Content uploaded by Bengt Sturesson
Author content
All content in this area was uploaded by Bengt Sturesson on Sep 14, 2019
Content may be subject to copyright.
Idiopathic Pelvic Girdle Pain as it Relates to the Sacroiliac Joint
Building a Collaborative Model of Sacroiliac Joint Dysfunction and
Pelvic Girdle Pain to Understand the Diverse Perspectives of
Experts
Paul W. Hodges, PhD, MedDr, DSc, FAA, FACP, BPhty(Hons), Jacek Cholewicki, PhD,
John M. Popovich Jr. DPT, PT, PhD, Angela S. Lee, MPH, Payam Aminpour, MS,
Steven A. Gray, PhD, Michael T. Cibulka, PT, DPT, FAPTA, OCS,
Mel Cusi, MBBS, FACSEP, FFSEM, PhD, Brian F. Degenhardt, DO,
Gary Fryer, PhD, BSc(Osteo), Annelie Gutke, PT, PhD, David J. Kennedy, MD,
Mark Laslett, PhD, NZRPS, FNZCP, DipMT, Dip MDT, Diane Lee, BSR, PT,
Jan Mens, MD, PhD, Vikas V. Patel, MD, Heidi Prather, DO, Bengt Sturesson, MD, PhD,
Brit Stuge, PT, PhD, Andry Vleeming, PhD
Abstract
Background: Pelvic girdle pain (PGP) and sacroiliac joint (SIJ) dysfunction/pain are considered frequent contributors to low back pain (LBP).
Like other persistent pain conditions, PGP is increasingly recognized as a multifactorial problem involving biological, psychological, and social
factors. Perspectives differ between experts and a diversity of treatments (with variable degrees of evidence) have been utilized.
Objective: To develop a collaborative model of PGP that represents the collective view of a group of experts. Specific goals were to
analyze structure and composition of conceptual models contributed by participants, to aggregate them into a metamodel, to ana-
lyze the metamodel’s composition, and to consider predicted efficacy of treatments.
Design: To develop a collaborative model of PGP, models were generated by invited individuals to represent their understanding of
PGP using fuzzy cognitive mapping (FCM). FCMs involved proposal of components related to causes, outcomes, and treatments for
pain, disability, and quality of life, and their connections. Components were classified into thematic categories. Weighting of connec-
tions was summed for components to judge their relative importance. FCMs were aggregated into a metamodel for analysis of the
collective opinion it represented and to evaluate expected efficacy of treatments.
Results: From 21 potential contributors, 14 (67%) agreed to participate (representing six disciplines and seven countries). Partici-
pants’models included a mean (SD) of 22 (5) components each. FCMs were refined to combine similar terms, leaving 89 components
in 10 categories. Biomechanical factors were the most important in individual FCMs. The collective opinion from the metamodel
predicted greatest efficacy for injection, exercise therapy, and surgery for pain relief.
Conclusions: The collaborative model of PGP showed a bias toward biomechanical factors. Most efficacious treatments predicted by
the model have modest to no evidence from clinical trials, suggesting a mismatch between opinion and evidence. The model enables
integration and communication of the collection of opinions on PGP.
Introduction
Pain in the lumbopelvic region is an enormous issue
globally and the leading cause of disability in the
developed and developing world.
1
Despite the enormity
of this problem, much remains to be learned about the
underlying causes for the condition, and the most effec-
tive preventative and treatment strategies. A major issue
PM R xx (2019) 1–13 www.pmrjournal.org
© 2019 American Academy of Physical Medicine and Rehabilitation
https://dx.doi.org/10.1002/pmrj.12199
is that in most cases the underlying cause/mechanism is
unknown.
2
Many different structures may be involved
2–4
and pain may be maintained by central sensitization
rather than by ongoing nociceptive input from the periph-
ery.
5
Of the many potential sources of nociceptive input
that may contribute to the pain experience, one struc-
ture that has been particularly controversial is the sacro-
iliac joint (SIJ). Although some argue that dysfunction of
the SIJ is a common contributor to low back pain (LBP)
(45% of individuals with chronic LBP below L5),
6
others
consider it to have infrequent involvement (10%-13% of
individuals with chronic LBP).
7–9
Dysfunction and pain associated with the SIJ has been
suggested to have specific characteristics (eg, pain loca-
tion and provoking activities)
9–11
and is considered to be
prevalent in conditions such as pelvic girdle pain (PGP:
defined as pain experienced between the posterior iliac
crest and the gluteal fold, particularly in the vicinity of
the SIJ), often in association with pregnancy.
10
Differen-
tial diagnosis for other causes of LBP has been based on
responses to specific pain provocation
11
and movement
tests.
12
Multiple disorders and mechanical dysfunctions
have been proposed. From a movement perspective, dys-
functions such as joint laxity,
13
failed load transfer
14
and
abnormalities of joint alignments
15
have been proposed,
with variable evidence.
16,17
The SIJ can also be a charac-
teristic site for specific rheumatological conditions (eg,
ankylosing spondylitis).
18
Because of the diversity of pro-
posed mechanisms, there is parallel diversity of treat-
ments offered, again with inconsistent evidence.
2,10,19
Current literature lacks consensus regarding mecha-
nisms, contributing factors, and treatments, and vastly
different views are held by different professional groups.
It was the contention of our study group that advances
could be made in our understanding of PGP (the term
selected in this study to include SIJ dysfunction/pain) by
building a model that included the diversity of conceptu-
alizations of the condition.
Collaborative modeling is a participatory method that
aims to gather the diverse opinions of individuals to build
a single model inclusive of all ideas that delineate the
scope of a problem.
20
Fuzzy cognitive mapping (FCM) is
a collaborative modelling technique that elicits partici-
pants’mental models about a problem through a
networked structure of concepts and their causal inter-
dependency.
21
The term “fuzzy”relates to the fact that
each connection was given a weighting, based on expert’s
opinion, to indicate the strength and direction of effect.
This approach incorporates a wide range of conceptu-
alizations into a standardized format that can be used to
illustrate and interpret the problem, and even to simu-
late possible solutions.
20
Although the main objective of
collaborative modeling is to synthesize and share knowl-
edge, the analysis of the structure, composition, and
functionality of FCM models enables identification of core
assumptions, evaluation of the relative importance
placed on different concepts and testing of various
scenarios, such as the impact of a treatment. This
approach has recently been used to build a collaborative
model of LBP using opinions of a broad range of experts.
22
The resultant model has highlighted that despite various
disciplinary backgrounds of contributors, psychological
features are considered to have the strongest importance
in LBP.
The overall objective was to develop a collaborative
model of PGP/SIJ dysfunction with contribution of
experts across a diverse range of disciplines. The specific
aims of this study were to analyze the structure and com-
position of the models generated by individual contribu-
tors, to aggregate them into a metamodel, to analyze
the composition of the metamodel, and to use the meta-
model to identify the group’s overall estimation of the
relative efficacy of treatments when all contributors’
opinions are combined.
Methods
To build a collaborative model, 14 individuals with
research and clinical expertise in PGP (purposefully
drawn from different disciplines), generated individual
FCMs that incorporated all of the components that they
considered to be relevant for the problem (eg, causes,
consequences, and treatments) and the connections
between them.
FCMs are semi-quantitative models that allow for the
analyses of the composition, structure, and behavior.
23
The composition of the FCMs includes the qualitative
concepts participants used to characterize the problem.
Accordingly, researchers frequently suggest the compari-
son of composition of the models to quantify similarities
or differences of contents.
