|Home | About | Journals | Submit | Contact Us | Français|
The objective of this study was to determine the prevalence of neck pain (NP) in the world population and to identify areas of methodological variation between studies. A systematic search was conducted in five databases (MEDLINE, EMBASE, CINAHL, OSH-ROM, and PsycINFO), followed by a screening of reference lists of relevant papers. Included papers were extracted for information and each paper was given a quality score. Mean prevalence estimates were calculated for six prevalence periods (point, week, month, 6 months, year, and lifetime), and considered separately for age, gender, quality score, response rate, sample size, anatomical definition, geography, and publication year. Fifty-six papers were included. The six most commonly reported types of prevalence were point, week, month, 6 months, year, and lifetime. Except for lifetime prevalence, women reported more NP than men. For 1-year prevalence, Scandinavian countries reported more NP than the rest of Europe and Asia. Prevalence estimates were not affected by age, quality score, sample size, response rate, and different anatomical definitions of NP. NP is a common symptom in the population. As expected, the prevalence increases with longer prevalence periods and generally women have more NP than men. At least for 1-year prevalence Scandinavian countries report higher mean estimates than the rest of Europe and Asia. The quality of studies varies greatly but is not correlated with the prevalence estimates. Design varies considerably and standardisation is needed in future studies.
Neck pain (NP) is a major public health problem, both in terms of personal health and overall well-being [22, 25, 61] as well as indirect expense [5, 10, 22]. For instance the total cost of NP in the Netherlands in 1996 was estimated to about 1% of the total health care expenditure or 0.1% of the Dutch gross domestic product .
Accurate prevalence estimates are desirable to serve as a basis for etiological studies and health care evaluation, and to assess the effect of NP in general populations [10, 22]. Unfortunately, prevalence studies on NP show great variation in both quality and results. For instance, the point prevalence varies between 6%  and 22%  and 1-year prevalence between 1.5%  and 75% .
Different results of observational studies may be due to varying definitions, for example, the neck region, NP, and the duration of pain. Methodological differences, such as non-comparable population samples, differing response rates, and the overall quality of the studies, may also cause bias and explain the discrepancies .
Although a few authors have tried to incorporate small reviews of NP prevalence in their papers [3, 21], this literature has never been systematically and critically reviewed. We, therefore, conducted a systematic and critical literature review in order to determine the prevalence of NP in the world population and to identify areas of methodological variation between studies.
A systematic search was conducted in the MEDLINE (Silverplatter, 1966–2002), EMBASE (Science Direct, 1975–2002), CINAHL (Silverplatter, 1967–2002), PsycINFO (Silverplatter 1967–2002), and OSH-ROM databases (RILOSH, NIOSHTIC2, MHIDAS, HSELINE, CISDOC all completed 2002/12). The search terms were: ‘neck’, ‘cervical’, ‘spinal’, ‘back’, ‘musculoskeletal’, ‘pain’, ‘ache’, ‘problem’, ‘complaint’, ‘prevalence’, ‘incidence’, ‘survey’, and ‘epidemiology’ (truncated when appropriate) (the full search strategy can be requested from the corresponding author). The search on MESH/EMTREE terms were equivalent to the free text search. Due to the different databases in the OSH-ROM databases only a free text search was possible. The bibliographic databases were searched, focusing on titles and abstracts, and relevant papers were retrieved. Reference lists of all the included and excluded papers were systematically screened for additional papers.
The following criteria were endorsed:
To illustrate our completeness of the search strategy, the numbers of retrieved and included papers from each database were tabulated. Capture–recapture analysis has previously been used in a systematic literature review , but is not considered appropriate in literature reviews due to the non-independency between the bibliographic databases .
Information on core items were extracted from each paper independently by two reviewers (R.F., J.H.) (Table 1). Any discordance or disagreement was resolved by discussion or by majority (K.O.K.). The crude response rate (numbers of responders out of the total number of subjects contacted) for each study is presented. The total and the gender prevalence estimates are presented or calculated if not provided.
