Does back and neck pain become more common as you get older? A systematic literature review

Definitions

Back pain in this review includes pain anywhere in the spine, including the neck, either in a single spinal area or more widespread across the spine/neck. The older population is defined as people aged 60 and over according to the United Nation’s cut-off criterion [1]. Countries included in this review are defined as countries with an advanced economy according to the International Monetary Fund, which includes 35 countries [7].

Search

A systematic literature search was conducted in Pubmed (http://www.pubmed.org) for a number of musculoskeletal diseases [8] published between January 1^st 2000 and July 1^st 2011. Search terms included both free text and MeSH terms and were combined by Boolean terms (AND, OR, NOT) (see Additional file 1). The following MESH terms included in this review were “neck pain”, “back pain”, and “low back pain” and were limited to include only studies containing “epidemiology”, “etiology”, or “diagnosis”. The search was restricted to English language only. No additional hand search was conducted. The retrieval of potentially relevant articles was conducted in two steps by one examiner. The first step focused on identifying relevant studies through the title and abstract. This was followed by retrieval of all full-text articles for further eligibility (see below).

Eligibility criteria

In the present literature review, only observational studies from developed countries that reported some type of BP on people aged 60 and over were included.

Extraction of information

All core information from the included articles was extracted by an unblinded examiner and collated into two tables. The most relevant information was: Article details, method of data collection, and quality issues (sample size, response rate, method of data collection, generalisability of the study, selection bias, and how participants had been informed of the definition of the anatomical area as well as definition of pain). Further information was retrieved on relevant BP variables (BP definition, severity of BP and chronicity), relevant age ranges (age groups immediately before and after the age of 60 and subsequent age groups after 60), information with regards to the objectives: 1) If significantly higher prevalence of BP immediately after than before 60 was found (yes/no), and 2) if prevalence of BP increased significantly after the age of 60 (yes/no).

Data analysis

Each article was reviewed independently by the two authors in order to address the two objectives. For objective 1, the cut-point of 60 years of age was sought out, but if not reported, then the two age groups closest before and after 60 were used for comparisons. Significant differences in prevalence rates for those younger and older than 60 were determined, either by observing graphs with confidence intervals, or by obtaining information in the text or tables on statistically significant differences regarding the target groups. When no such information was available but size of subgroups and prevalence proportions were reported, the 95% confidence intervals were calculated and differences considered significant if the intervals did not overlap. For objective 2, the same analysis was performed in relation to the age groups after 60 to observe if there was a statistically significant increase of prevalence estimates with increasing age.

No attempts were made at pooling prevalence estimates, not even for fairly similar BP definitions. Each study would use its own definition, and the possible increase in prevalence was considered within each study individually, thus making it unnecessary to group similar definitions or to comment on their lack of conformity. The “outcome” variable in the analysis was therefore “statistically significantly increased prevalence” yes/no, first in relation to the cut-point of 60 and thereafter for the increasing increments of age.

Search results

In total, 3511 articles were found through the search strategy (Figure 1). Based on their titles and abstracts, 60 articles were retrieved for further reviewing. Among these potentially relevant articles, 48 were not accepted for the final reviewing process, mainly because no prevalence estimates were reported on elderly aged 60 and over or because the sample size of the age groups were too small (for a full list of excluded articles, please see Additional file 2). Hence, only twelve papers fulfilled all criteria and were thus included in this review [9–20].

Study characteristics

Nine of the twelve articles had been conducted in Europe and the remaining three in the USA. The studies used a variety of different definitions of BP, either as a single entity or a combination of pain sites and also used different recall periods, duration of pain, and intensity. Neck pain was studied nine times, low back pain eight times, BP three times, upper back twice and neck/shoulders once. Occasionally, it was not clear exactly which one of the definitions was used in relation to the age specific prevalence estimates. Age intervals ranged from 18 to 102. However, usually the upper limit was not defined. Only one study had used the World Health Organization-defined cut-point for “old”, namely 60 [19].

Five studies used either home or telephone interviews, while the rest used a questionnaire (Table 3). Overall, the generalisability was high, but some selection bias was probably likely in most of the studies, as some individuals would have been excluded for various health reasons (i.e. being weak, ill or demented). Most studies described how they defined BP for the study participants, except for four studies [12, 16, 18, 20].

Is there a significant increase in the prevalence of back pain after age of 60?

In all studies but one [20] was it possible to find information on or calculate statistically significance differences between the relevant age groups. It was possible to compare prevalence estimates in people just below the age 60 with those just above 60 in all but one study, which did not include study subjects younger than 70 [13]. The age intervals immediately before the age of 60 varied slightly, but, for the purpose of this review, started at 45 years of age and onwards.

As can be seen in the 7^th column in Table 3, none of the studies revealed a statistically significant increase of BP at this time in life and this was the case regardless the definition of BP. There was no difference in outcome between the different definitions, severity, or chronicity.

Is there is a significant increasing prevalence of back pain with old age?

In one of the studies, significance was neither provided nor possible to calculate [20].