24,25
Moreover, examining the
structural characteristics of FCMs demonstrates how peo-
ple view the interconnectedness of system components
through a network of nodes and connections. These ana-
lyses aim to obtain valuable information from the struc-
ture of the maps through a set of network metrics (eg,
number of nodes, number of connections, centrality of
concepts, density, and complexity).
21,26
These metrics
can be interpreted as indices for comparing the structure
of FCMs and therefore highlighting the cognitive diversity
of participants.
FCM models also enable the quantitative assessment of
the behavior of the system using simulations.
27
Scenarios
can be run using FCM computation that enable simulation
of impact of a particular input (eg, the impact of a treat-
ment or change in a risk factor). The differences in key
elements of the system when specific aspects of the
model are changed are represented in the results of sce-
nario analyses
23,26,28,29
which characterize how partici-
pants perceive the behavior of the system.
Potential participants were identified by members of the
investigative team (P.H., J.C., J.P.) through extensive sea-
rch of the literature and speaker’s lists of relevant confer-
ences, and through discussion with other experts in PGP.
2Collaborative Model of SIJ Dysfunction
Potential contributors were considered eligible for
inclusion if they represented major disciplines in res-
earch/management of PGP and there was evidence that
they had made a substantial and ongoing contribution to
the literature related to PGP, as evidenced by at least
two of the following: (1) Contribution to at least three
published works in the preceding 3 years; (2) Keynote/
invited presentations at major meetings related to
LBP/PGP; (3) Contribution to major working groups/
committees of LBP organizations; (4) Contribution to
organization of major LBP/PGP meetings/conferences;
(5) Contribution to LBP/PGP texts; and (6) Contribution
to clinical practice guidelines/systematic reviews.
From a total of 21 invited contributors, 14 (67%) agreed
to participate (Table 1). The study was granted exemp-
tion from the Michigan State University Institutional
Review Board.
The individual FCMs were built during a 1–1.5-hour
semi-structured interview using videoconferencing
and the freely available Mental Modeler software
developed by SG.
30
Each participant was initially pres-
ented with three components:-“Pain,”“Disability,”
and “Quality of Life,”representing main outcomes of
living with pain associated with PGP. Participants were
then asked to name additional components (major fac-
tors contributing to PGP) that they considered would
“directly affect”these three outcome components,
and to consider all possible interactions between com-
ponents (including feedback loops) in their model. As
anewcomponent was added, the participant was
required to confirm the direction of the relationship,
whether it caused an increase or decrease in the com-
ponents it was connected to, and the strength of each
connection between −1 and 1. After completion of the
inclusion of components, participants were asked to
identify the treatments that they considered would
impact directly or indirectly the three main outcomes
of PGP, identify pathways for this impact (connections),
and to nominate the strength of these connections.
Sessions were recorded with the consent of the
participants for later clarification of meaning of ele-
ments of the model.
The study core team (P.H., J.C., J.P., A.L.) reviewed
the components present in the initial 14 FCMs and modi-
fied them into a standardized format either by using the
terms selected by the participants or a term of synony-
mous meaning from a list of standardized terms. This
was done to enable aggregation of FCMs contributed by
individual experts. The standardized terms were defined
based on outcomes from several phases of consultations
and a consensus meeting with the participants in the sim-
ilar study concerning LBP.
22
As a result of this process, the
structure of the original FCMs (structural features such as
the number of components and connections)orcomposi-
tion (the issues represented by the components)didnot
change, but the components with similar meaning in dif-
ferent FCMs were combined to form a smaller number of
standardized unique terms (from the original 312 to 89).
These terms were then allocated to 10 categories
(Table 2). (For a more detailed description of the process
of refining terms and categories, see reference 22.)
The structure and composition of FCMs were analyzed
using the graph theory.
20
The following metrics were cal-
culated (adopted from Cholewicki et al
22
).
FCM structure:
1. Total Components (N) –number of components in-
cluded in an FCM
2. Total Connections (C) –total number of connections in
either direction included in an FCM
3. Density (D) –number of connections as a proportion of
the number of all possible connections in both direc-
tions (see Appendixs S1 for equation)
4. Connections per Component –average number of con-
nections in either direction per component
5. Complexity Score –calculated as the ratio of Receiver/
Driver components (total number of components that
only have inputs/total number of components that only
have outputs) and provides a measure of the degree to
which effects of Driversare considered.
FCM Composition:
1. Sum of Centrality (Sc) –centrality (c
i
) measures the
weighted contribution of each component within the
FCM. Sc is then calculated as the sum of centralities
of all components in a category. A standardized Sc
(NSc) score was calculated by normalizing the Sc for
each category to the total Sc for all categories,
excluding “Outcomes”and “Treatment/ Intervention”
for each FCM (see Appendix S1 for equation). Compo-
nents in the FCM with the highest centrality values
are considered the most important.
2. Cognitive Color Spectrum –color bar chart that dem-
onstrates the sequence of dominance of categories in
a participant’s FCM. It is generated by sorting the
NSc of each category by their color starting from the
most central category.
Table 1
Disciplines and countries of participants (10 males and 4 females)
Discipline Subdiscipline Country Number
Physical
Therapy
5
Clinical (3) Canada; New Zealand;
United States
Musculoskeletal
research (2)
Norway; Sweden
Orthopedic Surgery Sweden; United States 3
Physical Medicine &
Rehabilitation
The Netherlands;
United States
3
Anatomy The Netherlands 1
Osteopathy Australia 1
Sports Medicine Australia 1
3P.W. Hodges et al. / PM R xx (2019) 1–13
3. Cognitive Diversity Index (CDI) –quantitative measure
that reflects how many different categories are repre-
sented in an FCM, and simultaneously considers how
evenly the components are distributed among those
categories. A higher value indicates that an FCM has
components representing more categories and con-
tributing more evenly to these categories, whereas a
lower value indicates fewer categories and bias
toward specificcategories (see Appendix S1 for
equation).
The individual FCMs were aggregated into a meta-
model that represented the group’s view. Because we
did not have any data regarding the credibility of the con-
cepts by which to weigh them during the aggregation, a
simple FCM averaging method with zeros was used.
31,32
In this method, each individual FCM was converted to
the adjacency matrix and augmented to include all
unique components present in all FCMs after the refine-
ment of terms, resulting in the same matrix size 89 ×89
for all FCMs. The connections between components not
mentioned in the original FCM created by a participant
were given “zero”weights in his/her individual FCM, rep-
resenting “dummy”concepts added to the model. Subse-
quently, the metamodel was constructed by averaging
connections across the adjacency matrices:
a=1
MΣ
M
i=1ai,
where M is the number of participants, ais the connection
matrix of the aggregated model, and a
i
is the connection
matrix of the model developed by ith participant (in this
case, 14 participants).
The metamodel was used to study the relative empha-
sis placed by the group on various categories (eg, Psycho-
logical vs Biological factors) by computing Sc for each
category and it was also used to evaluate the group’s col-
lective view of relative efficacy of different treatments
by performing scenario simulations. The metamodel can
receive an input concept by initializing one or more of
its components to a value between 0 and 1. During simu-
lation, metamodel state is iteratively calculated until it
converges by propagating the initial values throughout
the metamodel network according to the weights
between its components and a threshold function.
33
The
final values of the components representing the output
concept are examined to assess the relative effect of var-
ious input concepts. To assess the relative efficacy of dif-
ferent treatments, they were individually initialized to
1 in each simulation and the resultant values of “Pain,”
“Disability,”and “Quality of Life”outcome components
were recorded. The simulations were performed using a
custom-written software in Python (Python Software
Foundation, www.python.org) with a sigmoid threshold
function.
