Article Text

Prevalence and risk factors for back pain in sports: a systematic review with meta-analysis
  1. Fiona Wilson1,
  2. Clare L Ardern2,3,
  3. Jan Hartvigsen4,5,
  4. Kathryn Dane1,
  5. Katharina Trompeter6,7,
  6. Larissa Trease8,
  7. Anders Vinther9,
  8. Conor Gissane1,
  9. Sarah-Jane McDonnell10,
  10. JP Caneiro11,
  11. Craig Newlands12,
  12. Kellie Wilkie13,
  13. David Mockler14,
  14. Jane S Thornton15,16
  1. 1 Discipline of Physiotherapy, School of Medicine, Trinity College, Dublin, Ireland
  2. 2 Division of Physiotherapy, Karolinska Institute, Stockholm, Sweden
  3. 3 Sport & Exercise Medicine Research Centre, La Trobe University, Melbourne, Victoria, Australia
  4. 4 Department of Sports Science and Clinical Biomechanics, University of Southern Denmark, Odense, Denmark
  5. 5 Nordic Institute of Chiropractic and Clinical Biomechanics, Odense, Denmark
  6. 6 Department of Sports Medicine and Sports Nutrition, Ruhr University Bochum, Bochum, Germany
  7. 7 Department of Applied Health Sciences, Division of Physiotherapy, Hochschule für Gesundheit (University of Applied Sciences), Bochum, Germany
  8. 8 Healthcare in Remote and Extreme Environments program, School of Medicine, University of Tasmania, Hobart, Tasmania, Australia
  9. 9 Department of Physiotherapy and Occupational Therapy and QD-Research Unit, Copenhagen University Hospital, Herlev and Gentofte, Copenhagen, Denmark
  10. 10 Sport Ireland Institute, Dublin, Ireland
  11. 11 School of Physiotherapy and Exercise Science, Faculty of Health Science, Curtin University, Perth, Western Australia, Australia
  12. 12 Body Performance, Cambridge, New Zealand
  13. 13 Bodysystem Physiotherapy, Hobart, Tasmania, Australia
  14. 14 John Stearne Library, Trinity College Dublin, Dublin, Ireland
  15. 15 Fowler Kennedy Sports Medicine Clinic, Western University, London, Ontario, Canada
  16. 16 Western Centre for Public Health and Family Medicine, Schulich School of Medicine and Dentistry, Western University, London, Ontario, Canada
  1. Correspondence to Dr Fiona Wilson, Discipline Of Physiotherapy, Trinity College Dublin, Dublin 2, Ireland; wilsonf{at}tcd.ie

Abstract

Objectives We aimed to determine the prevalence of low back pain (LBP) in sport, and what risk factors were associated with LBP in athletes.

Design Systematic review with meta-analysis.

Data sources Literature searches from database inception to June 2019 in Medline, Embase, Cumulated Index to Nursing and Allied Health Literature (CINAHL), Web of Science and Scopus, supplemented by grey literature searching.

Eligibility criteria Studies evaluating prevalence of LBP in adult athletes across all sports.

Results Eighty-six studies were included (30 732, range 20–5958, participants), of which 45 were of ‘high’ quality. Definitions of LBP varied widely, and in 17 studies, no definition was provided. High-quality studies were pooled and the mean point prevalence across six studies was 42%; range 18%–80% (95% CI 27% to 58%, I2=97%). Lifetime prevalence across 13 studies was 63%; range 36%–88% (95% CI 51% to 74%, I2=99%). Twelve-month LBP prevalence from 22 studies was 51%; range 12%–94% (95% CI 41% to 61%, I2=98%). Comparison across sports was limited by participant numbers, study quality and methodologies, and varying LBP definitions. Risk factors for LBP included history of a previous episode with a pooled OR of 3.5; range 1.6–4.0 (95% CI 1.9 to 6.4). Statistically significant associations were reported for high training volume, periods of load increase and years of exposure to the sport.

