2.1 Design

The present study is a population-based prospective study with 17 years follow-up. The cohort consists of adult women. The women received a questionnaire by mail in 1990, 1995 and 2007. Two reminders were sent out on each occasion.

2.2 Materials

All women born between 1st January 1940 and 31st December 1969, and registered in Arendal municipality in 1989, a total of 2498 women, were included in the study at baseline. In 1995 and 2007, as many women as possible were retraced by the Norwegian National Registry. Of these, 1338 women responded on all three occasions and were defined as triple responders.

2.3 Screening questionnaire

The questionnaire consisted of two parts (Appendix A). Part one dealt with pain issues. There were four questions about pain and/or stiffness in muscles, joints, back and the whole body during three continuous months or longer the last year. A woman who gave at least one affirmative answer in the first part was classified as a positive responder and was assumed to suffer from CP, while a woman who gave no affirmative answer was classified as a negative responder and assumed to have no chronic pain (NP).

The second part consisted of 13 questions about NHC, sleep and pain-modulating factors that are often reported by individuals with CWP [3,6,12,13]. A time frame was not specified. Four questions were related to pain, of which three were about modulating factors for pain (physical activity, weather changes or feeling nervous, anxious, mentally stressed or depressed) and one about having pain at night. The remaining nine questions were about disrupted sleep, non-restorative sleep and seven NHC. The NHC were fatigue, feeling anxious, frightened or nervous, regular headaches, rumbling stomach, alternately hard/loose stools, numbness/tingling in the skin or joints/muscles feeling swollen. The questions could be answered with ‘yes’ or ‘no’. Those who failed to answer one or more questions about NHC or sleep problems as instructed were defined as non-completers.

2.4 Statistics

2.4.1 Samples and variables

The responders were distributed into three samples (Fig. 1); Sample 1: all pain free individuals in 1990 (n = 577). Sample 2: all individuals with any kind of CP in 1990 (n = 761). Sample 3: a subsample of sample 2 including only individuals with CRP (n = 607), and excluding 154 with CWP. Due to non-completers of the questionnaire, missing data appear in 12-15% in both univariate and multivariate analysis.

Fig. 1

Study population, response and samples of responders.

We have used four explanatory variables; age, NHC, sleep problems and pain modulating factors.

Age: The responders were categorized in three; 20-29 years, 30-39 years and 40-49 years.

NHC: There are seven questions about NHC on the questionnaire. They were analyzed separately in the univariate analysis to clarify if they were associated with CP or not. However, the questions are not sufficiently extensive to draw conclusions about the casual relationship between each symptom and CP. Thus, in the multivariate model, NHC are expressed as a continuous simple count variable (from 0 to 7).

Sleep problems: The questions about disrupted sleep and non-restorative sleep are presented as two dichotomous (yes/no) variables.

Modulating factors for pain: there were three questions about modulating factors for pain; physical activity, weather changes or feeling nervous, anxious, mentally stressed or depressed. In addition, there was one question about having pain during night. The variables are dichotomous.

3.1 Response rates, missing values and non-responders

In 1990, 2438 of 2498 women (82%) returned the questionnaire. In 1995, 1894 of 2435 (78%) responded. In November 2006, 2261 of the initial 2498 women were traced by the Norwegian National Registry. During the follow-up period 89 women had died, 36 had moved abroad and 112 were not found in the Registry. In total, 1654 (73%) returned the questionnaire; of these, 1338 responded for the third time and were defined as triple responders (Fig. 1). The triple responders were representative of the total responder group in age, pain status, sleep problems and health complaints. As the triple responders provide a virtually complete dataset at all three registrations, the following analyses and results refer to this group. The baseline data for the triple responders are presented in Table 2.

There were 460 non-responders in 1990. These individuals were somewhat younger than the whole group of responders at the same registration point (data not shown). Beyond that, we have no information on this group.

3.2 Prevalence, incidence and recovery of CP

The prevalence of CP in this population at the three registration points was 57% in 1990, 53% in 1995 and 61% in 2007. The corresponding numbers for CWP are 12%, 17% and 23%. CWP increased stepwise from the youngest to the oldest age group and did not seem to stabilize during the follow-up period (Figs. 2 and 3). It seems that the prevalence of chronic pain stabilizes from 50 years onward at about 70%.

Fig. 2

No pain, all kind of chronic pain and chronic widespread pain in different age groups in 1990.

Fig. 3

No pain, all kind of chronic pain and chronic widespread pain indifferent age groups in 2007.

During follow-up, 44% of the initially pain free women developed chronic pain. This corresponds to an annual incidence of 2.6%. At the last registration, 25% of the individuals who reported chronic pain at baseline reported no pain. This corresponds to an annual recovery rate of 1,5%.

