Background

Pelvic Girdle Pain (PGP) has been defined as “pain between the posterior iliac crest and the gluteal fold, particularly in the vicinity of the sacroiliac joints, and pain may radiate to the posterior thigh and can also occur in conjunction with/or separately in the symphysis” [1] (pp797). In the past, it has sometimes been considered a subgroup of low back pain (LBP); however, PGP includes also pain at the pubic symphysis and is therefore considered a different entity. The term lumbopelvic pain (LPP) is a broader term that has been used to describe LBP and/or PGP without differentiation between the two groups [2].

Pelvic Girdle Pain is a common complaint during pregnancy, affecting 23 to 65% of women depending on how it is measured and defined [3, 4]. Although many women recover after the birth, 17% have continuing symptoms 3 months postpartum [2] and 8.5% have not recovered 2 years postpartum [5]. In Sweden, in a cohort of 371 women with PGP, 10% of women still had symptoms 11 years after the birth [6]. In another Swedish cohort, 40.3% had long term pain in the low back or pelvic girdle area 12 years postpartum [7]. Additionally, PGP is one of the leading causes of sick leave during pregnancy [7,8,10], resulting in large economic costs to families and society.

Studies examining the effectiveness of interventions for PGP measure different outcomes, making it difficult and sometimes impossible to pool data in meta-analysis to increase the certainty of effect measures [11, 12]. To address this issue, an international consensus-based Core Outcome Set (COS) for PGP is being developed (registration: http://www.comet-initiative.org/studies/details/958) [13]. The systematic review presented here forms the first key part of the PGP-COS (Pelvic Girdle Pain – Core Outcome Set) study and provides a structured overview of the outcomes and outcome measurements that are used across PGP as well as LPP (since this includes PGP) intervention studies and systematic reviews. It will feed into the larger PGP-COS study by providing a preliminary list of outcomes that will be included into an online Delphi survey and face-to-face consensus meeting to identify a final COS for PGP.

The objective of this systematic review was to identify and examine what outcomes are measured and reported, and how outcomes are measured, in intervention studies and systematic reviews of interventions for PGP.

Methods

The protocol for this systematic review was published as part of the PGP-COS study protocol [13]. Criteria for considering papers for inclusion in the systematic review are outlined in Table 1. A second objective (To compare outcomes measured in intervention studies and systematic reviews on PGP to outcomes measured in studies on LPP) was added post hoc, since many studies that we identified in preliminary searches did not differentiate between LBP or PGP, and it was considered important to compare outcomes measured in these studies since LPP includes PGP. We analysed and have presented the results by the subgroups PGP and LPP.

Table 1 Inclusion criteria
Full size table

Search methods & study selection

The following databases were searched on the 11th May 2018 (from inception): PubMed, the Cochrane Library, PEDro and Embase. Details of search terms used for each database can be found in Additional file 1. No language or time filters were applied. We screened reference lists of included studies for further relevant studies. Citations were exported to Endnote and duplicates were removed. Two review authors (FW, MO) reviewed each citation independently against the inclusion criteria in two stages: (a) title and abstract screening and (b) full text screening, using Covidence software [14]. Disagreement was resolved through discussion.

Data collection and synthesis

All outcomes (and their verbatim definitions) examined in the included studies were extracted by two reviewers (FW, MO) independently and their corresponding outcome measurement instruments/methods, where reported, were also recorded. The quality of outcome reporting was assessed using the six questions proposed by Harmen et al. [15] and this was conducted by two independent reviewers (FW, MO). The outcomes were then grouped into core outcome domains using the OMERACT (Outcome measures in rheumatology) filter 2.0 framework: (a) life impact; (b) resource use/economic impact; (c) pathophysiological manifestations and (d) death [16]. This framework aims to provide a structure for measuring outcomes and developing core outcome sets. Within the OMERACT framework ‘adverse events’ should also be flagged alongside the core domains. We therefore grouped adverse events into a separate domain [16]. The findings are synthesised and reported by these core domains, for PGP and LPP separately, for comparison. We have reported this systematic review according to the PRISMA guideline [17].