Conclusion LBP in sport is common but estimates vary. Current evidence is insufficient to identify which sports are at highest risk. A previous episode of LBP, high training volume, periods of load increase and years of exposure are common risk factors.

  • athlete
  • epidemiology
  • lower back
  • lumbar spine
  • sport

Statistics from Altmetric.com

Background

Low back pain (LBP) is common in the general and athletic population1 2 and is a leading cause of disability globally. Most LBP episodes are short-lasting and self-limiting; however, persistent and recurrent pain is common and for some it becomes disabling, yielding substantial personal, community and economic impact.1 3 Up to 9 in every 10 Olympic athletes experience LBP in their lifetime; at any 1 time, up to 2 in every 3 athletes might be experiencing LBP.2 The number of athletes affected varies substantially between different sports, and there is limited information regarding LBP in sports other than Olympic sports.2

A previous episode of LBP is a consistent predictor of future episodes.4 In the general population, LBP is associated with heavy physical work, smoking, obesity, poor general health and low socioeconomic status.3 Although athletes are less likely to be affected by such factors, they may experience stress, fatigue, anxiety and impaired sleep and mood5 6; psychosocial factors that may predispose them to LBP. Regular exercise can reduce the risk of LBP,7 although this relationship has been reported to be to be U-shaped8 and high levels of physical activity have been proposed to increase risk.9

LBP in athletes can not only be career ending but may also contribute to lifelong pain episodes. Nonetheless, the question ‘What sports have the highest prevalence of LBP?’ remains unanswered. This evidence gap and lack of research synthesis means that professionals dealing with athletes have limited resources to guide prevention and treatment programmes. Knowledge from high-quality observational studies as well as from randomised clinical trials is building blocks for sports injury prevention programmes.10

Our primary aim was to examine the prevalence of LBP in adult athletes in sport. The secondary aim was to explore risk factors associated with LBP in sport.

Methods

Protocol and registration

Reporting of this systematic review was guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses recommendations. Methods of the analysis and inclusion criteria were prespecified and documented in the protocol (PROSPERO database ID: CRD42018094382).

Study eligibility criteria

We considered observational studies (cross-sectional, case control and cohort) evaluating the prevalence of LBP in athletes for inclusion. There was no limit on how LBP was defined, publication date or status. We excluded case studies and case series where N≤5, opinion pieces, letters to editors and studies not written in the English language. Participants had to be athletes of any sporting discipline aged ≥18 years. The primary outcome of interest was the prevalence of LBP in athletes which included condition specific, non-specific and acute and chronic LBP. Secondary outcomes such as exposure data and risk factors were included.

Sources and study selection

We searched five electronic databases (Medline, Embase, CINAHL, Web of Science and Scopus) from inception to June 2019. All search strategies were created by an experienced librarian (DM) and are summarised in online supplemental appendix 1. The last search was run on 12 June 2019. Two authors (FW and LT) independently screened titles and abstracts for eligibility using Covidence review software (Veritas Health Innovation, Melbourne, Australia). The same reviewers independently screened full-text articles as required to determine eligibility. All screening was performed independently and at each stage disagreement (as indicated by Covidence) was resolved by discussion until consensus was reached. Grey literature searching included searching reference lists of included studies and conference proceedings of the following organisations: American College of Sports Medicine (2011–2019), American Physical Therapy Association (2012–2019), World Confederation for Physical Therapy (2011–2019).

Supplemental material

Data extraction and management

After duplicate removal, studies were exported to Endnote software and reviewers (FW, S-JM, AV, LT) extracted data using a customised data extraction form. Discrepancies were resolved by discussion and agreement. The data items extracted were: author(s), country, year, sports discipline, final sample size, age and sex distributions, sports level, response rate, study methods, prevalence of LBP expressed as the number of participants reporting LBP out of the total study sample and potential risk factors (extracted as ORs or relative risk where possible). There was no blinding to study author, institution or journal throughout the data extraction process.

