Article Text

Download PDFPDF

Systematic review of prognostic factors for work participation in patients with sciatica
  1. Teddy Oosterhuis1,
  2. Veerle R Smaardijk1,
  3. P Paul FM Kuijer1,
  4. Miranda W Langendam2,
  5. Monique H W Frings-Dresen1,
  6. Jan L Hoving1
  1. 1 Coronel Institute of Occupational Health, Amsterdam UMC, University of Amsterdam, Amsterdam Public Health Research Institute, Amsterdam, The Netherlands
  2. 2 Department of Clinical Epidemiology, Biostatistics and Bioinformatics, Amsterdam UMC, University of Amsterdam, Amsterdam Public Health Research Institute, Amsterdam, The Netherlands
  1. Correspondence to Teddy Oosterhuis, Academic Medical Center, University of Amsterdam, Department Coronel Institute of Occupational Health, Amsterdam Public Health Research Institute, Amsterdam 1100 DD, The Netherlands; t.oosterhuis{at}


Sciatica impacts on the ability to work and may lead to a reduced return to work. This study reviewed and summarised prognostic factors of work participation in patients who received conservative or surgical treatment for clinically diagnosed sciatica. We searched MEDLINE, CINAHL, EMBASE and PsycINFO until January 2018. Cohort studies, using a measure of work participation as outcome, were included. Two independent reviewers performed study inclusion and used the Quality In Prognosis Studies tool for risk of bias assessment and GRADE to rate the quality of the evidence. Based on seven studies describing six cohorts (n=1408 patients) that assessed 21 potential prognostic factors, favourable factors for return to work (follow-up ranging from 3 months to 10 years) included younger age, better general health, less low back pain or sciatica bothersomeness, better physical function, negative straight leg raise-test, physician expecting surgery to be beneficial, better pain coping, less depression and mental stress, less fear of movement and low physical work load. Study results could not be pooled. Using GRADE, the quality of the evidence ranged from moderate to very low, with downgrading mainly for a high risk of bias and imprecision. Several prognostic factors like pain, disability and psychological factors were identified and reviewed, and these could be targeted using additional interventions to optimise return to work. PROSPERO registration number: CRD42016042497.

  • epidemiology
  • musculoskeletal

This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See:

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Key messages

What is already known about this subject?

  • Sciatica impacts on the ability to work and may lead to a reduced return to work.

  • Prognostication is important for physicians but guidance in the prognostication process regarding return to work in patients with sciatica is lacking.

What are the new findings?

  • Physicians can assess whether a worker is more likely to return to work by assessing prognostic factors. Favourable factors for return to work include younger age, better general health, less low back pain or sciatica bothersomeness, better physical function, negative SLR-test, physician expecting surgery to be beneficial, better pain coping, less depression and mental stress, less fear of movement and low physical work load.

How might this impact on policy or clinical practice in the foreseeable future?

  • Prognostic factors like pain, disability and psychological factors can be used in the prognostication process. More importantly, these prognostic factors can be targeted by referring for additional interventions in order to promote return to work.


Lumbosacral radicular syndrome, often called sciatica, is commonly caused by a herniated lumbar disc.1 The syndrome is characterised by lower limb pain radiating below the knee in an area of the leg served by one or more lumbosacral nerve roots. There may be other neurological findings such as sensory and motor deficits. Sciatica is usually self-limiting with pain and disability decreasing over time,2 but not all patients fully recover.2–4 Surgical treatment is usually offered in more severe cases when severe radiating leg pain persists after a period of conservative management.5 In a large study (n=782), 34% of conservatively treated patients experienced very or extremely bothersome symptoms at 6 months follow-up.3 Similarly, a systematic review (n=13 883) showed that surgically treated patients reported, despite decreased pain and disability scores 3 months after surgery, on average mild to moderate pain and disability 5 years after surgery.4

The direct and indirect costs of patients suffering from sciatica are high,6 and an important cost driver is work absenteeism.7–9 In the acute phase, most people with sciatica will stop working and some will resume work in the short time. Return to work (RTW) rates vary from 66% after 2 years10 to between 67% and 85% after 10 years.11 The high socioeconomic impact of sciatica and its impact on the ability to work in patients raise the need to identify factors that predict reduced RTW. Prognostic evidence could assist clinicians to better define high risk groups and inform both clinicians and patients with regard to counselling and treatment choices to promote RTW. The objective of this study was to review and summarise prognostic factors of work participation in patients with sciatica.


