SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. What's already known about this topic?
  4. What does this study add?
  5. 1. Introduction
  6. 2. Method
  7. 3. Results
  8. 4. Discussion
  9. 5. Conclusion
  10. Author contributions
  11. Acknowledgements
  12. References

Background

Pain from the musculoskeletal system often occurs in more than one site. This appears to affect prognosis negatively. Knowledge about specific pain patterns is lacking.

Objectives

To define specific patterns of musculoskeletal co-complaints occurring alongside a primary musculoskeletal complaint.

Methods

Using data from an interview-based health survey of a nationally representative sample of the adult Danish population in 1991 (n = 4817), we describe the co-occurrence of musculoskeletal complaints. Using latent class analysis, we identify clusters of musculoskeletal complaints.

Results

Forty percent reported a complaint during a 2-week period; the most common being the low back, neck, shoulder, and knee, and 40% reported more than one complaint. Two latent classes were found for each of the nine primary pain sites except for the low back where three were found. For all primary pain areas, the largest class had site-specific pain only, whereas the smallest class had diffuse pain covering large parts of the body. For participants with a primary musculoskeletal complaint in the spine, the highest probabilities for co-complaints were at other sites in the spine. For primary complaints in the extremities, co-complaints occurred most commonly at adjacent areas. One noticeable exception was a primary complaint of knee pain where co-complaints were found in more remote areas as the neck and low back.

Conclusions

Unique clusters of musculoskeletal co-complaints can be determined based on primary pain site. These patterns are different for persons with a primary complaint in the spine compared with persons with a primary complaint in the extremities.


What's already known about this topic?

  1. Top of page
  2. Abstract
  3. What's already known about this topic?
  4. What does this study add?
  5. 1. Introduction
  6. 2. Method
  7. 3. Results
  8. 4. Discussion
  9. 5. Conclusion
  10. Author contributions
  11. Acknowledgements
  12. References
  • Pain in the musculoskeletal system often manifests itself in more than one anatomical site.
  • Pain in more than one site affects prognosis negatively.

What does this study add?

  1. Top of page
  2. Abstract
  3. What's already known about this topic?
  4. What does this study add?
  5. 1. Introduction
  6. 2. Method
  7. 3. Results
  8. 4. Discussion
  9. 5. Conclusion
  10. Author contributions
  11. Acknowledgements
  12. References
  • In spite of having pain in multiple sites, participants were able to identify a site of primary complaint.
  • Based on this primary pain site, unique clusters of musculoskeletal co-complaints can be determined.
  • These patterns are different for persons with a primary complaint in the spine compared with persons with a primary complaint in the extremities.

1. Introduction

  1. Top of page
  2. Abstract
  3. What's already known about this topic?
  4. What does this study add?
  5. 1. Introduction
  6. 2. Method
  7. 3. Results
  8. 4. Discussion
  9. 5. Conclusion
  10. Author contributions
  11. Acknowledgements
  12. References

Pain and disability from the musculoskeletal (MSK) system is a public health problem all over the world, resulting in suffering of millions of people and exorbitant costs to societies (Maniadakis and Gray, 2000; Kotlarz et al., 2009; Wikman et al., 2011). Over the past decades, it has become apparent that often, MSK pain occurs in more than one site in an individual, and therefore traditional approaches in research where MSK pain has been studied as a narrow site-specific problem are failing (Macfarlane, 1999; Birrell, 2004; Peat et al., 2006; Kamaleri et al., 2008). This is important because local MSK pain is more severe and disabling if pain is also present in other body regions (Peat et al., 2006). Furthermore, an increasing number of pain sites magnify the risk of a poor prognosis and poor overall function, poor fitness, negative emotions, and it has a negative impact on both daily physical and social activities (Kamaleri et al., 2008). In fact, the risk of an attack of low back pain (LBP) becoming chronic increases sixfold if the person is also suffering from pain at other sites in addition to the back (Thomas et al., 1999), and Dunn et al. (2011a) found that having pain in the upper body alongside having pain in the low back nearly quadrupled the risk of a poor prognosis 1 year later in primary care back pain patients. Office workers suffering from neck pain have greater pain intensity if they also have pain in other body parts (Andersen et al., 2010). The same is true for knee pain where the presence of pain in other areas such as the foot, low back or elbow negatively affects scores for knee pain and disability (Suri et al., 2010). Finally, in intervention studies, patients who have undergone total knee arthroplasty do worse if they are also suffering from LBP (Escobar et al., 2007; Novicoff et al., 2009).