34
Results
The individual FCMs ranged from 14 to 32 components
that were linked by between 25 and 125 connections for
an average of 2.2 (SD = 1.1) connections per component
(Table 3). In general, density was inversely related to
the number of components, that is, when more compo-
nents were included, fewer of the total possible connec-
tions were made. This relationship reached statistical
significance when one outlier (#9) with a large number
of connections was omitted from the calculation (R =
−0.80, P= .001). Examples of two models with substan-
tially different complexity and density scores are pres-
ented in Figure 1.
Table 2
Categories for allocation of FCM components
Category Definition
Behavioral/Lifestyle Lifestyle “choices”including: smoking;
sleep; physical activity; diet; insufficient
time.
Biomechanical Factors that determine/cause/relate to
tissue loading including lifting; posture;
motor control; muscle imbalance; etc.
Comorbidities Conditions that are comorbid with PGP and
pain such as: rheumatoid arthritis;
cancer; or diabetes.
Individual Factors that are part of the “make-up”of
the person including: age; body weight;
physical capacity; strength; genetics; and
individual features thought to predispose
to PGP and pain such as prior history.
Nociceptive detection
and processing
Biological factors related to
pain/nociception including:
sensitization; neuroimmune interaction;
“neuromatrix”, etc.
Psychological All aspects related to psychology including:
fear of pain/(re)injury; catastrophizing;
self-efficacy; etc.
Social/Work/Contextual Factors related to work and relationships
including: work support; family
environment; social status;
spirituality/religion. Includes factors that
are external to the person such as
environmental/policy, access to
treatment; political, physical
environmental, social, cultural context.
Tissue injury or
pathology
Biological factors of tissue/systems
including: tissue injury; disease;
pathology; cytokines; and consequences/
outcome of loading rather than the
mechanisms that cause loading, which
are categorized as “Biomechanical.”
Outcomes Core outcome measures included in every
model were “Pain,”“Disability,”and
“Quality of Life.”
Treatment/Intervention Any intervention for treatment and
prevention of SIJ pain.
Adapted from Cholewicki et al.
22
ICF = International Classification of
Functioning, Disability and Health.
61
4Collaborative Model of SIJ Dysfunction
The most frequent category with the highest Sc in the
individual FCM models was “Biomechanics”(five
models) followed by the “Social/Work/Contextual”cat-
egory (three models) (Figure 2). When thematically
related categories of “Biomechanics”with “Tissue
injury or pathology”(biophysical factors) and “Psychol-
ogy”with “Social/Work/Contextual”(psychosocial fac-
tors) were grouped, seven and four FCMs gave the
highest Sc to the biophysical and psychosocial factors,
respectively.
The metamodel, an aggregate of all individual FCMs,
which reflected collective group thinking, consisted of
89 components and 372 connections, and is presented in
Figure 3A (Interaction version final metamodel is available
in Appendix S2). In this metamodel, the “Biomechanical”
factors (Figure 3B) had the highest Sc, followed by “Psycho-
logical”(Figure 3C), and “Behavioral/Lifestyle”factors
(Figures 2 and 4). The “Comorbidities”and “Nociceptive
detection and processing”categories had the two lowest
Sc scores in the metamodel. The centrality attributed to
each component of the metamodel is presented in Table 4.
These data show the relative weighting placed on each com-
ponent within each category and overall. The components
with greatest centrality overall (other than outcomes) were:
Cognitive (3.123) (Psychological) (eg, expectations, beliefs
and perceptions concerning pain); Good Physical Activity
(1.724) (Behavioral/lifestyle); Poor Sleep (1.094)
(Behavioral/lifestyle); Inflammation (1.079) (Tissue injury
or pathology); Poor Anatomical/Structural characteristics
(1.012) (Biomechanical); Motor Impairment (0.982)
(Biomechanical); Poor Posture and Alignment (0.876)
(Biomechanical); Employment (0.870) (Social/Work/Con-
textual factors); Access to Support Networks (0.838)
(Social/Work/Contextual factors); and Tissue Damage
(0.794) (Tissue injury or pathology).
Although none of the individual FCMs included compo-
nents from all eight categories (half of the models
included components from five or fewer categories), the
metamodel components represented all categories
(Figures 2, 4, and 5). Therefore, as expected, the CDI of
the metamodel (6.39) was higher than CDIs of any of the
individual FCMs (Figure 5).
The results from the metamodel simulation of various
treatment interventions identified by the participants
are presented in Figure 6 as the effects on “Pain,”
“Disability,”and “Quality of Life”relative to the most
effective treatment. These metamodel simulation
results, which summarize the collective opinion of all
contributors, suggest that the interventions expected to
be most effective for reducing “Pain”are injection, exer-
cise therapy, and SIJ surgery. Exercise therapy, cognitive
behavioral therapy, and advice/education were consid-
ered the most effective interventions for reducing “Dis-
ability.”Exercise therapy, cognitive behavioral therapy,
and SIJ surgery were the interventions considered to be
the most effective for improving “Quality of life.”Accep-
tance therapy (psychological therapy that teaches mind-
fulness skills to deal with the uncontrollable experience
of pain
35
) had the smallest expected effects on the three
outcomes.
Discussion
This study produced a collaborative model of PGP that
represents the collective view of the experts across a
range of disciplines. The model shows how expert opin-
ions differ between consideration of PGP and LBP, and
how opinions relate to current evidence for treatments.
Table 3
Metrics describing structure of FCMs for each participant
FCM no. Total comp. Total connections Density Connections per comp. Complexity score Cognitive diversity index
#1 20 46 0.121 2.3 0.111 6.21
#2 20 39 0.103 2.0 0.077 5.99
#3 26 51 0.078 2.0 0.053 5.65
#4 25 41 0.068 1.6 0.111 5.04
#5 21 37 0.088 1.8 0.077 4.88
#6 26 41 0.063 1.6 0.067 4.74
#7 16 28 0.117 1.8 0.111 4.62
#8 21 54 0.129 2.6 0.000 4.61
#9 23 128 0.253 5.6 0.000 3.96
#10 14 38 0.209 2.7 0.125 3.74
#11 18 25 0.082 1.4 0.083 3.72
#12 23 40 0.079 1.7 0.000 3.67
#13 27 49 0.070 1.8 0.067 3.63
#14 32 51 0.051 1.6 0.043 3.61
Mean 22 48 0.108 2.2 0.066 4.58
SD 5 25 0.057 1.1 0.043 0.90
Min. 14 25 0.051 1.4 0.000 3.61
Max. 32 128 0.253 5.6 0.125 6.21
Comp. = component; No. = number. Order of FCM numbers is identical to that used in figures.
5P.W. Hodges et al. / PM R xx (2019) 1–13
Comparison of Structure and Composition of
Individual Models of PGP and LBP
Individual FCMs for PGP were diverse in their structure
with 14–32 components and 25–128 connections,but
were, on average, less for FCMs of PGP than LBP (compo-
nents: 22 vs 25; connections: 48 vs 77) with fewer connec-
tions per component (2.2 vs 3.1).
22
Although this implies
that PGP was generally considered to be less complex
than LBP, most of the same categories were considered
to be relevant. A notable exception was that only one
FCM for PGP included components related to “Nocicep-
tive detection and processing”(eg, central sensitization)
in comparison to the 18 of 29 FCMs for LBP. Limited recog-
nition of this issue is at odds with the growing recognition
of such processes in maintenance of pain,
36
including
PGP
5
and appears consistent with the tendency toward
biomechanical conceptualization of SIJ dysfunction (see
below).