This review is reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines.12

Eligibility criteria

We included full-text original articles of studies concerning adults (≥18 years) clinically diagnosed with sciatica, who received either conservative treatment or surgical treatment. Studies with participants having stenosis or cauda equina syndrome were excluded, if no separate data were available for participants without stenosis or cauda equina syndrome. We included cohort studies that evaluated any possible prognostic factor associated with RTW as a measure of work participation.


We searched relevant cohort studies using MEDLINE via PubMed, CINAHL via EBSCOhost, EMBASE and PsycINFO via OVID, from inception until January 2018. Specific search terms for the population, work participation and prognostic filters were used. The search strategy was developed with input from the review team and a clinical librarian, based on search strategies for sciatica using the search strategy for the 2016 NICE guideline on lumbosacral radicular syndrome,13 the published MEDLINE filter for prognostic studies14 and Yale University’s methodological research filter for prognosis and natural history.15 The clinical librarian developed the string for work participation. The search strategy was adapted for each database. Furthermore, references in relevant reviews and in identified cohort studies were screened. We did not apply any language restrictions. Online supplementary appendix 1 shows the search strategy used in MEDLINE.

Supplementary file 1

Study selection

Pairs of review authors (TO, VRS, PK, MHWFD and JLH) independently selected the studies to be included by applying the selection criteria. First, title and abstract screening was performed using Covidence ( Subsequently, full-text articles of potentially relevant studies were retrieved and assessed. Disagreements were resolved using consensus.

Data collection process

Using a standardised form, one reviewer (TO or VRS) extracted data from the included studies. A second reviewer (TO, VRS, PK or JLH) checked the results. Data that were extracted included first author, year, country; case definition; source population; characteristics of the study population; inclusion and exclusion criteria; sample size, including number of complete cases; prognostic factors and potential confounders including their measurement method; definition of RTW as the work participation outcome; description of the content of treatment (eg, surgery, rehabilitation and other conservative); length of follow-up; analysis used (univariable or multivariable regression); extracted or calculated ORs or HRs with 95% CIs, if sufficient data were available and source of funding. Study authors were contacted in case of insufficient information on any of these items. In case of multiple follow-up moments per study, the latest follow-up was used.

Risk of bias in individual studies

Pairs of review authors (TO, VRS, PK, MWL and JLH) independently assessed the risk of bias of the included studies by using the Quality In Prognosis Studies (QUIPS) tool.16 At the study level, six domains were rated as high, moderate or low risk of bias according to the QUIPS guidelines. Finally, an overall risk of bias was determined per study: low, moderate or high risk. Consensus was used to resolve disagreements. If no agreement was reached, a third reviewer was consulted. Study authors were contacted in case of insufficient information to assess the risk of bias.

Synthesis of results

Meta-analyses were planned with a random-effects meta-analysis model, but only if populations, prognostic factors, outcomes and time points were sufficiently homogeneous across studies.17 We considered RTW outcomes of 6 months or more as a long-term follow-up. We planned separate analyses for (1) studies reporting ORs and HRs, (2) studies assessing surgical and non-surgical populations and (3) different non-surgical approaches. If meta-analyses were not feasible, we performed a narrative synthesis.18

The Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach was used to assess the overall quality of the evidence.19 Evidence from explorative cohort studies started as moderate quality evidence. Evidence from confirmative cohort studies started as high-quality evidence.20 The quality of the evidence was downgraded according to the performance of the studies against five domains: risk of bias (<75% of participants from studies with a low risk of bias), inconsistency (point estimates of both OR >1.0 and OR <1.0 in meta-analyses; no overlap in CIs in meta-analyses), indirectness and imprecision (fewer than 10 participants per prognostic factor or category in case of categorical variables; non-significant results; CIs crossing OR=0.5 or 2.021; fewer than 100 cases reaching endpoint). Publication bias was assessed through the construction of funnel plots only if 10 studies or more were included in the meta-analyses.


The search yielded 2953 articles: MEDLINE 627, CINAHL 853, EMBASE 1396 and PsycINFO 77. After removal of duplicates, 2583 articles remained (figure 1). After screening titles and abstracts, 64 full-texts were read. Of these, six articles (that described five unique cohorts) fulfilled all eligibility criteria. Screening of reference lists of included studies identified one more eligible study.22 This resulted in seven studies included in total. Reasons for exclusion of 58 full-texts were (>1 reason per study possible): no work participation measure or not using RTW as a work participation measure (33), no prognostic study (23), no data or no separate data for patients with sciatica (14) and sciatica was the prognostic factor or outcome (3). Online supplementary appendix 2 provides an overview of all excluded studies.

