in the non-surgical group resumed working. Of these five, three returned to the same profession and two changed careers for less physically demanding employment. On average, those in the surgical group returned to work sooner after their accident than those in the non-surgical group (6.7 months (range 1 to 18 months) versus 13.8 months (range 6 to 33 months)). In Wood 2003, 10 of the 24 patients treated surgically returned to work within six months and a further four returned to work subsequently; of these, eight returned to a similar job and six changed to a less physically demanding occupation. In the non-surgical group, 17 of 23 patients resumed work within six months and a further two returned subsequently; of these, 15 returned to a similar job.

Rate of subsequent surgery and complications

In Wood 2003, two patients in the non-surgical group who were excluded from the analysis died from causes unrelated to their injuries before the final follow-up could be performed.

Only surgically-treated patients in both trials had subsequent surgery (implant removal) for complications (8/41 versus 0/38; RR 8.39, 95% CI 1.12 to 62.87) (see Analysis 1.6). Implant removal for the remaining surgical group patients in Siebenga 2006 was scheduled after a period of nine to 12 months. In Wood 2003, the complications rate in the surgical group was higher than in the non-surgical group (16/24 versus 3/23; RR 5.11, 95% CI 1.71 to 15.24). Siebenga 2006 found no statistically significant difference between two groups (5/17 versus 3/15; RR 1.47, 95% CI 0.42 to 5.14). Pooled data using the random-effects model because of heterogeneity (Chi² = 2.20, df = 1 (P = 0.14); I² = 55%) showed a non-statistically significant trend favouring non-surgical treatment for overall complications (21/41 versus 6/38; RR 2.85, 95% CI 0.83 to 9.75) (see Analysis 1.7). The specific complications reported by the two trials are described in Table 4.

Table 4. Complications
  1. Note there are repeated counts in this table.

Complications Wood 2003 Siebenga 2006
SurgicalNon-surgicalSurgicalNon-surgical
Wound dehiscence2   
Instrumentation/bone failure2 1 
Wound infection1 deep 2 (1 deep) 
Pseudarthrosis1   
Neurapraxia1   
Ketoacidosis1   
Instrumentation break2 1 
Urinary tract infection21  
Seroma1   
Instrumentation removal (for complication)6 2 
Skin blisters 1  
Severe pain at the bone harvest site  1 
Conus medullaris syndrome   1
Continued use of orthosis reflecting irrational fear
of neurologic deterioration. Severe depression.
   1
Scoliosis (and signs of nerve root compression)   1
Total19273

Secondary outcomes

Sagittal plane kyphosis

Because of the considerable statistical heterogeneity between the two trials (Chi² = 8.05, df = 1 (P = 0.005); I² = 88%) we did not pool the data for local kyphosis (see Analysis 1.8). Siebenga 2006 reported local and regional sagittal angles; we have only reported on the former here. In the surgical group of this trial, the mean local sagittal angle decreased in the surgical group from 16.8 ° at admission to 8.6 ° at 4.3 years follow-up. Conversely, the mean local sagittal angle increased in the non-surgical group from 15.7 ° at admission to 19.8 ° at follow-up. This trial found significantly less kyophotic deformity after surgery at long-term follow-up (MD -11.20 °, 95% CI -15.50 ° to -6.90 °). In the group managed surgically in Wood 2003, the mean kyphosis (according to the picture shown in the article the angles may have been the local sagittal angles) was 10.1 ° on admission and 13 ° at the time of the final follow-up (minimum of two years). In the non-surgical group, the mean kyphosis was 11.3 ° on admission and 13.8 ° at final follow-up. There was no statistical difference at the time of final follow-up between the two groups in Wood 2003 (MD -0.81 °, 95% CI -6.56 ° to 4.94 °).

Spinal canal compromise

In both groups of Wood 2003, the degree of spinal canal compromise, estimated using computer tomography, was significantly reduced at two-year follow-up compared with that at presentation. There was no statistically significant difference between the two treatment groups at two-year follow-up in canal occlusion (% of normal anteroposterior canal diameter estimated from the average of the adjacent uninjured vertebrae) (MD 2.62%, 95% CI -6.57% to 11.81%) (see Analysis 1.9). Siebenga 2006 did not report this outcome.