The categories with greatest centrality in individual
FCMs differed from that reported for LBP. Whereas “Psy-
chology”was the category with greatest centrality for
nearly half of LBP FCMs (14/29), this was identified for
only 1 of 14 FCMs for PGP. This concurs with the general
view of the literature regarding PGP. For instance, the
European Guidelines for the Diagnosis and Treatment of
PGP state that “based on the present limited knowledge,
the impression is that yellow flags (psychosocial features)
are less common among PGP patients than among LBP
patients.”
10
In contrast to LBP, the category that most frequently
had the highest centrality in PGP FCMs was “Biomechan-
ics”(5/14 compared with 4/29 for LBP), and if considered
along with the related category of “Tissue injury or
pathology,”this accounted for 7/14 of the FCMs. The
greater bias toward biomechanics and tissue injury may
have several explanations. First, the term “PGP”attri-
butes the condition to a specific anatomical structure,
which contrasts the case for LBP. This could have led to
the participants’interpretation of a more mechanical
foundation and stronger attribution to tissue-level
effects. Second, there has been considerable emphasis
on biomechanical models of SIJ function and
dysfunction,
13,14
which has strongly influenced both
Structure
Total Comp. 16
Total
Connect. 28
Density 0.117
Connect. per
Comp. 1.75
No. Driver
Comp. 9
No. Receiver
Comp. 1
No. Ordinary
Comp. 6
Complexity
Score 0.111
Total Comp. 21
Total
Connect. 54
Density 0.129
Connect. per
Comp. 2.57
No. Driver
Comp. 10
No. Receiver
Comp. 0
No. Ordinary
Comp. 11
Complexity
Score 0
Sum of Centrality
0 5 10 15
Tissue…
Social/W…
Psych.
Nocicep.
Individual
Comorbid.
Biomech.
Behavior.
024
Tissue injury
Social/Work
Psych.
Nocicep.
Individual
Comorbid.
Biomech.
Behavior.
Figure 1. Fuzzy Cognitive Maps (FCMs) for two representative participants with different structure and composition. The FCMs that were generated
by the participants are shown (left) in their original form, prior to refinement of the component terminology. The Sum of Centrality (Sc) (middle) and
structural features are shown for the final FCM after refinement. Note the different Sc and range of structural features that characterize the models of
different participants considering the problem of PGP and pain. Comp. = component; Connect. = connection. (See Table 3 for full titles of categories.)
6Collaborative Model of SIJ Dysfunction
conservative
37
and surgical management.
38
This empha-
sis is exemplified by strong statements such as “PGP is
related to non-optimal stability of the pelvic girdle
joints”in clinical guidelines.
10
Third, mostly biomechan-
ical factors have been considered to predispose an indi-
vidual to PGP (eg, falls, repetitive stress, scoliosis, and
leg length discrepancy).
2
Fourth, differential diagnosis
of PGP has generally involved response to mechanical
tests for the SIJ
10,11
and SIJ diagnostic anesthetic
blocks.
39,40
Fifth, the greater emphasis on psychosocial
rather than biomechanical features in LBP is likely to be
largely explained by the limited success of interventions
that target the latter.
41
Although it is possible that a
mechanical interpretation of PGP is accurate, this does
not reflect recent work that has identified associations
between several psychological features and persistence
of PGP (eg, self-efficacy; anxiety and depression; pain
catastrophizing).
42
Composition of the Metamodel
Consistent with the features of the FCMs, the composi-
tion of the metamodel reflected bias toward the “Biome-
chanics”category. This category had the highest sum of
centrality and included three components within the
10 highest centralities (“Poor anatomical/structural
characteristics,”“Motor impairment,”and “Poor posture
and alignment”). The related category of “Tissue injury
or pathology”also ranked highly and included the compo-
nent of “Inflammation”(category: Tissue injury or
pathology), which had high centrality related to the asso-
ciation between PGP and arthritic conditions including
ankylosing spondylitis.
Although “Psychology”was the fourth ranked category
in the metamodel, the highest ranked individual compo-
nent was the psychological feature of “Cognition (expec-
tations, beliefs, and perceptions concerning pain),”and
the centrality attributed to it was almost double that of
Pain
Disability
Quality of life
Negatve cognitions
Good Physical Activity
Life Demands
Poor Sleep
Inflammation
Exercise Therapy
Employment
Impaired Body Structure and Function
Poor Posture and Alignment
Relaxation
Motor Impairment
General health
Nagative emotion
Joint Instability
Cognitive behavioral therapy
(A) (B)
(C)
Physical capacity
Denervation interventions
Exercise
Access to Support Networks
Participation Limitation
Movement restriction
Reduced strength
Tissue Damage
Manual therapy
Positive behaviour
Activity Limitation
Obesity
Medical Comorbidities
Pregnancy
Posture and movement training
Negative Psychological factors
Evidence Based Care Pathway
Central Sensitization
Pain medication
Advice/Education
Positive Life Social Factors
Mechanical Overloading
Injection
Social Roles Fulfilled
Counseling and Education about the aerobic exercise
Work Satisfaction
Socioeconomic status
SIJ Surgery
Negative Life Social Factors
Negative Lifestyle Factors
Poor patient activation
Peripheral Sensitization
Patient's Adherence to Care
Diseases
Good Posture and Alignment
Educational level
Taping and braces
Poor Anatomical/Structural characteristics
Manipulation
Positive Psychological factors
Duration of pain
Anti-inflammatory medication
Cumulative Mechanical overload over time
Sleep restoration
Hormonal
Health Literacy
Negative genetics
Social Roles Unfulfilled
Pathology
Physical treatment
Smoking
Compromised myofascial fascial integrity
Massage
Negative Life Enviornment
Past history of pain
Flexibility
Psychological Intervention
Pain relieving intervention
Nutritional Counselin
g
Regenerative medicine
Good Nutrition
Acceptance Therapy
Dry needling
Secondary Gain
Tissue Degeneration
Acupuncture
Work Dissatisfaction
Modalities
Pelvic floor therapy
Heat/Ice
Optimal Motor Control
Figure 3. Metamodel of PGP/SIJ dysfunction. (A) Complete metamodel of PGP/SIJ dysfunction. All Components and Connections are shown. Catego-
ries are identified by color of circles (Components) and outgoing Connections. Size of circles indicates normalized sum of centrality. Treatment com-
ponents are distributed around the outside of the model.(B) “Biomechanical”components displayed at higher resolutions. (C) “Psychological”
components displayed at higher resolution.
Cognitive Color Spectrum
#1
#2
#3
#4
#5
#6
#7
#8
#9
#10
#11
#12
#13
#14
Metamodel
Indiv
Tiss
Com
Beha
Soci
Psyc
Biom
Biom
Psyc
Beha
Com
Soci
Indiv
Beha
Biom
Indiv
Tiss
Com
Soci
Biom
Tiss
Psyc
Noc
Soci
Com
Tiss
Biom
Soci
Beha
Psyc
Soci
Biom
Psyc
Tiss
Beha
Com
Tiss
Psyc
Soci
Beha
Biom
Biom
Beha
Tiss
Indiv
Soci
Psyc
Psyc
Beha
Biom
Soci
Tiss
Beha
Soci
Biom
Indiv
Biom
Beha
Psyc
Soci
Indiv
Soci
Psyc
Beha
Com
Biom
Psyc
Indiv
Soci
Beha
Soci
Biom
Indiv
Psyc
Com
Beha
Biom
Soci
Beha
Psyc
Tiss
Indiv
Com
Noci
Behavioral/Lifestyle
Biomechanical
Comorbidities
Individual factors
Nociceptive detection & processing
Psychological
Social/Work/Contextual
Tissue injury or pathology
MM
Figure 2. Cognitive Color Spectra for the individual participants and a
metamodel. Each category is ranked by magnitude of the Normalized
Sum of Centrality (NSc) for each participant (#1 to #14) and the meta-
model (MM). Note that “Biomechanical”category has the highest mean
NSc for the MM and is the highest ranked category for 5 of
14 participants.
