Article Text
PDF
Abstract
Objective Neck pain is a common musculoskeletal disorder, but little is known about which individuals develop neck pain. This systematic review investigated factors that constitute a risk for the onset of non-specific neck pain.
Design and setting A range of electronic databases and reference sections of relevant articles were searched to identify appropriate articles. Studies investigating risk factors for the onset of non-specific neck pain in asymptomatic populations were included. All studies were prospective with at least 1 year follow-up.
Main results 14 independent cohort studies met the inclusion criteria for the review. Thirteen studies were assessed as high quality. Female gender, older age, high job demands, low social/work support, being an ex-smoker, a history of low back disorders and a history of neck disorders were linked to the development of non-specific neck pain.
Conclusions Various clinical and sociodemographic risk factors were identified that have implications for occupational health and health policy. However, there was a lack of good-quality research investigating the predictive nature of many other variables.
- Neck pain
- risk factors
- systematic review
- epidemiology FQ
- public health FQ
- risk prediction
Statistics from Altmetric.com
Neck pain is one of the four most commonly reported musculoskeletal disorders.1–5 It is estimated that in the adult ‘world population’ there is a mean lifetime prevalence of 50%, year prevalence of 37%, month prevalence of 25%, and point prevalence of 10%.6 About one-fifth of adults who were previously pain free report a new episode of neck pain in a 1-year period.7
Although not associated with high morbidity, the high prevalence and episodic nature of neck pain incurs substantial costs. In the Netherlands in 1996, the cost of neck pain was estimated to be US$686 million, about 0.1% of the gross domestic product at that time.8 A minority of this cost (23%) was related to direct healthcare costs such as hospitalisation, medical specialists and medical procedures, while 77% was incurred from indirect costs such as work absenteeism and disability.8
Risk factors are variables associated with a greater likelihood of developing specific pathology. As with other musculoskeletal disorders, the development of non-specific neck symptoms is thought to be related to multiple factors.9–11 The identification of factors that predispose individuals to develop neck pain may contribute to recognition of at-risk groups and measures of primary prevention. Primary prevention is directed toward reducing the risk of initial onset of a problem.12 For example, occupational health departments may be interested in addressing workplace factors that predispose their employees to neck injury.
Several reviews have investigated the risk factors related to the onset of neck pain.10 11 13–15 The conclusions from these reviews are of limited value since they have relied almost exclusively on the results of observational and case–control studies. Recently more prospective studies have been conducted and added to the body of literature relating to the onset of neck pain. The aim of this systematic review is to summarise current good-quality literature in order to identify factors that have been linked to the onset of a new episode of non-specific neck pain.
Methods
Data sources and search strategy
The systematic review strategy described by the NHS Centre for Reviews & Dissemination is used for this review.16 Online searches were conducted on AMED (1985–Aug 2007), CINAHL (1982–Aug 2007), EMBASE (1974–Aug 2007), Medline (1966–Aug 2007), PsychINFO (1806–Aug 2007), Cochrane Register of Systematic Reviews and PEDro. Google and Google Scholar were also searched to ensure that no articles were overlooked. Examples of keywords used were ‘neck or cervical pain’, ‘OR or predictors or risk factors or probability'. The search was restricted to include observational, prospective, cohort or follow-up studies in the English language. The reference lists of primary studies identified through database search were scanned to identify relevant additional citations. Key journals (ie, Journal of Epidemiology and Community Health, Journal of Clinical Epidemiology, Occupational and Environmental Medicine) were hand searched to identify relevant articles that may not yet have been indexed on the online databases.
Selection of studies
A study was included if: (1) the study population was free from neck pain at baseline assessment; (2) it was a prospective study with a follow-up period of 1 year or more; (3) it focused on determinants for onset of non-specific neck pain (where non-specific neck pain is defined as pain without a specific systemic disease being identified as the underlying cause of the complaint); (4) it related to human subjects; (5) it was a full, peer-reviewed report published in the English language.