These pain patterns thus appear to have an impact on prognosis for individuals with MSK disease, and it is therefore of interest to study this comprehensively at the population level in order to reliably determine just how common these patterns of co-occurring MSK pain are and exactly what they look like. Traditionally, the distribution of MSK pain has been studied by mapping the prevalence of co-occurring pain, disregarding which pain sites are most important to the individual. However, because MSK symptoms are so common, no specific patterns, other than generalized pain, have emerged in previous research. In addition, there is a high risk of misclassification, i.e., symptoms occurring simultaneously by coincidence and not because they are part of a pattern. If, on the other hand, specific and mutually exclusive patterns based on primary complaint can be identified, such clusters can form the basis for follow-up studies of prognosis or treatment effect modification to determine their clinical relevance. Thereby, such analytic approaches are useful in classifying heterogeneous populations into unobserved latent classes (Adamson et al., 2007) and patterns of pain have been shown to better predict physical and social functions than number of pain sites (Schmidt and Baumeister, 2007). To our knowledge, descriptive and analytic studies comprising whole body assessment of MSK pain in a large population-based adult sample have never been published.

The overall aim of this study is therefore to (1) describe MSK complaints and co-occurring MSK complaints in a large population-based sample and (2) determine and compare specific patterns of co-occurring MSK in groups defined by primary complaints.

2. Method

  1. Top of page
  2. Abstract
  3. What's already known about this topic?
  4. What does this study add?
  5. 1. Introduction
  6. 2. Method
  7. 3. Results
  8. 4. Discussion
  9. 5. Conclusion
  10. Author contributions
  11. Acknowledgements
  12. References

2.1 Setting and study sample

Since 1987, the National Institute of Public Health in Denmark has conducted nationally representative surveys of the adult Danish population aged 16 years or older (Ekholm et al., 2009). Using the Danish Civil Registration System, 5986 persons chosen at random from the adult Danish population were invited to a MSK health survey in 1991.

2.2 Variables

In the current analysis, the following MSK variables were used:

A MSK complaint: Have you during the past 2 weeks had trouble (pain or discomfort) in any of these locations: neck, shoulders, upper back, elbows, low back, hands/wrist, hip, knee, feet/ankles, have not had any trouble. To aid the respondent in answering this question, the interviewer showed a diagram with the anatomical areas clearly shaded and labeled. Each person could respond ‘yes’ or ‘no’ to any of the complaints, i.e., more than one confirmatory answer was allowed.

A primary MSK complaint was determined with the following question: If ‘yes’ was answered to any of the above, in which area have you had the most trouble? This question had the same response options and again the respondent was shown the diagram. However, the person had to choose only one area as being the primary area.

A MSK co-complaint was defined as any other MSK complaint that occurred alongside the primary MSK complaint identified as above.

Age and gender were derived from the Central Person Register number.

2.3 Analysis

Data have been carefully validated, cross-checked and cleaned (Ekholm et al., 2009).

First, we performed a descriptive analysis. The overall prevalence of MSK complaints and the prevalence of primary MSK complaints in the different regions were tabulated by age (16–44 and 45+) and gender. Then, for each primary MSK complaint, the proportion of MSK co-complaints (i.e., MSK complaints that were not primary) was calculated.

Second, for each primary MSK complaint, we used latent class analysis (LCA) to identify latent classes of MSK co-complaints. LCA is built on the assumption that the association between MSK co-complaints is explained by an underlying variable with an unknown number of classes known as latent classes. The optimal number of classes can be determined using goodness-of-fit tests, which, by definition, must be as small as possible (Lanza et al., 2011). We used Consistent Akaike's Information Criterion (CAIC) and Bayes Information Criterion (BIC), which have been recommended to determine the optimal number of classes in situations such as this (Lin and Dayton, 2007). We gauged the appropriateness of the final LCA model by calculating the average posterior probability for each derived class as recommended by Nagin (2005). For a given class, the estimated posterior probability for being in that class should be high while the probability of being in any other class hence should be low.

To determine the complaint profile for a given class, we also used estimated class-specific posterior probabilities for having each MSK co-complaint given membership of that class. We arbitrarily divided probabilities for MSK co-complaints into quartiles so that they were considered to be very high for probabilities greater than 0.75, high for probabilities between 0.50 and 0.75, moderate for probabilities between 0.25 and 0.50, and low if the probability was below 0.25. We tested for homogeneity of class-specific parameters across gender and age groups (16–44 and 45+ years).

