Are Older Adults Missing From Low Back Pain Clinical Trials? A Systematic Review and Meta-Analysis

Authors


Abstract

Objective

There is evidence to suggest that older adults are underrepresented in randomized controlled trials of health interventions. The aim of this review was to systematically examine the age-related inclusion criteria distribution of participants in randomized controlled trials (RCTs) of low back pain (LBP) interventions and to investigate if this distribution pattern changes over time.

Methods

We identified, in PubMed, 1,047 RCTs on nonspecific LBP published since 1992, of which a random sample of 400 was assessed for inclusion in the review. Included studies were grouped according to treatment type. Data were extracted on year of publication, types of treatment, total number of participants, age inclusion criteria, and age of included participants.

Results

A total of 274 RCTs published between 1992 and 2010 met the inclusion criteria. A total of 41.6% (n = 114) of the included trials excluded people ages >65 years. The pooled mean age of participants was 44.3 years (95% confidence interval 42.4–46.3); the pooled minimum age for inclusion was 19.8 years and the pooled maximum age for inclusion was 65.4 years. We found no trend of including older participants in RCTs on LBP over time.

Conclusion

Despite an aging population around the globe, older adults are largely excluded from RCTs evaluating management of LBP, and there is no evidence of change in this practice over the last 2 decades.

INTRODUCTION

Low back pain (LBP) is a common and costly condition. A recent systematic review of the prevalence of LBP estimated its point prevalence to be approximately 12%, with a 1-month prevalence of 23% ([1]). The review also demonstrated that this condition is more common in women and those ages 40–80 years. LBP is currently ranked the number one cause of years lived with disability in the world and number 6 in terms of disability-adjusted life-years ([2]). The costs associated with LBP are substantial; for example, Australia spends in excess of $1 billion each year on direct costs and $8 billion on indirect costs ([3]).

While it is clear that LBP affects men and women of all ages, it was previously believed that adults of working age were the most vulnerable ([4-6]). However, Dionne et al ([7]) recently ascertained that even though the prevalence of benign back pain indeed appears to decrease with increasing age, the prevalence of severe back pain continues to increase. Moreover, LBP is the third most commonly reported musculoskeletal condition among older patients visiting primary care in the US ([6]) and the third leading cause of chronic disability in people ages ≥65 years in Canada ([8]). These data indicate that LBP is a significant health condition in older adults as well. Outside the LBP field, there is compelling evidence suggesting that older adults are significantly underrepresented in randomized controlled trials (RCTs) of health interventions, despite their higher levels of morbidity, greater consumption of prescription medications, cognitive impairment, attitudinal barriers, and greatest potential for clinical benefit compared to younger subjects ([9, 10]). For example, a systematic review by Levy and colleagues revealed that 53% of the identified clinical trials excluded people ages >65 years and the exclusion percentage increased to 72% for those ages >75 years ([11]). While this issue has not been directly studied in the LBP field, there is some preliminary evidence that the situation is likely to be the same. For example, a systematic review on the prevalence of LBP in older adults confirmed that this population is underrepresented in the literature on back pain ([6]).

This has important implications, since it creates uncertainty about whether treatments, shown to be effective in trials with highly selected patient groups, are effective under normal conditions of care with a broad age range of patients. Moreover, clinicians treating older patients attempt to extrapolate findings from studies of a much younger sample. Translating results of younger to older participants, especially when drug treatments are used, can have harmful consequences ([10]).

Therefore, the aims of this study were to systematically examine inclusion criteria of RCTs on LBP based on the age of participants and the age distribution of participants included in RCTs of LBP, and to describe if this distribution pattern has changed over time.

Box 1. Significance & Innovations

  • Low back pain (LBP) is a commonly reported musculoskeletal condition among older patients visiting primary care.
  • However, there is compelling evidence suggesting that older adults are significantly underrepresented in randomized controlled trials (RCTs) of health interventions.
  • This review has identified the age used for inclusion and exclusion of participants in 274 trials on LBP interventions published in the last 20 years.
  • Our results show that older adults are largely excluded from RCTs evaluating management of LBP, and this reality has not changed over the last 20 years.