Assessment of methodological quality

Two authors (KT and KD) independently assessed study quality using the critical appraisal checklist developed by Leboeuf-Yde and Walker11 12 and modified by Trompeter et al 2 (box 1). The tool comprised 12 criteria related to the representativeness of the study sample, quality of data and definition of LBP. Each criterion was judged as ‘criterion fulfilled’ (+) or ‘criterion not fulfilled’ (−). A score was then assigned to each study depending on proportion of fulfilled criteria. Disagreements were resolved via consensus discussion. We selected a cut-off point of 65% as previously reported with studies achieving a score over 65% considered to be of high quality.1

Box 1

Study methodological quality critical appraisal tool

A: Is the final sample representative of the target population?

1.At least one of the following must apply to the study: an entire target population, randomly selected sample or sample stated to represent the target population

2.At least one of the following: reasons for non-response described, non-responders described, comparison of responders and non-responders or comparison of sample and target population

3.Response rate and, if applicable, drop-out rate reported

B: Quality of the data?

4.Were the data primary data of back pain or were they taken from a survey not specifically designed for that purpose?

5.Were the data collected from each adult directly or were they collected from a proxy?

6.Was the same mode of data collection used for all subjects?

7.At least one of the following in the case of a questionnaire: a validated questionnaire or at least tested for reproducibility

8.At least one of the following in the case of an interview: interview validated, tested for reproducibility or adequately described and standardised

9.At least one of the following in the case of an examination: examination validated, tested for reproducibility or adequately described and standardised

C: Definition of back pain

10.Was there a precise anatomic delineation of the back area or reference to an easily obtainable article that contains such specification?

11.Was there further such specification of the definition of back pain; and were participants questioned on the frequency, duration or intensity and character of the pain. Or was there reference to an easily obtainable article that contains such specification?

12.Were time periods clearly stated, for example, 1 week, 1 month or lifetime?

Data synthesis

The prevalence data were separated into five categories: 12-month retrospective, 12-months prospective, point, life time and combined (high and non-high-quality studies together). For each of these categories, a combined prevalence proportion was estimated using the R Metafor package (meta-analysis for R) and the MedCalc Software Package V.19.3.1 (MedCalc Software, Ostend, Belgium, www.medcalc.org). We used a Freeman-Tukey transformation (arcsine square root transformation) to calculate the weighted summary proportion under a random effects model. We used the random effects model because of significant heterogeneity between studies. We applied the software filter to remove individual studies in sequence as a sensitivity analysis to explore the reason for heterogeneity as measured by I2 and Cochran’s Q. Publication bias was assessed with funnel plots using MedCalc Software.

In addition, a further five prevalence estimates were calculated in each category for those studies that met the criteria of ‘high quality’. We compared prevalence data across different sports by narrative comparison of data at different time points where this was possible. Finally, for meta-analysis of specific risk factor data, the Mantel-Haenszel method was used to calculate the weighted pooled OR under the fixed effects model using MedCalc Software. Next the heterogeneity statistic was incorporated to calculate the summary OR under the random effects model which is reported below. When pooling if risk factor data were not possible it was reported narratively.

Differences between the protocol and review

We originally planned to use the Newcastle Ottawa Scale to assess study quality. During pilot testing, we judged that the one that we ultimately selected was superior because it specifically related to LBP. We did not originally plan to examine the use of radiology but during data extraction, we realised that radiological assessment was often included as part of LBP screening, so we decided to include it as part of the risk factor synthesis.

Results

The initial search identified 518 records for screening (after duplicates were removed). Figure 1 summarises the study selection process and reasons for excluding studies. The full texts of 125 articles were assessed for eligibility; 39 were excluded and we extracted data from 86 studies.

Figure 1

Study selection process.

