Health Coaching for Low Back Pain and Hip and Knee Osteoarthritis: A Systematic Review with Meta-Analysis

Abstract Background Health coaching aims to empower people to reach their goals and is increasingly used in health care settings. Whether health coaching improves pain and disability for people with hip or knee osteoarthritis (OA) or low back pain (LBP) is unknown. Methods Six databases were searched for randomized controlled trials assessing health coaching or motivational programs in adults with hip or knee OA or LBP, with each condition investigated independently. Meta-analyses were performed with random-effects models in the Cochrane Collaboration Review Manager 5.3 program. Results Seventeen eligible studies were found. No studies analyzing hip OA alone were found. Pooled analyses found statistically significant decreases in mid-term pain (mean difference [MD]: –7.57; 95% confidence interval [CI]: –10.08 to –5.07; P < 0.001, I2 = 0%), short-term disability (standard mean difference [SMD]: –0.22; 95% CI: –0.41 to –0.03; P = 0.02, z = 2.32, I2 = 0%), and mid-term disability (SMD: –0.42; 95% CI: –0.75 to –0.09; P = 0.01, z = 2.49, I2 = 60%), favoring the intervention for chronic LBP. There were significant improvements in knee OA long-term functional disability (MD: –3.04; 95% CI: –5.70 to –0.38; P = 0.03; z = 2.24; I2 = 0%). Conclusion Meta-analyses provide evidence that health coaching reduces both disability and pain in people with chronic LBP and reduces disability in people with knee OA, though the clinical significance is unknown. There is currently no evidence supporting or refuting the use of health coaching for hip OA.


Introduction
The growth and aging of the world population has led to significant increases in the incidence and prevalence of lifestyle-related musculoskeletal disorders, such as low back pain (LBP) [1] and hip and knee osteoarthritis (OA) [2]. These conditions are the leading causes of adult musculoskeletal pain and economic burden worldwide [3]. The magnitude of this burden has been highlighted by the estimated global number of years lived with disability: 64.9 million associated with LBP, 8.3 million for knee OA, and 1.3 million for hip OA [4]. High prevalence has a direct impact on the financial burden of these conditions, with the annual Australian direct health care expenditure for LBP being AUD4.8 billion in 2013 [5]. Reported figures are likely to underestimate the true out-of-hospital expenditure because of the extensive use of over-the-counter pain relief medication and the use of private services, such as physiotherapy [6]. In comparison, the U.S. total direct and indirect costs attributed to LBP were estimated to be US$84.1 billion and US$624.8 billion [7]. Australian health care costs for OA in 2015 came to a total of AUD5.5 billion, with indirect costs estimated to be at least twice this value [8]. These costs are expected to continue to rise, and the burden on already overloaded health care systems is unsustainable. Finding ways to alleviate this pressure and reduce costs and dependency on health care systems while maintaining or improving health outcomes is therefore of paramount importance globally.
Lifestyle behaviors are primary contributors to chronic disease development and influence their progression and severity through noncompliance with recommended treatment [9]. Although they differ in anatomic position and function, LBP and knee and hip OA have similar guideline management recommendations that encompass nonpharmacological and pharmacological interventions, including education, advice to remain active, exercise, analgesics, and nonsteroidal anti-inflammatories [10][11][12][13]. Implementation of these guidelines in health care and patient adherence to evidence-based programs have been demonstrated to be suboptimal [14,15].
In recent years, health coaching (HC) has been identified as a way of empowering individuals to self-manage their conditions. It draws on many behavior change theories, such as the "trans-theoretical model of change" and "motivational interviewing," which aim to help patients remain healthy and prevent costly hospital admissions resulting from failing to follow recommended treatment plans [16]. A typical HC model involves identifying how an individual's ideal health status differs from their current position and then exploring ways to close this gap [17]. With support, the individual works toward a more sustainable, health-promoting change in behavior [17]. The success of HC has been demonstrated in other chronic conditions, with increased patient adherence to physical exercise, weight loss programs, and recommended therapeutic regimes in hypertension and diabetes [18]. It has also been shown to help patients with cancer self-manage pain [19] and improve quality of life for patients with chronic obstructive pulmonary disease [20]. It is an area of increasing interest for LBP and hip and knee OA, with the hope that it can improve health outcomes and reduce costs, but to date, the effectiveness of HC for these conditions has been unclear.
The present review aimed to systematically appraise the literature investigating the effectiveness of HC to reduce pain and disability and inform on the recommended dosage for people with LBP and hip or knee OA.

