The age- and sex-specific occurrence of bothersome neck pain in the general population – results from the Stockholm public health cohort

The Stockholm Public Health Cohort is a recently established prospective study, set within the framework of the Stockholm County Council Public Health Surveys with the aim of collecting data for regional public and occupational health surveillance in the general population [12]. Between October 29^th in 2002 and March 22^nd in 2003, 49,914 individuals (age 18–84 years) were randomly selected after stratification for municipality from 1.4 million eligible residents in Stockholm County (the source population), and sent an extensive postal questionnaire (Figure 1). Responders (n = 31,182; 62%) were resurveyed in 2007 via an extensive mixed mode postal/web-based questionnaire. Participants in both surveys (n = 23,794, corresponding to a 76% retention rate) constitute the study population used for the analyses in this study. A comparison between the study population in this study and Stockholm County census data shows that men, those under the age of 45, those born outside Sweden, and those unemployed or in the lowest quartile of income, were more likely to be non-responders to the surveys (unpublished data). To describe the characteristics of the participants in the cohort, we used information from the baseline questionnaire in 2002/2003. To describe the occurrence of BNP, retrospective information from the follow-up questionnaire in 2007 was used. The most valid information on the occurrence of BNP was judged to be from 2006 since that was the last full year before the questionnaire was answered in 2007. We used information from 2005 to study the incidence proportion.

We measured BNP in 2005 and 2006 using retrospective questions in the follow-up questionnaire sent out in 2007. The exact wording of the questions was: 1) “During the last five-year period, have you had neck pain for at least three consecutive months that has bothered you considerably?”. 2) “If yes, which year/years did you have such pain? a. 2002, b. 2003, c. 2004, d. 2005, e. 2006, f. 2007”. 3) “During the last five-year period, have you had neck pain, for at least seven consecutive days but less than three consecutive months, that has bothered you considerably?” 4) “If yes, which year/years did you have such pain? a. 2002, b. 2003, c. 2004, d. 2005, e. 2006, f. 2007”. The full questionnaire used in the survey is attached in the Appendix 1.

Our case definitions are accordingly survey-based bothersome neck pain (BNP) of short or long duration in the general population. This is based on the classification proposed by the Bone and Joint Decade 2000–2010 Task Force on Neck Pain regarding Axis I (source of subjects), Axis II (setting and sampling) and Axis IV (duration) [13]. To describe the duration of the NP we used the terminology short duration (at least seven consecutive days but less than three consecutive months) or long duration (at least three consecutive months) as suggested by the Bone and Joint Decade 2000–2010 Task Force on Neck Pain and Its Associated Disorders [13, 14]. To describe the severity of the NP (Axis III) we used the word bothersome in the questionnaire, which has been shown to be valid to measure severity of low back pain, and to identify people in the highest category of low back pain and disability [15].

We defined five sub-cohorts of the Stockholm Public Health Cohort for the analyses in this study, based on the retrospective questions in the follow-up questionnaire described above.

How we selected study participants in the cohort is also described in Figure 1.

We considered the sample of the Stockholm County population used in this study to be an actual random sample, since weighted analyses according to the stratified sampling scheme showed only negligible differences in the results.

Figure 1 and Table 1 provide a detailed description of the disorder estimates (the prevalence, the incidence proportion and the incidence proportion of recovery) and give the cohorts used for the analyses.

In a systematic review, the one-year prevalence of activity-limiting NP in the general population was 1.7%-11.5% [2] referring to one survey among Hong Kong residents [17], and one survey among patients from British general practices [18]. This is lower than in our study and might be explained by a higher prevalence in Stockholm County [19], or by the fact that the definition of NP and recall periods differs across the surveys. Another explanation might be differences in response rate in different surveys. When comparing our results to other Swedish surveys [5, 12] and a Danish survey [7], our results regarding prevalence are similar.

The oldest individuals in our cohort (70–84 years) had a one-year prevalence of short-and/or long-duration BNP of 16% among women and 11% among men, clearly lower than in the younger age groups. In Denmark, a higher one-month prevalence among 100-year-olds was found; 23% and 19% for women and men respectively [6]. This difference might be due to a more stringent definition of NP in our survey or that the prevalence is higher in the oldest age group, which is not included in our study. Differences in recall periods in the surveys might be another explanation. In another study, getting older did not increase the NP burden even though pain was found to persist longer in older groups [7]. Neither short- nor long-duration BNP was more common among the elderly than among other age groups in our study.