Characteristics of included studies

We included 86 studies with a total of 30 732 (range 20–5958) participants (online supplemental table 2). Six studies did not report their sample size.13–18 Studies were conducted across 21 countries and included 58 different sports. Questionnaires were the most common data collection tool (61 studies); 24 of the 61 studies used validated questionnaires most commonly the Standardized Nordic Questionnaire for LBP that was used in 10 studies.19–27 Twenty-five studies reported a survey response rate and those varied between 21%28 and 100%13 20 29 with the mean response rate being 68%. Recall periods varied from point to lifetime. Characteristics of included studies are presented in online supplemental table 2 and prevalence data with risk factors are presented in online supplemental table 3.

Supplemental material

Supplemental material

Methodological assessment

The methodological quality assessment of the 86 studies is shown in online supplemental table 4. A score over 65% (high quality) was reached by 45 studies. Studies commonly did not fulfil criteria for representativeness of the sample (37%), information about non-response (57%), response and drop-out rates (58%), no valid or adequately described and standardised method in case of a questionnaire (73%) or examination (42%). In 67% of the studies, there was no clear anatomic delineation of the lumbar spine and in 47% no mention of frequency, duration, intensity or character of the pain.

Supplemental material

Definition of LBP

Overall, there was no consensus regarding LBP definition with differences in specific location, and lack of measurement of intensity, frequency and duration. Some studies that used the term ‘back pain’ either failed to identify which part of the spine was involved or used it to mean LBP.

In 17 studies, no definition of LBP was reported.30–46 Ten studies collected information via the Standardised Nordic Questionnaire19–27 47; 33 studies included time loss from sport or activity modification13–18 28 29 48–72; 16 studies defined LBP with no reference to time loss/activity modification.73–88 Four studies collected information on disability associated with LBP34 40 89 90 using the Oswestry Disability Index (ODI), the Roland-Morris Disability Questionnaire. Two studies91 92 used the Osaka University Test to define the characteristics of LBP.

Prevalence of LBP in athletes

Prevalence data are presented in online supplemental table 3.

Lifetime prevalence

The pooled lifetime prevalence of LBP in athletes (19 studies, see figure 2)19 21 22 25–27 31 37 42 47 50 57 61 68 70 77 79 81 89 was 63% (95% CI 54% to 73%, I2=99%) and was 63% (95% CI 51% to 74%, I2=99%) when only the 13 high-quality studies were pooled (range 25%–89%). In three studies that also collected information on lifetime prevalence in age-matched physically active (not an ‘athlete’ participating in organised sport) controls, prevalence was 38% (95% CI 32% to 44%, I2=46%).19 22 27

Figure 2

Forest plot of lifetime prevalence of low back pain.

12-Month prevalence

The pooled 12-month prevalence of LBP (35 studies, see figure 3)19–21 23–28 30 31 36 39 43 44 47 49 55 58 61 64 66 72 74 76 81 82 86 87 93–95 was 44% (95% CI 36% to 52%, I2=99%) and was 51% (95% CI 41% to 61%, I2=98%) when only 22 high-quality studies19–28 39 47 49 58 61 64 72 76 81 86 87 95 were pooled (range 7%–94%). The pooled 12-month prevalence of LBP in 28 retrospective studies was 47% (95% CI 39% to 55%, I2=98%) and was 53% (95% CI 43% to 63%, I2=98%) when only 19 high-quality studies19–28 39 47 49 58 61 76 81 86 87 were pooled (range 11%–94%). Pooled 12-month prevalence in seven prospective studies was 28% (95% CI 14% to 44%, I2=97%) and was 33% (95% CI 11% to 60%, I2=93%) when three high-quality studies64 72 95 were pooled (range 7%–78%).

Figure 3

Forest plot of 12-month prevalence of low back pain.

Point prevalence

The pooled point prevalence of LBP (12 studies,22 38 40 41 46 52 65 73 88 91 92 96 see figure 4) was 33% (95% CI 22% to 44%, I2=98%, range 6%–81%) and was 42% (95% CI 27% to 58%, I2=97%) when six high-quality studies22 38 40 41 65 92 were pooled (range 18%–81%).