Methods
This systematic review was registered with PROSPERO, the International prospective register of systematic reviews (registration number CRD42018086072).

Search Strategy
The search strategy was developed in collaboration with a research librarian and consisted of four separate components that used Medical Subject Headings (MeSH) and keywords (Supplementary Data Appendix 1). These were then combined with the HC component. The following databases were searched electronically up to November 24, 2021: Ovid MEDLINE V R >1946; Ovid EMBASE >1947; Cochrane Central Register of Controlled Trials; Ovid PsycINFO >1967; EBSCO CINAHL; and Web of Science. A hand search of the included studies' reference lists and any relevant systematic reviews found was also conducted.

Study Selection
Two independent reviewers (JLP and GV) screened all records by titles and then abstracts in accordance with the eligibility criteria. Full texts were obtained and screened, with any disagreement referred to a third independent reviewer (MLF) for resolution via a blinded vote count.
Randomized controlled trials published in peerreviewed journals in any language were included if they contained at least one of the target populations, with participants 18 years of age or older and with symptoms of any duration. Diagnosis of knee or hip OA was in accordance with the American College of Rheumatology clinical criteria, radiographic diagnosis, or medical practitioner diagnosis. LBP studies were excluded if the target population included pregnancy-related LBP, serious spinal pathologies (e.g., fracture, malignancy, inflammatory disease), or LBP associated with neural tissue involvement (e.g., radiculopathy, spinal canal stenosis). To be included, studies needed to involve HC or motivational programs that aligned with HC models. Studies based on cognitive behavioral therapy were excluded because of the complex, structured nature of this psychological treatment that targets maladaptive thought patterns. Comparators could include usual care (e.g., general practitioner management or physiotherapy, including exercise, advice, manipulative therapy, or electrotherapy), no treatment, or any other conservative treatment. Although the registered protocol stated that only studies in English would be included, to minimize selection bias, the search was not limited by language, with Google Translate (Google LLC, Mountain View, CA, USA) used as required.

Outcomes
Our primary outcomes were pain and disability. The secondary outcome measures of exercise compliance, physical activity, quality of life, sleep, weight loss, health care use, adverse events, and costs are listed in Supplementary Data Appendix 2.

Data Extraction
Two reviewers (JLP and GV) independently extracted data from the selected full texts into a standardized data extraction spreadsheet. Disagreements were resolved by discussion between the two reviewers; the third reviewer was not required. Descriptive data pertaining to study setting, population (sample size, gender, age), and intervention characteristics (duration, number of sessions, content, mode of delivery, additional intervention components) were collected. For each study, relevant outcome measures were collected at each time point. Data were extracted as means with standard deviations (SDs). If only 95% confidence intervals (CIs) were available, SDs were calculated with the Cochrane Collaboration Review Manager (RevMan) 5.3 Software. Outcome data were extracted for the shortterm ( 3 months), mid-term (>3 but <12 months), and long-term follow-ups (!12 months).

Methodological Quality Assessment
The methodological quality of included studies was assessed with the Physiotherapy Evidence Base Database (PEDro) scale [21]. This scale assesses the internal and external validity of studies and scores them via 11-point criteria. Studies were rated from 0 to 10, as the first criterion is excluded from the final score.

Strength of the Body of Evidence
The strength of the body of evidence after meta-analysis was appraised with the GRADE approach (Grading of Recommendations Assessment, Development, and Evaluation) [22]. The quality of the evidence was rated down according to the following GRADE criteria: risk of bias (if 50% studies had low methodological quality that scored <7/10 on the PEDro scale); inconsistency (unexplained heterogeneity [I 2 > 50%]); imprecision (small sample size [ 400 participants]); indirectness (participants, intervention, or outcomes differ from those of interest); and publication bias (industry-funded studies or all small studies included).