All analyses were performed using PROC LCA (Lin and Dayton, 2007) software and SAS version 9 (Cary, NC, USA).

3. Results

  1. Top of page
  2. Abstract
  3. What's already known about this topic?
  4. What does this study add?
  5. 1. Introduction
  6. 2. Method
  7. 3. Results
  8. 4. Discussion
  9. 5. Conclusion
  10. Author contributions
  11. Acknowledgements
  12. References

3.1 Descriptive analysis

Of the 5986 invited, 4817 (80.5%) participated in the interview-based survey. Of the 4817 respondents, 2836 (58.9%) reported having no complaints during the previous 2 weeks and 20 (0.4%) reported at least one complaint but did not report which one was primary.

The most common sites for MSK complaints were the low back, neck, shoulder, and knee, and these were also the most common primary complaint sites (Table 1). Overall, and as expected, a higher proportion of women reported MSK pain compared with men regardless of site with the exception of knee and ankle/foot pain in the younger age group, and a greater proportion of the 45+ age group reported MSK pain compared with the 16–44 age group (Table 1). Sixty percent of participants reporting a primary MSK complaint reported no MSK co-complaints beyond the primary. For the remaining 40%, 20.3% of these reported one MSK co-complaint, 10.5% reported two MSK co-complaints, 5% reported three MSK co-complaints, 2.3% reported four MSK co-complaints, 1.2% reported five MSK co-complaints, and < 1% reported six to nine MSK co-complaints. Women, but not men, reported a higher number of MSK co-complaints with increasing age (data not shown).

Table 1. Prevalence of musculoskeletal complaints and primary musculoskeletal pain during the previous 14 days, n (%), in a nationally representative sample of the Danish population aged 16 and older
 Overall prevalenceaPrevalence of primary paina
♂ + ♀♂ + ♀
16–44b45+16–4445+16+16–4445+16–4445+16+
1298c100713351177481712981007133511774817
  1. a

    All % are calculated based on the number of respondents (n) in age/gender group.

  2. b

    Age.

  3. c

    n.

No pain during the past 14 days864 (66.6)612 (60.7)798 (59.8)582 (49.4)2,856 (59.3)     
Neck127 (9.8)127 (12.6)246 (18.4)242 (20.5)742 (15.4)72 (5.6)63 (6.3)139 (10.4)105 (8.9)379 (7.9)
Upper back64 (4.9)55 (5.5)91 (6.8)96 (8.2)306 (6.4)28 (2.2)18 (1.8)37 (2.8)22 (1.9)105 (2.1)
Low back181 (13.9)169 (16.8)216 (16.2)283 (24.0)849 (17.6)134 (10.3)123 (12.2)157 (11.8)170 (14.4)584 (12.1)
Shoulder108 (8.3)111 (11.0)169 (12.7)227 (19.3)615 (12.8)44 (3.4)33 (3.3)65 (4.9)82 (6.9)224 (4.7)
Elbow21 (1.6)33 (3.3)23 (1.7)58 (4.9)135 (2.8)12 (0.9)9 (0.9)11 (0.8)12 (1.0)44 (1.0)
Wrist/hand36 (2.8)47 (4.7)70 (5.2)145 (12.3)298 (6.2)23 (1.8)20 (1.9)44 (3.3)43 (3.7)130 (2.7)
Hip22 (1.7)74 (7.4)34 (2.6)128 (10.9)258 (5.4)13 (1.0)32 (3.2)18 (1.4)48 (4.1)111 (2.3)
Knee101 (7.8)109 (10.8)78 (5.8)185 (15.7)473 (9.8)67 (5.2)60 (5.9)45 (3.4)73 (6.2)245 (5.1)
Ankle/foot58 (4.5)79 (7.9)42 (3.2)128 (10.9)307 (6.4)41 (3.2)37 (3.7)21 (1.6)40 (3.4)139 (2.9)

Results in Table 2 are based on the 1961 persons (40.7% of sample) who reported a primary MSK complaint during the previous 2 weeks and illustrate the commonness of co-complaints in persons with a MSK complaint.