According to a recently published review, the one-year incidence of NP ranges between 10 and 21% [1]. The one-year incidence in our study was lower (4% for men and 7% for women), reflecting the more important NP surveyed in our study. By assuming that all individuals in the cohort free from BNP in 2005 contributed one year at risk to develop BNP in 2006, and that those reporting NP in 2006 had only one episode each, we calculated the incidence rate of BNP to be 44 per 1,000 person-years among men and 68 per 1,000 person-years among women. According to a systematic review, the incidence rate per 1,000 person-years ranges from 0.055 (for disc herniation with radiculopathy) to 179 (for self-reported NP) [2].

Our results indicate that younger age predicts a better course for men as well as for women. We have not found sex-specific estimates of the prognostic effect of age in the literature, but in the general population as a whole, younger age was reported to be prognostic of better recovery [11].

The evidence for the prognostic role of sex in NP outcome varies across studies [11]. We found that the prognosis was worse for women than for men.

Our case definitions included information on four of the five axes in the classification proposed by the Bone and Joint Decade 2000–2010 Task Force on Neck Pain for enhancing comparisons between studies [13]. Our information was survey-based (Axis I) and referred to the general population (Axis II). We had some information on severity (Axis III) and we had full information on duration of the episodes (Axis IV). Like many other studies, we lacked information on each individual’s pattern of NP (Axis V), that is, we could not differentiate between single episodes, recurrent, and persistent pain. Accordingly we could not base our calculations of incidence proportions on those free of pain at baseline, but had to use the more demanding criterion of study participants being free from pain during the entire preceding year. Most probably this implies an underestimation of the true incidence proportion. In our opinion, the survey questions allowed us to arrive at a fairly reasonable estimate of the one-year period prevalence, although they do not formally meet the textbook definition of period prevalence (i.e. the sum of the number of cases prevalent at a point in time (prevalence) plus the number that occurs in a subsequent period (e.g. the subsequent year) [16].

The lack of information on the true number of episodes and their specific durations in our study also implies that only very crude estimates of incidence and recovery rate may be reported. However, the pattern of NP over time is usually not measured in detail in surveys, due to economical and practical reasons. We had to define recovery as at least one-year duration of no BNP, which is an arbitrary definition that may be uninformative especially if an individual case of BNP is recurrent with longer time intervals.

Crombie et al. suggest that a pain condition, such as NP is to be regarded as a dynamic process, since pain is characterized more by change than by stability [20], and that transitions between different NP states and changes in pain intensity and severity are to be expected. These limitations are the same as in most other studies in this area. We strongly believe that there is room for improvements in the musculoskeletal epidemiological research area, not only regarding the case definitions, but also regarding more details in the reporting of prevalence and incidence.

A major strength of this study is the large population-based sample, enabling analyses in subgroups of age and sex. In addition, we study NP that is bothersome, which is of high importance for the individual as well as for the society.

The main threat to the validity of our findings is the risk of bias in the one- to two-year retrospective recall of BNP from 2007 to 2006 and 2005 resulting in a misclassification of disease. Individuals with long-duration BNP might have overestimated the severity of prior pain periods, and individuals with recurrent pain might have forgotten about the severity of the pain in earlier episodes. The design of the cohort is prospective but since the follow-up time was five years, we used information about BNP from the follow-up questionnaire to report on the occurrence. The most valid information on BNP was judged to be from 2006 since that was the last full year before the questionnaire was answered in 2007. Nevertheless, additional analyses showed that the prevalence of BNP was two to three percentages points lower in 2005 than in 2006, indicating an underestimation of the prevalence that may have biased estimates of incidence proportion of getting BNP (overestimation) and incidence proportion of recovery (underestimation) between 2005 and 2006.

Another threat to the validity is the single generic question about BNP used on the questionnaire. It is probable that some individuals did not read the question carefully enough, and therefore have reported NP of less magnitude. This might overestimate the burden of BNP. Further the question used to define BNP are not validated to capture important neck pain that relates to disability, treatment requirements, sick leave and all associated disability, etc., even though the question has been shown to be valid for capturing important low back pain [15]. If it is not valid for BNP, our results might be biased and the neck pain reported on might be less important than stated.

The study population constitutes 48% of the 49,914 persons asked to take part in the baseline survey, and there were some internal dropouts, which threatens the external validity of the results. Since a higher proportion of the non-respondents were men under the age of 45, our results may be biased with a probable overestimation of the overall occurrence of BNP. The analyses of the smallest subgroups (e.g. in the youngest and oldest) were somewhat hampered by the lack of statistical power and should hence be interpreted with caution.

In summary we consider our results to be valid and that the study provides important information about age and sex differences relating to prevalence, incidence proportion and prognosis of BNP.