Correlation between the final amount of kyphosis or canal compromise and the reported pain or disability

In both trials, there was no correlation between the final amount of kyphosis and the reported pain or disability (see Table 5).

Table 5. Correlation between the final amount of kyphosis or canal compromise and the reported pain or disability
  1. LSA = local sagittal angle
    P = P value
    r = correlation coefficient
    RMDQ-24 = Roland Morris Disability Questionnaire (score 0 to 24)
    RSA = regional sagittal angle
    VAS = visual analogue score or scale

Siebenga 2006RMDQ-24VAS spine scoreVAS pain
LSARSALSARSALSARSA

r = -0.30,

P = 0.09

r= -0.29,

P = 0.11

r = 0.20,

P = 0.29

r = 0.17,

P = 0.38

r = 0.20,

P = 0.29

r = 0.17,

P = 0.38

Wood 2003kyphosis and pain scoreRoland and Morris questionnaireOswestry questionnaire
Surgical groupr = 0.05; P = 0.8r = 0.05; P = 0.8r = 0.3; P = 0.14
Non-surgical groupr = 0.22; P = 0.29r = 0.19; P = 0.39r = 0.25; P = 0.27
Mean duration of hospitalisation

In Wood 2003, the mean duration of hospitalisation was 10.7 days (range 6 to 27 days) in the surgical group and 7.9 days (range 2 to 17 days) in the non-surgical group. In Siebenga 2006, the mean hospital stay was 14.6 days (9 to 21 days) in the surgical group and 12.2 days (6 to 25 days) in the non-surgical group. While length of hospital stay was longer in the surgical group in both trials, there were insufficient data for testing the significance.

Economic evaluation

Wood 2003 reported that for 32 patients (15 surgical; 17 non-surgical) who had an isolated thoracolumbar burst fracture without other substantial trauma requiring specific treatment during the initial hospitalisation, the mean charge for surgically-treated patients was 'approximately' USD 49,063 (range USD 26,517 to USD 102,583). This was reported as significantly higher (P < 0.01) than the mean charge of USD 11,264 (USD 4686 to USD 20,891), covering hospitalisation and cast or brace treatment, for non-surgically treated patients. No economic data were reported by Siebenga 2006.

Discussion

Summary of main results

The available evidence for comparing surgical with non-surgical treatment for thoracolumbar fractures without neurological deficit was limited to two small trials with a total of 79 participants available at a minimum of two years follow-up. Both trials were at risk of bias. Additionally, these trials produced contrasting results and the significant heterogeneity between the results for outcomes in common rendered meta-analysis inappropriate for all but overall complications and surgery for complications. Siebenga 2006 found superior pain, function and return to work outcomes for the surgical group, whereas the converse was true for Wood 2003. There were greater numbers of participants with early complications in the surgical groups and only those in this group had subsequent surgery, involving implant removal either for complications or as a matter of course. One trial reported that surgery was over four times more costly than non-surgical treatment.

Overall completeness and applicability of evidence

The available evidence is sparse and the two included trials are clinically heterogeneous. For instance, there was a difference in the range of fracture sites, although the majority fell into the same categories, and in the types of surgical and non-surgical interventions. It is impossible to judge the applicability of the contrasting results for pain and function from the two trials. However, the risk of subsequent surgery, entailing implant removal, and of early complications and higher initial costs after surgery are likely to apply more generally.

Quality of the evidence

There were some limitations in the design, implementation and reporting of the two included trials. Notably, both were at high risk of bias relating to performance and detection. The small sample sizes, clinical heterogeneity and contrasting results for patient-reported outcomes of pain and function of the two trials further limit the quality of the evidence.