Table 2. Musculoskeletal co-complaints by musculoskeletal primary complaints during the previous 14 days for 1961 Danes aged 16+
 Primary complaint
NeckUpper backLow backShoulderElbowWrist/handHipKneeAnkle/foot
379a10558422444130111245139
  1. a

    n.

  2. b

    n (%).

Neck 31 (29.5)b133 (22.8)92 (41.1)7 (15.9)33 (25.4)16 (14.4)35 (14.3)10 (7.2)
Upper back50 (13.2) 70 (12.0)32 (14.3)5 (11.4)18 (13.8)8 (7.2)11 (4.5)6 (4.3)
Low back76 (20.1)35 (33.3) 42 (18.8)6 (13.6)21 (16.2)32 (28.8)32 (13.1)23 (16.5)
Shoulder164 (43.3)28 (26.7)111 (19.0) 5 (11.4)29 (22.3)14 (12.6)25 (10.2)11 (7.9)
Elbow15 (4.0)5 (4.8)25 (4.3)17 (7.6) 13 (10.0)5 (4.5)6 (2.4)3 (2.2)
Wrist/hand29 (7.7)12 (11.4)47 (8.0)24 (10.7)7 (15.9) 15 (13.5)19 (7.8)15 (10.8)
Hip14 (3.7)8 (7.6)65 (11.1)17 (7.6)3 (6.8)10 (7.7) 20 (8.2)8 (5.8)
Knee30 (7.9)15 (14.3)71 (12.2)26 (11.6)7 (15.9)17 (13.1)29 (26.1) 31 (22.3)
Ankle/foot23 (6.1)12 (11.4)49 (8.4)16 (7.1)3 (6.8)19 (14.4)16 (14.4)27 (11.0) 

3.2 Latent class analysis

Results of the LCA showed that for all primary MSK pain sites except LBP, two latent classes provided the best fit for the data. Average posterior probabilities were 0.90 or higher for all latent classes, except class 1 for primary neck pain, where it was 0.84 and thus indicating a low chance of misclassification. Shown in Table 3 are posterior probabilities for MSK co-complaint for the four most common primary MSK complaints, and in Table 4 for the five least common primary MSK complaints. Fig. 1 illustrates pain patterns for all latent classes.

figure

Figure 1. Patterns of musculoskeletal co-complaints for latent classes of musculoskeletal primary complaints in a population-based sample of Danes aged 16+. Shading indicates class-specific posterior probabilities for musculoskeletal co-complaint. *Conditional probability for proportion of participants in class.

Download figure to PowerPoint

Table 3. Latent class-specific posterior probability of musculoskeletal co-complaints for the four most common primary musculoskeletal pain sites in the Danish population aged 16+
 NeckLow backShoulderKnee
1a (0.89)2 (0.11)1 (0.81)2 (0.12)3 (0.08)1 (0.81)2 (0.19)1 (0.88)2 (0.12)
  1. a

    Latent class and conditional probability for proportion of participants in class.

  2. b

    Average posterior probability for latent class.

Neck  0.090.880.580.360.630.070.66
Upper back0.080.540.060.270.530.110.260.010.28
Shoulder0.380.890.040.850.76  0.030.59
Elbow0.020.220.020.000.390.020.290.020.09
Wrist/hand0.050.330.030.120.510.010.530.030.41
Low back0.130.77   0.120.460.090.36
Hip0.010.290.040.130.820.020.310.070.14
Knee0.040.410.070.070.720.050.40  
Ankle/feet0.020.380.040.040.560.020.300.080.31
Ave post probb0.840.980.890.910.970.960.900.970.93
Table 4. Latent class-specific posterior probability of musculoskeletal co-complaints for the five least common primary musculoskeletal pain sites in the Danish population aged 16+
 Upper backElbowWrist/handHipAnkle/foot
1a (0.66)2 (0.34)1 (0.81)2 (0.19)1 (0.77)2 (0.23)1 (0.85)2 (0.15)1 (0.93)2 (0.07)
  1. a

    Latent class and conditional probability for proportion of participants in class.

  2. b

    Average posterior probability for latent class.