A scoring system was developed, based on two quality scoring systems previously used for assessing prevalence studies on low-back pain [39, 44] (Table 2). The quality criteria focused on representative population samples, valid and reliable outcome measures, and precision of the prevalence estimates. Quality scores were assigned to each paper independently by two reviewers (R.F., J.H.) and subsequently compared. Any disagreement was resolved by discussion, and if still unsolved a third reviewer was included (K.O.K.). No attempt at blinding the reviewers was made. Each criterion was weighted equally, as we could not differentiate which criterion is more important for the overall quality assessment. The quality criteria were clearly defined a priori to avoid any bias. Studies with a score of 1 or 2 points were considered to be of poor quality, between 3 and 5 points of medium quality, and 6 or 7 points of higher quality.
For each prevalence period, the mean prevalence estimates from homogeneous study samples based on adult populations were calculated (the total number of subjects with NP divided by the total number of participants) and visualised graphically. Prevalence was considered separately for age, gender, quality score, response rate, sample size, anatomical definition, geography, and publication year, where applicable. The average gender ratio was calculated for each prevalence period (i.e. the total number of females with NP divided by the total number of males with NP).
In total, 56 original papers were included (Table 3) [1, 2, 4, 6–9, 11–21, 23, 24, 26–36, 38, 41–43, 46–48, 51, 52, 54–57, 59, 60, 62–71]. Fifty-five papers were found by search in electronic databases and one paper was found via reference lists. All electronically retrieved papers could be retrieved in MEDLINE or EMBASE. No additional papers were added after searching other databases.
Thirty-five papers fulfilling the inclusion criteria were subsequently excluded, mainly because data had already been presented in another paper (16 papers), the type of prevalence was not stated (11 papers), or too broad anatomical definition was used (i.e. neck–shoulder–upper-limb) (eight papers). The full list of excluded papers is available from the corresponding author.
Table 4 lists all the included studies. Almost half the studies (46%) were from Scandinavia, 23% from the rest of Europe, 16% from Asia, and 11% from North America. Two papers were from Australia and one from Israel.
Most studies (79%) had unbiased, randomized population samples. The sample sizes varied from 300  to 51,050  participants. Thirty-seven (66%) of the studies had sample sizes of more than 1,000 subjects. The crude response rates varied between 15%  and 100% [14, 41, 42, 57]. Twenty had inadequate (<70%) response rates [2, 6, 7, 9, 16, 19, 21, 23, 27, 28, 30–32, 36, 42, 43, 48, 53, 56, 71]. In four studies the response rate was not reported [15, 26, 47, 59].
The six most commonly reported types of prevalence were 1-year (39%), point (13%), lifetime (13%), 6-months (11%), 1-month (10%), and 1-week (10%). In many studies, extra criteria were added to the prevalence definitions (e.g. ‘Pain lasting for more than 3 months’). The definition of NP (i.e. pain, ache, troublesome, soreness) and the anatomical definition of the neck region also varied between studies.
The prevalence estimates and confidence intervals for all included studies are presented in Fig. 1. Generally and as expected, the ranges increase with longer prevalence periods. Thus, the mean prevalence estimates for the adult populations show a steady increase with increased prevalence periods.
In eight studies [1, 4, 7, 21, 47, 48, 59, 63], the point prevalence was presented ranging from 5.9%  to 38.7% . For the adult population (15–74 years), the prevalence ranged from 5.9%  to 22.2% , with a mean prevalence of 7.6%. One study  focused specifically on an elderly population (65+ years) with 38.7% point prevalence.
The 1-week prevalence was presented in six studies [15–17, 19, 30, 54], ranging from 1.4%  to 36% . The NP definition in the latter study was ‘waking pain and/or stiffness’, which to our knowledge is the only study using this kind of definition. The mean 1-week prevalence for the remaining of the studies (15–90 years) was 12.5%, ranging from 1.4%  to 19.5% .
In six studies [23, 26, 32, 46, 62, 66], the 1-month prevalence was presented. The range for the adult population (16–79 years) was between 15.4%  and 41.1% , with a mean of 23.3%. One study  focused specifically on children, with a prevalence of 6.9%.
The 6-month prevalence was reported in seven studies [5, 9, 19, 21, 27, 33, 65]. For the entire adult population (18–80 years), the prevalence was between 6.9%  and 54.2% , with a mean prevalence of 29.8%. The prevalence for the three studies on children/adolescents ranged between 6% (12-year-old males) and 45% (18-year-old females) .