3.3 Initially pain-free women (Sample 1); course and risk factors for onset of CP

Of the initially pain-free 577 triple responders, 56% were still pain free in 2007, while 34% reported CRP and 10% reported CWP (Fig. 4). There was a crude association between disrupted sleep, non-restorative sleep, fatigue, feeling anxious, frightened or nervous, regular headaches, rumbling stomach, alternately hard/loose stools, increasing number of NHC and development of CP (Table 3, first column). Age had no significant influence on development of CP in the univariate analyses. In the multivariate model, disrupted sleep and number of NHC remained significantly associated with development of CP, respectively ([OR 2.0 CI (1.1-3.5)]) and ([OR 1.5 CI(1.2-1.8)]) (Table 4). However, age seems to be a confounding factor for the effect of disrupted sleep on development of CP, yet not a very strong one.

Fig. 4

Report of no pain, chronic regional pain and chronic widespread pain in the triple responders (n =1338) in 1990 and 2007.

3.4 Women with initially CP (Sample 2); course and predictive factors for persistent pain

In 1990, 761 individuals reported CP. Of these, 25% were pain free in 2007, while 75% still had CP (Fig. 4). Each of the health complaints, disrupted sleep, non-restorative sleep, increasing number of NHC and age were significantly associated with persistence of CP in the univariate analyses (Table 3, second column).

As in analyses of Sample 1, only disrupted sleep ([OR 1.8 (1.1-2.7)]) and increasing number of NHC ([OR 1.2 (1.1-1.4)]), remained significantly associated with persistence of CP in the multivariate model (Table 4). Age seems to be a moderate confounding factor for the effect of disrupted sleep on the risk of having persistent pain.

Analysing the modulating factors for pain and having pain during night, we found that pain after physical activity ([OR 1.6 (1.0-2.4)]) and pain worsened by weather changes ([OR 1.9 (1.2-3.0)]) were associated with persistent CP. Feeling nervous, anxious, mentally stressed or depressed and having pain during night did not reach a significant level (Table 5).

3.5 Women initially reporting CRP (Sample 3); course and risk factors for developing CWP

As described in Section 2, this is a subsample from Sample 2. In 1990, 607 individuals reported CRP. Of these, 46% reported CRP again in 2007, while 25% had developed CWP (Fig. 4).

Disrupted sleep, feeling anxious, frightened or nervous, rumbling stomach, numbness/tingling in skin/muscles and joints/muscles feeling swollen were significantly associated with increased extent of pain in the univariate analyses (Table 3, third column).

Disrupted sleep ([OR 2.1 (1.2-3.4)]) and increasing number of NHC ([OR 1.3 (1.1-1.5)]) remained significantly associated with increasing extent of pain in the multivariate model (Table 4). In this sample, age does not seem to act as a confounder.

Analysing the modulating factors for pain and having pain during night, we found that pain during night turned out to be a confounding factor as it increased the risk of worsening of pain ([OR 1.9 (1.0-3.4)]) (Table 5) and sleeping problems reached a slightly lower risk estimate than in the other analysis ([OR 1.6 (0.9-2.8)]).

A multivariate model with an interaction term between age and disrupted sleep were used in all three samples. None of the interaction terms are statistical significant, and age does not seem to be an effect modifier (data not shown).

4.1 Prevalence of CP

At all three registrations the prevalence of self-reported CP was higher than 50%. This prevalence is among the highest presented in European studies [1,2,3,4,5,6,10]. There are several factors that can contribute to this.

We have used a time frame of one year for reporting CP. The same definition has been used in a Swedish [2] and a Norwegian study [5]. In the Swedish study, a prevalence of CP of 35% was found, while the prevalence of CP in the Norwegian study was reported to be 45% and 48% at two different registration points. Two other studies asked for persistent pain the last three months ahead of the distribution of the questionnaires. The prevalence found in these studies was 45 and 54% [4,7]. Based on these figures, we assume that the time frame did not affect the results significantly. However, Breivik et al. used a stricter definition requiring among others >6 months duration of pain and intensity of 5 or more on a scale from 1 (no pain) to 10 (worst pain imaginable), and probably this resulted in a lower prevalence [10].

The present study is performed on a population consisting solely of females aged 20 50 years at baseline and 37–67 years at the last registration. The prevalence of CP, and especially CWP, is higher in females than in men, and the prevalence of CP tends to increase until the age of 65 years [5,7,9]. This may result in a higher prevalence of CP in this population compared to mixed populations aged 20 74 years which are the most commonly used study populations.

The prevalence of CP increased by 4.3% from 1990 to 2007. However, due to the likely magnitude of error introduced by under- and over-reporting (Table 1), we conclude that the prevalence is rather stable throughout the follow-up period. On the other hand there was a marked increase in women with CWP, especially in the older age group. Thus, the most prominent change in this population was that the pain problem intensified with respect to pain extent.

4.2 Risk factors

We have studied age, health complaints and sleep problems as possible risk factors for onset of and for different courses of CP.