Data Analysis
Studies were considered for meta-analysis if the comparisons were similar in terms of content, dosage, and follow-up time points. Meta-analyses were performed through the use of random-effects models in the Cochrane Collaboration Review Manager 5.3 program.
Each condition was analyzed independently, with separate analyses performed for each outcome and time point where available.
Pain intensity measures were 0-to 100-mm visual analog scales or 0-10 numerical rating scales and were converted for the analysis to a common 0-100 scale. Results are presented as mean differences (MDs) with 95% confidence intervals (95% CIs). Different scales were used to measure disability (e.g., Roland-Morris Disability Questionnaire, Oswestry Disability Index), and data were pooled and presented as standardized mean differences (SMDs) with 95% CIs. Clinical significance was defined as an MD of at least 20 points on a 0-to 100point scale [23]. An SMD of 0.2 was considered a small treatment effect, whereas 0.4 was considered a moderate and 0.8 a large treatment effect [24,25]. Between-study statistical heterogeneity was assessed with the I 2 , with values <40% considered low heterogeneity, values of 40% to 75% moderate heterogeneity, and values >75% considerable heterogeneity [26].

Study Selection
A total of 2,880 studies were identified. After removal of duplicates, 1,350 titles and 237 abstracts were screened by the two independent reviewers. A total of 61 studies were identified for full-text review; of those, 22 articles reporting on 17 randomized controlled trials (six studies on knee OA and 11 studies on LBP) were included in the review ( Figure 1). Studies identified that required translation via Google Translate were subsequently excluded on the basis of the inclusion and exclusion criteria.

Intervention
Knee OA HC dosage varied considerably, ranging from six sessions over 6 months [33] to nine sessions over 2 years [40], and additional intervention components differed extensively among studies. One study compared the addition of six HC telephone sessions and five additional physiotherapy sessions vs five physiotherapy sessions alone [33]. The second study compared an intensive supervised walking program (65 minutes, three times a week for 1 year) vs the same program with an additional 20 education sessions, six group counseling sessions, and 12 monthly phone counseling sessions, vs the provision of an information pamphlet alone [39]. Two studies compared one session of counseling vs 8-10 telephone HC sessions over 6 to 24 months [37,40]. One study compared the use of a Fitbit (Fitbit, San Francisco, CA, USA) device vs four individual HC calls over an 8-week time frame vs no treatment [41]. The final study compared one or more sessions of an existing nurse-led phone education program vs 5-10 HC calls over 6 months [35] (Table 1).
Low Back Pain HC provided in the LBP studies was generally of shorter duration but greater intensity than that provided in knee OA studies. HC dosage varied from one to three sessions [28,31] to as many as 10 sessions over a 5-to 8-week period [27,29,43]. Two studies provided HC for 6 months, ranging from 10 to 16 sessions [30,38]. Five studies incorporated HC into the usual physiotherapy sessions [27,29,31,42,43]. Four studies conducted HC remotely via telephone [30,32,36,38]. One of these conducted an initial face-to-face physiotherapy session to provide back-specific education, advice, and behavioral change techniques, with all other contacts occurring via telephone [38]. Another provided inpatients with two smallgroup sessions followed by two or more HC calls once they were outpatients [32]. In most studies, physiotherapists provided the HC (Table 1).

Methodological Quality of the Studies
The methodological quality of the included studies is shown in Table 2. All studies used random allocation of participants, with concealed allocation in 11 studies [27, 28, 30, 32, 33, 35-38, 40, 41]. Baseline characteristics were similar in all studies, except one that used Zelen's design in which participants were randomized before consent, but analyses were adjusted for baseline  differences [30]. PEDro scores ranged from 5/10 to 8/10, with an average PEDro score of 6.71 (SD: 1.05). The overall quality of the studies was considered to be moderate.

Intervention Effectiveness
The results of individual studies are listed in Supplementary Data Appendix 2. Meta-analyses were considered possible for knee OA pain and disability at mid-and long-term follow-up (see Figures 2, 3, 6, and 7) and chronic LBP pain and disability at short-and midterm follow-up (see Figures 4, 5, 8, and 9).

Strength of the Body of Evidence
The studies included in the meta-analyses were rated according to the GRADE criteria [22] in Table 3. The quality of evidence for LBP was considered very low for short-term pain and short-term to mid-term pain and disability and low for short-term disability. The evidence was marked down for risk of bias (50% studies of low methodological quality with individual PEDRo scores <7/10); imprecision (pooled analyses with combined sample sizes of 400); and suspected publication bias, as evidenced by the small available study sample sizes and absence of published study protocols or clinical trial registrations. The quality of the evidence was rated low for disability at short-term follow-up, as described above, but was not marked down for imprecision.
The quality of evidence for knee OA was considered low for pooled analyses on pain and long-term disability and very low for the pooled analysis of mid-term disability. The evidence was marked down because of    imprecision (meta-analyses contained combined sample sizes 400), and publication bias was not detected. The pooled analysis on disability at the mid-term follow-up was also marked down because of inconsistency (unexplained heterogeneity [I 2 > 50%]).