An increase in prevalence estimates with increasing age was reported in two studies [11, 18] and possibly in a third study [20] (8th column Table 3). However, in the majority of the studies, no significant increases were found. In fact for the oldest group, estimates were generally declining in several of the studies (see last column Table 3) and significant declines were reported in three of the articles (last column Table 3). There was no difference in outcome between the different definitions, severity, or chronicity.

Possible explanations

These findings could be explained by a cohort effect, with people born during certain periods of time having a more or less robust physique. Only long-term (more or less life-long) longitudinal studies would be able to provide truly valid answers in relation to the course of BP, but no such studies exist. Other relevant explanations would be an increased tolerance to pain with age [21] and survival of the fittest; both resulting in lower reporting rates. A fourth explanation could be a decreased need to be physically active in old age, i.e. a reduction of pain provoking circumstances. This appears a likely explanation, as BP seems actually to diminish in the oldest old, who would be least likely to perform heavy physical activities. In three studies [10, 11, 18], was it obvious that the researchers had taken steps not to exclude the frail elderly (i.e. including also those with terminal illness, severe psychiatric disorders or severe dementia). It is not known if frail old people are more or less likely to have BP, so it is not known, in what direction – if any – this imbalance may push the results.

Comparisons with other reviews

Only two previous reviews on the prevalence of BP in the elderly population have been identified [6, 22]. Bressler et al. included articles between 1966 and 1999 [22]. As this current review includes articles from 2000 and onwards it would thus be interesting to compare their findings with those from the current review to see if any age related trends have changed over time. Unfortunately their objective was not to examine any trends in prevalence of low back pain with increasing age, but merely to try to establish overall prevalence estimates in the elderly population. Therefore no relevant comparisons were possible.

Age trends were, however, the main objective in Dionne et al.’s review [6]. They grouped studies by definition of BP and country of origin and dealt with any age including 65 but did not look specifically at the cut-point between middle age and old age. They reported a number of different types of curves (an increased prevalence with age, a decreased prevalence with age, a curvilinear relationship, and a flat curve) taking into account all age groups, including the younger people. In relation to middle age to old age, they found a curvilinear relationship (highest at 55) particularly for one-year period prevalence measures and for some measures of chronic pain. They also reported on the results of a weighted analysis for those aged 60 and more (with all types of BP problems); showing a relatively flat trend. A graph of a weighted calculation of severe BP predicted an increase of a few percent between the ages of 54 and 60 and about 5% between 60 and 90 years. Their study was published in 2004, and since then some additional large scale studies have been published and included in the current review. Nevertheless, our results were no different from theirs, i.e. no positive association between age and BP after middle age and into old age was found.

Methodological considerations of the review

This review had some potential and real limitations but also some strengths. Only one electronic database (Pubmed) was included in the search and thus some relevant articles may have been missed. Based on other reviews on similar musculoskeletal conditions, who have included other electronic databases (i.e. EMBASE, CINAHL, etc.), It is therefore possible that some potentially relevant articles may have been missed [6, 22]. However, unless such articles had completely different results, the occasional missing report would not affect the results, as the results were largely homogeneous.

The search strategy was also limited to the elderly population through MeSH terms. This may have lead to exclusion of some studies if for some reason they were not properly indexed in Pubmed. Finally, as only English language articles were included, any articles published in national non-English medical journals are missing in this literature review.

The review was uncritical, in the sense that although some methodological issues relating to bias were included, no attempt was made at grading the articles in relation to risk of bias. Also there were no predefined minimal criteria for acceptance in relation to quality. Had the results varied between studies, it would have been necessary to analyze study methodology and definitions, in order to understand such differences. However, this simplistic approach was apparently acceptable, as there were no difference in results across the review material, indicating that the results are fairly robust, at least in epidemiologic studies that made the effort of obtaining a relatively representative study sample.

The review was performed by two independent persons. It was always possible to reach consensus, meaning that the results probably are reproducible. Another advantage was the simplistic approach; that of dealing only with obvious difference of prevalence estimates, without confusing the picture with numerous different definitions of BP.

The heterogeneity of pain definitions is already a well known problem as it makes it difficult to compare results from different studies. It is not realistic to expect complete conformity in this area, as researchers may have had specific reasons for why they use a unique pain definition. Standardized and well defined questions are obviously important for study subjects so they know what they answer to. However, it is not important in the present review that BP definitions are all similar. Specific BP definitions would only have been necessary if there had been a need to relate prevalence estimates to the different age groups. However, as this study only attempted to find an increased prevalence over age, no attempts were made to pool results in relation to specific definitions of BP. The reason for this was that an increased prevalence of BP would be apparent regardless its definition. The decision not to concentrate on the various definitions of BP was shown to be an acceptable approach, as the findings in the present review were unequivocal across all definitions and regardless the different age-group definitions.

However, this review did not deal with the consequences of BP, nor of the duration and accompanying symptoms. Only very little information was also found on severity and chronicity. It would not be unreasonable to assume that recovery of BP is slower in old age or that some specific symptoms develop as the spine continues to degenerate. Therefore, although the prevalence of BP does not increase in old age, it is possible that specific symptoms and consequences may develop, giving a different profile of BP as the years roll on. To investigate further this concept, epidemiologic studies would have to relate age to more detailed descriptions of symptoms and consequences of BP.