Results

Screening and selection of included papers

A total of 7092 studies were identified from the initial search after removal of duplicates. We excluded 6842 studies during title and abstract screening, and the full texts of 250 articles were reviewed. A total of 145 studies were excluded at full text selection. Reasons for exclusion were: duplicates (n = 30), the wrong study design (n = 67), published in a language other than English (n = 7), examining LBP only (n = 30), or the wrong study population (n = 11). A further two studies were identified for inclusion from screening reference lists of included studies, with a total of 107 studies being included in the analysis. Figure 1 provides a flow diagram detailing the results of the search and selection process.

Fig. 1
figure 1

Screening and selection process of articles

Full size image

Characteristics of included studies

Of the 107 studies included in the review, 31 were systematic reviews, 61 were Randomised Controlled Trials (RCTs) (including one follow up study [18] of another included study [19]), 11 were non-controlled intervention studies, two studies were non-randomised controlled studies [20, 21] and one study identified itself as a quasi-randomised design because no blinding of participates took place [22]. A total of 33 studies on PGP and 74 studies on LPP were included. Studies were published between the year 1991 and 2018, with 54% of studies published in the last 5 years. Studies were undertaken in a variety of geographical locations, across Europe, North and South America, Australia, New Zealand, Asia and Africa, with the highest percentage in Europe (66%), particularly in Sweden and Norway (30% of all included studies). Of the studies that focused on PGP, 24 studies (72.7%) included a physical examination as a requirement for the diagnosis of PGP. In comparison, only 18 (24.3%) of the studies focusing on LPP included a physical examination as a requirement for a diagnosis of LPP. Additional details of the characteristics of included studies can be found in Additional file 2.

An overview of the quality of reporting of the included studies [15] is presented in Tables 2 and 3, with higher quality reporting indicated by a yes vote, where applicable. All PGP studies (100%) and most LPP studies (94%) clearly reported and defined the primary outcome(s). About two thirds of studies did not differentiate between primary and secondary outcomes, making questions three and four not applicable. For transparency, the full quality of reporting assessment of each study determined by the six questions outlined by Harmen et al. [15] can be found in Additional file 3.

Table 2 Quality of reporting of included studies on PGP
Full size table
Table 3 Quality of reporting of included studies on LPP
Full size table

Outcomes and outcome measurements

A total of 46 outcomes were identified and categorised into core domains using the OMERACT filter 2.0 framework: ‘life impact’, ‘resource use/economic impact’, ‘pathophysiological manifestations’ and ‘death’. No outcomes were identified in the core domain ‘death’, but ‘adverse events’ outcomes were identified. Outcomes and their corresponding outcome measurements are presented separately for studies that focused on PGP or focused on LPP in the Tables 4, 5, 6 and 7. Of the 46 outcomes identified, 26 were in the life impact core domain (Table 4), five were in the resource-use/economic impact domain (Table 5), 11 were in the pathophysiological domain (Table 6), and four outcomes were classified in the adverse events domain (Table 7).

Table 4 Outcomes and outcome measurements identified in the ‘Life impact’ core domain for PGP and LPP respectively
Full size table
Table 5 Outcomes and outcome measurements identified in the ‘Resource-use/ economic impact’ domain for PGP and LPP respectively
Full size table
Table 6 Outcomes and outcome measurements identified in the ‘Pathophysiological manifestations’ domain for PGP and LPP respectively
Full size table
Table 7 Outcomes and outcome measurements identified in the ‘Adverse events’ domain for PGP and LPP respectively
Full size table

The differences in the number of outcomes reported in studies on PGP and studies on LPP by core domain are outlined in the Table 8. Notable, psychological outcomes and economic outcomes were more commonly measured in LPP studies compared to PGP studies. A further comparison of the different outcomes reported in each domain between PGP and LPP studies is outlined in Additional file 4.

