Article Text

Extended Report
The global burden of neck pain: estimates from the Global Burden of Disease 2010 study
  1. Damian Hoy1,
  2. Lyn March2,
  3. Anthony Woolf3,
  4. Fiona Blyth4,
  5. Peter Brooks5,
  6. Emma Smith6,
  7. Theo Vos7,
  8. Jan Barendregt8,
  9. Jed Blore9,
  10. Chris Murray10,
  11. Roy Burstein10,
  12. Rachelle Buchbinder11,12
  1. 1University of Queensland, Herston, Queensland, Australia
  2. 2Department of Rheumatology, University of Sydney Institute of Bone and Joint Research, Royal North Shore Hospital, St Leonards, New South Wales, Australia
  3. 3Department of Rheumatology, Royal Cornwall Hospital, Truro, UK
  4. 4School of Public Health, University of Sydney, Camperdown, New South Wales, Australia
  5. 5Australian Health Workforce Institute, University of Melbourne, Carlton, Victoria, Australia
  6. 6Northern Clinical School, Sydney Medical School, University of Sydney, Royal North Shore Hospital, St Leonards, New South Wales, Australia
  7. 7University of Queensland, School of Population Health, and Institute for Health Metrics and Evaluation, University of Washington, Seattle, Washington, USA
  8. 8School of Population Health, University of Queensland, Herston, Queensland, Australia
  9. 9Department of Epidemiology and Preventive Medicine, School of Public Health and Preventive Medicine, Monash University, Melbourne, Victoria, Australia
  10. 10Institute for Health Metrics and Evaluation, University of Washington, Seattle, Washington, USA
  11. 11Department of Epidemiology and Preventive Medicine, School of Public Health and Preventive Medicine, Monash University, Victoria, Australia
  12. 12Monash Department of Clinical Epidemiology, Cabrini Institute and Monash University, Melbourne, Victoria, Australia
  1. Correspondence to Dr Damian Hoy, University of Queensland, School of Population Health, Herston Rd, Herston, QLD 4006 Australia; damehoy{at}


Objective To estimate the global burden of neck pain.

Methods Neck pain was defined as pain in the neck with or without pain referred into one or both upper limbs that lasts for at least 1 day. Systematic reviews were performed of the prevalence, incidence, remission, duration and mortality risk of neck pain. Four levels of severity were identified for neck pain with and without arm pain, each with their own disability weights. A Bayesian meta-regression method was used to pool prevalence and derive missing age/sex/region/year values. The disability weights were applied to prevalence values to derive the overall disability of neck pain expressed as years lived with disability (YLDs). YLDs have the same value as disability-adjusted life years as there is no evidence of mortality associated with neck pain.

Results The global point prevalence of neck pain was 4.9% (95% CI 4.6 to 5.3). Disability-adjusted life years increased from 23.9 million (95% CI 16.5 to 33.1) in 1990 to 33.6 million (95% CI 23.5 to 46.5) in 2010. Out of all 291 conditions studied in the Global Burden of Disease 2010 Study, neck pain ranked 4th highest in terms of disability as measured by YLDs, and 21st in terms of overall burden.

Conclusions Neck pain is a common condition that causes substantial disability. With aging global populations, further research is urgently needed to better understand the predictors and clinical course of neck pain, as well as the ways in which neck pain can be prevented and better managed.

  • Outcomes research
  • Epidemiology
  • Health services research

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.


Neck pain occurs commonly throughout the world and causes substantial disability and economic cost.1 The pain and disability associated with neck pain have a large impact on individuals and their families, communities, healthcare systems and businesses.2–4 Economic consequences include the cost of healthcare, reduced work productivity, work absenteeism and insurance. As part of the Global Burden of Disease 2010 Study (GBD 2010), the global burden of musculoskeletal (MSK) conditions was estimated. Burden was expressed in disability-adjusted life years (DALYs).

This paper details the methods and results for estimating the global burden of neck pain for GBD 2010. It is part of a series of articles. The main overall articles for GBD 2010 were published in the Lancet,5–9 and the MSK-specific papers are published in this issue of Annals of Rheumatic Diseases.10–18 One of these papers give an in-depth description of the methods used for estimating the global burden of the MSK conditions15 and this should be read in conjunction with the current paper.


