Categories of factors associated with pain, disability and/or return-to-work/work absence in adults with low back or neck pain

 * While the terms ‘pain intensity’ and ‘pain severity’ are sometimes used interchangeably, Mansell (2021) described how pain severity can be seen as an overarching term, encompassing different dimensions of pain (such as intensity, duration, and impact)

Prognostic factors for pain or perceived non-recovery in patients with acute or sub-acute musculoskeletal, non-traumatic neck pain

Prognostic models for recovery in patients with neck pain

Prognostic factors

Mansell (2021) performed an umbrella review of systematic reviews to present an overview of evidence on factors associated with pain, disability, and/or return-to-work/work absence outcomes in patients presenting with low back or neck pain in ambulant or occupational healthcare. This information could be used to inform patients about their prognosis and to make individualized decisions about optimal management. Since this was an umbrella review, data were analyzed at the level of systematic reviews, and it was not clear from the results section whether patients included in the original studies received treatment. Therefore, although Mansell (2021) uses the term ‘prognostic factors’, we cannot exclude the possibility that some of these studies may in fact have reported predictive factors.

A systematic search was performed in MEDLINE (Ovid), EMBASE (Ovid), CINAHL-Plus (EBSCO) and PsychINFO. The search involved systematic reviews published between 2008 and 28 September 2020. Studies fulfilling the following criteria were included in the review: (1) systematic review or overview of systematic reviews; (2) results for adult populations (18 years or older) with neck and/or low back pain of any duration, including whiplash, sciatica and radiculopathies, in any occupational or ambulant health setting. Systematic reviews were excluded if they focused on neck or low back pain following severe trauma, on patients admitted to hospital, or did not include factors that were self-reported by the patient. Relevant factors included any self-reported psychological, social or clinical variable and relevant outcomes included any measure of pain, functional disability and/or return-to-work/work absence.

In stage I of the review, data were extracted from the included systematic reviews. If a factor was reported in more than one review and was found to be associated with one of the selected outcomes, the factor was selected for stage II of the review.

In stage II, additional information was extracted, including strength of the association between factor and outcome, consistency of effects, and risk of bias. Risk of bias was assessed using the ROBIS (Risk of Bias in Systematic reviews tool). The level of confidence in the robustness of the associations was assessed using GRADE:

• High confidence: prognostic factors for which consistent high-quality evidence was presented with at least one high quality SR and no conflicting SRs;
• Moderate confidence: prognostic factors for which consistent findings from at least one recent medium quality SR were available, with the majority of findings from other concurrent SRs (where applicable) in the same direction of effect;
• Low confidence: prognostic factors for which summary findings were of low or unclear risk of bias from the majority of SRs and with conflicting results, or when only a single SR reported significant but only moderate-level findings for that predictor;
• Very low confidence: when none of the above conditions were met.

Forty-one systematic reviews reporting 90 factors were included in stage I of the umbrella review. Of these, ten reviews focused on neck pain only, and four reviews included both patients with neck pain and low back pain. In stage II, 35 reviews reporting 25 factors were included. The risk of bias in these reviews was judged to the low (n=10), high (n=8) or unclear (n=17).

In the narrative evidence synthesis of this umbrella review, more weight was placed on reviews conducted in the last five years and reviews with an overall low risk of bias. For seven categories of factors, the level of evidence was judged to be moderate (for the remaining categories, the level of evidence was low). These categories of factors included (1) disability/activity limitation; (2) mental health; (3) pain intensity; (4) pain severity; (5) coping; (6) expectation of outcome/recovery; and (7) fear-avoidance. Even for these factors, conflicting evidence was seen with regard to the strength of the associations.