MATERIALS AND METHODS

Search methods for identification of RCTs

We conducted a search by running a clinical query in PubMed from the earliest record to August 2010 and combined the search term “low back pain [MeSH]” with the category “therapy” and the scope “narrow, specific search.” We randomly selected 400 RCTs from this list by allocating all studies to a random number and, after sorting them in ascending order, extracting the first 400 titles. Presuming that at least 50% would meet the eligibility criteria, this approach would provide a minimum of 200 records for data extraction. We expected that a total of 200 trials would yield precise pooled results.

To select the studies, 2 authors (TP, CS) independently screened all search results for eligibility using the title first, then the abstract, and then the full article. All full articles identified were accompanied by a screening form. Differences were resolved by consensus, and if consensus could not be reached, then an independent third author (MLF) arbitrated.

Inclusion criteria for RCTs

RCTs recruiting participants of any age with nonspecific LBP of any duration, with or without sciatica, were included. Published protocols of RCTs or secondary analyses using data collected from an RCT were excluded. RCTs where participants had a specific pathology (including vertebral infection, cancer, fracture, ankylosing spondylitis, or cauda equina syndrome) were not considered for the review. RCTs on back pain associated with pregnancy were also excluded from the review. Studies in languages other than English were included whenever an appropriate translation was available. For studies separately reporting long-term followup outcomes and where short-term outcomes had been previously published, we acquired the original publication for data extraction and excluded the other article. Methodological quality appraisal of the included studies was not assessed because the aims did not include the reporting of treatment effects.

Data extraction

One reviewer (CS) extracted data from the eligible studies by using a standard data extraction form. All data were double checked by a second independent reviewer (TP). Differences were resolved by consensus, and if consensus could not be reached, then an independent third author (MLF) arbitrated.

Data were extracted on the year of publication, types of treatment, total number of participants, minimum and maximum ages for inclusion, sex, and mean ± SD age of the participants across all intervention groups as well as within each intervention group. If data were missing, we assigned them as not available. For those studies that did not provide an SD of the mean age of their participants, the SD was estimated from the reported range, 95% confidence interval (95% CI), or SE, as recommended in the Cochrane handbook for systematic reviews ([12]). In trials that only reported the mean age by intervention group but did not provide a total mean age, this was estimated by combining the data of each intervention group.

Statistical analysis

For the analyses, included RCTs were grouped into treatment categories, according to the intervention reported in the experimental arm of the trial. The mean age of included participants was pooled across studies with sufficient data reported and then by treatment category, where pooled means were weighted by variance (SE). Likewise, the pooled minimum and maximum ages for inclusion were calculated and weighted by the randomized sample size. A dot graph was used to plot the pooled minimum and maximum ages as well as the mean age of all participants by treatment category. Fractional polynomial graphs were built to plot the minimum and maximum ages for inclusion and the mean age of participants by publication year to observe if these parameters changed over time. All analyses and graphs were conducted using Stata, version 12.

RESULTS

Search results

The search on PubMed identified 1,047 studies and of the randomly selected 400 RCTs, 95 studies were excluded after title and abstract screening and 31 studies were excluded after full-text screening. The main reasons for exclusion on the full-text screening were articles describing secondary analyses of an RCT (n = 11) and studies separately reporting long-term followup outcomes (n = 8). A total of 274 studies (65.8%) met the inclusion criteria and were included in the review. A summary of the retrieval and processing of the studies is shown in Figure 1.

Figure 1.

Retrieval of studies for the review. RCTs = randomized controlled trials.

Included trial/study characteristics

The 274 RCTs included a total of 46,725 participants. Of these studies, 94.9% (n = 260) enrolled <500 patients, 3.3% (n = 9) enrolled 500–1,000 participants, and 1.8% (n = 5) enrolled >1,000 participants per study. All included RCTs were published between 1992 and 2010. Each of the 274 included RCTs was allocated to 1 of 14 treatment categories according to the experimental arm (where the control group received no or minimal treatment or placebo). Trials comparing ≥2 interventions were categorized in the pragmatic trials category (n = 15), where pragmatic trials are defined as those comparing 2 complex treatments given under similar practical conditions and no treatment inferiority is implied ([13]). For instance, the RCT by Giles and Muller ([14]) compared medication, acupuncture, and spinal manipulation and was therefore categorized as a pragmatic trial. With a total number of 60 studies, the exercise category comprised the largest number of RCTs, followed by the drugs category, with 39 RCTs. The heat wrap therapy, massage, and orthosis categories had the smallest number of trials, with 2 allocated RCTs each.