Figure 1

Flow diagram. RTW, return to work.

Study characteristics

Table 1 shows the characteristics of the seven included studies describing six cohorts. All 1408 patients were diagnosed with a lumbar disc herniation. Three publications, describing two cohorts,23–25 included a mixed population consisting of both conservatively and surgically treated patients. In these cohorts, 30%25 and 53%23 of the patients received surgery, at 2 and 4 years follow-up, respectively. One of these studies controlled the analysis for initial type of treatment.24 The four remaining studies included surgical patients only22 26–28 including one that consisted of patients with reoperation for recurrent herniation.27 Three cohorts were from North-America and three were from Europe. The number of participants varied from 46 to 394 per cohort, mean age ranged from 35 to 46 years, 28%–78% were male and all analyses included working populations. Two cohorts measured short-term follow-up at 328 and 6 months.26 Long-term outcomes ranged from 2,22 25 3,27 423 to 10 years24 follow-up, with the majority measured between 2 and 4 years. All studies used self-reported RTW which was measured in various ways: being employed,23 24 28 return to usual number of work hours per week,26 return to full-time work,25 return to ‘any’ work,22 the ability to work at least 6 months.27 All studies used multiple regression analysis (adjusted ORs) and all but one reported ORs. This one study22 reported betas (ln), which we converted to ORs. Clinical heterogeneity, differences in RTW measures and the use of different sets and measurements of prognostic factors, confounders and follow-up time points precluded pooling of data or performance of any subgroup analyses given the limited number of studies.

Table 1

Study characteristics of the seven included studies

Risk of bias within studies

Table 2 shows the results of the risk of bias assessment, using QUIPS.16 Four studies had an overall low risk of bias,23 25–27 of which two studies scored a low risk of bias on all six domains.25 26 Three studies had an overall moderate risk of bias.22 24 28

Table 2

Risk of bias assessment of the seven included studies

Prognostic factors

In total, 21 potential prognostic factors for RTW were assessed. All factors were derived from multiple regression models. Some factors were assessed at different time points. Age and sex were included in three studies, general health and fear avoidance beliefs were included in two studies. Four pain measures were used in four studies: back pain intensity, back pain frequency, sciatica bothersomeness and opioid use. The results of all studies are summarised below. All studies measured RTW, but reported prognostic factors for either RTW or reduced RTW. This lead to ORs both >1 and <1 for similar prognostic factors, despite all associations being in the same direction for the same factors.

RTW in both mixed and surgical populations

Workers with less fear avoidance beliefs were more likely to RTW at 6 months (OR 1.09 more fear avoidance - less RTW, SE 0.04, estimated 95% CI 1.01 to 1.18, surgical population)26 and 2 years (OR 0.93 less fear avoidance - more RTW; 95% CI 0.90 to 0.97, mixed population).25

RTW in mixed populations

Younger age did not predict RTW at 2 years (OR 0.97; 95% CI 0.93 to 1.00),25 but predicted RTW at 4 years (OR 0.7; 95% CI 0.60 to 0.8023) and 8 years (OR 0.42; 95% CI 0.30 to 0.5824). Better general health predicted RTW at 2 years (OR 1.03; 95% CI 1.01 to 1.0525) and 4 years (OR 1.10; 95% CI 1.00 to 1.2023). Less sciatica bothersomeness predicted RTW at 2 years (OR 0.89; 95% CI 0.82 to 0.9725), lower low back pain intensity predicted RTW at 4 years (OR 0.80; 95% CI 0.60 to 0.9023) and better physical function predicted RTW at 10 years (OR 1.40; 95% CI 1.10 to 1.8024). A positive SLR test predicted reduced RTW at 2 years (OR 0.44; 95% CI 0.20 to 0.9525). The physician expecting surgery to be beneficial predicted RTW at 10 years (OR 5.00; 95% CI 1.65 to 17.7024). No association with RTW was found for: sex (female OR 0.61; 95% CI 0.31 to 1.2225; male OR 0.33; 95% CI 0.09 to 1.0024), receiving workers’ compensation (OR 0.60; 95% CI 0.30 to 1.2024), initial surgical treatment (OR 1.40; 95% CI 0.46 to 4.6024), low back pain frequency (no data presented24), the Quebec classification (no data presented24) and mental health (no data presented24).