7P.W. Hodges et al. / PM R xx (2019) 1–13
Table 4
Centrality of individual components in the metamodel
Category Component Centrality
Biomechanical Poor anatomical/structural characteristics 1.012
Biomechanical Motor impairment 0.982
Biomechanical Poor posture and alignment 0.876
Biomechanical Joint instability 0.666
Biomechanical Strength (reduced) 0.568
Biomechanical Movement restriction 0.473
Biomechanical Physical capacity 0.461
Biomechanical Optimal motor control 0.386
Biomechanical Cumulative mechanical overload over time 0.250
Biomechanical Mechanical overloading 0.204
Biomechanical Good posture and alignment 0.091
Biomechanical Flexibility 0.057
Social/Work/Contextual factors Employment 0.870
Social/Work/Contextual factors Access to support networks 0.838
Social/Work/Contextual factors Socioeconomic status 0.348
Social/Work/Contextual factors Evidence-based care pathway 0.257
Social/Work/Contextual factors Negative life social factors 0.21
Social/Work/Contextual factors Positive life social factors 0.204
Social/Work/Contextual factors Work satisfaction 0.200
Social/Work/Contextual factors Social roles fulfilled 0.179
Social/Work/Contextual factors Social roles unfulfilled 0.107
Social/Work/Contextual factors Negative life environment 0.086
Social/Work/Contextual factors Secondary gain (eg, work environment, motivation, legal) 0.064
Social/Work/Contextual factors Work dissatisfaction 0.036
Behavioral/Lifestyle Good physical activity 1.724
Behavioral/Lifestyle Poor sleep 1.094
Behavioral/Lifestyle Life demands 0.714
Behavioral/Lifestyle Exercise 0.547
Behavioral/Lifestyle Negative lifestyle factors 0.196
Behavioral/Lifestyle Patient’s adherence to care 0.164
Behavioral/Lifestyle Smoking 0.049
Behavioral/Lifestyle Good Nutrition 0.021
Psychological Cognitive (expectations, beliefs & perceptions concerning pain) 3.123
Psychological Emotional (distress, anxiety and depression) 0.598
Psychological Negative psychological factors 0.355
Psychological Behavioural (coping, pain behavior & activity/activity avoidance) 0.346
Psychological Positive psychological factors 0.136
Tissue injury or pathology Inflammation 1.079
Tissue injury or pathology Tissue damage 0.794
Tissue injury or pathology Compromised myofascial fascial integrity 0.252
Tissue injury or pathology Pathology 0.100
Tissue injury or pathology Tissue degeneration 0.043
Individual factors General health 0.571
Individual factors Pregnancy 0.380
Individual factors Poor patient activation (ability to participate in health care) 0.186
Individual factors Hormonal 0.150
Individual factors Genetics (negative) 0.096
Individual factors Health literacy 0.096
Individual factors Educational level 0.059
Individual factors Past history of pain 0.043
Comorbidities Medical comorbidities 0.462
Comorbidities Overweight (obesity) / BMI 0.344
Comorbidities Diseases (infections, rheumatoid arthritis, malignancies) 0.121
Nociceptive detection and processing Central sensitization 0.204
Nociceptive detection and processing Peripheral sensitization 0.131
Outcomes Pain 7.923
Outcomes Disability 5.480
Outcomes Quality of life 5.258
Outcomes Activity limitation 1.136
Outcomes Body structure and function (Impaired) 0.688
Outcomes Participation limitation 0.582
Outcomes Duration of pain 0.054
8Collaborative Model of SIJ Dysfunction
the next ranked component. Thus collectively the expert
contributors placed weight on psychological issues, but
this was focused on a single main feature, in contrast to
the multiple separate features in the “Biomechanics”
category.
Several components in the “Behavioral/lifestyle”
category ranked in the top 10 highest centralities.
These were “Poor sleep”and “Good physical activity.”
In the absence of data in PGP, this is likely to be
explained by strong evidence for an association with
LBP.
43,44
Somewhat surprisingly, “Pregnancy”(category: Indi-
vidual) was only ranked 23rd in terms of centrality in
the metamodel. This is unexpected considering the high
point prevalence of PGP during and after pregnancy
(~20%
45
) and the focus on this group in many studies of
epidemiology and differential diagnosis.
10
The reason
for the limited explicit inclusion of the term “pregnancy”
in the metamodel is not clear but is likely explained by
inclusion of terms that describe factors associated with
pregnancy rather than the term itself.
Interpretation of Relative Efficacy of Treatments
Based on the Metamodel
The metamodel generated from the individual contrib-
utor FCMs enables investigation of the collective opinion
with respect to the expected efficacy of different treat-
ments on “Pain,”“Disability,”and “Quality of Life.”This
approach considers the overall weight from all contribu-
tors. The treatments predicted to have the greatest
expected efficacy differed between outcomes of “Pain,”
“Disability,”and “Quality of Life.”
“Injection,”“Exercise therapy,”and “SIJ surgery”
were predicted to have the greatest impact on “Pain.”
Does this match the best available evidence? “Injection”
was used by participants to refer to the intra- or
extraarticular injection of several different agents (eg,
steroid injections, analgesic agents, prolotherapy, or
combination). Although this intervention was considered
to have the greatest effect on pain, the efficacy of any
injected agent has very limited evidence. Reports of only
a few small randomized controlled trials (RCTs; 10–24 par-
ticipants) with short term follow-up have been publi-
shed.
46,47
These studies show temporary relief in a
subset of individuals, predominantly those with active
inflammation (eg, ankylosing spondylitis).
47
Local anes-
thetic injection is also used for diagnostic purposes.
39
Considering this limited evidence, the strong endorse-
ment of this treatment in the model is somewhat surpris-
ing, although it may be explained by the absence of
strong evidence for any treatment for PGP. This finding
does align with the increasing use of injection procedures
in clinical practice.
48
Evidence for the efficacy of exercise therapy as a
treatment for PGP is modest.
10,19
In terms of PGP, exer-
cise therapy has been examined primarily in management
of PGP in association with pregnancy, and patients
included in trials have commonly not been differentiated
from those with LBP, which may be unrelated to PGP.
Results are conflicting; some studies show large
37
and
long-term effects,
49
whereas others show no effect.
50
Meta-analysis of trials is generally not possible because
of heterogeneity of exercise approach and included
patient populations. The consensus is that exercise can
reduce pain and improve function, but there is little basis
to determine which exercise approach is the best.
10
A
common approach tested in the literature has been motor
control interventions.
19
This concurs with the high cen-
trality in the metamodel of “Motor impairment”and
“Poor posture and alignment,”which are targeted by
these approaches.
Surgery, which includes at least 17 different
approaches (most involving fusion).
51
has historically
had low evidence.
10
A small number of randomized con-
trolled trials
52,53
and cohort studies are available.
10
Results are variable, with good
52
to poor outcomes
reported.
54
Recent trials of a minimally invasive surgi-
cal approach have shown promising outcomes for care-
fully selected individuals with positive response to
intra-articular diagnostic anesthetic injections and
other diagnostic tests.
52,53
Most authors suggest that
surgery is a potential option when nonoperative man-
agement has failed.