Age-related inclusion and exclusion criteria

Of the 274 included RCTs, 194 (70.8%) provided data on a minimum age for inclusion and 147 (53.6%) documented a maximum age for inclusion. A summary of the minimum and maximum ages for inclusion by treatment category is shown in Table 1. Figure 2 shows a graphical representation of the pooled minimum age, maximum age, and mean age of participants by treatment category. The results of the pooled minimum and maximum ages for inclusion by treatment category show the lowest pooled minimum age for inclusion was in the injection/infusion category (17.7 years) and the lowest pooled maximum age for inclusion was in the back school category (57.7 years), with the pooled minimum age for inclusion across all trials of 19.8 years and the pooled maximum age for inclusion of 65.4 years (Table 1).

Table 1. Pooled minimum and maximum ages for inclusion, pooled mean age, 95% CI (years), and number of allocated randomized clinical trials by treatment category*
 Pooled minimum age, years (no.)Pooled maximum age, years (no.)Pooled mean age, years (no.; 95% CI)
  1. 95% CI = 95% confidence interval.
Acupuncture22.6 (11)76.2 (7)52.7 (13; 47.8–57.8)
Back school19.8 (8)57.7 (7)44.7 (13; 40.1–49.3)
Pain management17.6 (17)62.7 (16)41.3 (18; 38.5–44.2)
Electrotherapy19.7 (11)64.1 (5)47.9 (11; 39.0–56.0)
Exercise19.3 (35)58.6 (33)38.5 (43; 34.6–42.4)
Heat wrap therapy18.0 (2)80.0 (1)48.0 (2; 30.4–65.5)
Massage18.0 (1)81.0 (1)43.7 (2; 38.2–49.3)
Drugs20.0 (33)71.2 (25)47.0 (32; 42.0–51.0)
Injection/infusion17.7 (12)70.1 (5)47.8 (17; 44.0–51.6)
Spinal manipulative therapy19.9 (23)59.1 (13)44.6 (15; 42.2–47.0)
Surgery19.9 (12)59.8 (12)43.3 (7; 40.9–45.6)
Orthosis20.0 (1)59.0 (1)45.0 (1; 41.2–48.8)
Alternative medicine20.9 (16)65.0 (10)48.7 (13; 44.0–53.5)
Pragmatic trials21.1 (12)59.6 (11)42.2 (11; 42.4–46.3)
Total19.8 (194)65.4 (147)44.3 (198; 42.4–46.3)
Figure 2.

Pooled mean age of participants and pooled minimum and maximum ages for inclusion by treatment category.

A frequency histogram was built to display the cumulative percentage of studies by the minimum and maximum ages used for inclusion (Figures 3A and B). The participants had to be age ≥20 years to be included in 61.7% (n = 169) of the trials and age ≤65 years in 41.6% (n = 114) or age ≤75 years in 50.4% (n = 138) of the trials to be included. The most commonly used minimum age for inclusion was 18 years (n = 115) and the most commonly used maximum age for inclusion was 65 years (n = 39), followed by 60 years (n = 29).

Figure 3.

Frequency histograms showing the cumulative percentage of studies by A, the minimum age for inclusion and B, the maximum age for inclusion.

Age distribution of participants

One hundred ninety-eight RCTs (72.3%) with a total of 35,625 participants provided data to estimate the mean age of included participants. A pooled mean age of 44.3 years (95% CI 42.4–46.3) was found across these 198 RCTs. Table 1 shows the mean age of the participants and the 95% CIs across these RCTs categorized by treatment. In general, the mean ages across the types of treatments were similar. RCTs in the exercise category had the youngest pooled mean age (38.5 years; 95% CI 34.6–42.4). With a pooled mean age of 41.3 years (95% CI 38.5–44.2) and 18 included studies, category 3, pain management, was the second-youngest category. The highest pooled mean age (52.7 years; 95% CI 47.8–57.8) was identified in the acupuncture category.