RTW in surgical populations

Older age (OR 0.92; 95% CI 0.85 to 0.99) predicted reduced RTW at 3 months.28 Passive pain coping (OR 1.08, SE 0.04, estimated 95% CI 1.00 to 1.1726 and higher physical work load (OR 1.19, SE 0.06, estimated 95% CI 1.06 to 1.34)26 predicted reduced RTW at 6 months. Depression (estimated OR 1.5422) and occupational mental stress (estimated OR 1.3222) predicted reduced RTW at 2 years. No association with RTW was found for sex (OR 0.22; 95% CI 0.04 to 1.09), BMI (OR 0.90; 95% CI 0.78 to 1.04), general health (OR 1.03; 95% CI 0.98 to 1.08) and physical function (OR 1.06; 95% CI 0.997 to 1.13), smoking status (OR 4.37; 95% CI 0.82 to 23.27)28; neither for a combined measure of pain and disability presurgery (estimated OR 1.4222), job satisfaction (OR 0.9826) and duration of sick leave (OR 1.2626). The latter two studies22 26 did not present CIs. In patients who underwent revision surgery, surgery with fusion (OR 0.56; 95% CI 0.33 to 0.97), psychiatric comorbidity before revision surgery (OR 0.19; 95% CI 0.05 to 0.68) and opioids use within 1 month of revision surgery (OR 0.44; 95% CI 0.26 to 0.75) predicted reduced RTW at 3 years.27

Quality of evidence

Table 3 shows the quality of evidence for all prognostic factors based on the GRADE criteria. Using these criteria, we assessed whether the quality of the evidence should be downgraded (or upgraded). First, all studies included were explorative studies. Therefore, the starting point for the quality of evidence was moderate. Second, the quality was further downgraded for moderate risk of bias in 16 factors, and third, for imprecision in 19 factors. Factors were only assessed in one study each (ie, one study with the same population and follow-up); therefore, the GRADE item inconsistency was not applicable. We only included studies that investigated prognostic factors of RTW in sciatica populations. Therefore, indirectness, the last item, was never a reason for downgrading. Publication bias was not assessed due to the insufficient number of studies. The quality of evidence of prognostic factors included was graded as either moderate (10 factors), low (six factors) or very low (11 factors). For age, the quality of evidence varied between moderate, low to very low, and for sex between low and very low. This depended on the cohort and time point.

Table 3

Quality of the evidence and reasons for downgrading (in bold)


Work participation is an important goal for sciatica patients of working age. In this study, we reviewed and summarised the prognostic factors of RTW in these patients in both short and long terms, up to 10 years. We found moderate to very low quality GRADE evidence for a wide range of factors to predict RTW: general health, pain and disability, psychological factors, other health-related factors, care and work-related factors. There was insufficient data to observe any trends or differences between factors over time.

Several prognostic factors were also identified in two systematic reviews in non-surgically treated populations with sciatica, though these used recovery,29 or pain and disability30 as outcomes, as opposed to RTW in the current review. The two sciatica reviews found that physical symptoms like pain intensity and leg pain were prognostic factors, whereas we found that bothersomeness and low back pain predicted reduced RTW. These physical symptoms may be used to identify patients with both an increased risk of reduced recovery and RTW. In contrast, the physical factors better health and functional status predicted RTW, which has been found in low back pain populations as well.31 Our study also found that age, sex, job satisfaction and neurological findings showed no association with RTW, confirming earlier findings of no association of these factors with clinical outcomes in sciatica.29 30

Psychological factors in prognostic research are useful as these can potentially be modified but can also be used to select patients for specific interventions. Fear of movement is a modifiable psychological factor that predicted reduced RTW in the current review and pain and disability in low back pain patients31 and pain at long-term follow-up in sciatica.32 Mental stress29 30 and passive pain coping28 30 were predictors also previously identified. Finally, depression has been shown to predict application for early retirement in sciatica.32 These findings underline the conclusion of the North American Spine Society clinical guideline for lumbar disc herniation with radiculopathy that psychosocial variables are important factors that influence recovery.33 Screening for these psychological factors may therefore be considered, with subsequent referral to interventions targeting these factors, such as multidisciplinary biopsychosocial rehabilitation.34 Psychological therapies, with or without exercise, using a cognitive behavioural approach are also recommended in a recent NICE guideline on low back pain and sciatica to target psychosocial barriers in patients who avoid normal activities by discussing inappropriate beliefs about their condition.13