Neck0.080.710.030.710.080.820.080.530.040.44
Upper back  0.000.580.050.430.030.310.010.46
Shoulder0.010.760.000.580.050.780.050.570.040.49
Elbow0.000.14  0.040.280.010.250.000.28
Wrist/hand0.040.260.030.70  0.080.430.070.55
Low back0.210.560.110.240.090.370.200.780.120.61
Hip0.030.170.030.230.020.26  0.030.33
Knee0.040.340.080.470.090.260.180.710.180.76
Ankle/feet0.030.270.030.230.100.290.050.68  
Ave post probb0.980.950.991.000.960.940.900.980.990.97

Except for primary neck and shoulder pain, the largest latent class had a small probability of any co-complaint, and thus had pain mainly at the primary pain site. In contrast, the smallest class exhibited a diffuse pattern of pain with MSK co-complaints spread over the entire body (Fig. 1).

Specifically in these smaller classes, i.e., the smallest latent classes for the primary complaints, we found for primary complaints in the spine (neck, upper back, lower back) very high or high probabilities for co-complaints at other spine sites and lower probabilities for co-complaints in the extremities (Fig. 1). For the low back, for which three latent classes were found, the largest class had only LBP, one class had pain in the spine only, and the smallest class had very high and high probabilities for pain across the entire body. For the smaller classes with primary pain in the extremities, the highest probabilities for co-complaints were at adjacent body sites. One noticeable exception was for primary knee pain where a high probability for co-occurring neck pain but not for hip pain was found (Fig. 1).

An effect of gender (p = 0.01) was seen for latent classes related to primary complaints in the hip where women reported more pain sites, and an effect of age (p = 0.01) was seen for primary complaints in the knee where more people in the younger group had knee pain only (data not shown).

4. Discussion

  1. Top of page
  2. Abstract
  3. What's already known about this topic?
  4. What does this study add?
  5. 1. Introduction
  6. 2. Method
  7. 3. Results
  8. 4. Discussion
  9. 5. Conclusion
  10. Author contributions
  11. Acknowledgements
  12. References

To our knowledge, the current study is the first to employ advanced statistical techniques to specifically characterize patterns of MSK symptoms in the adult population based on the primary pain site. The results from this study will provide the basis for prognostic research in this population-based cohort, which has 20 years of follow-up in several official Danish registers (Davidsen et al., 2011). We found that the overall 2-week prevalence of MSK pain in this population-based sample of Danes aged over 16 years was 40%, the low back, the neck/shoulder area, and the knees being both the most common overall pain sites and also the most common sites of primary pain. Results of the latent class analyses allow us to make these general conclusions regarding pain patterns:

  1. For all primary pain sites, the largest latent classes, i.e., the latent classes with the highest probability, had low probabilities for co-complaints, indicating that most often a MSK complaint represents an isolated and site-specific problem/condition.
  2. For all primary pain sites, a minority of participants, i.e., the latent classes with the lowest probabilities, could be classified as having high probabilities for co-complaints at other body sites.
  3. If primary complaints were in the spine, i.e., neck, upper back, low back, very high or high probabilities for co-complaints were in the other areas of the spine and not the extremities.
  4. If these primary complaints were in the extremities, the highest probabilities for co-complaints were in neighbouring areas with one noticeable exception: Primary complaint in the knee showed high probability for co-complaint in the neck and low probability for the hip and foot.

Determining clusters of pain using data-driven analytic approaches, as in LCA, has several advantages over merely observing the prevalence of co-occurring pain. First, and most importantly, the clustering of pain sites may be driven by patterns of pain that are idiosyncratic to specific pathoanatomical conditions, or are idiosyncratic to central nervous system adaptations to persistent pain. These pain patterns may not be detectable by simply summing the number of pain sites, as some instances of multisite pain may co-occur simply by chance. The underlying notion is that both the sites of pain and the number of pain sites may be required to identify clinically important pain clusters. It may be that people with pain in the same body site belong to highly different pain groups. An example of this is the study by Schmidt and Baumeister (2007), which showed that pain clusters identified by LCA were associated with physical function, even when controlling for the number of pain sites, pain intensity and medical conditions. Just as factor analysis can identify latent factors that are explanatory of the patterns of scoring on questionnaire instruments, LCA appears to be capable of identifying latent classes that are explanatory of the patterns of scoring in clinical measures such as the distribution of pain.