Figure 1 outlines the steps taken in estimating the burden of neck pain. The GBD Neck Pain Expert Group performed steps 1 to 3, and the GBD Core Team performed the remaining steps.

Figure 1

Steps taken in estimating the global burden of neck pain, GBD 2010.

Established case definition

The initial case definition for neck pain was: ‘Activity-limiting neck pain (± pain referred into the upper limb(s)) that lasts for at least 1 day’. The anatomical region of the neck was defined according to the recommendation by The Bone and Joint Decade 2000–2010 Task Force on Neck Pain and its Associated Disorders.3 ‘Neck or shoulder’ pain was assumed to be a proxy for ‘neck’ pain. For the final analysis, ‘activity-limiting’ was removed from the case definition because: (1) this provided a more robust analytical model given that relatively few data points from the systematic review conformed to the case definition of neck pain that was activity-limiting; and (2) this definition aligned better with the neck pain definition used in national health surveys that were included in the final analysis.

Established health states

A series of sequelae were developed to characterise the different levels of severity and take into account the variation in functional loss associated with acute and chronic neck pain with or without arm pain (table 1).19 Each sequela was defined in lay terms.

Table 1

Sequelae for neck pain in GBD 2010

Performed systematic reviews

The systematic reviews were described elsewhere20 ,21—see online supplementary material (appendix 1) for further detail. For incidence, a small number of studies were found, but all counted the number of people as the numerator rather than the number of incident episodes. This number could not be converted to episode incidence as no data were found on the average number of episodes a person with neck pain experiences over time. Thus, incidence could not be used as a parameter in the burden estimates.

For duration and remission, only one population-based study met inclusion criteria.22 However, data from this study could not be included because at baseline, subjects were asked if they had experienced neck pain in the past month, but no data were captured on cumulative duration at that point; and similarly at the 12 month follow-up survey, subjects were again asked if they had experienced neck pain in the past month, but it was unclear whether or not those that did report neck pain had had symptoms from baseline. For mortality, no consistent and conclusive evidence was found that neck pain is associated with an increased risk of mortality.

For prevalence, 94 published studies providing 447 age and/or sex-specific estimates were included. All included studies were assessed for risk of bias using a tool specifically developed for GBD 2010.21 High risk of bias estimates (n=189) and estimates with a prevalence period greater than 1 year (n=37) were excluded, leaving a total of 221 estimates from 58 studies. The data were then checked for outliers and three estimates (from two studies) were removed. One of these was the World Mental Health Survey estimate from Ukraine as its reported 1 year prevalence of chronic back or neck pain of 42.2% was three SDs higher than the mean of 18 World Mental Health Survey estimates while the next highest estimate was 1.2 SD units above the mean.23 The other study removed was an Australian study which was a postal survey with low response rate, and had an undue upward effect on the region's estimates.24 This left a total of 218 estimates from 56 studies.

There was substantial heterogeneity between studies with respect to prevalence period and case definition (ie, the minimum episode duration), anatomical location, and whether or not cases had to experience activity limitation. To make data points more comparable, adjustments were made in DisMod-MR, a Bayesian meta-regression tool developed for GBD 2010 by predicting the value of a data point as if the study had used the reference definition. To do so, DisMod-MR estimates coefficients for study-level covariates by comparing the values of prevalence measured by various methods in the global dataset. For the purpose of these analyses, it was necessary to reduce the number of categories of case definition and prevalence period. This was done by merging some of the categories on the basis of overlapping CIs or expert opinion (on the basis of proximity to overlapping CIs) for prevalence and/or regression coefficients. To determine how best to reduce the number of categories, a multivariate regression was done with prevalence (log transformed plus 0.2 to achieve normality) as the dependent variable and the following independent variables: age, sex, prevalence period, minimum episode duration, anatomical location, activity limitation, coverage, urbanity and risk of bias (see online supplementary material appendix 2).

Three groups were formed for prevalence recall period: (1) point; (2) short-term (1 week to 3 months); and (3) longer-term (6 months to 1 year). These comprehensively captured all variations of recall period. Two groups were formed for the anatomical case definition: (1) ‘neck’; and (2) ‘neck or arm pain’, ‘neck or upper thoracic pain’, and ‘neck or back pain’. For the minimum episode duration definition variations, four groups were formed: (1) one day; (2) ‘not specified’; (3) 1 week; and (4) 3 months and chronic. Note, the first category in each of the above groups was considered the reference category in DisMod-MR.