Potential biases in the review process

There were some limitations in the review process that may have resulted in potential biases. Firstly, we restricted study inclusion to those published in English or Chinese. Secondly, we have been unsuccessful in our attempts to get missing information from the trial investigators. Thirdly, similar to the first version of this review, our presented primary outcomes differ from those listed in Types of outcome measures. This is unavoidable in terms of the available data but it also indicates that some appraisal of the outcomes may be warranted before the next update.

Agreements and disagreements with other studies or reviews

A meta-analysis by Gnanenthiran 2012 included four trials, of which two were included in our review. One (Shen 2001) was excluded because of excessively compromised analysis, and the fourth study was not a randomised or quasi-randomised trial. Although acknowledging the clinical and statistical heterogeneity in the four studies, Gnanenthiran 2012 decided to present the results for pain and functional outcomes. They concluded that there were no differences between the two intervention groups in pain, Roland Morris Disability Questionnaire scores, and return to work rates. Similar to our findings, they found surgery was associated with higher complication rates and costs.

Authors' conclusions

Implications for practice

The available evidence provided by two small and potentially biased randomised controlled trials is insufficient to conclude whether surgical or non-surgical treatment yields superior pain and functional outcomes for people with thoracolumbar burst fractures without neurological deficit. There is some evidence, however, that surgery is associated with more early complications and the need for subsequent surgery as well as greater initial healthcare costs.

Implications for research

In order to inform practice in this area, there is a need for a large, multicentre, high quality and adequately reported randomised controlled trial that assesses the relative effects of surgical versus non-surgical treatment for thoracolumbar burst fractures without neurological deficit. In determining the inclusion criteria for such a trial, consideration should be given to the use of a contemporary, internationally recognised classification system for thoracolumbar fractures, such as the Thoracolumbar Injury Classification and Severity Score (TLICS) (Vaccaro 2005). The focus should be on collecting patient-reported outcome measures of pain and function, and severe complications leading to disability.

Acknowledgements

We are very grateful for valuable comments and support from Bill Gillespie, Helen Handoll and Cumhur Kilincer on drafts of this updated review.

We thank Janet Wale, Peter Herbison, Rajan Madhok and Jeremy Fairbank for their comments on the first version of the review. We also thank Marc Swiontkowski for helpful comments on the protocol. We are also grateful to Lindsey Elstub, Joanne Elliott and Lesley Gillespie of the Cochrane Bone, Joint and Muscle Trauma Group.

Thanks too to Yuan Hao for preparing the first version of this updated review. In addition, the review authors who prepared the protocol and first version of the review are gratefully acknowledged: Liao Yi, Bai Jingping, Jin Cele and XiLin Baoleri.

Data and analyses

Download statistical data

Comparison 1. Surgical versus non-surgical treatment
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Visual Analogue Pain Scores (0 to 100 mm: worst) at follow-up2 Mean Difference (IV, Fixed, 95% CI)Totals not selected
2 Roland and Morris Questionnaire at follow-up (0 to 24: worst result)2 Mean Difference (IV, Fixed, 95% CI)Totals not selected
3 Visual Analogue Scale Spine Score (0 to 100 mm: best) at follow-up1 Mean Difference (IV, Fixed, 95% CI)Totals not selected
4 Oswestry Questionnaire at minimum 24 months follow-up1 Mean Difference (IV, Fixed, 95% CI)Totals not selected
5 Return to work2 Risk Ratio (M-H, Fixed, 95% CI)Totals not selected
6 Subsequent surgery (for complications)279Risk Ratio (M-H, Fixed, 95% CI)8.39 [1.12, 62.87]
7 Complications279Risk Ratio (M-H, Random, 95% CI)2.85 [0.83, 9.75]
8 Kyphosis at follow-up (degrees)2 Mean Difference (IV, Fixed, 95% CI)Totals not selected
9 Degree of canal compromise (% occlusion) at 2 years1 Mean Difference (IV, Fixed, 95% CI)Totals not selected
Analysis 1.1.

Comparison 1 Surgical versus non-surgical treatment, Outcome 1 Visual Analogue Pain Scores (0 to 100 mm: worst) at follow-up.