Secondly, the relative size of the latent classes can be determined and misclassification quantified by posterior probabilities. Misclassification is very common in MSK epidemiology and is a product of both the clarity/strength of the underlying data structure and the ability of the analysis/descriptive approach to detect that data structure. In LCA, each participant is placed in the latent class with the highest probability for his or her particular profile and our very high posterior probabilities indicate that our model had low risk of misclassification (Nagin, 2005). In very prevalent conditions such as MSK pain, complaints can invariably co-occur simply by chance, and when chance patterns are included in analyses, estimates of co-occurrence become imprecise and diluted. Therefore, in contrast to descriptive approaches, the likelihood of misclassification in LCA can be quantified and estimated using goodness-of-fit tests and average posterior probabilities. We used CAIC and BIC to determine the number of latent classes but acknowledge that other criteria exist (Lanza et al., 2011), and the use of such alternative methods may have resulted in a different class structure.

Many authors have described the prevalence of co-occurring patterns of MSK symptoms and collectively they report that co-occurrence of symptoms is common. For instance, Natvig et al. (2001) found that in their population, 25% of persons reporting pain in the low back also reported pain up to four other body areas. Data from the North Staffordshire Osteoarthritis Project show that in middle-aged and older persons, 74% had more than one painful joint in the lower limb (Peat et al., 2006), and in persons with symptomatic knee pain 62% had pain in two or more body areas (Croft et al., 2005). Using LCA in a population-based sample, Schmidt and Baumeister (2007) found seven classes exhibiting a range of pain patterns. These results, however, are difficult to compare with our results, because pain patterns were not grouped by primary pain site. LCA has also been used to model pain patterns in the spine and extremities among adolescents (Adamson et al., 2007), Beales et al. (2012) identified clusters of back pain, co-morbidities, and quality of life in 17-year-olds, and in another study of teenagers, Auvinen et al. (2009) found that the number of pain sites and the probabilities for pain were highly correlated. Lastly, LCA has been used to model the longitudinal course of back pain among patients presenting for primary care in the United Kingdom (Dunn et al., 2011a), as well as the trajectory of back pain, headache, stomach pain and facial pain in an adolescent cohort (Dunn et al., 2011b).

Even though patterns for pain are remarkably similar for the primary pain sites, consequences of the pain are likely different. For example, a primary pain in the hip may prevent a person from being physically active, which would not be the case for pain in the elbow. The fact that pain has high probabilities for occurring at adjacent sites could indicate that the basis for the pain is biomechanical, i.e., pain results in altered function and use in an area which then irritates adjacent structures that in turn become painful. One exception to the pattern is a primary complaint in the low back where a ‘spine specific class’ was found (LC2). Recent studies have shown that pain in the three spinal regions behaves very similarly with respect to occurrence and consequences which may reflect the same underlying principle (Hartvigsen et al., 2009; Leboeuf-Yde et al., 2011, 2012).

Some may argue that self-reported pain is an imprecise measure. In the current study, risk of misunderstanding and misclassification was minimized by interview- and not questionnaire-based data collection, and by showing the participants a diagram with the different body areas clearly marked, misunderstandings and thus misclassifications were likely to have been kept to a minimum. An exception may be the neck/shoulder region where it may be possible to distinguish pain originating in the glenohumeral joint from pain originating in the cervical spine, while it may be more arbitrary whether pain in the area of the trapezius muscle is classified as neck or shoulder pain. Probably for some analytic purposes, these pain sites should be collapsed into one group. Another limitation in this analysis is the sample size, which for some of the primary complaints resulted in strata with few persons for the age, gender, and co-complaint groups, and therefore the most reliable results are for the most common primary complaints as presented in Tables 3 and 4. This may explain why we did not find statistically significant age and gender effects for less common primary complaints; however, this finding should be studied further in future analyses of specific complaints in this sample. These data were collected in 1991 and patterns of MSK complaints may have changed in the population since then. Finally, we consider the 2-week recall period a strength, because probably longer recall periods result in misclassification (Johansen and Wedderkopp, 2010).

So far, these data have not been utilized in scientific analyses. They are, however, ideally suited to study patterns of MSK pain in the population and their consequences, because they have been included in the Danish National Cohort Study, which now has 20 years of follow-up data on a range of register-based variables such as the seeking of treatment in primary care, hospitalizations, surgeries, work absence, disability pension, early retirement and a range of other socio-economic factors (Davidsen et al., 2011).

The potential clinical implications of this type of data-driven research are far reaching. Besides minimizing misclassification and aiding in prognosis research, it helps clinicians recognize specific patterns of co-complaints in patients presenting with a primary complaint, and potentially such classes can inform inclusion criteria for clinical trials evaluating the effectiveness of tailored interventions. It may be that for some patients, a clinician will be able to decide whether a patient needs any treatment at all, and if so, to tailor the type and intensity of such treatment based largely on the pain pattern.