Established disability weights

Surveys were conducted in five countries for GBD 2010 and complemented by an open access internet survey; pairwise comparison questions were used, in which respondents were asked to indicate which of two health states presented as brief lay descriptions they considered ‘the healthier’. Results were used to derive disability weights (DWs).9

Added information from National Health Surveys

Additional information on prevalence of neck pain was derived from the US National Health Information surveys (2002–2009, 168 data points)25 and National Health and Nutrition Examination Survey (NHANES) (2009; 20 data points).26 Data from these surveys were not included in the neck pain prevalence systematic review as they did not fulfil our inclusion criteria at that time.

Bayesian meta-regression

DisMod-MR is a Bayesian meta-regression tool that has a number of functions, including: (1) pooling heterogeneous data and adjusting data for methodological differences; (2) checking data on incidence, prevalence, duration, remission and mortality risk for internal consistency; and (3) predicting values for countries and regions with little or no data using disease-relevant country characteristics and random effects for country, region and super-region. In the absence of usable incidence and remission data, a ‘prevalence-only’ model was run (see online supplementary material appendix 3).

Severity distribution

To estimate the distribution of neck pain cases across the GBD 2010 health states, the USA Medical Expenditure Panel Survey (MEPS) from 2000 to 2009 was used. This had information on the prevalence of 158 disorders included in the GBD as well as health status information provided by all individuals using the Short Form-12 (SF-12) questionnaire.27

In order to provide a translation of SF-12 values into a scale comparable with that used by the GBD 2010 DWs, the GBD Core Team conducted a small study on a convenience sample of respondents who were asked to fill in SF-12 to reflect 62 lay descriptions covering a wide range of severity that were used in the GBD DW surveys. With regression methods, the proportion of an individual's SF-12 score, translated into a GBD DW, that could be attributed to neck pain was calculated, while controlling for any other comorbid condition.

The MEPS DW values for all respondents reporting neck pain were grouped into four categories (table 2). The proportion of cases in MEPS that after comorbidity adjustment had no symptoms was ignored because the DisMod-MR results were for those with point prevalent neck pain. It was assumed that the 41% of cases without symptoms reflect people who had neck pain over the 1-year period of reporting in MEPS but who did not experience the pain at the time of responding to the health status questions in SF-12. The average DWs were calculated by age.

Table 2

The four sequelae categories for neck pain (with disability weights and proportional distributions), GBD 2010

Final burden estimates

The DALY is the standard metric used to quantify burden.28 DALYs are calculated by combining years of life lost due to premature mortality and years lived with disability (YLD). As no evidence for mortality from neck pain was found, YLDs and DALY estimates are the same. The average DW was multiplied by the age/sex/region-specific prevalence for the years 1990, 2005 and 2010 to derive YLDs. The uncertainty interval (UI) around each quantity of interest was calculated from SEs around all data inputs and the uncertainty from all steps of data manipulations, including the use of country and region fixed effects in DisMod-MR and the severity distributions. Uncertainty ranges are bounded by the 2.5 and 97.5 centile values. Further details on how uncertainty was calculated can be found elsewhere.5 Prevalence estimates were standardised using the 2001 WHO standard population.29

As disability weights were derived for single health states, simple addition of YLDs for all conditions would assume that disability is additive if a person has comorbid health states. Thus, a person with a number of more severe health states could be awarded a cumulative disability weight that exceeds 1, which equates to greater health loss than ‘being dead’. Assuming a multiplicative function between DWs for comorbid health states assures that a combined DW can never be greater than 1. To make a correction for comorbidity, hypothetical populations were simulated for each age, sex, country and year. Individuals in these hypothetical populations were assigned to have no, one or more health states based on the prevalence figures for each health state. The multiplicative function was applied to any individual with comorbid health states and the average DW for each component health state reduced proportionately. This allowed an estimate of the reduction in DW for any health state in an age and sex group by country and year: the comorbidity correction.


Description of included data

There were 406 data points included in the final DisMod-MR models. These were from 28 countries and 11 of the 21 GBD 2010 regions. The majority of studies used for these data included both sexes and a broad age range in the adult population.