A study was excluded if: (1) it did not focus on onset of non-specific neck pain as a key outcome; (2) it evaluated whiplash-associated disorder; (3) neck pain had not been analysed separately from other musculoskeletal disorders; (4) it evaluated therapeutic or surgical intervention; (5) it focused on prognostic factors that predict the progression of neck pain in a symptomatic population; (6) it was a case–control or cross-sectional study; (7) it concerned a population with specific underlying pathology such as tumours, fractures, infection, inflammatory disorders and osteoporosis.
After the initial search, a three-phase screening strategy was used to identify the articles to be reviewed. Firstly, the titles and abstracts identified by the search strategy were screened by one investigator (SMc) who retained all papers related to neck pain. Secondly, two independent reviewers (SMc and SM) reviewed the remaining titles and abstracts using the inclusion and exclusion criteria to select potentially relevant studies. Finally, the full-text articles were retrieved, and both reviewers independently reviewed each of the retrieved articles to ensure that they met the inclusion/exclusion criteria for the review. In the event of disagreement between the two reviewers, a third reviewer (JKM) reviewed the article and arbitrated until agreement for inclusion or exclusion was reached.
Quality assessment of studies
The quality assessment tool used here, and reflected in table 1, is adapted from two similar assessment tools that have been used in previous systematic reviews of prognostic factors for whiplash-associated disorders and non-specific neck pain.9 17 The assessment tool consists of 17 items, each having a ‘yes’, ‘no’ and ‘don't know’ option. The original assessment tool had 16 items. Point B2 was added to allow assessment of studies based on cohort size. This has been used as a point of assessment in previous systematic reviews of cohort studies and is an important consideration in studies where multivariate analysis has been undertaken.9 18 Smaller studies, with large numbers of predictive variables, allow less confidence in the results of the analysis.19 There is no universal method of calculating sample size for multivariate analysis,19 but sample sizes of 300 have been described as fair.18 Accordingly, our review gave studies with sample sizes of more than 300 subjects a higher rating.18
Results of methodological assessment
To determine the methodological quality of the studies in the review, each criterion was evaluated for the presence of sufficient information and the likelihood of bias. If sufficient information was available, the criterion was rated positive (yes) and given a 1-point score. If bias was considered likely, the criterion was rated negative (no) and given a zero score. When information was not given or was unclear, the criterion was rated inconclusive (don't know) and given a zero score. For each study, a total quality score was computed by counting all positively rated criteria (maximum score 17 points). If articles were based on the same cohort, one quality score was given based on the information from all available publications. The assessment tool was piloted on three studies that were not included in the review, and minor adjustments were made. After this, the review studies were assessed and scored independently by two independent reviewers (SMc, SM). In the case of disagreement, consensus was sought, but when disagreement persisted, a third independent reviewer (JKM) made the final decision. A study was considered to be of good quality if it scored ≥9 points on the quality assessment scale. This is in line with a previous review.17
Data extraction
The two independent reviewers (SMc, SM) used a standardised form to extract information and data regarding the study population, inclusion and exclusion criteria, type of prognostic factors, follow-up period, outcomes and data on risk. If consensus could not be reached, a third reviewer (EG) made the final decision.