5. Conclusion

  1. Top of page
  2. Abstract
  3. What's already known about this topic?
  4. What does this study add?
  5. 1. Introduction
  6. 2. Method
  7. 3. Results
  8. 4. Discussion
  9. 5. Conclusion
  10. Author contributions
  11. Acknowledgements
  12. References

We have introduced LCA as a method of grouping MSK pain patterns in the population. Persons with MSK complaints are able to identify which one is their primary complaint in spite of 40% reporting MSK pain in more than one site. Overall, the highest probability for MSK co-complaints in persons with a primary MSK complaint in the spine is at other sites in the spine, whereas for persons with a primary MSK complaint in the extremities, the highest probability is for co-complaints in neighbouring areas with a primary complaint of knee pain as a noticeable exception, where concurrent pain areas were the neck and low back. Phenotypes identified in this cohort can be used to study potentially different clinical courses.

Author contributions

  1. Top of page
  2. Abstract
  3. What's already known about this topic?
  4. What does this study add?
  5. 1. Introduction
  6. 2. Method
  7. 3. Results
  8. 4. Discussion
  9. 5. Conclusion
  10. Author contributions
  11. Acknowledgements
  12. References

All authors participated in the conceptualization of the study in brain storming sessions. M.D. carried out the analysis and all authors provided feedback. J.H. drafted the manuscript and all authors contributed suggestions and approved the final manuscript. J.H. is the guarantor.

Acknowledgements

  1. Top of page
  2. Abstract
  3. What's already known about this topic?
  4. What does this study add?
  5. 1. Introduction
  6. 2. Method
  7. 3. Results
  8. 4. Discussion
  9. 5. Conclusion
  10. Author contributions
  11. Acknowledgements
  12. References

The authors gratefully acknowledge Jacob Toft Vestergaard for help with preparation of Fig. 1, and Peter Kent, PhD, for helpful suggestions to the revised manuscript.