The global age-standardised point prevalence of neck pain (from 0 year to 100 years of age) in 2010 was estimated to be 4.9% (95% UI: 4.6 to 5.3). It was higher in women (mean: 5.8%; 95% UI: 5.3 to 6.4) than in men (mean: 4.0%; 95% UI: 3.7 to 4.4). The age and sex distribution across regions was similar. DisMod-MR assumes a similar age pattern for all regions unless there are sufficient data points in a region to indicate a variation from the global age pattern. The large heterogeneity in the neck pain dataset meant that there was no departure from the default of a common age pattern (figure 2). Prevalence peaked at around 45 years of age.

Figure 2

DisMod-generated 2010 prevalence of neck pain by age, sex, year and region, GBD 2010.

Age-standardised prevalence in 2010 was highest in the North America high income region (mean: 6.5%; 95% UI: 5.6 to 7.5) followed by western Europe (mean: 6.3%; 95% UI: 5.8 to 6.8), and lowest in South Asia (mean: 3.3%; 95% UI: 2.8 to 4.0) followed by South-East Asia (mean: 3.8%; 95% UI: 3.4 to 4.3) (figure 2). Prevalence did not change significantly between 1990 and 2010.


Globally, and out of the 291 conditions studied, neck pain was ranked as the 4th greatest contributor to global disability (measured in YLDs), and the 21st in terms of overall burden (measured in DALYs)—tables 3 and 4. DALYs increased from 23.9 million (M) (95% UI: 16.5M to 33.1M) in 1990 to 33.6M (95% UI: 23.5M to 46.5M) in 2010. Over the 20-year period, population growth contributed 30% to this 47% increase in DALYs from neck pain, and population ageing the remaining 17%, while the epidemiological rates remained relatively constant. DALYs were higher in women (19.9M; 95% UI: 13.6M to 27.8M) than in men (13.7M; 95% UI: 9.4M to 18.9M). DALYs were highest in the 40–45 years age group.

Table 3

Age-standardised prevalence and DALYs (with 95% CIs) for neck pain in the age range 0–100 years, by region and sex, 2010, GBD 2010

Table 4

Regional neck pain YLD and DALY rankings in 2010 (out of 291 conditions), GBD 2010


Estimates of the global burden of neck pain

This is the first time the global burden of neck pain has been estimated. The methods were complex and the process took almost 6 years. The results show that the prevalence and burden from neck pain is high around the world. Out of the 291 conditions studied in GBD 2010, neck pain was found to rank 21st in terms of overall burden and 4th in terms of overall disability. In addition, the study has identified that neck pain prevalence peaks in the middle age groups. With improved child survival and aging populations throughout the world, especially in low-income and middle-income countries, the number of people experiencing neck pain will increase substantially over the coming decades.

Strengths and limitations

The greatest strength of the neck pain burden estimates in GBD 2010 is the extensive series of systematic reviews that were undertaken to obtain data for making the estimates and the large number of captured studies. Risk of bias was also considered and studies considered to be at high risk of bias were excluded from the analysis. Further strengths included: (1) the development of a new case definition and set of functional health states for neck pain; (2) the development of a set of disability weights for these health states, which were derived through community-based and health professional surveys in a number of countries; and (3) use of a new, more advanced version of DisMod that can (A) pool all data rather than rely on a ‘pick and choose’ method, (B) perform meta-regression to make data points from different studies more comparable, (C) use data to fill in missing information, and (D) carry forward uncertainty throughout the analysis.

Despite these strengths, there were some limitations. The functional domains in GBD 2010 refer to body functions and structures (eg, vision) as well as more complex human operations (eg, mobility), but they are not as inclusive as the WHO International Classification of Functioning, Disability and Health. They do not refer to broader aspects of life such as participation, well-being, carer burden, economic impact and burden of disease from the individual's perspective. It is important that burden of disease estimates are supplemented with this information to consider the full impact of a condition in a population.

Health state valuations are complex and can lead to very different answers. GBD 2010 endeavoured to respond to previous criticism that existing DWs were derived by a small group of international public health experts. Consequently, GBD 2010 undertook large-scale population and internet surveys to ask the general public to provide the health state valuations. That meant each health state had to be described in lay terms. Pilot testing revealed that descriptions of more than 30–40 words could not be absorbed. There will be ongoing work on DWs to explore what difference small changes in the wording of health state descriptions can make to the ultimate health state valuation.