The other factors, for which no robust and/or consistent associations were found included: (1) awareness of impending collision; (2) general health; (3) previous treatment; (4) previous pain; (5) lifestyle: alcohol intake; (6) pain duration; (7) multisite pain; (8) radiating/leg pain; (9) anger; (10) illness perceptions; (11) somatization; (12) PTS symptoms; (13) support at work; (14) job control; (15) job dissatisfaction; (16) previous sick leave; (17) work demands: physical; and (18) participation restriction.

The studies included in the systematic review presented associations between self-reported factors and treatment outcomes. However, it is not clear how these factors should be used in clinical practice, e.g. no cut-off values are described, the factors are potentially correlated, and no data are available about the predictive performance of these factors. Therefore, as also recommended by Mansell (2021), further research is needed to explore the impact of using prognostic information on treatment and referral decisions, patient outcomes and costs.

No formal GRADE assessment was undertaken in this analysis of the literature because the authors of the umbrella review used the GRADE approach to assess confidence in the evidence. In addition, it would have been difficult to perform a GRADE assessment since no effect sizes were reported in the umbrella review.

Verwoerd (2019) performed a systematic review of prognostic factors to provide an overview of modifiable and non-modifiable factors associated with the development of chronic musculoskeletal pain in patients who experienced a first episode of idiopathic, non-traumatic neck pain. The review was specifically focused on prognostic factors, studies that reported predictive factors for the outcomes of specific treatment were excluded. The development of chronic musculoskeletal pain was operationalized by the outcomes ‘pain intensity’ and ‘perceived non-recovery’ at three, six, and twelve months. Insight in prognostic factors could be used to improve clinical decision-making and to inform patients.

A systematic search was performed in Medline (PubMed), PsychINFO, EMBASE, SPORTdiscus, and CINAHL. The search included publications up to 21 October 2017. The authors also checked the reference lists of all included articles and performed a grey literature search. Studies fulfilling the following criteria were included in the review: (1) prospective cohort studies; (2) univariable or multivariable analysis to identify prognostic factors; (3) adults with idiopathic, non-traumatic acute (0-3 weeks) and/or sub-acute (3-12 weeks) neck pain; (4) follow-up period of at least three months; (5) papers published in English, Dutch, French or German; (6) outcomes included pain or perceived non-recovery. Studies were excluded if: (1) any of the following applied to patients: neck surgery, radiculopathy and myelopathy, headache, widespread pain, no neck pain at baseline; (2) pain not due to musculoskeletal pain (affecting bone(s), joint(s), muscle(s), or related soft tissue(s)); (3) >40% of the sample had whiplash related neck pain. No meta-analysis was performed as the study populations and outcomes were judged to be too heterogeneous. The authors assessed both the strength and the quality of the evidence. Strength of the evidence was assessed as follows: strong evidence was defined as more than one low risk of bias study providing consistent evidence (>75% of studies showing the same direction of effect). Moderate evidence was defined as more than one high risk of bias study and/or one low risk of bias study providing consistent evidence. A prognostic factor was judged to be of limited evidence if it was only reported in a single study. Quality of the evidence was assessed using a modified GRADE approach.

Six studies were included in the review. Four studies were conducted in the Netherlands, one was conducted in the UK, and one was conducted in Switzerland. A total of 2446 patients were included in these studies, 39% were male and 61% were female. Four studies were prospective cohort studies, the remaining two studies reanalyzed data from RCTs. Patients were recruited from general primary care practices (four studies), chiropractice practices (one study), and the general population (one study). Three studies measured pain intensity and five studies measured self-perceived non-recovery. In all six studies, the risk of bias was judged to be high.

For the outcome pain intensity, 47 prognostic factors were analyzed across the included studies. In the multivariable analysis, moderate evidence was found for the factors older age (>40 years) and concomitant back pain to be prognostic for higher pain intensity at 12 months follow-up (Hill, 2004; Hoving, 2004).

Studies analyzed a total of 43 prognostic factors for the outcome ‘perceived non-recovery’.