55
Although failure of nonoperative
care has not yet been shown to predict outcome, a
recent meta-analysis found better outcomes for the
aforementioned carefully selected individuals if they
had long duration pain and older age.
56
Outcomes were
worse for individuals with a history of opioid use and
smoking.
One notable omission from the high-ranking treat-
ments considered to be effective for pain was denerva-
tion interventions. Although this might be surprising
considering the large literature investigating this
approach,
2
it concurs with recent discussion of the lim-
ited evidence of clinical efficacy.
57
Not surprisingly, interventions considered to be effec-
tive for “Disability”and “Quality of Life”differed from
0 102030405060708090100
Nociceptive detection & processing
Comorbidities
Individual factors
Tissue injury or pathology
Social/Work/Contextual factors
Behavioral/Lifestyle
Psychological
Biomechanical
Relative Sum of Centrality (%)
Figure 4. Relative Sum of Centrality (Sc) for the metamodel. The Sc
values are presented as relative to the “Biomechanical”category that
had the highest Sc value.
9P.W. Hodges et al. / PM R xx (2019) 1–13
those for “Pain.”Exercise therapy was considered the
most effective treatment for “Disability”and “Quality
of Life,”and there is some evidence for this from
RCTs.
10,19
Cognitive behavioral therapy was the second
most favored option for both disability and quality of life.
Although this appears logical, it has not been investigated
in PGP and the perceived potential efficacy of
this approach is probably based on evidence from
RCTs in LBP.
58,59
Likewise, perceived efficacy of
advice/education for PGP is likely to be based on work
in LBP,
60
as no RCTs have tested this intervention in
patients with PGP. Of interest, “SIJ surgery”was consid-
ered the third most likely intervention to impact “Quality
of Life.”This has some evidence but will likely be rele-
vant for a small subset of patients.
52
Limitations
By its nature, collaborative modeling aims to summa-
rize the diverse opinions of contributors into a single
00.51
Acceptance therapy
Psychological intervention
Nutritional counseling
Pain relieving intervention
Heat/Ice
Pelvic floor therapy
Modalities
Regenerative medicine
Dry needling
Denervation interventions
Massage
Acupuncture
Manipulation
Taping and braces
Physical treatment
Advice/Education
Anti-inflammatory medication
Couns. & ed. about aerobic exercise
Injection
Posture and movement training
Sleep restoration
Pain medication
Manual therapy
Relaxation
SIJ surgery
Cognitive behavioral therapy
Exercise therapy
Quality of Life
00.51
Acceptance therapy
Psychological intervention
Heat/Ice
Pelvic floor therapy
Nutritional counseling
Modalities
Pain relieving intervention
Sleep restoration
Regenerative medicine
Dry needling
Denervation interventions
Massage
Acupuncture
Anti-inflammatory medication
Manipulation
Taping and braces
SIJ surgery
Manual therapy
Couns. & ed. about aerobic exercise
Pain medication
Relaxation
Injection
Posture and movement training
Physical treatment
Advice/Education
Cognitive behavioral therapy
Exercise therapy
Disability
00.51
Acceptance therapy
Nutritional counseling
Pain relieving intervention
Psychological intervention
Heat/Ice
Pelvic floor therapy
Modalities
Regenerative medicine
Couns. & ed. about aerobic exercise
Sleep restoration
Dry needling
Denervation interventions
Massage
Acupuncture
Anti-inflammatory medication
Relaxation
Posture and movement training
Physical treatment
Manipulation
Manual therapy
Pain medication
Advice/Education
Cognitive behavioral therapy
Taping and braces
SIJ surgery
Exercise therapy
Injection
Pain
Figure 6. Metamodel simulations of the effects of various interventions on Pain, Disability, and Quality of Life. The effects are presented as relative
to the most effective intervention and are ranked from the most effective at the top to the least effective at the bottom of each panel.
0 0.2 0.4 0.6 0.8 1
#1
#2
#3
#4
#5
#6
#7
#8
#9
#10
#11
#12
#13
#14
MM
0.2 0.4 0.6 0.8 1.00
Relative Cognitive Diversity Index
Normalized Sum of Centrality
Figure 5. Normalized Sum of Centrality (NSc) and Cognitive Diversity Index (CDI) for individual participants Fuzzy Cognitive Maps (FCM) and the
metamodel (MM). The order of FCMs is identical to that in Figure 3. Colors in the NSc refer to the categories (see Figure 3 for definitions of color).
A high CDI indicates that a participant considers components across a broad range of categories with relatively similar NSc between categories. A
low CDI indicates that a participant considers components across a few categories with a bias of NSc to only some categories. This analysis does not
imply that one model of considering PGP and pain is better or worse, but characterizes the different ways that participants consider the problem.
10 Collaborative Model of SIJ Dysfunction
representative model. This will necessarily involve some
simplification and, thus, some limitations. First, the
metamodel clusters all presentations of PGP together
(eg, ankylosing spondylitis, pregnancy-related condi-
tions, and so on), and although specific treatments may
be expected to be effective for specific groups, this can-
not be reflected in this model. Second, we collapsed some
similar terms from individual FCMs into a smaller group of
components that were established via extensive consul-
tation with experts (including some who participated in
this study) during the development of a collaborative
model of LBP.
22
In some cases, the final terminology and
grouping may require further consideration. For instance,
“Exercise therapy”was nominated as a stand-alone treat-
ment, but several other treatments could also be consid-
ered to be forms of exercise (eg, “Posture and movement
training,”“Counseling and education about aerobic exer-
cise”). Alternatively, it could be argued that exercise
therapy is heterogeneous and should be further sub-
divided into subtypes to better reflect their independent
roles. The same issue could be considered for “Injection”
and “SIJ surgery,”which have multiple forms.
Additional limitations relate to the group of experts
who contributed to the model. The group was relatively
small and involved mainly individuals from medical and
physical therapy backgrounds. Despite the relatively
small sample, it has been reported that the number of
new variables accumulated per FCM beyond 12 FMCs is
relatively small.
26
Although this reflects the bias to these
fields in the published literature, greater involvement of
individuals from psychology and other disciplines might
have changed the centrality of the metamodel compo-
nents. Finally, it is necessary to recognize that this model
reflects the opinions of the expert group that we selected
and further work is needed to determine whether it
reflects opinions more broadly.
Conclusion
This paper presented a collaborative model of PGP.
Inspection of the model has provided insight into the com-
plexity of this condition and the relative importance
placed by the experts/contributors on different domains
in this condition and how this differed from that observed
in LBP. The model also exposed a disconnect between per-
ceived relative efficacy of different interventions and the
available evidence. The metamodel provides some direc-
tions for future research, such as testing some of the pro-
posed connections between the components and helps in
identification of the interventions that should be evalu-
ated as a matter of priority.
Supporting Information
Additional supporting information may be found online
in the Supporting Information section at the end of the
article.
References
1. GBD 2016 DALYs and HALE Collaborators. Global, regional, and
national disability-adjusted life-years (DALYs) for 333 diseases and
injuries and healthy life expectancy (HALE) for 195 countries and
territories, 1990-2016: a systematic analysis for the global burden
of disease study 2016. Lancet. 2017;390(10100):1260-1344.
2. Cohen SP, Chen Y, Neufeld NJ. Sacroiliac joint pain: a comprehensive
review of epidemiology, diagnosis and treatment. Exp Rev Neu-
rother. 2013;13(1):99-116.
3. Szadek KM, Hoogland PV, Zuurmond WW, De Lange JJ, Perez RS. Pos-
sible nociceptive structures in the sacroiliac joint cartilage: an
immunohistochemical study. Clin Anat. 2010;23(2):192-198.