Table 8 Outcome count by core domain for PGP and LPP studies
Full size table

Discussion

A large number of primary intervention studies (n = 76) and systematic reviews (n = 31) were identified. A total of 46 outcomes were measured across all studies. The majority of outcomes related to the ‘life impact’ core domain of the OMERACT framework. This would be expected considering the nature and main symptoms of PGP and LPP. Within the life impact core domain, pain intensity was the most commonly reported outcome in both PGP and LPP studies, followed by the outcomes function and disability. Fifteen (20%) studies on LPP included psychological outcomes versus only three (9%) PGP studies. This is likely because LPP includes LBP, which has had a strong psychosocial focus within the literature the past few decades, including on aspects such as fear avoidance and catastrophising. It might be that PGP is often perceived as a transient condition related to pregnancy and researchers therefore assess fewer psychosocial factors that are involved in developing chronicity. However, not all women recover and PGP can persist postpartum [2, 5-7, 118]. Moreover, PGP has been associated with psychological factors including emotional distress [119], depression [118, 120] and anxiety [118]. Only 14 studies/reviews (13%) examined any adverse events. This is contrary to current recommendations to always assess adverse events for any intervention study or systematic review [121, 122].

A range of outcome measurements were used across studies to measure certain outcomes. For example, pain intensity alone was measured using 10 different outcome measurement instruments, and function was examined using 13 different tools across the studies. This emphasises not only the need for a COS but also for consensus on how to measure the identified COS. This systematic review will contribute to a list of initial outcomes to be included in a multistakeholder, international Delphi survey to reach consensus on a PGP-COS. Subsequently, the next part of the PGP-COS study will determine ‘how’ best to measure the developed COS [13].

This systematic review also showed that the included intervention studies/reviews often use different terminology to describe the same outcomes. For example, when examining the measurement tools for the outcomes ‘function’ and ‘disability’, the same tools are frequently used. While some studies use the term ‘function’ and others ‘disability’, most studies do not provide a clear definition of the terms. Another example of where there is clearly inconsistency in terminology and a lack of definitions in original manuscripts is for the outcomes ‘quality of life’ and ‘health status’. Again, the same measurement instruments tend to be used and terms seem to be used interchangeably across different studies. This observed inconsistency strengthens the rationale for the development of an agreed PGP-COS.

Chiarotto et al. [123] published a COS for non-specific LBP in 2015 and, while there was some overlap in findings, the list of outcomes they identified from the LBP literature differed significantly from our findings of the outcomes measured in the PGP/LPP literature. They identified the following outcomes in LBP studies that were not identified in our review of PGP/LPP studies: death, cognitive functioning, social functioning, sexual functioning, satisfaction with social role and activities, pain quality, independence (Life impact); informal care, societal services, legal services (Resource-use/ economic impact); muscle tone, proprioception, spinal control, and physical endurance (Pathophysiological manifestations). Outcomes that we identified in this review of PGP/LPP studies but that were not identified in the review of the outcomes measured in the LBP literature [123] were: Self-efficacy, confidence, patient expectations of treatment (Life impact); and anthropometric measures (weight/height), pregnancy and maternal outcomes, surgical outcomes (Physiological manifestations). Some of the observed differences could be put down to differences between PGP and LBP. However, differences in outcomes seem largely arbitrary instead of relating to the distinguishing features of PGP and LBP. Similarly, when comparing studies examining PGP only with studies examining LPP in this systematic review, the reason for the observed discrepancies in the outcomes chosen by studies’ authors are mostly unclear. This supports using the outcomes identified in this review only as an initial list for the consensus process to develop a PGP COS, allowing for other outcomes to be added by all stakeholders including patients, clinicians, researchers, service providers and policy makers.

Conclusions

Studies and systematic reviews examining the effectiveness of interventions for PGP and LPP assess a range of outcomes, predominantly pain intensity and disability/function, and use a large variety of outcome measurement instruments. Few studies examine adverse events and economic outcomes. Not only do different studies often measure different outcomes, authors also rarely define outcomes and terminology for outcomes varies, making comparison of study findings very difficult.