The 1-year prevalence was estimated in 22 studies [8, 11, 14, 20, 28, 29, 31, 34–36, 38, 41–43, 52, 54–56, 60, 64, 67, 71]. Based on fairly homogeneous definitions of NP, the prevalence ranged from 16.7%  to 75.1%  for the entire adult population (17–70 years), with a mean of 37.2%. In two studies on adolescents, prevalences of 15.8%  and 22.1%  were reported. Three studies focused specifically on elderly populations, with a prevalence range between 8.8%  and 11.6% .
Eight studies presented the lifetime prevalence [2, 12, 21, 38, 46, 57, 68, 69]. In two studies on lifetime, NP among natives from small islands in the South Pacific Ocean was reported (0.2% and 2.1%) [68, 69]. The lifetime prevalence for the rest of the adult population (18–84 years) ranged from 14.2%  to 71% , with a mean of 48.5%. One study focused specifically on elderly women, with a prevalence of 17% .
In studies dealing with pain of longer duration (e.g. pain lasting for more than 3 months), lower prevalence estimates were presented, except for the point prevalence [1, 48, 59]. Studies where NP was reported as the most painful site also generally showed low prevalence estimates [6, 9, 62]. Otherwise, the various definitions of NP (e.g. pain, ache, soreness, etc.), different type of questions, or difference in response format (i.e. yes/no, graded pain, etc.) did not systematically affect the prevalence estimates (data not shown).
In studies on elderly populations, low estimates were reported in 1-year prevalences [14, 35, 71]. However, in the point prevalence elderly people reported more NP than for the remaining adult populations .
The criteria scores for all included studies are shown in Table 4. Seven studies (13%) [1, 8, 15, 18, 21, 41, 51] were of higher quality, 42 studies (75%) were of medium quality, and seven studies (13%) [2, 6, 9, 30, 47, 57, 60] were of poor quality. Only two studies were assigned maximum points [1, 18], whereas one paper did not score any points at all .
No pattern between quality scores and prevalence estimates was seen in any of the prevalence periods (i.e. studies with low quality scores did not show greater differences in prevalence estimates compared to studies with high scores). In fact, in the study obtaining zero point  and the study assigned maximum points  nearly identical point prevalence estimates were reported.
No consistent pattern of the effect of inadequate (<70%) response rate on the prevalence estimate was found (Table 4). Some studies with low response rates reported high prevalence estimates (point [21, 48], 1-week [30, 54], 6-months , and 1-year prevalence [42, 56]), whereas other studies reported low prevalence estimates (1-week  and 1-year prevalence [31, 36, 71]).
Studies with inadequate sample sizes (less than 1,000 participants) did not differ from studies with higher sample sizes in relation to the prevalence estimates (Table 4).
The anatomical definition varied between studies, typically either including or excluding the shoulder region. Twelve studies included the shoulder in the anatomical definition of the neck region (Table 4). However, there were no differences between studies including or excluding shoulders in any of the prevalence estimates.
The mean 1-year prevalence estimates from different geographical regions based on comparable studies are presented in Table 5. The 1-year prevalence was higher in Scandinavian countries than in the rest of Europe and Asia. However, this was not statistically significant. The lifetime prevalence estimates found in two studies from the Tokelau Islands (small islands in the South Pacific Ocean) were very low and close to zero [68, 69]. However, in the remaining studies geographical/regional differences did not affect prevalence estimates.
All studies were published between 1980 and 2002, with the majority (87.5%) of publications from 1991 and onwards. None of the prevalence estimates showed any distinct pattern of change over time (data not shown).
This is the first comprehensive systematic and critical review on NP prevalence and we therefore provide reference data for future studies on NP. Although considerable heterogeneity in prevalence estimates was found, two trends are evident: first, the average NP prevalence estimates increase with longer prevalence periods; second, in nearly all studies women reported more NP than men.
Differences in prevalence estimates could be a result of several factors. First, wording of the questions and use of different manikins may affect the results [37, 49, 50]. In the majority of studies self-developed questionnaires were used and this may explain some of the observed variation in the prevalence estimates. Second, the anatomical definition varies between studies (i.e. including or excluding the shoulder region). Interestingly, no general differences between studies including or excluding the shoulder region were seen, and it has been questioned whether neck and neck/shoulder pain can actually be clearly distinguished from each other [53, 61]. Finally, methodological quality of a study may affect the outcome. Surprisingly, this was not the case in our review. In fact, two studies with very different quality scores showed roughly the same prevalence estimates [1, 59], indicating that estimates of NP prevalence could be regarded as independent of the quality of individual studies.