Analysis 1.2.

Comparison 1 Surgical versus non-surgical treatment, Outcome 2 Roland and Morris Questionnaire at follow-up (0 to 24: worst result).

Analysis 1.3.

Comparison 1 Surgical versus non-surgical treatment, Outcome 3 Visual Analogue Scale Spine Score (0 to 100 mm: best) at follow-up.

Analysis 1.4.

Comparison 1 Surgical versus non-surgical treatment, Outcome 4 Oswestry Questionnaire at minimum 24 months follow-up.

Analysis 1.5.

Comparison 1 Surgical versus non-surgical treatment, Outcome 5 Return to work.

Analysis 1.6.

Comparison 1 Surgical versus non-surgical treatment, Outcome 6 Subsequent surgery (for complications).

Analysis 1.7.

Comparison 1 Surgical versus non-surgical treatment, Outcome 7 Complications.

Analysis 1.8.

Comparison 1 Surgical versus non-surgical treatment, Outcome 8 Kyphosis at follow-up (degrees).

Analysis 1.9.

Comparison 1 Surgical versus non-surgical treatment, Outcome 9 Degree of canal compromise (% occlusion) at 2 years.

Appendices

Appendix 1. Search strategies

The Cochrane Library (Wiley Online Library)

#1   MeSH descriptor Spinal Injuries explode all trees (597)
#2 MeSH descriptor Lumbar Vertebrae, this term only (1738)
#3   MeSH descriptor Thoracic Vertebrae, this term only  (255)
#4   (#2 OR #3)  (1870)
#5   (#1 AND #4) (137)
#6   (burst or compression) :ti,ab,kw  (3990)
#7   (#5 AND #6) (36)
#8   (thoracolumbar or (thoraco NEXT lumbar)) :ti,ab,kw (131)
#9   ((burst NEAR fract*) or (burst NEAR injur*) or (compression NEAR fract*)) :ti,ab,kw (267)
#10 (#8 AND #9) (37)
#11 (#7 OR #10) (39) 

MEDLINE (Ovid Online)

1    exp Spinal Injuries/ (16163)
2    Thoracic Vertebrae/ (14419)
3    Lumbar Vertebrae/ (35605)
4    or/2-3 (43539)
5    and/1,4 (4248)
6    burst.tw. (30639)
7    and/5-6 (552)
8    (thoracolumbar or thoraco lumbar).tw. (5133)
9    ((burst adj5 (fract$ or injur$)) or (compression adj5 (fract$ or injur$))).tw. (5670)
10  and/8-9 (632)
11  or/7,10 (808)
12  randomized controlled trial.pt. (336136)
13  controlled clinical trial.pt. (85111)
14  Randomized Controlled Trials/ (82897)
15  Random Allocation/ (75700)
16  Double Blind Method/ (116905)
17  Single Blind Method/ (16674)
18  or/12-17 (563705)
19  Animals/ not Humans/ (3686416)
20  18 not 19 (521244)
21  clinical trial.pt. (473453)
22  exp Clinical Trials as topic/ (260609)
23  (clinic$ adj25 trial$).tw. (214213)
24  ((singl$ or doubl$ or trebl$ or tripl$) adj25 (blind$ or mask$)).tw. (118294)
25  Placebos/ (31302)
26  placebo$.tw. (138905)
27  random$.tw. (574532)
28  Research Design/ (67613)
29  or/21-28 (1186313)
30  29 not 19 (1095797)
31  30 not 20 (618404)
32  Comparative Study.pt. (1602137)
33  Evaluation Studies.pt. (171413)
34  Follow Up Studies/ (454503)
35  Prospective Studies/ (326935)
36  (control$ or prospectiv$ or volunteer$).tw. (2560395)
37  or/32-36 (4253242)
38  37 not 19 (3297829)
39  38 not (20 or 31) (2689468)
40  20 or 31 or 39 (3829116)
41 11 and 40 (348)

EMBASE (Ovid Online)