Pain
Knee OA All studies (n ¼ 891) assessed pain, with a meta-analysis performed on three studies (n ¼ 463) at mid-and longterm follow-up. These studies compared 5-10 sessions of HC over 6 months to existing physiotherapy or advice, education, and physical activity guidance [33,35,37]. The results show that the addition of HC provided no significant improvement in pain at mid-term follow-up ( Figure 2; MD: -0.29; 95% CI: -1.10 to 0.51; P ¼ 0.48; z ¼ 0.71; I 2 ¼ 69%) or at long-term follow-up (Figure 3; MD: -0.29; 95% CI: -0.82 to 0.24; P ¼ 0.29; z ¼ 1.07; A short-term, statistically significant difference favoring HC was found in one study providing baseline physician recommendation and nine HC calls over 2 years (MD: -0.97; 95% CI: -1.77 to -0.17; P ¼ 0.02) vs physician recommendation alone [40]. The other studies found no difference between groups at any time point [41,45]. Studies were excluded from the meta-analysis because of incomparable intervention dosages [40,41] or incomparable additional therapies provided [39].

Low Back Pain
Eight studies (n ¼ 2,606) assessed pain [28-32, 38, 42, 43]. Short-term outcomes of three studies (n ¼ 295) that compared physiotherapy vs physiotherapy plus HC / motivational strategies for chronic LBP were pooled [29, 42, The results show that HC added to usual physiotherapy care provided no further benefit compared with physiotherapy alone for pain reduction in the short term ( Figure 4; MD: -2.01; 95% CI: -5.81 to 1.79; P ¼ 0.30, z ¼ 1.04; I 2 ¼ 0%). Pooled analysis of mid-term followup (two studies; n ¼ 219), however, found the addition of HC to usual physiotherapy care resulted in a small but statistically significant effect on pain reduction compared with usual physiotherapy alone ( Figure 5; MD: -7.57; 95% CI: -10.08 to -5.07; P ¼ <0.001, z ¼ 5.93, Three studies provided longer follow-up data (12 months [34], 24 months [30], and 5 years [46]), but no meta-analysis could be performed because of incompatible outcome measures (pain intensity vs number of days in pain) and different time points. One that compared HC and physiotherapy vs physiotherapy alone [29] found continued significant improvements in pain at 12 months compared with the control (MD: -15.50; 95% CI: -27.82 to -3.18; P ¼ 0.02). Data collected at 5year follow-up showed that the intervention group continued to have significantly less pain than at baseline, whereas the control group did not [46]. Reasons studies were excluded from the meta-analysis included having a shorter intervention length [27,31], including acute LBP [36], or using interventions that were significantly different [30,38].

Secondary Outcomes
No significant differences were found in any other secondary outcome that could be pooled in a meta-analysis at any time point for either knee OA or LBP. Only one study reported a significant improvement in quality of life for LBP at 12 months [28], and one study reported a significant increase in the number of days on which participants exercised [30] favoring the intervention.
Only two studies reported the cost and savings for the intervention [35,47]. One knee OA study reported that the direct cost of the intervention in addition to the existing nurse-led support was AUD514, with no evidence of savings in other health service resources [35]. One LBP study provided an economic evaluation of the intervention [47] and reported a mean intervention cost of AUD708 per person. Although that study found no difference between the intervention and control groups for any clinical outcomes (pain, disability, weight, or body mass index), the study found significantly lower costs in terms of health care services used for LBP (AUD292; 95% CI: À872 to À33), medication costs (AUD30; 95% CI: À65 to À4), and work absenteeism (AUD1,000; 95% CI: À3,573 to À210) favoring the intervention group [50]. The incremental cost-effectiveness ratios for quality-adjusted life years indicated that one quality-adjusted life year gained was associated with a societal cost saving of $31,087, demonstrating that the intervention was on average less costly and more effective than usual care [47]. Society's willingness to pay per unit of effect gained was not determined [47].

Discussion
The aim of this systematic review was to investigate the effect of HC on pain and disability for knee OA, hip OA, and LBP. A comprehensive search strategy across multiple databases identified 17 original studies. No study was found that included participants with hip OA that performed subgroup analysis of the individual musculoskeletal condition. Consequently, there is currently no evidence for the effect of HC specifically for the population with hip OA.