This lack of correlation between study–quality and outcome estimates may, however, be true for our scoring system only, but, since no evaluation of the appropriateness of quality criteria for epidemiological studies on the prevalence of musculoskeletal disorders such as NP exists, this remains unknown. Lebeouf-Yde and Lauritsen  developed a comprehensive set of quality criteria for studies assessing low-back pain prevalence, whereas Loney and Stratford  developed a less detailed scoring system for this purpose. However, no consensus exists as to what quality scoring system should be used when evaluating prevalence studies. It is our opinion that scoring systems should be kept as simple as possible and be easy to use. Our quality criteria cover basic issues that should be considered in any epidemiological study. Nevertheless, more work is needed in this area.
Previous attempts at reviewing the literature on NP prevalence also showed wide prevalence ranges [3, 21]. However, these reviews included very few papers. We included 56 papers due to a more comprehensive search strategy and different inclusion/exclusion criteria. More than one database should be included in a comprehensive search and one paper was not retrieved in any database. Therefore, screening of reference lists must be performed for complete retrieval of all relevant literature.
The results of this review should be interpreted in the light of several limitations. Our search strategy was broad and resulted in many potentially relevant papers. However, since most of the included studies had other primary objectives than to study NP prevalence, the broad search strategy was necessary in order to retrieve all relevant studies. We did not conduct a search of journals by hand, as prevalence studies are reported in many different journals. For example, the 56 included papers in our review were retrieved from 29 different journals.
Although no prior training with regards to applying the quality assessment was conducted, no substantial disagreements between the raters were seen for any of the papers. Any initial divergence was based merely on interpretations of the papers, and in fact the third reviewer was never involved in the quality assessment.
The mean prevalence estimates should be regarded with caution, as our review showed a great heterogeneity among the included studies. For example, the pain definition and duration of pain varied, typically by including the shoulder region or extending the duration of pain (i.e. pain lasting more than 3 months). Also, upper and lower age ranges were not identical between studies. For instance, some studies included 15-year-olds in the adult group, whereas others defined adulthood from 20 years of age and onward.
Reviewer bias cannot be ruled out, as we did not attempt to blind reviewers. This was because some studies were already known to the reviewers, hence making a fully blinded assessment was impossible.
We found that women consistently report more NP than men and this is in agreement with other reviews dealing with NP [3, 21]. In fact, women appear to report more musculoskeletal pain than men, and it has been suggested that this is based on different physiological mechanism for pain perception between the sexes . No other review has included studies specifically on children or elderly populations.
Homogeneity is a key issue when comparing results from different studies. Some of the requirements in studies on prevalence are: sufficiently large and unbiased study samples with non-responder analysis, uniform and valid anatomical definitions, and precise outcome measures. Our systematic review clearly indicates a lack of homogeneity in prevalence studies on NP. This is also apparent in other smaller reviews on NP prevalence where the variation in the definitions and the durations of NP [3, 21], as well as the sample source and age distribution have been emphasised as explanations for the large variation in the NP prevalence estimates .
Knowledge about prevalence does not in itself inform about the impact of NP on individuals and on society at large. To assess the impact on individuals, information such as pain intensity, influence on daily performance, general health status, care seeking, and co-morbidities should be available. To assess the impact of NP on society at large, both direct costs (e.g. use of medication and/or health care providers) and indirect costs (e.g. number of sick days, decreased daily performance) should be provided. Such clinically and sociologically relevant information is difficult to obtain, and this may in part explain the diversity in questionnaires used.
Time has come to reach a consensus on these issues and develop new standardised instruments that entail more relevant outcome measures based on our present knowledge. This would facilitate comparisons between different countries and cultures, and, most importantly, provide clinicians, researchers, and politicians with a relevant and detailed picture of both occurrence and impact of NP in the population.
NP is a common symptom in the population. The prevalence increases with longer prevalence periods and generally women have more NP than men. At least for 1-year prevalence, Scandinavian countries report higher mean estimates than in the rest of Europe and Asia. The quality of studies varies greatly but is not correlated with the prevalence estimates. Design varies considerably between studies and standardization is needed in future studies.