1    Spine Fracture/ or Spine Injury/ (15502)
2    Vertebra Fracture/ (8813)
3    or/1-2 (23778)
4    Thoracolumbar Spine/ (3154)
5    and/3-4 (1235)
6    burst.tw. (35332)
7    and/5-6 (297)
8    (thoracolumbar or thoraco lumbar).tw. (6440)
9    (burst adj5 (fract$ or injur$)).tw. (1260)
10  and/8-9 (571)
11  or/7,10 (589)
12  exp Randomized Controlled Trial/ (328650)
13  exp Double Blind Procedure/ (110736)
14  exp Single Blind Procedure/ (16360)
15  exp Crossover Procedure/ (34922)
16  Controlled Study/ (3858969)
17  or/12-16 (3937818)
18  ((clinical or controlled or comparative or placebo or prospective$ or randomi#ed) adj3 (trial or study)).tw. (643947)
19  (random$ adj7 (allocat$ or allot$ or assign$ or basis$ or divid$ or order$)).tw. (156994)
20  ((singl$ or doubl$ or trebl$ or tripl$) adj7 (blind$ or mask$)).tw. (146813)
21  (cross?over$ or (cross adj1 over$)).tw. (62664)
22  ((allocat$ or allot$ or assign$ or divid$) adj3 (condition$ or experiment$ or intervention$ or treatment$ or therap$ or control$ or group$)).tw. (196768)
23  or/18-22 (961563)
24  or/17,23 (4412295)
25  limit 24 to human (2676479)
26  and/11,25 (185)

Chinese Biomedical Literature database

1. exp Spinal Injuries/
2. Thoracic Vertebrae/
3. Lumbar Vertebrae/
4. or/2-3
5. and/1,4
6. burst.tw.
7. and/5-6
8. (thoracolumbar or thoraco lumbar).tw.
9. ((burst and (fract? or injur?)) or (compression and (fract? or injur?))).tw.
10. and/8-9
11. or/7,10
12. randomised controlled trial.pt.
13. controlled clinical trial.pt.
14. Randomized Controlled Trials/
15. Random Allocation/
16. Double-Blind Method/
17. Single-Blind Method/
18. or/12-17
19. Animal/ and not Human/
20. 18 and not 19
21. clinical trial.pt.
22. exp Clinical Trials/
23. (clinic? and trial?).tw.
24. ((singl? or doubl? or trebl? or tripl?) and (mask? or blind?)).tw.
25. Placebos/
26. placebo?.tw.
27. random?.tw.
28. Research Design/
29. (latin and square).tw.
30. or/21-29
31. 30 and not 19
32. 31 and not 20
33. Comparative Study/
34. exp Evaluation Studies/
35. Follow-Up Studies/
36. Prospective Studies/
37. (control? or prospectiv? or volunteer?).tw.
38. Cross-Over Studies/
39. or/33-38
40. 39 and not 19
41. 40 and not (20 or 32)
42. or/20,32,41
43. and/11,20
44. and/11,32
45. and/11,41
46. or/43-45 (1)

What's new

DateEventDescription
8 May 2013New search has been performedFor this version of the review, published in June 2013, the following changes were made:
1. the search was updated to October 2012;
2. one trial (Siebenga 2006), with 34 participants, was newly included;
3. the assessment of risk of bias now replaces the previous methodological quality assessment.
8 May 2013New citation required but conclusions have not changedThe authorship was changed. In addition, the title has been changed to 'Surgical versus non-surgical treatment for thoracolumbar burst fractures without neurological deficit' (from 'Operative versus non-operative treatment for thoracolumbar burst fractures without neurological deficit').

History

Protocol first published: Issue 1, 2005
Review first published: Issue 4, 2006

DateEventDescription
11 September 2008AmendedConverted to new review format.

Contributions of authors

Minawaer Abudu: protocol development, searching for trials, quality assessment of trials, data extraction, data analysis, review development.
Xingyu Kong: searching for trials, quality assessment of trials, data extraction, data analysis.
Wu Taixiang: quality assessment of trials, statistical advice, review development.
Chen Xueyi: supervisor for development of the review.