Figure 4

Forest plot of point prevalence of low back pain.

Other periods

Further data were reported in 15 studies and are reported in online supplemental table 3. Eleven studies13–17 33 45 54 78 97 98 reported LBP prevalence using retrospective analysis of medical data ranging from 1.5% to 10.2%. Four studies reported LBP prevalence according to exposure14 15 29 95 ranging from 0.41/100029 to 11/1000 athlete exposures.14

Funnel plots and sensitivity analyses

Funnel plots were constructed for lifetime, 12-month and point prevalence time points (see online supplemental figures 5–7). Asymmetry was demonstrated in all funnel plots, indicating lack of precision in prevalence estimates, high heterogeneity and possible publication bias. Sensitivity analyses in all meta-analyses, conducted by sequential removal of studies using the MedCalc filter, did not substantially alter heterogeneity in any analysis.

Supplemental material

Supplemental material

Supplemental material

Severity of LBP

Severity or burden of LBP was defined by time loss or by using tools specifically designed to characterise LBP associated disability. Generally, severity of the LBP was poorly reported or not at all. Seven studies quantified the severity of LBP in athletes according to number of days lost from training or competition.14 49 54 63 64 67 85 Specific injuries (eg, ‘stress fractures’ or ‘disc injury’) resulted in the most days missed of any injury ranging from a mean of 110.5 to a mean of 1244 days54 63; median time loss from disc injury was 23 days.14 Two studies in cricket reported a time loss of 44.9 and 55.1 days for each LBP episode63 67 and in the same sport, the time lost per lumbar stress fracture was 169 days.85 The shortest symptom duration was a median 3 days.25 One in four triathletes reported LBP that lasted more than 3 months.70

Three studies used ODI to assess back-related disability34 40 71 but only one of these71 provided data as a disability score of 4.9 (5.4) out of 50 points where 21/50 indicates high disability. Külling et al 90 reported ‘no abnormal results’ for the Roland Morris Disability Questionnaire (RMDQ). One study reported a mean visual analogue scale (VAS) score for LBP at 5.75 out of 10.43

LBP prevalence in different sports

The 12-month prevalence of LBP in athletes was investigated in 38 different sports and is reported in table 1.

Table 1

12-month prevalence of low back pain across different sports

Point prevalence

The point prevalence of LBP in athletes was investigated in 10 different sports and are reported in table 2.

Table 2

Point prevalence of low back pain across different sports

Risk factors and other factors associated with LBP

Risk factor data and data reporting associations between variables and LBP are reported in online supplemental table 3. It was not possible to pool findings for most variables because of the heterogeneity of data analyses and reporting methods. The most common risk factors or statistically significant associations for LBP reported across studies were higher training volumes (all high-quality studies),21 22 47 77 83 95 with OR at 1.122 and 1.2,83 and periods of increased training; particularly going from a training or preparation to competition period with incidence risk ratio (IRR) per season reported by one study at 1.815 when pre is compared with within season (3/4 high-quality studies).15 19 27 76 Athletes who had competed in the sport for longer were at increased risk of LBP (1/3 high-quality studies),34 48 70 with Clay et al 34 reporting 58% greater years of participation in rowing for rowers who sustained LBP in a season compared with those who did not. A history of LBP presented increased risk of reporting an episode for the study period with an OR ranging from 1.6 to 3.9 across five studies (2/5 high quality).34 51 56 68 95 Because of insufficient data in reporting, only two studies could be pooled51 68 resulting in a pooled OR of 3.5; 95% CI 1.9 to 6.4 (I2=0%).

Increasing age was significantly associated with LBP prevalence in seven studies (five high quality)13 21 24 66 67 95 98 and in eight studies (two high quality) there was no relationship between age and LBP.28 30 43 52–54 83 96 Thus, the relationship between increasing age and LBP is inconsistent. The relationship between sex and LBP prevalence in athletes is also inconsistent; three studies (one high quality)77 82 88 reported higher prevalence in women and three (one high quality)15 25 45 reported higher prevalence in men.