4. Szadek KM, Hoogland PV, Zuurmond WW, de Lange JJ, Perez RS.
Nociceptive nerve fibers in the sacroiliac joint in humans. Reg
Anesth Pain Med. 2008;33(1):36-43.
5. Apte G, Nelson P, Brismee JM, Dedrick G, Justiz R III, Sizer PS Jr.
Chronic female pelvic pain--part 1: clinical pathoanatomy and
examination of the pelvic region. Pain Pract. 2012;12(2):88-110.
6. van der Wurff P, Buijs EJ, Groen GJ. A multitest regimen of pain
provocation tests as an aid to reduce unnecessary minimally inva-
sive sacroiliac joint procedures. Arch Phys Med Rehabil. 2006;87
(1):10-14.
7. Manchikanti L, Singh V, Pampati V, et al. Evaluation of the relative
contributions of various structures in chronic low back pain. Pain
Physician. 2001;4(4):308-316.
8. Schwarzer AC, Aprill CN, Bogduk N. The sacroiliac joint in chronic
low back pain. Spine. 1995;20(1):31-37.
9. Petersen T, Olsen S, Laslett M, et al. Inter-tester reliability of a new
diagnostic classification system for patients with non-specific low
back pain. Aust J Physiother. 2004;50(2):85-94.
10. Vleeming A, Albert HB, Ostgaard HC, Sturesson B, Stuge B. European
guidelines for the diagnosis and treatment of pelvic girdle pain. Eur
Spine J. 2008;17(6):794-819.
11. Laslett M, Aprill CN, McDonald B. Provocation sacroiliac joint tests
have validity in the diagnosis of sacroiliac joint pain. Arch Phys
Med Rehabil. 2006;87(6):874.
12. Mens JM, Vleeming A, Snijders CJ, Koes BW, Stam HJ. Validity of the
active straight leg raise test for measuring disease severity in
patients with posterior pelvic pain after pregnancy. Spine. 2002;27
(2):196-200.
13. Vleeming A, Schuenke MD, Masi AT, Carreiro JE, Danneels L,
Willard FH. The sacroiliac joint: an overview of its anatomy, func-
tion and potential clinical implications. J Anat. 2012;221(6):
537-567.
14. Snijders CJ, Vleeming A, Stoeckart R. Transfer of lumbosacral load
to iliac bones and legs. Part 1: biomechanics of self bracing of the
sacroiliac joints and its significance for treatment and exercise. Clin
Biomech. 1993;8:285-294.
15. Dontigny RL. Dysfunction of the sacroiliac joint and its treatment.
J Orthop Sports Phys Ther. 1979;1(1):23-35.
16. Sturesson B, Selvik G, Uden A. Movements of the sacroiliac joints. A
roentgen stereophotogrammetric analysis. Spine. 1989;14(2):162-165.
17. Aldabe D, Milosavljevic S, Bussey MD. Is pregnancy related pelvic
girdle pain associated with altered kinematic, kinetic and motor
control of the pelvis? A systematic review. Eur Spine J. 2012;21(9):
1777-1787.
18. Braun J, Sieper J. The sacroiliac joint in the spondyloarthropathies.
Curr Opin Rheumatol. 1996;8(4):275-287.
19. Gutke A, Betten C, Degerskar K, Pousette S, Olsen MF. Treatments
for pregnancy-related lumbopelvic pain: a systematic review of
physiotherapy modalities. Acta Obstet Gynecol Scand. 2015;94
(11):1156-1167.
20. Gray SA, Zanre E, Gray SRJ. Fuzzy cognitive maps as representations
of mental models and group beliefs. In: Papageorgiou E,
ed. Fuzzy Cognitive Maps for Applied Sciences and Engineering.
Intelligent Systems Reference Library. Vol 54. Heidelberg: Springer;
2014.
11P.W. Hodges et al. / PM R xx (2019) 1–13
21. Kosko B. Fuzzy cognitive maps. Int J Man-Machine Stud. 1986;24(1):
65-75.
22. Cholewicki J, Popovich JM Jr, Aminpour P, Gray SA, Lee AS,
Hodges PW. Development of a collaborative model of low back pain:
report from the 2017 NASS consensus meeting. Spine J. 2019;19(6):
1029-1040.
23. Yoon BS, Jetter AJ. Comparative analysis for fuzzy cognitive map-
ping. Paper presented at: PICMET ’16: Technology Management for
Social Innovation 2016.
24. Markóczy L, Goldberg J. A method for eliciting and comparing causal
maps. J Management. 1995;21(2):305-333.
25. Langfield-Smith K, Wirth A. Measuring differences between cogni-
tive maps. J Oper Res Soc. 1992;43(12):1135-1150.
26. Ozesmi U, Ozesmi S. Ecological models based on people’s knowl-
edge: a multi-step fuzzy cognitive mapping approach. Ecol Model.
2004;176(1–2):43-64.
27. Giabbanelli PJ, Gray SA, Aminpour P. Combining fuzzy cognitive
maps with agent-based modeling: frameworks and pitfalls of a pow-
erful hybrid modeling approach to understand human-environment
interactions. Environ Model Softw. 2017;95:320-325.
28. Gray SA, Gray S, De kok JL, et al. Using fuzzy cognitive mapping as a
participatory approach to analyze change, preferred states, and
perceived resilience of social-ecological systems. Ecol Soc. 2015;
20(2):11.
29. Gray SRJ, Gray SA, Gagnon AS, et al. Are coastal managers detecting
the problem? Assessing stakeholder perception of climate vulnera-
bility using fuzzy cognitive mapping. Ocean Coast Manage. 2014;
94:74-89.
30. Gray S. MentalModeler; 2018. http://www.mentalmodeler.org.
Accessed October 3, 2018.
31. Kosko B. Hidden patterns in combined and adaptive knowledge net-
works, international journal of approximate reasoning. Int J Approx
Reason. 1988;2:337-393.
32. Stach W. Learning and Aggregation of Fuzzy Cognitive Maps - an
Evolutionary Approach. Edmonton: University of Alberta; 2010.
33. Iakovidis DK, Papageorgiou E. Intuitionistic fuzzy cognitive maps for
medical decision making. IEEE Trans Inf Technol Biomed. 2011;15
(1):100-107.
34. Aminpour P. PyFCM: python for fuzzy cognitive mapping; 2018.
https://github.com/payamaminpour/PyFCM/wiki. Accessed
October 3, 2018.
35. Veehof MM, Trompetter HR, Bohlmeijer ET, Schreurs KM. Accep-
tance- and mindfulness-based interventions for the treatment of
chronic pain: a meta-analytic review. Cogn Behav Ther. 2016;45
(1):5-31.
36. Woolf CJ. Central sensitization: implications for the diagnosis and
treatment of pain. Pain. 2011;152(3 Suppl):S2-S15.
37. Stuge B, Laerum E, Kirkesola G, Vollestad N. The efficacy of a treat-
ment program focusing on specific stabilizing exercises for pelvic
girdle pain after pregnancy: a randomized controlled trial. Spine.
2004;29(4):351-359.
38. Sturesson B, Uden A, Onsten I. Can an external frame fixation
reduce the movements in the sacroiliac joint? A radiostereometric
analysis of 10 patients. Acta Orthop Scand. 1999;70(1):42-46.
39. Kennedy DJ, Engel A, Kreiner DS, Nampiaparampil D, Duszynski B,
MacVicar J. Fluoroscopically guided diagnostic and therapeutic
intra-articular sacroiliac joint injections: a systematic review. Pain
Med. 2015;16(8):1500-1518.