Declarations of interest

None known.

Differences between protocol and review

  1. Reappraisal of Types of outcome measures led to the downgrading of economic evaluation to a secondary outcome and removal of the secondary outcome 'correlation between the final amount of kyphosis/canal compromise and the pain reported or disability' - this was still reported but not as a measure of treatment effect.

  2. We assessed risk of bias instead of study methodological quality.

  3. In the protocol, we planned to synthesise data but as only two trials with substantial heterogeneity were included, only limited meta-analysis was possible.

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Siebenga 2006

Methods

Multicentre (three university hospitals) prospective randomised, parallel, controlled trial

The randomisation procedure was not described

Length of follow-up: mean 4.3 years (range 2.0 to 6.6 years)

Participants

Participants: 34 participants recruited between 1998 and 2002. 18 patients received surgical treatment and 16 patients received non-surgical therapy. There were 7 women and 10 men in the surgical group, the mean age was 45.7 years (27 to 59); 5 women and 10 men in the non-surgical group, the mean age was 37.3 years (18 to 58).

It is very likely that one of two hospitals was in the Netherlands and the other in Germany.

Inclusion criteria: Traumatic fracture of T10 to L4, AO Type A (compression fracture), no neurologic deficit (ASIA/Frankel E), age 18 to 60 years, and period between trauma and surgical treatment less than 10 days.

Exclusion criteria: Patients with AO Type A1.1 fracture, pregnancy, pathological or osteoporotic fracture, patients with end-stage disease (ASA IV), patients with a history of previous back surgery, patients with a recent psychiatric history, patients using drugs or other illegal substances, or patients presenting with any accompanying injury that might interfere with the treatment of the spine fracture or the mobilisation scheme after hospital discharge (i.e. lower extremity injuries prohibiting early weight-bearing motion).

Loss to follow-up: Two patients were lost to follow-up. One left the hospital against medical advice and could not be contacted again. The other one was treated surgically but refused to co-operate further in this study. The remaining 32 patients (94%) completed at least 2-year functional and radiologic follow-up.

Interventions

Surgical treatment (intervention): The patients were managed with a short-segment posterior stabilisation using the titanium version of the Universal Spine System. Three days post-surgery, the patients were mobilised wearing a Jewett hyperextension orthosis for the next 3 months. A standardised physiotherapy scheme was followed. The same instructions with regard to activity and mobilisation were given as in the non-surgical treated group.

Non-surgical treatment (control): The patients were treated with a rehabilitation course on a Circ-O-Lectric or Rotorest bed. After resting on a horizontal bed for a minimum period of 5 days, the patients began a standardised physiotherapy scheme to train trunk musculature. Nadroparine was given daily as an anticoagulant until the patient was discharged to the rehabilitation clinic. A Jewett hyperextension orthosis was fitted and the patients were instructed to wear the brace at all times for 3 months. Patients were advised not to engage in heavy work and sports for 3 months.

Outcomes

Visual analogue scale (VAS) for pain (0 to 100 mm: worst outcome)

VAS Spine Score (0 to 100: worst outcome) (Knop 2001). This validated patient-rated outcome score consists of 19 questions on patient's perception of pain and back-related activities, each of which are rated using a 100 mm VAS. The overall score is the mean of all items, and can be any value between 0 (severe disability) and 100 (no disability)

RMDQ-24 (Roland and Morris disability questionnaire - 24 questions)

Percentage of patients who returned to work

Time interval before return to work

Complications and subsequent surgery

Radiographical results include both local and regional kyphotic deformity

Length of hospitalisation

NotesFigure 1 in the article shows how the authors defined local and regional sagittal angles. The authors measured the outer rather than the more usual inner angle for local deformity, which resulted in negative angles. We have converted these to positive values for presenting in this review.
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo description about the randomisation method used to generate the allocation sequence. Unable to communicate with the original authors to identify the methods of randomisation
Allocation concealment (selection bias)Unclear riskNot mentioned
Blinding (performance bias and detection bias)
All outcomes
High riskNot mentioned
Incomplete outcome data (attrition bias)
All outcomes
Low riskLoss to follow-up was described in detail, and the rate was 6% (2/34)
Selective reporting (reporting bias)Unclear riskUnclear
Other biasUnclear riskUnclear