Data synthesis
The interobserver agreement of quality assessment was derived by calculating percentage agreement and a κ coefficient to correct for chance agreement.20 Extracted information is presented in table format to highlight differences and similarities between the studies. The asymptomatic populations differed in terms of age, occupation and country of recruitment. There were different follow-up periods, different outcomes, different baseline variables and different methods of data analysis. Given this heterogeneity of selected studies, a meta-analysis was not attempted, but instead a qualitative synthesis was conducted with conclusions based on levels of evidence used in previous systematic reviews (table 2).17 21 22
Levels of evidence for predictive factors
Results from multivariate analysis were used to establish the levels of evidence. If only univariate results were available, these were used to determine the levels of evidence instead.17 23 Outcomes, in the way of p values, RR, OR or HRs were extracted from each study. Significant associations (p<0.05) or clinically relevant estimates of risk, defined as RR, OR or HR ≤0.5 or ≥2.0 were deemed relevant.17 24 A positive effect of a risk factor implied a decreased risk of the development of neck symptoms. A negative effect of a risk factor implied an increased risk of the occurrence of the neck symptoms. No effect of a risk factor implied that the presence of the factor neither increased nor decreased the risk of developing neck symptoms.13
Results
Selection of studies
The flow chart (figure 1) shows the process of study selection. The initial search yielded 385 citations of which 273 articles were not related to the topic, 112 were independently screened by two reviewers (SMc and SM), and 66 were included for screening of the full articles. After a consensus meeting, 51 studies were excluded from further review and 15 studies were included in the final review, two of which were produced from the same cohort of patients.25 26 As a result, 14 independent cohort studies were included in this systematic review.7 25–38 Details of the studies excluded from this review are available from the lead author on request.
Flow diagram of selection process of studies.
Methodological quality
The reviewers reached agreement on 85% of the quality items assessed, the overall interobserver agreement being κ=0.61 with a SE of 0.1.20 This represents good agreement between the two reviewers.39 Disagreements were mainly related to reading errors or interpretation of the quality criteria list. These disagreements were easily resolved. Disagreement persisted on only one item (item O).34 The third reviewer (EG) made the final decision in this case. The results of the quality assessment are in table 1.
The cohorts are ranked by their methodological quality score, with higher scores indicating better methodological quality. The overall quality scores ranged from 8 to 15, with all but one study achieving high-quality status. The most common methodological shortcomings were related to poor description of the inception cohort (item A), drop-out rates (item E), information about drop-outs (item F), and poor use of outcome measures (item L). All the studies presented crude or adjusted estimates of risk. In some studies, it was unclear which confounding variables had been included in the multivariate analysis.
Study characteristics and risk factors
The main characteristics of the study populations, the range of risk factors and outcomes for each cohort, including the univariate and multivariate statistical results supporting each factor are detailed in supplementary appendix 1 (available online only). Of the 14 cohorts, 12 were recruited from occupational groups,25–27 29–38 one study was recruited from the general population,7 and one study was recruited from the general population of schoolchildren.28 The sample sizes varied from 124 to 21378.32 35 Twelve (85%) of the studies enrolled more than 300 subjects, and six (43%) enrolled over 1000 subjects. The shortest follow-up period was set, a priori, at 1 year, and the longest follow-up period was 12 years.33 The percentage of subjects lost to follow-up varied between 0% and 50.5%,38 40 although, in two studies, this figure was not made clear.27 32 The percentage of subjects lost to follow-up was less than 20% at 1 year in only four cohorts.25 26 35 37 38
Summary of risk factors
Table 3 presents a qualitative summary of the available evidence for the different risk factors for the onset of neck pain. Forty-five risk factors were identified. For ease of interpretation, these risk factors were grouped into one of four categories: physical (n=20), psychological (n=6), sociodemographic (n=13) or clinical (n=6). There is strong evidence that high job demands (n=2363, two studies),25 26 30 female gender (n=3666, two studies),7 37 low social or work support (n=3229, two studies),25 26 31 being an ex-smoker (n=2432, two studies),31 33 a history of low back problems (n=3937, two studies)7 29 and a history of neck problems (n=4569, three studies)7 29 37 represent a risk for the future onset of neck symptoms. There is also moderate evidence that older age represents a risk for the future onset of neck symptoms (n=942, two studies).27 37 However, there was strong evidence that age is a risk factor for men (n=21915, two studies),33 35 whereas for women there was conflicting evidence, with one study suggesting that age has a protective effect.33 All the remaining variables highlighted in table 3 have limited evidence of predictive ability for outcome as they have only one high-quality study supporting them.