Overall, no consistent associations were found between imaging findings and LBP in athletes. Sixteen studies (7/16 high quality) used MRI, X-rays or ultrasound imaging as an adjunct to other assessments to examine associations between imaging findings and LBP across 15 different sports.37 38 41 42 46 57–59 85 89–93 96 98 Abnormalities were found in a range from 14% of athletes in NFL96 to 90.9% in judo.92 Two studies compared radiological abnormalities between athletes and non-athletes.37 89 Hangai et al 37 reported the prevalence of disc degeneration in athletes at 50% compared with 31% in non-athletes although Kraft et al 89 reported that athlete prevalence was no higher than controls. Three studies (none high quality) reported that LBP prevalence was positively associated with radiological abnormality37 59 91 whereas four studies (one high quality) found no association.42 46 89 92 Four studies used imaging to explore the relationship between LBP and muscle function or muscle cross-sectional area (CSA)38 57 58 93 in football and cricket, reporting an association with lower trunk muscle CSA and LBP in two studies58 93 (one high quality) and no association in another.38 Hides et al 57 reported reduced contraction capacity (muscle thickness) of trunk muscles associated with LBP in cricketers.

Movement tests (musculoskeletal (MSK) screening) were included in some studies to examine if they were associated with, or could predict LBP prevalence. An association between reduced trunk muscle function and LBP in tennis players,20 golfers35 and football players71 was found whereas no association was found in seven studies (2/7 high quality) examining football, swimming, rowing, racquet sports, basketball, sailing and cricket.38 46 55 66 73 75 87

A number of studies reported that sport-specific activities such as weight lifting or team positions such as blocking in the NFL were associated with a report of LBP. Other factors which were associated with LBP were family history,32 56 increased body mass35 37 51 and competition level.22 All are reported in online supplemental table 3.

Discussion

LBP is common in athletes, presenting from high-quality studies in our review as point prevalence at 42%, lifetime at 63% and 12-month 53% (retrospective studies) and 33% (prospective studies). We are unable to confidently highlight those sports which are likely to have a higher prevalence than others, because of broad variation in research methods, definitions and reporting styles across studies. The most commonly reported risk factors and variables associated with LBP were higher athlete training volumes and following a change to increased training/competition load. A history of LBP was also associated with risk of a new episode within the study period and years of participation in sport increased LBP risk. There was inconsistent evidence to demonstrate an association/risk increase with sex, increasing age, the presence of radiological abnormalities and musculoskeletal screening test batteries.

Around half of the 86 studies in our review were not of high methodological quality, for reasons including poor characterisation of LBP and use of non-validated survey methods. LBP prevalence was collected at many time points, the most popular of which was 12-month retrospective. There was a notable difference in prevalence when 12-month retrospective reporting was compared with prospective data, highlighting the issue of recall bias. The disparity in these prevalence estimates means that it is likely that longer recall times (such as lifetime prevalence) are influenced by participants’ ability to remember events over lengthy time periods. Future research should focus on time periods where recall is likely to be more accurate.

LBP definition

Studies varied widely in definitions, from simply referring to ‘the back’ to more comprehensive definitions adapted from non-athlete LBP studies. The term ‘low back injury’ rather than LBP was widely used, often as part of a larger injury surveillance programme where nuances of presentation were not captured. Forty-three of the included studies focused only on match and training time-loss injuries whereas 18 studies focused on LBP without any need to report time loss from sports participation to be included. Fourteen studies focused on a presentation to medical staff definition which is likely to be subject to bias because of variation in accessibility of medical staff, particularly in amateur sports. Only 47% of studies precisely defined the anatomical location of LBP and only 66% specified the frequency, duration or intensity and character of the LBP episode. However, there is currently no standardised questionnaire validated for exploring athlete LBP specifically and this should be considered for the future. These results highlight an urgent need to create a definition of LBP for the athlete population that can be used in surveillance research. This definition should be clear about timescale to avoid capturing pain which reflects lower back sensitisation as a response to load (which may be a temporary training response) rather than a true episode. It is possible that those studies which reported very high prevalence (100%) may have been influenced by this. It should be considered that training is often likely to be modified rather than lost and this should be reflected in the definition.