Wood 2003

Methods

Randomised controlled trial, "a blind, computer-generated, randomisation process"

Length of follow-up: mean 44 months (range 24 to 118 months)

Participants

Participants: 53 (26 treated surgically and 27 managed non-surgically) patients (between 1992 to 1997). Characteristics provided for the 47 participants who were followed-up: 16 men and 8 women in the surgical group, the mean age was 43.3 years (19 to 68); 16 men and 7 women in non-surgical group, the mean age was 39.4 years (20 to 66). Thoracolumbar (T10 to L2) burst fracture without neurological deficit.
Minneapolis, Minnesota.

Inclusion criteria: All of the following, an isolated burst fracture within the thoracolumbar region demonstrated on anteroposterior and lateral radiographs; a computed tomography scan revealing a burst-type fracture with retropulsion of vertebral body bone posteriorly into the spinal canal; no new neurological abnormality of the lower extremities or abnormality of bowel or bladder function; presentation less than 3 weeks after the time of the injury; aged between 18 and 66 years; no medical illnesses that would preclude an surgical intervention; no ongoing cancer, infection, bleeding disorder, or skin disease; an understanding of the English language. Concomitant stable compression fractures elsewhere in the spine were permitted if they did not require specific treatment.

Exclusion criteria: a closed-head injury (a score of <14 points on the Glasgow coma scale on admission); an open vertebral fracture; a neurological deficit related to the fracture; a loss of structural integrity within the posterior osteo-ligamentous complex (such as facet fracture-dislocation or flexion-distraction ligament disruption); a senile, osteopenic, or insufficiency fracture. A laminar fracture was neither an exclusion criterion nor a contraindication for non-surgical treatment. No absolute degree of kyphosis, canal encroachment by bone, or anterior loss of height was a criterion for exclusion.

Loss to follow-up: 4 (2 from each group) could not be contacted, 2 from non-surgical group died from unrelated causes before the final follow-up. 47 patients were followed for a minimum of two years (mean 44 months).

Interventions

Surgical treatment (intervention): managed with either a short-segment posterolateral spinal arthrodesis with pedicle screw-hook instrumentation and autologous iliac crest bone-grafting or an anterior fibular and rib-strut construct arthrodesis with local autogenous bone-grafting and instrumentation.

Non-surgical treatment (control): application of a body cast or orthosis.

Outcomes

10 cm (0 = no pain) VAS for pain

"Modified" Roland and the modified Morris disability questionnaire

Oswestry back-pain questionnaire

SF-36 health survey

Rate of return to work

Complications

Radiographical results include sagittal plane kyphosis

Degree of canal compromise (data from a computed tomography scan - calculated by dividing the available anteroposterior diameter of the canal space at the injured level by the average of the diameter of the two uninjured vertebrae below and above the injured level)

Length of hospitalisation

Cost of treatment

Notes 
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskComputer-generated randomisation process
Allocation concealment (selection bias)Unclear riskRandomisation described as "blind". No other details
Blinding (performance bias and detection bias)
All outcomes
High riskNo description
Incomplete outcome data (attrition bias)
All outcomes
Low riskLoss to follow-up was described in detail, and the rate was 11% (6/53)
Selective reporting (reporting bias)Unclear riskUnclear
Other biasUnclear riskUnclear

Characteristics of excluded studies [ordered by study ID]

StudyReason for exclusion
Shen 2001Seven participants were assigned to the surgical group refused surgery and were reassigned to the non-surgical group. Three participants in the non-surgical group did not complete the two-year study period and were excluded. This study was excluded because the randomisation of patients had been compromised by inappropriate analysis; intention-to-treat analysis was not possible
Wan 2005The randomisation was applied only in a proportion of participants

Ancillary