There are some differences between our results and earlier studies. Although two previous reviews29 30 showed limited evidence for no association between high work load and poor outcome in terms of pain and disability, we found an association between high work load and reduced RTW in the current study. Considering the difference in outcomes, we hypothesise that work load might be influencing RTW more, as per the current review, than recovery, pain or disability in previous reviews. Also, in our review, a positive SLR test strongly predicted reduced RTW,26 and Ashworth et al 30 found a positive SLR test to predict ‘worse outcome’ in terms of pain and disability. Verwoerd et al reported inconclusive findings with one study finding no association, and another finding a negative association with recovery.29 Specificity of the SLR has been found to be limited for diagnostic use, when used in isolation.35 Neurological tests, often used in conjunction with SLR testing, did not show that neurological deficits or signs were predictive of RTW in our review.25 26 Although the use of SLR testing for diagnostic reasons may be limited, it may serve useful for prognostic reasons for RTW.

RTW rates across the studies included in this review ranged between 67% and 87% and were surprisingly similar, with the exception of the study that assessed RTW after revision surgery.27 The RTW rates in the three studies we included for surgically treated patients were 66.9% at 3 months,2878% at 6 months26 and 81% at 2 years follow-up.22 For mixed surgical and conservative populations, the rates were 73% at 2 years,2580%–87% at 4 years22 and 78% at 10 years follow-up.24Patients who underwent revision surgery had lower RTW rates with 40.2% (without fusion) and 27.0% (with fusion).27 Apparently, patients undergoing revision surgery represent a different group, with poorer prognostic outcomes, especially when discectomy was combined with fusion. Based on the data presented in the studies included, it is not possible to define to what extent RTW was reduced in patients with unfavourable scores on predictors of RTW, compared with those with favourable scores. To facilitate clinical impact, it is important for future prognostic studies to report separate RTW rates for those with favourable and unfavourable scores on predictors of RTW.

It is suggested that factors influencing recovery may differ between surgically and conservatively treated populations.30 In the studies included in this review with mixed populations, the percentage of patients treated surgically were 30%25 and 53%.23 In the latter, initial treatment did not significantly predict RTW. Most prognostic factors in this review were tested in either mixed or surgical populations, which precludes drawing conclusions on comparability of prognostic factors between these populations. Based on this review and previous reviews,29 30 pain intensity seems to be a prognostic factor across all populations, that is, conservative, surgical and mixed, and psychological factors may be important in all patient groups as well.

This study has various strengths and limitations. The data were collected in a systematic way and analysed following current standards for risk of bias assessment, by means of the QUIPS tool,16 and grading of the quality of the evidence, by applying the GRADE method.19 Most included studies reported imprecise measures. The quality of the prognostic evidence ranged from moderate to very low, meaning that estimates for these latter factors are likely to change when more studies will be available. The results need to be interpreted with caution, as estimates are likely to change when future studies will be available. These future studies should preferably include larger samples of either conservatively or surgically treated patients (or analyse data from these groups separately), and test combinations of factors that have been found to be significant in the current review and previous reviews. These factors are preferably measured with instruments from the core outcome set for low back pain,36 and include a standardised instrument to measure RTW, that would allow for meta analyses.37

The prognostic evidence from this review, although partially of low to very low quality, may be used to identify potential high risk patients for delayed or no RTW. This information may assist clinicians and occupational healthcare professionals in guiding these high risk patients, in advising or referring them for additional care or vocational rehabilitation, or in managing and counselling patients’ expectations regarding RTW. Monitoring physical and psychological factors also seem relevant as these predict recovery and RTW in the limited number of studies on sciatica in the current review and also in low back pain studies and several clinical guidelines.13 31 To enhance work participation, physicians could consider monitoring prognostic factors in patients with sciatica that might benefit from additional clinical management or work-directed care. Given the importance of work participation, we recommend more and well-conducted prognostic studies on this important societal outcome of care.



  • Contributors Drafting the manuscript: TO, PK and JLH; revising the manuscript for content: TO, VRS, PK, MWL, MHWFD and JLH; study concept and design: TO, PK and JLH; data collection: TO, VRS, PK, MWL and JLH; analysis or interpretation of data: TO, PK, MWL and JLH.

  • Funding The study was funded by the Dutch Employee Insurance Agency (UWV).

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Patient consent for publication Not required.