Prevalence of LBP

The prevalence of LBP in the sporting population is high, especially considering that the athlete group is likely to have fewer comorbidities compared with in the general population. Although the prevalence percentages we report in this review may appear higher than those reported in studies of the general population,3 4 methodological issues and wide variation across studies mean we cannot be confident in these estimates. Meta-analyses demonstrated high heterogeneity and funnel plots showed asymmetry due to methodological issues which could include selection and publication bias, as well as very large variation in study sample sizes. Furthermore, there is likely a strong selection process in sports meaning that only individuals able to withstand pain and injuries remain in their sport, the so-called ‘healthy-worker effect’. Therefore, it is important in future studies to collect data on both incidence and prevalence of LBP in athletes. The prevalence of LBP in athletes varied depending on methodologies and case definitions. Generally, studies did not report a minimum symptom period or if the episode of LBP was ‘first ever’ or recurring so incidence could not be established. There may have been bias towards participants self-selecting for inclusion in a study because of their LBP experience. It is also likely that athletes have easier access to medical care and may be more likely to report an episode. These issues many lead to overestimation of prevalence. On the other hand, some athletes may be reluctant to report LBP out of fear for repercussions. The high heterogeneity associated with all pooling highlighted that combined estimates should be interpreted with caution.

Risk factors

Many studies that reported ‘risk factor’ data were of inadequate design to make this conclusion and rather reported associations, meaning that data should be interpreted accordingly (see online supplemental table 3). Research of adequate design is needed in this population to confirm risk factors for athlete LBP. Despite current knowledge that risk of LBP is a multidimensional health symptom that is influenced by the interaction of multiple factors across the biopsychosocial spectrum,3 surprisingly, none of the included studies addressed this. It is essential that future studies address both biological, psychological and social issues as triggers of LBP episodes as well as risk factors for various consequences and disability.99

History was associated with increased risk of reporting a new episode in sport which agrees with findings in the general population.4 This highlights the importance of prevention of the first episode if possible (considering the complexities of non-modifiable risk). LBP was more common with high training volumes, especially when associated with workload spikes. This is supported by some studies that associate the influence of repetitive and heavy loading on the risk of LBP in support of dose response and temporality,100 although this association is not consistent. It may be also explained by an imbalance in a work to recovery ratio that is associated with injury across different sports which supports the risk associated with workload spikes.101 Technical issues in many sports were linked to causation but no studies had appropriate methodology to confirm as a risk factor.

A focus should be on education and load management and to improve the paucity of research into prevention programmes. Sports associated with high biomechanical loading of the spine should examine strategies to improve spine load tolerance in these athletes and to modify biomechanics appropriately to minimise risk of LBP.

Severity of LBP

Some of the studies in this review indicated that LBP in athletes has a protracted recovery period in comparison to other injuries in the same population16 54 76 although reporting limitations limit conclusions being drawn. Although LBP is complex in terms of recovery and is influenced by individual context and beliefs about LBP and management approaches, this should however raise concern. LBP in athletes, as in the general population, can be the cause of considerable and long-standing functional impairment, both to sporting performance and participation (including retirement from sport). However, many studies lacked reports relating to the burden of LBP, particularly in terms of the recovery trajectory. To better understand the burden of LBP on the athlete, future studies should examine condition severity, length of recovery time, loss of career as a result of LBP.