Clinical Implications
The pooled results suggest that HC could have favorable outcomes for both the chronic LBP and knee OA populations at different time points, though the level of evidence remains low to very low. As the CIs crossed the predetermined threshold for clinical significance for both disability and pain (i.e., an MD of 20 points on a 0-to 100-point scale [23] and an SMD of 0.2 [25]), we conclude that it is uncertain whether HC provides small but clinically important benefits for disability or pain (see Table 4 for assistance in the interpretation of effect sizes expressed as an SMD). Moreover, we lack information on the long-term effect and whether this strategy works better alone or in combination with other active management strategies.
Only two studies reported on health service costs after the provision of HC alone for LBP [47] and knee OA patients on a specialist waiting list [37]. No cost savings were found for the population with knee OA [37]. Conversely, significant savings were found for the population with LBP [47], which might be more meaningful and crucial to the health care system for this population and requires further investigation.
Six LBP studies [27,29,31,36,42,43] and one on knee OA [33] reported small between-group differences favoring HC in addition to physiotherapy vs physiotherapy alone, suggesting that adding HC to active interventions might not result in substantial additional improvements. Interestingly, two studies compared HC vs minimal care in patients with LBP [38] and knee OA [37] with high body mass index on an orthopedic waiting list. The studies found no difference between groups in any outcome at any time point except for the mental health component of quality of life at 6-month follow-up in participants with knee OA. The authors of these studies acknowledged the use of "non-disease-specific healthy lifestyle intervention," though training for LBP and knee OA was given to the health coaches providing the intervention [37,38]. They further identified that to adequately support and manage pain and facilitate lifestyle changes in the populations with chronic LBP and knee OA, a more intensive, disease-specific telephone intervention might be required [37,38].
It is important to acknowledge that HC was usually found to be delivered simultaneously with physiotherapy management for LBP [27,29,31,34,36,42,43]. Providing HC when adherence levels tend to drop after completion of physiotherapy could prove more effective by providing greater accountability and support for selfmanagement. A pilot study reviewed the use of HC for consistent pain after completion of hospital-based physiotherapy to reduce care-seeking, pain, and disability [48]. It found that HC was well accepted by patients and might help to reduce care-seeking in this population, but the outcome data could not be analyzed because the study was underpowered for such evaluation. Studies on knee OA delivered HC with a diverse range of simultaneous management strategies, including physiotherapy, nurse-led support, supervised walking programs, and physical activity tracking devices. Despite the range of additional management strategies, the present review has shown that the addition of HC could also be of benefit to this population.
Optimal HC dosage continues to require further clarification and standardization. Previous systematic reviews for telephone-delivered physical activity and dietary behavior change found that 12 or more phone calls over a 6-to 12-month period have been shown to improve behavior [18,49]. Studies in the present review reporting significant improvements in pain favoring HC varied significantly in their frequency and duration for both LBP and knee OA. The programs for LBP, though intense, might have been too short, and some of the knee OA programs might have lacked effective intensity. Consequently, no recommendations on optimal frequency and duration for either population can be made.
The pooled knee OA studies, however, provided 5-10 HC sessions over a 6-month period, which could help inform future knee OA HC studies and clinical programs.
All the knee OA studies used telephone-based HC, with only four LBP studies favoring this delivery mode [30,32,36,38], despite telephone-based interventions demonstrating the potential to provide greater patient access and overcome barriers to accessing care, including time and travel requirements for face-to-face appointments [50]. Their value has been further highlighted during the global coronavirus disease 2019 (COVID-19) pandemic. A previous report suggested that telephone counseling hampered the ability to establish patient rapport [33], yet provision of face-to-face consultations via online methods might alleviate this issue. A proof-ofconcept study offered fortnightly motivational interviewing via videoconference to participants with knee OA to improve physical activity [51]. That study (n ¼ 61) found significant improvements in the intervention group in moderate to vigorous physical activity, daily steps, Knee Injury and Osteoarthritis Outcome Score activity of daily living subscale score, and quality of life in the short term. Despite this apparent favorable finding, long-term effects of delivering this type of intervention via videoconferencing remain unknown.