40. Simopoulos TT, Manchikanti L, Singh V, et al. A systematic evalua-
tion of prevalence and diagnostic accuracy of sacroiliac joint inter-
ventions. Pain Phys. 2012;15(3):E305-E344.
41. Foster NE, Anema JR, Cherkin D, et al. Prevention and treatment of
low back pain: evidence, challenges, and promising directions. Lan-
cet. 2018;391(10137):2368-2383.
42. Elden H, Gutke A, Kjellby-Wendt G, Fagevik-Olsen M, Ostgaard HC.
Predictors and consequences of long-term pregnancy-related pelvic
girdle pain: a longitudinal follow-up study. BMC Musculoskelet Dis-
ord. 2016;17:276.
43. Klyne D, Barbe M, van den Hoorn W, Hodges PW. ISSLS PRIZE IN CLIN-
ICAL SCIENCE 2018: longitudinal analysis of inflammatory, psycho-
logical and sleep-related factors following an acute low back pain
episode: the good, the bad and the ugly. Eur Spine J. 2018;27:763-
777. in press.
44. Lin CW, McAuley JH, Macedo L, Barnett DC, Smeets RJ, Verbunt JA.
Relationship between physical activity and disability in low back
pain: a systematic review and meta-analysis. Pain. 2011;152(3):
607-613.
45. Albert HB, Godskesen M, Westergaard JG. Incidence of four syn-
dromes of pregnancy-related pelvic joint pain. Spine. 2002;27
(24):2831-2834.
46. Luukkainen RK, Wennerstrand PV, Kautiainen HH, Sanila MT,
Asikainen EL. Efficacy of periarticular corticosteroid treatment of
the sacroiliac joint in non-spondylarthropathic patients with
chronic low back pain in the region of the sacroiliac joint. Clin Exp
Rheumatol. 2002;20(1):52-54.
47. Maugars Y, Mathis C, Berthelot JM, Charlier C, Prost A. Assessment of
the efficacy of sacroiliac corticosteroid injections in
spondylarthropathies: a double-blind study. Br J Rheumatol. 1996;
35(8):767-770.
48. Manchikanti L, Hirsch JA, Pampati V, Boswell MV. Utilization of facet
joint and sacroiliac joint interventions in medicare population from
2000 to 2014: explosive growth continues! Curr Pain Headache Rep.
2016;20(10):58.
49. Stuge B, Veierod MB, Laerum E, Vollestad N. The efficacy of a treat-
ment program focusing on specific stabilizing exercises for pelvic
girdle pain after pregnancy: a two-year follow-up of a randomized
clinical trial. Spine. 2004;29(10):E197-E203.
50. Mens JM, Snijders CJ, Stam HJ. Diagonal trunk muscle exercises in
peripartum pelvic pain: a randomized clinical trial. Phys Ther.
2000;80(12):1164-1173.
51. Stark J, Fuentes J, Fuentes T, Idemmili C. The history of sacroiliac
joint arthrodesis: a critical review and introduction of a new tech-
nique. Curr Orthop Prac. 2011;22(3):545-557.
52. Polly DW, Swofford J, Whang PG, et al. Two-year outcomes from a
randomized controlled trial of minimally invasive sacroiliac joint
fusion vs. non-surgical management for sacroiliac joint dysfunction.
Int J Spine Surg. 2016;10:28.
53. Sturesson B, Kools D, Pflugmacher R, Gasbarrini A, Prestamburgo D,
Dengler J. Six-month outcomes from a randomized controlled trial
of minimally invasive SI joint fusion with triangular titanium
implants vs conservative management. Eur Spine J. 2017;26(3):
708-719.
54. Schutz U, Grob D. Poor outcome following bilateral sacroiliac joint
fusion for degenerative sacroiliac joint syndrome. Acta Orthop
Belg. 2006;72(3):296-308.
55. Polly DW Jr. The sacroiliac joint. Neurosurg Clin N Am. 2017;28(3):
301-312.
56. Dengler J, Duhon B, Whang P, et al. Predictors of outcome in conser-
vative and minimally invasive surgical Management of Pain Originat-
ing from the sacroiliac joint: a pooled analysis. Spine. 2017;42(21):
1664-1673.
57. Juch JNS, Maas ET, Ostelo R, et al. Effect of radiofrequency dener-
vation on pain intensity among patients with chronic low Back pain:
the mint randomized clinical trials. JAMA. 2017;318(1):68-81.
58. Henschke N, Ostelo RW, van Tulder MW, et al. Behavioural treatment
for chronic low-back pain. Cochrane Database Syst Rev. 2010;7:
CD002014.
59. Hajihasani A, Rouhani M, Salavati M, Hedayati R, Kahlaee AH. The
influence of cognitive behavioral therapy on pain, quality of life,
and depression in patients receiving physical therapy for chronic
low Back pain: a systematic review. PMR. 2018;11(2):167-176.
60. Engers A, Jellema P, Wensing M, van der Windt DA, Grol R, van
Tulder MW. Individual patient education for low back pain. Cochrane
Database Syst Rev. 2008;1:CD004057.
61. WHO. International Classification of Functioning, Disability and
Health. Geneva, Switzerland: WHO Press; 2001.
12 Collaborative Model of SIJ Dysfunction
Disclosure
P.W.H. The University of Queensland, NHMRC Centre of Clinical Research Excel-
lence in Spinal Pain, Injury and Health, School of Health and Rehabilitation Sci-
ences, Brisbane, Australia. Address correspondence to: P. W. H.; School of
Health and Rehabilitation Sciences, The University of Queensland, Brisbane, QLD
4072 Australia; e-mail: p.hodges@uq.edu.au
J.C., J.M.P. Jr., A.S.L. MSU Center for Orthopedic Research, Department of Oste-
opathic Surgical Specialties, Michigan State University, East Lansing, MI
P.A., S.A.G. Department of Community Sustainability, Michigan State University,
Natural Resource Building, East Lansing, MI
M.T.C. Physical Therapy Program, Maryville University, St. Louis, MO
M.C. School of Medicine, Sydney,University of Notre Dame Austral ia, Darlinghurst,
Australia
B.F.D. A.T. Still University, Kirksville, MO
G.F. College of Health & Biomedicine, Victoria University, Melbourne, Australia
A.G. Department of Health and Rehabilitation, Institute of Neuroscience and Phys-
iology, University of Göteborg, Göteborg, Sweden
D.J.K. Department of Physical Medicine and Rehabilitation, Vanderbilt University
Medical Center, Nashville, TN
M.L. Health and Rehabilitation Research Institute, AUT University, Auckland,
New Zealand; Southern Musculoskeletal Seminars, New Zealand
D.L. Diane Lee & Associates, South Surrey, Canada
J.M. Department of Rehabilitation Medicine & Physical Therapy, Erasmus
University Medical Center, Rotterdam, The Netherlands
V.V. P. Department of Orthopaedic Surgery, University of Colorado, Denver, CO
H.P. Departments of Orthopaedic Surgery and Neurology, Washington University
School of Medicine, St Louis, MO
B.S. Department of Orthopedics, Aleris, Ängelholm Hospital, Ängelholm, Sweden
B.S. Division of Orthopaedic Surgery, Oslo University Hospital, Oslo, Norway
A.V. Department of Anatomy, Medical Osteopathic College of the University of
New England, Biddeford, ME; and Department of Rehabilitation Sciences and Phys-
iotherapy, Faculty of Medicine and Health Sciences, Ghent University, Belgium
13P.W. Hodges et al. / PM R xx (2019) 1–13