Prevalence data for many countries and a number of regions were unable to be found. In GBD, no individual data point is adopted as reflecting the truth. Instead, statistical models are employed to all available data sources with corrections built in for risk of data bias in order to best predict true values of any epidemiological entity of interest. In regions with lots of data, estimates reflect the available data or at least find reasonable middle ground between what often are rather heterogeneous data sources. For regions with little or no data, the estimates will borrow strength from all other data and any available predictors of disease outcome. The level of uncertainty around each estimate reflects the strength of the underlying data to make a particular region's estimate.

It is a guiding principle in GBD to make estimates even if data are sparse. The alternative of not including some conditions that are known to be important contributors elsewhere is that it would give policy makers the impression that the condition is not important in their country. Also, in the studies that were captured, there was considerable methodological variation, especially relating to the prevalence period and case definition used making pooling of the data difficult. While ‘neck or shoulder’ pain was assumed to be a proxy for ‘neck’ pain, future GBD efforts may benefit from adjusting ‘neck or shoulder’ to ‘neck’ pain using the same Bayesian meta-regression method as for the other anatomical definition variations. In GBD 2010, the burden of neck pain may be overestimated due to the inclusion of shoulder pain.

Researchers are encouraged to adopt recent recommendations on defining neck pain in epidemiological studies to assist future reviews, enable comparisons between countries and improve our understanding of neck pain.3

While using the MEPS study had the advantage of estimating the distribution of severity while taking comorbidity into account, it also had limitations. There is likely to have been some level of recall bias despite there being five follow-up points over 2 years. Also, MEPS may not be representative of the health state experience for neck pain across the globe. In low-income and middle-income countries, where services for the prevention and management of neck pain are not as extensive as the USA, the health state experience could be different.

Suggested further research

There is clearly a need for further research on the natural history of neck pain. Long-term longitudinal population-based studies would provide important information on the natural history, average duration and disability associated with an episode of neck pain. Incorporating this research with pain diaries to track the daily patterns of pain and disability would add greater depth to this research. With expanding and aging populations in many low-income and middle-income countries, the burden from neck pain in these areas will grow significantly over coming decades. There is an urgent need to increase our understanding and attempt to mitigate the growing burden of neck pain in low/middle-income countries.


Neck pain is one of the main causes of disability throughout the world and requires greater attention from governments, health service providers and researchers. Further research is urgently needed to better understand the predictors and clinical course of neck pain across different settings, particularly in low-income and middle-income countries, and the ways in which neck pain can be prevented and better managed.


The authors are grateful for the co-operation of the following individuals who were kind enough to provide data upon request: Professors Tim Carey, Fereydoun Davatchi and Atiqul Haq, and Drs Arash Tehrani and Rowsan Ara. In addition, we are thankful to Dr Rungthip Puntumetakul, Ms Melinda Protani and Dr Rumna De in the testing of the risk of bias tool, and Ms Melinda Protani and Dr Rumna De for their assistance with data extraction. The MSK Neck Pain and Low Back Pain Expert Group comprised of the following individuals: Rachelle Buchbinder, Damian Hoy, Peter Brooks, Lyn March, Anthony Woolf and Fiona Blyth. The GBD Core Team members included on this paper are: Christopher Murray and Theo Vos.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Handling editor Tore K Kvien

  • Contributors All authors had substantial contribution to: conception and design, or analysis and interpretation of data; drafting the article or revising it critically for important intellectual content; and final approval of the version to be published.

  • Funding Supported by the Bill and Melinda Gates Foundation (to Dr Hoy, Dr Blore and Prof Vos), the Australian Commonwealth Department of Health and Ageing (to Dr Smith and Prof March), University of Sydney Institute of Bone and Joint Research (to Dr Smith), the Australian National Health and Medical Research Council (Postgraduate Scholarship 569772 to Dr Hoy and Practitioner Fellowships 334010 (2005–2009) and 606429 (2010–2014) to Prof Buchbinder) and the Ageing and Alzheimer's Research Foundation (Ass Prof Blyth).

  • Competing interests None.

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

Linked Articles