Comparison with Previous Literature
To the present authors' knowledge, there have been three previous systematic reviews in this area. The first was limited to motivational interviewing and included general musculoskeletal conditions (including fibromyalgia, osteoporosis, and LBP) [52]. The authors of that review identified the need for further well-designed randomized controlled trials to measure effectiveness. The second reviewed HC for LBP and concluded that there was a small and inconsistent body of literature supporting HC for this population [53]. The third reviewed the effect of HC delivered by physical therapists to patients with a range of disorders (LBP, diabetes, coronary artery disease, Parkinson's disease, rheumatoid arthritis, and stroke) and found that HC had positive effects on physical activity in 6 of the 11 included studies [54]. Because of the heterogeneity of the studies, however, no metaanalysis could be performed. The secondary outcomes in this present review suggest that the addition of HC to usual physiotherapy care might result in improved exercise compliance in the population with LBP in the short [31,43] and middle term [29] and in the population with knee OA in the short term [45] (Results in online appendix). These results align with a previous systematic review that showed improved compliance with therapeutic regimes in other non-musculoskeletal conditions [18]. Our systematic review identified seven additional LBP studies [29-32, 34, 38, 42] and a total of 13 studies not previously included in reviews [29][30][31][32][33][34][35][37][38][39][40][41][42]. All studies in our review were of moderate to high methodological quality, and meta-analyses have provided more precise pooled estimates of treatment effect.

Strengths and Limitations
Strengths of the present review include the extensive search strategy across multiple databases, followed by hand searches of included studies and relevant systematic reviews, with thorough screening and data extraction by independent reviewers who followed a pre-registered protocol. Alerts from each database were set to ensure the literature search was valid up to the date of acceptance for publication. We were also able to identify a larger number of studies involving comparable populations that enabled the pooling of data.
The limitations of this review should also be acknowledged. First, despite the extensive search involving multiple databases, it is possible that some studies were missed. The incompatible outcome measures used in the included studies (e.g., number of days in pain vs pain levels, differences in dosage, and additional interventions) prevented the remaining studies from being included in the meta-analysis. In addition, although the minimum number of studies with which to perform a meta-analysis was met [55], the studies included in the meta-analyses contained small participant numbers, meaning that they were potentially underpowered, with the possibility of producing a false negative result and therefore a Type II error [56].
Finally, there is no standardized content for HC interventions within the health care setting, and the studies in this review used numerous underpinning theories. Consequently, the content of the interventions differed. This limitation, however, is not dissimilar to interventions included in other systematic reviews, such as exercise for LBP [57] and lower limb OA [58]. Exercise programs in these reviews varied significantly and included individual and group programs, specific muscle strengthening, global strengthening, graded activity programs, and the addition of advice or behavioral techniques [57,58]. It is likely that the differing intervention strategies in the studies included in the present review reflect differences that occur in real practice. There was also heterogeneity in the content of the control or comparator program in each study. Differences in the "usual physiotherapy care" given, whether this involved active or passive strategies or advice, also reflect the differences that occur in managing these conditions in real practice.
As previously found, few studies adequately described the training of the health professionals delivering the intervention, preventing replication [52,53]. Assessments of treatment fidelity were reported in only three studies [27,31,33]. Holden et al. [53] observed that the efficacy of treatment was difficult to assess because of the lack of accurate treatment fidelity assessment, which continues to be valid in the present review. Future randomized controlled trials need to adequately report intervention content, training of professionals delivering the intervention, and treatment fidelity measures.
The authors of the present review acknowledge that the use of Google Translate for studies not published in English was likely open to misinterpretation and errors in translation that might have resulted in some studies being incorrectly excluded from this review.

Conclusion
There is evidence that HC could provide pain relief in the middle term and improvements in disability in the short and middle term for patients with chronic LBP. Although the body of literature for knee OA remains small, there is low-level evidence that HC could provide improvements in disability in the long term. Studies evaluating the effects of HC for hip OA are notably lacking. To date, evidence appears unclear as to whether improvements from adding HC to management programs is clinically important.

Authors' Contributions
JLP: protocol design, database search, identification of eligible studies, data extraction, assessments of quality and analysis of results, extracting relevant data and conducting quality assessments, synthesizing results, and authoring of manuscript original draft. GV: duplicated eligibility screening, data extraction, and quality assessments. JAM: synthesis and interpretation of results. PHF: interpretation of results. MLF: review inception, protocol development, and interpretation of results. All authors reviewed prior drafts of the systematic review, discussed and commented on the results, and contributed to the manuscript.