Physical, psychological, sociodemographic and clinical risk factors for the development of neck pain
Discussion
This review summarised the findings of 15 prospective studies from 14 independent cohorts investigating the predictive nature of around 50 physical, psychological, socio-demographic and clinical factors for the onset of non specific neck pain. The methodological quality of the studies was high with all but one study exceeding our quality standard of nine or more points out of a possible 17.
Overall analysis revealed evidence that older age, female gender, high job demands, low social or work support, being an ex-smoker, a history of low back disorders and a history of neck disorders predicts the future onset of neck pain. Many of the variables in table 3 have been investigated by only one study, making it impossible to be sure of their predictive nature. Many physical, psychological, socio-demographic and clinical variables have not been investigated. This is an important area of further research since multiple variables have been shown to be important in the development of other musculoskeletal disorders.41–43
In addition, this review provides evidence from two high quality studies and one low quality study that older age, up to a point, predicts the future onset of neck pain.27 35 37 These studies investigated subjects in a younger age range (20–52 years of age at baseline) followed up over a maximum of 5 years. One high quality study provided conflicting evidence which suggested that older age has a protective effect in women.33 They investigated an older population (45–55 years of age at baseline) followed up over 12 years. There is strong evidence from other good quality longitudinal studies to suggest that subjects in the age range 45–55 are twice as likely to develop neck pain as their younger counterparts.44–47 The incidence of neck pain with age is thought to steadily increase until about the age of 55 and then begin levelling out.1 3 4 The findings of this review support the above hypothesis, but further research would be required to confirm this.
Limitations of the study
The possibility of publication bias cannot be excluded. Studies with significant results are more likely to be published or result in multiple publications, making identification of these papers more likely.39 48 Studies in languages other than English, unpublished studies, studies from non indexed journals and relevant studies from lesser known databases may have been missed.
There is no clear definition of what constitutes a new episode of neck pain for example ‘never previously had neck pain’, ‘no neck pain in the past year’ etc. Similarly there are no clear guidelines defining an asymptomatic population. For the purposes of this review, the study population was pain free at baseline regardless of how long they had been pain free before this. The onset of neck pain was considered to be the onset of any reported neck symptom, regardless of severity of symptoms and regardless of whether the patient had experienced similar symptoms before. This was done for pragmatic reasons since the majority of people have experienced neck pain at some point in their lives.6 We excluded several studies in which the asymptomatic group contained patients who were mildly symptomatic. Only four of the studies in this review investigated subjects who had been pain free for at least 1 year prior to recruitment.27 32–34
The levels of evidence in this review were drawn from the findings of univariate and multivariate analyses. The use of univariate analyses, when multivariate analyses were not available, could have biased the conclusion of the levels of evidence for the risk factors, since univariate analysis does not adjust for confounding factors. This approach has been used before.17 Three studies failed to provide multivariate analysis, and two of these contributed to the levels of evidence table.29 38 If we had not used data from their univariate analysis only one of our conclusions would have changed, there no longer being strong evidence to support back pain as a risk factor.
Findings from previous systematic reviews
Several reviews and systematic reviews have identified risk factors for the onset of neck pain. An early review of psychosocial work factors for the development of musculoskeletal disorders identified 13 cross-sectional studies and five longitudinal studies focusing on the development of symptoms in the neck/shoulder region.14 They found some evidence that monotonous work, time pressure, high work demands, physical load at work and stress was linked to the development of neck/shoulder symptoms. A second review examined psychosocial variables that are associated with neck disorders in 26 studies, only two of which were prospective.15 They found a consistent association between neck disorders and physical risk factors such as repetition, forceful exertion, and awkward or static postures. A third review of psychosocial risk factors for the development of neck pain included 28 cross-sectional studies and one prospective study.10 They found limited evidence for a wide range of factors namely: high quantitative job demands, low social (co-worker) support, low job control, high and low skill discretion and low job satisfaction. A fourth systematic review of physical risk factors identified 22 cross-sectional studies and two prospective studies.13 They found limited evidence for a positive relationship between development of neck pain and the following physical factors: neck flexion, arm force, arm posture, duration of sitting, twisting or bending of the trunk, hand-arm vibration and workplace design. One final systematic review of 37 prospective studies identified a range of psychological and cognitive variables as risk factors for back and neck pain.11 Only five of these studies related to neck pain populations and none were related to the onset of non specific neck pain.44 46 49–51
These reviews support the existence of associations between a range of physical and psychosocial factors and neck disorders. However the majority of primary studies were cross-sectional and do not demonstrate a temporal relationship between the factor and the outcome. Therefore causality is difficult to establish.52 53 Consequently the results of these reviews must be interpreted with caution. Despite the limited number of prospective studies and the reliance on cross-sectional studies, these reviews reached similar conclusions to our own review. However, none of the previous reviews support our findings that female gender, older age, history of low back or neck disorders are strong risk factors for the development of neck pain. All of the reviews, including our own, were limited in their ability to draw conclusions due to the small number of good quality studies investigating risk factors for neck pain.