It is known that prevalence of CP is higher in older age groups [1]. Our results indicate that increasing age is not a risk factor for onset of pain. However, age turned out to be a risk factor for CRP developing to CWP and it was a moderate confounder on the effect of disrupted sleep. There was no indication of age being an effect modifier on the effect of sleep.

Health complaints were reported significantly more often at baseline by pain-free women who developed CP during follow-up than by those who remained pain free, and by women who had an unfavorable course of their pain condition. Similar results are reported by others [12,21]. In this study, one question covered each complaint (Appendix A). This is not sufficient to draw conclusions with regards to the importance of each complaint being an independent risk factor for development, persistence or worsening of pain. The number of health complaints however, was significantly associated with both development of CP and a negative course of CP.

Disrupted sleep was an independent risk factor for development of chronic pain in initially pain free women in this study (OR 2.0, 95% CI 1.1-3.5). Of the pain-free women who reported sleep problems in 1990, 62% had developed CP in 2007. It has been assumed that there is a correlation between CP and sleep problems [22,23,24]. Experimental studies have supported the assumption that there is a bidirectional relationship between sleep disturbances and pain [25,26,27,28].

There are a few prospective studies in which sleep problems have been studied as a possible risk factor for the development of pain [29,30]. A 14-year follow-up study from Norway concluded that sleep problems, in addition to sex, age, education and number of painful sites at baseline, predicted the number of painful sites at the follow-up [31]. CWP is a main part of the fibromyalgia syndrome. A newly published study documented that sleep problems in pain free individuals is a risk factor for development of fibromyalgia [32]. Results from the present study support the assumption that the quality of sleep plays an important role in development of CP.

Sleep problems were also associated with persistence of CP compared to becoming pain free and also with worsening of pain. However, when pain during night was added to the model, we found that there were some confounding between pain during night and sleep problems in the analyses concerning worsening of pain. The risk estimate for sleep problems was however, only slightly reduced. In the analyses concerning persistence of pain, there was no confounding between these two variables.

As many as 53% of the women reported changes in pain severity. Most individuals moved only one category (from NP to CRP or from CRP to CWP or vice versa). Among women with CP, CWP was the most stable condition; 79% of those who reported CWP in 1990 and 1995 still had CWP in 2007 while only 4% had become negative responders. This is in contrast to individuals with CRP; of those who reported CRP in both 1990 and 1995, 24% were pain free in 2007. These findings indicate that an increased pain extent is a risk factor for persistent pain in addition to pain per ce. Our results support findings from earlier studies [2,31].

4.3 Common considerations

Our findings of an association between increasing number of NHC and sleep problems and both development and worsening of CP may indicate related pathogenetic mechanisms for these conditions. However, CP is a complex condition and it may co-exist with other conditions like depression [33]. Depression also includes NHC and sleep problems. There are significant deficiencies in the understanding of the relationship between pain, NHC, sleep problems and depression. It is suggested that CP and depression share some pathological pathways, and that this may be the explanation to the similarities in complaints [34,35].

In the present study, the screening questionnaire did not comprise questions about self-reported depression, and thus, we cannot outline if depression preceded pain. However, we believe that our finding of an association between NHC, sleep problems and CP is valuable in identifying individuals with an increased risk of developing CP.

4.4 Considerations concerning registration of pain

Our use of the term “CWP” in results from the questionnaire based survey can be criticized as we have not used a mannequin or asked detailed questions about localizations of pain. However, results from the comparison of answers given on the questionnaire with answers given in the interview showed that the mean number of painful sites reported by women classified as having CRP, was 2.55. Referring to the ACR-90 criterion, the smallest possible number of painful sites in individuals with CWP is three. This supports our assumption that individuals classified as having CRP in this study actually have localized pain. The mean number of painful sites reported by individuals we classified as having CWP was 6.25, which reflects a marked increase in pain extent compared to CRP, and referring to the ACR-90 criterion of widespread pain, we decided to use the term CWP.

In questionnaire-based surveys, there is a risk of under- and over-reporting the degree of pain. In 1990, 21% of the women over-reported pain. In 2007, 20% of the negative responders underreported pain and 20% of the positive responders over-reported pain (Table 1). This illustrates that questionnaire-based surveys about CP must be interpreted with caution.

Another source of error may be that the non-responders did not match the responder group. The non-responders from 1990 were somewhat younger than the rest of the cohort. Restricting the analysis to the triple responders may result in selection bias. However, the triple responder group equalizes the total responder group, and we therefore believe that the selection of triple responders will not affect the results.

4.5 Considerations concerning registration of risk factors

The questions about the different health complaints and sleep problems are rather simple and this simplicity might be criticized [24]. As the questionnaire opted only for ‘yes’ and ‘no’ answers, this must be seen as coarse screening. On the other hand, related to clinical practice where time for each patient is often limited, meaningful screening questions can be of importance. A time frame was not stated for these questions. The questions were asked in the present tense. We assumed that the individuals answered the questions about health complaints on the basis of their experience of their general health.