References

  1. Top of page
  2. Abstract
  3. What's already known about this topic?
  4. What does this study add?
  5. 1. Introduction
  6. 2. Method
  7. 3. Results
  8. 4. Discussion
  9. 5. Conclusion
  10. Author contributions
  11. Acknowledgements
  12. References
  • Adamson, G., Murphy, S., Shevlin, M., Buckle, P., Stubbs, D. (2007). Profiling schoolchildren in pain and associated demographic and behavioural factors: A latent class approach. Pain 129, 295303.
  • Andersen, L.L., Christensen, K.B., Holtermann, A., Poulsen, O.M., Sjogaard, G., Pedersen, M.T., Hansen, E.A. (2010). Effect of physical exercise interventions on musculoskeletal pain in all body regions among office workers: A one-year randomized controlled trial. Man Ther 15, 100104.
  • Auvinen, J.P., Paananen, M.V., Tammelin, T.H., Taimela, S.P., Mutanen, P.O., Zitting, P.J., Karppinen, J.I. (2009). Musculoskeletal pain combinations in adolescents. Spine 34, 11921197.
  • Beales, D.J., Smith, A.J., O'Sullivan, P.B., Straker, L.M. (2012). Low back pain and comorbidity clusters at 17 years of age: A cross-sectional examination of health-related quality of life and specific low back pain impacts. J Adolesc Health 50, 509516.
  • Birrell, F.N. (2004). Patterns of joint pain: Lessons from epidemiology. Rheumatology 43, 408409.
  • Croft, P., Jordan, K., Jinks, C. (2005). ‘Pain elsewhere’ and the impact of knee pain in older people. Arthritis Rheum 52, 23502354.
  • Davidsen, M., Kjoller, M., Helweg-Larsen, K. (2011). The Danish National Cohort Study (DANCOS). Scand J Public Health 39, 131135.
  • Dunn, K.M., Jordan, K.P., Croft, P.R. (2011a). Contributions of prognostic factors for poor outcome in primary care low back pain patients. Eur J Pain 15, 313319.
  • Dunn, K.M., Jordan, K.P., Mancl, L., Drangsholt, M.T., Le Resche, L. (2011b). Trajectories of pain in adolescents: A prospective cohort study. Pain 152, 6673.
  • Ekholm, O., Hesse, U., Davidsen, M., Kjoller, M. (2009). The study design and characteristics of the Danish national health interview surveys. Scand J Public Health 37, 758765.
  • Escobar, A., Quintana, J.M., Bilbao, A., Azkarate, J., Guenaga, J.I., Arenaza, J.C., Gutierrez, L.F. (2007). Effect of patient characteristics on reported outcomes after total knee replacement. Rheumatology 46, 112119.
  • Hartvigsen, J., Nielsen, J., Kyvik, K.O., Fejer, R., Vach, W., Iachine, I., Leboeuf-Yde, C. (2009). Heritability of spinal pain and consequences of spinal pain: A comprehensive genetic epidemiologic analysis using a population-based sample of 15,328 twins ages 20–71 years. Arthritis Rheum 61, 13431351.
  • Johansen, B., Wedderkopp, N. (2010). Comparison between data obtained through real-time data capture by SMS and a retrospective telephone interview. Chiropr Osteopat 18, 17.
  • Kamaleri, Y., Natvig, B., Ihlebaek, C.M., Bruusgaard, D. (2008). Localized or widespread musculoskeletal pain: Does it matter? Pain 138, 4146.
  • Kotlarz, H., Gunnarsson, C.L., Fang, H., Rizzo, J.A. (2009). Insurer and out-of-pocket costs of osteoarthritis in the US: Evidence from national survey data. Arthritis Rheum 60, 35463553.
  • Lanza, S.T., Dziak, J.J., Huang, L., Xu, S., Collins, L.M. (2011). PROC LCA & PROC LTA Users' Guide, Version 1.2.6 (Pittsburgh: Pennsylvania State University).
  • Leboeuf-Yde, C., Fejer, R., Nielsen, J., Kyvik, K.O., Hartvigsen, J. (2011). Consequences of spinal pain: Do age and gender matter? A Danish cross-sectional population-based study of 34,902 individuals 20–71 years of age. BMC Musculoskelet Disord 12, 39.
  • Leboeuf-Yde, C., Fejer, R., Nielsen, J., Kyvik, K.O., Hartvigsen, J. (2012). Pain in the three spinal regions: The same disorder? Data from a population-based sample of 34,902 Danish adults. Chiropr Man Ther 20, 11.
  • Lin, T.H., Dayton, C.M. (2007). Model selection information criteria for non-nested latent class models. J Educ Behav Stat 22, 16.
  • Macfarlane, G.J. (1999). Generalized pain, fibromyalgia and regional pain: An epidemiological view. Baillieres Best Pract Res Clin Rheumatol 13, 403414.
  • Maniadakis, N., Gray, A. (2000). The economic burden of back pain in the UK. Pain 84, 95103.
  • Nagin, D.S. (2005). Group-Based Modeling of Development (Boston: Harvard University Press).
  • Natvig, B., Bruusgaard, D., Eriksen, W. (2001). Localized low back pain and low back pain as part of widespread musculoskeletal pain: Two different disorders? A cross-sectional population study. J Rehabil Med 33, 2125.
  • Novicoff, W.M., Rion, D., Mihalko, W.M., Saleh, K.J. (2009). Does concomitant low back pain affect revision total knee arthroplasty outcomes? Clin Orthop Relat Res 467, 26232629.
  • Peat, G., Thomas, E., Wilkie, R., Croft, P. (2006). Multiple joint pain and lower extremity disability in middle and old age. Disabil Rehabil 28, 15431549.
  • Schmidt, C.O., Baumeister, S.E. (2007). Simple patterns behind complex spatial pain reporting? Assessing a classification of multisite pain reporting in the general population. Pain 133, 174182.
  • Suri, P., Morgenroth, D.C., Kwoh, C.K., Bean, J.F., Kalichman, L., Hunter, D.J. (2010). Low back pain and other musculoskeletal pain comorbidities in individuals with symptomatic osteoarthritis of the knee: Data from the osteoarthritis initiative. Arthritis Care Res 62, 17151723.
  • Thomas, E., Silman, A.J., Croft, P.R., Papageorgiou, A.C., Jayson, M.I., Macfarlane, G.J. (1999). Predicting who develops chronic low back pain in primary care: A prospective study. BMJ 318, 16621667.
  • Wikman, A., Wardle, J., Steptoe, A. (2011). Quality of life and affective well-being in middle-aged and older people with chronic medical illnesses: A cross-sectional population based study. PLoS ONE 6, e18952.