Using univariable analysis, moderate evidence was found for a ‘previous period of neck pain’ and ‘accompanying headache’ to be prognostic for perceived non-recovery at 12 months (Hoving, 2004; Vos, 2008). However, these factors were not retained in more than one multivariable model. The authors judged the evidence to be of low to very low quality.

The authors recommend performing a high-quality cohort study to provide a higher level of evidence on prognostic factors, especially modifiable prognostic factors.

No formal GRADE assessment was undertaken because the authors of the review used a modified GRADE approach to assess confidence in the evidence and judged the evidence to be of low and mostly very low quality.

Prognostic models

The systematic review by Wingbermühle (2018) presents an overview of multivariable prognostic models for recovery in people with neck pain. The focus of this review was on models that could be used in primary care.

A systematic search was performed in MEDLINE, EMBASE, and CINAHL from inception up to 3 May 2016. Reference lists of all included studies were checked for additional relevant studies. Studies fulfilling the following criteria were included in the review: (1) studies constructed multivariable prognostic models; (2) the data for these models was derived from prospective cohort studies or RCTs; (3) participants in these studies included people with non-serious specific or non-specific neck pain at any stage; (4) factors included in the models included baseline characteristics at intake and needed to be applicable to and easily obtained in non-invasive musculoskeletal primary care; (5) outcomes included pain, disability, or perceived recovery. Studies were excluded if they focused on (cost)-effectiveness, side effects, or developing a questionnaire, or when they used clinical procedures involving skin penetration (e.g. injection). Risk of bias was assessed using the PROBAST (Prediction model study Risk Of Bias Assessment Tool).

Fifty-three studies were included in the review, this included 46 studies that described the development of prognostic models, four validation studies and three studies that described both the development and validation of prognostic models.

In the 49 studies in which the development of a prognostic model was described, the mean age of participants ranged between 30 and 65 years. The mean duration of symptoms ranged from 60 days to 108 months. In 29 studies, participants received some form of treatment, models developed in these studies may have been predictive models rather than prognostic models. Follow-up for outcome measurement ranged between one week and five years. Models included one to ten predictors. All 49 studies had a high risk of bias.

In the seven studies describing the validation of a prognostic model, mean age of participants ranged from 32 to 49 years. Follow-up for outcomes ranged between one week and 12 months. In six studies, participants were recruited from a physiotherapy setting, in the remaining study participants came from an insurance company population. Four studies focused on neck pain, while three studies involved prognostic models for whiplash-associated disorders. Four studies were judged to have high risk of bias, while three studies were judged to have low risk of bias. Of these three studies, two had low concerns regarding applicability and were judged to be usable (Schellingerhout, 2010; Ritchie, 2015).

The model by Schellingerhout (2010) was developed based on data from three RCTs performed in Dutch primary care settings (n=468) (Hoving, 2002; Vonk, 2009; Pool, 2006). External validation was performed using data from an RCT conducted in UK primary care settings (n=315) (Dziedzic, 2005). The population consisted of patients with non-specific neck pain. Nine factors were included in the final model: (1) age; (2) pain intensity; (3) headache; (4) radiation of pain to elbow/shoulder; (5) previous neck complaints; (6) cause of complaints; (7) low back pain; (8) employment status; and (9) quality of life (EuroQoL). The model predicted the probability of complaints persisting for at least 6 months and was presented in the form of a score chart. The discriminative ability was AUC 0.66 in the Dutch population and AUC 0.65 in the UK population.

Ritchie (2015) developed models to predict two potential recovery pathways: a moderate to severe pathway with outcome of NDI ≥30% and one full recovery pathway with outcome NDI ≤10%. The study was performed in Australia and participants were recruited through hospital accident and emergency departments, primary care practices, and general advertisement. The population consisted of patients with whiplash-associated disorders. The models were developed based on data from 53 patients and validated in a population of 48 patients. The model for the recovery pathway included two predictors: NDI ≤32 and age ≤35. Sensitivity was 55% and specificity was 86%. The model for the moderate to severe pathway included three predictors: NDI ≥40, age ≥35, and hyperarousal (PDS subscale ≥6). Sensitivity was 44% and specificity was 99%.