Implications for clinical practice

LBP prevalence in athletes is common across all times points. Findings of this review indicate that it is possible that a U-shaped relationship does exist and that higher levels of exercise and engagement in sport increase risk of a LBP episode.7 8 Risk factors for LBP in athletes are similar to the non-athlete population but it is possible that good load management in athletes may influence the risk of a LBP episode.8 Sports with high mechanical loading of the spine should ensure that training includes methods to build load tolerance. In line with the general population, the use of imaging in managing LBP is of low value unless specifically indicated and our review did not support imaging as a routine assessment in athletes.

Future research

Absence of high-quality epidemiology in athlete LBP means that further research is needed before robust conclusions should be drawn. Research should focus on a standard, sport-specific definition of LBP and survey instruments should be adapted and validated in an athlete population, and studies should include athlete specific outcomes. The burden of athlete LBP should be explored, particularly recovery trajectories, ongoing disability and the need to retire from sport. Prospective data collection and analysis across sports should be used to understand incidence of LBP in athletes; longitudinal studies will explore long-term outcomes and risk of repeated episodes.

Limitations of this review

We only included studies in this review that were in the English language or provided a translation. There was no blinding of reviewers to study authors at any stage. The tool that we used to assess quality has been used in LBP populations and has been used in assessing quality in Olympic sports although it has not been specifically developed for use in athletes. Furthermore, it may lack sensitivity to assess all aspects which influence study quality. Only adult athletes were included and we recognised that LBP is common in adolescent sports. Inclusion of a study required that the LBP that was reported was a result of sport participation. However, given the complexity of LBP presentation, it is not possible to confirm that sport was always related to causation. The term back pain was used in some studies rather than low back pain. At full text study screening and data extraction, the full paper was examined and a judgement was made that the study was examining the lower back. We note that the terms back pain and LBP are used interchangeably at times. Selection bias due to non-responses in individual studies and disproportionately large sampling of those who had experienced LBP is likely to have biased the pooled estimates. We defined ‘point prevalence’ as reported pain at the time of interview/study enrolment although we acknowledge that this time point commonly includes report of ongoing pain and pain in the last 7 days and reported these together. The studies across this review are biased to elite athletes and results may not be generalised to the general sporting population. Heterogeneity was very high when pooling data and this should be considered when observing prevalence estimates.

Conclusion

LBP is common in athletes, but little is known about its cause and true impact. Prominent risk factors are a previous LBP episode as well as high training loads and load spikes. Future studies should examine the interaction of multiple factors across the biopsychosocial spectrum when examining athlete LBP.

What is already known

  • Low back pain (LBP) is one of the most common problems in the general population and is a leading cause of disability.

  • Exercise is effective in reducing the risk of LBP episodes although this relationship has been proposed to be U-shaped and a high level of physical activity can be associated with increased risk.

  • LBP in athletes can be career ending and can result in considerable disability.

What are the new findings

  • LBP is prevalent in athletes, with prevalence similar or higher than the general population, and with a long recovery trajectory for some. This indicates that while exercise is generally protective for LBP, it may increase risk at higher volumes.

  • Risk factors for LBP in sport include history, high volume of training and competition with poor load management and years of exposure to sport.

  • Current screening programmes presently are insufficient to identify those at risk.

  • Future research in athlete LBP should provide a specific sport appropriate definition with a temporal component. Survey tools should be developed and validated in an athlete population.

Acknowledgments

Thank you to Alison McGregor who contributed to initial planning of this paper.

References

Supplementary materials

Footnotes

  • Twitter @fionawilsonf, @clare_ardern, @DrLarissaTrease, @janesthornton

  • Contributors FW, KW, JST, LT and AV designed concept, project managed and assisted with manuscript writing, editing and data extraction. S-JM and CN assisted with data extraction and paper reviews. JPC: data interpretation and manuscript writing. CG: data analysis. DM: search strategy. KD: editing, creating tables and quality review. KT: quality review. CLA and JH: project management and editing.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

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

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.