Implications
A previous history of neck pain is a strong risk factor for the development of further episodes of neck pain. The majority of people who experience neck pain may be expected to recover54 55 but there are indications that the clinical course of neck pain will follow a pattern of intermittent episodes of pain and disability over a period of years.7 Our earlier systematic review identified that a prior history of neck pain was also strongly predictive of the progression of neck pain to recurrent, persistent or disabling neck pain.23 Therefore, occupational health departments should consider strategies to prevent the development or the return of neck problems, for example minimising the impact of high work demands and ensuring that workers are appropriately supported at work. Clinicians need to ensure that they are taking relevant management approaches to ensure a speedy recovery from the current episode, prevent the return of further neck problems, and provide patients with self-management strategies should recurrences happen. It is incumbent upon researchers to investigate treatment strategies that reduce the risk of further episodes of neck pain.
Individuals with a history of back pain and pain at other sites are at risk of developing neck pain. It is not clear why pain at other sites should lead to the onset of neck pain, and current research offers few insights into this particular relationship; however these findings are consistent with previous research.56 57 It has been concluded by one review that diseases of all kinds tend to cluster in certain individuals and that back pain is part of this pattern.58 This may also be true for neck pain. The nature of this relationship is unclear, but speculatively may relate to genetic, psychosocial or socio-demographic variables. Further research in this area is required.
To date, the evidence for risk factors is either limited or lacking. Many factors are supported by evidence from one high quality study only. Others, particularly psychological and socio-demographic factors, lack any evidence at all. Identification of risk factors that predispose patients to develop further episodes of neck pain may be important to prevent the patient from developing recurrent, persistent or chronic neck pain and disability. Given the socioeconomic cost and work absence caused by neck pain,8 59 there is a need for further prospective studies which investigate clinical, physical, psychological and socio-demographic risk factors for development of neck pain.
Conclusions
Only a few clinically relevant risk factors were identified by this review. Strong evidence was found to link female gender, older age, high job demands, low social or work support, being an ex-smoker, a history of low back disorders and a history of neck disorders to the development of non specific neck pain. There was a lack of good quality research investigating the predictive nature of many clinical, physical, psychological and socio-demographic variables. These are important areas for further research.
What is already known on this subject
The existence of associations between a range of physical and psychosocial factors and neck disorders is well-recognised. Causal relationships between these factors and the development of neck pain have been previously difficult to establish due to the large body of cross-sectional primary studies available.
What this study adds
This study provides strong evidence that older age, female gender, high job demands, low social/work support, being an ex-smoker, a history of low back pain and a previous history of neck disorders predicts the future onset of neck pain. Occupational health policy should include strategies to prevent the development of neck problems.
Acknowledgments
Gwendolijne Scholten-Peeters who gave us permission to adapt her quality assessment tool, criterion list for quality assessment tool, and data extraction tool.
References
Footnotes
Funding Arthritis Research Campaign, Copeman House, St Mary's Court, St Mary's Gate, Chesterfield, Derbyshire S41 7TD, UK.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.