A consistent factor in these models was age; in both models older patients were more likely to develop a persistent pain problem. In the score chart by Schellingerhout (2010), for every 10 years of age, the score increases with 7 points, which roughly corresponds with an increase of a few percent in the probability that neck problems will still be present at six months. In the models by Ritchie (2015), being less than 35 years of age was a prognostic factor for recovery, while being 35 years or older was a prognostic factor for ongoing moderate to severe pain and disability.

Wingbermühle (2018) concluded that most studies developing prognostic models were at high risk of bias, and only few models have been validated. However, three externally validated models were judged to be at low risk of bias of which two were judged to be applicable and usable. Further refinement of these models and evaluation of the impact of using these prognostic models on patient outcomes is needed.

No formal GRADE approach was undertaken but the level of evidence is expected to be very low because of the high risk of bias in most studies and indirectness of the evidence (models were developed to be used in primary care, most models were not externally validated, and studies included patients with whiplash-associated disorders).

One broad search was conducted to identify English language studies published from 2000 that described factors associated with a persistent pain problem in adults with chronic musculoskeletal pain. Given the broad topic, involving all patients with chronic pain and both modifiable and non-modifiable factors, a scoping search was performed to answer the following question: Which modifiable and non-modifiable factors are associated with a persistent pain problem in adults with chronic musculoskeletal pain?

P: Adults (≥18 years) with chronic musculoskeletal pain.

I: Modifiable and non-modifiable factors that are associated with a persistent pain problem.

C: Absence of these factors.

O: Persistent pain problem.

Search and select (Methods)

The databases Medline (via OVID) and Embase (via Embase.com) were searched with relevant search terms on 18 December 2020. The detailed search strategy is depicted under the tab Methods. The initial systematic literature search resulted in 1508 hits, including 101 systematic reviews, 597 RCTs and 810 observational studies. Given the large body of literature, it was decided to select one systematic review for the four major indications within the field of pain rehabilitation: low back pain, neck pain, shoulder pain and fibromyalgia. An update of the search was performed on 30 June 2021, involving only systematic reviews. This update resulted in 291 hits.

Studies were selected based on the following criteria: (1) systematic review; (2) recent publication date; (3) preferably multiple predictors included in the prediction model (multivariable model); and (4) preferably externally or internally validated.

Fifteen studies were initially selected based on title and abstract screening. Four of these publications involved patients with neck pain (Mansell, 2021; Rashid, 2017; Verwoerd, 2019; Wingbermühle, 2018). The umbrella review by Mansell (2021) was a review of systematic reviews and includes two of the other reviews (Rashid, 2017; Verwoerd, 2019), therefore the guideline working group selected this review for an overview of factors associated with a persistent pain problem. Since the review by Mansell (2021) involved patients with neck pain and low back pain, the review by Verwoerd (2019) that focused specifically on neck pain was also selected. In addition, the guideline working group selected the review by Wingbermühle (2018) that was focused on multivariable prognostic models.

Results

Three systematic reviews were included in the analysis of the literature. Study characteristics and results are described below. The systematic review by Mansell (2021) included systematic reviews of patients with neck pain and low back pain. The authors used the GRADE approach to assess the confidence in the robustness of the associations found. The systematic review by Verwoerd (2019) presented an overview of modifiable and non-modifiable prognostic factors for persistent pain after a first episode of idiopathic, non-traumatic neck pain. The studies included in this review presented associations between patient characteristics and pain intensity and perceived non-recovery. The authors used a modified GRADE approach to assess the level of evidence. The systematic review by Wingbermühle (2018) focused on models that could be used in primary care to predict recovery. The authors did not assess the level of evidence, and no formal GRADE approach was undertaken in this analysis of the literature.