Effects of year of publication on age-related inclusion and exclusion criteria and the age distribution of participants

The graphs shown in Figures 4A–C were built using the two-way fractional polynomial approach and show the predicted minimum and maximum ages for inclusion and the mean age of participants by year of publication (1992–2010). Fractional polynomial regression analyses are used to fit and compare models other than linear (e.g., quadratic and cubic) and the best fit is then used to plot the 2 variables (e.g., mean age of the included participants and year of publication). The results showed a nonlinear (P = 0.271) and nonsignificant (P = 0.05) quadratic association between the mean age of participants and year of publication, which can be visualized in Figure 4C. Likewise, the minimum age for publication was shown to have a nonlinear (P = 0.63), nonsignificant (P = 0.19) quadratic association with publication year (Figure 4A). For the maximum age for publication, fractional polynomial regression models showed a linear (P = 0.001), but nonsignificant (P = 0.98), association with publication year, as shown in Figure 4B.

Figure 4.

Predicted A, minimum age, B, maximum age, and C, mean age of participants for inclusion by year of publication. 95% CI = 95% confidence interval.

DISCUSSION

In spite of back pain being a major problem in older age ([7]), older adults are underrepresented in trials evaluating treatments for LBP in terms of mean age of included participants as well as the age used for inclusion or exclusion of participants in a trial. More importantly, this scenario has not improved over the past 20 years. The paucity of including older adults in LBP trials has many important implications for both research and clinical practice. For instance, the exclusion of older adults limits the external validity of RCTs and creates uncertainties about whether treatments, shown to be effective in trials with highly selected patient groups, are effective under normal conditions of care.

Different barriers play important roles in the exclusion of older adults from clinical trials. Older adults present with more comorbidities and mobility limitations than younger adults, which might prevent them from attending or complying with the study protocols ([9, 15-17]). Other barriers include inadequate recruitment strategies or long trial durations, as well as cognitive and hearing impairment ([9, 15, 16, 18]). Moreover, older adults might be explicitly excluded from trials by investigators because they present with conditions such as renal insufficiency, or cardiovascular complications often seen as contraindications to a number of interventions being assessed. The relationship between the participants and the staff has to be comfortable for older adults, since interpersonal relationships have been considered crucial in motivating older adults to take part in clinical trials ([19]).

Our review has demonstrated that the specific LBP treatments evaluated in an RCT do not seem to influence the mean age of included participants, although we have observed an evident variation in the maximum age used in the inclusion criteria among different treatment categories (Figure 2). For instance, even though trials on medication (drugs and injection/infusion) have clearly adopted higher cutoffs for maximum age than those on spinal manipulative therapy and electrotherapy, the pooled mean age of included participants does not seem to vary to the same extent. This might be due to the fact that some treatment approaches, such as exercise, may present lower attrition among older adults, who are then excluded from these trials.

Our results are based on a random subsample of trials published since 1992. Although we acknowledge there are limitations in not including all published trials in our review, the pooled results show precise 95% CIs (Table 1), suggesting our study ascertainment has yielded a representative sample. A further limitation of our review is the lack of trial data on the percentage of included participants who were age >65 years. For this reason, we were only able to extract data on age used for inclusion and mean age of included participants, and therefore were unable to provide data on the proportion of older people who actually participate in trials that do not explicitly exclude them. Given the many barriers to participating in clinical trials older adults encounter, we would recommend future studies to report the percentage of participants included by age group, even when older adults are not explicitly excluded. Readers and consumers would then be able to confirm whether or not older adults are being well represented in clinical trials.

Despite the increasing age of the population globally and consequent increase in disability-related comorbidities, older adults are often excluded from RCTs in LBP and there is no evidence of improvement over the years. The maximum age for inclusion as a method to exclude older adults from RCTs is still frequently used (53.6%) in the field of LBP and limits the generalizability of trial results, producing uncertainties about the effectiveness of the studied treatments in that age category. Future studies in LBP should include a representative sample of older adults to make sure that the study sample reflects present demographic trends and to provide the best evidence to guide critical health decisions for the coming older society.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Ferreira had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Ferreira, Sun, Lin, Maher.

Acquisition of data. Ferreira, Sun, Lin, Maher.

Analysis and interpretation of data. Paeck, Ferreira, Lin, Tiedemann, Maher.

Ancillary