MAIN TEXT (3500/3500 words):
INTRODUCTION (394/400
words):
Vaginal morphology (angulation, dimensions) and position are key
indicators and determinants of pelvic organ support. Restoration of
normal vaginal anatomy and apical support is believed necessary for
successful repair of pelvic organ prolapse (POP). Normally, the lower
vagina is nearly vertical, the upper vagina is more horizontal and
directed posteriorly towards the sacral hollow, and the angle between
the upper and lower vagina is approximately 160° in older parous women.
Surgical repair of POP often abnormally alters vaginal anatomy, which
may increase the risk of prolapse recurrence.
Magnetic resonance imaging (MRI) has been used to assess vaginal anatomy
following prolapse surgery. However, little is known about how vaginal
morphology and position relate to prolapse recurrence after POP repair.
In our previous work, the De fining M echanisms ofA nterior Vagin al Wall D escent (DEMAND) study, we
showed that apical descent and—to a lesser extent—anterior vaginal
wall (AVW) elongation were mechanisms of prolapse recurrence following
vaginal surgery. Results suggested that these mechanisms may involve
postoperative changes in vaginal angulation and position (i.e.,
posterior-inferior deviation and straightening of the vagina) that can
promote vaginal mobility/distensibility by making the AVW more
susceptible to descent and elongation—especially with an enlarged
genital hiatus (GH).
As the primary DEMAND study and similar works were limited to 2D vaginal
analysis, this study assessed vaginal morphology and position in 3D to
gain more comprehensive insight into the relationship between vaginal
anatomy and prolapse recurrence after vaginal surgery. The primary aim
of this study was to identify postoperative vaginal morphology
(angulation, dimensions), and position factors associated with prolapse
recurrence following one of two vaginal surgeries for uterovaginal
prolapse in the DEMAND cohort: vaginal mesh hysteropexy with
sacrospinous fixation (hysteropexy) or vaginal hysterectomy with
uterosacral ligament suspension (hysterectomy).
The secondary aim was to correlate
postoperative vaginal morphology and position factors with measures
commonly associated with prolapse recurrence: vaginal
mobility/distensibility and GH size.
Because apical descent was the dominant mechanism of prolapse recurrence
in the DEMAND cohort, our primary hypothesis was that, within each
surgery group, women with prolapse recurrence had a more inferiorly
positioned (lower) vagina with an angulation distinct from that seen in
women with success. We further hypothesized that within the recurrence
and success groups, the same associations would be observed with
hysterectomy and hysteropexy. Our secondary hypothesis was that a lower,
straighter vagina would be associated with greater vaginal
mobility/distensibility and larger GH.
METHODS:
Study Design:
This was a secondary analysis of MRI data from the DEMAND study, a
multisite prospective supplementary study of the S tudy ofU terine P rolapse Procedure s-R andomized
(SUPeR) trial, designed to identify anatomic mechanisms and correlates
of prolapse recurrence based on MRI evaluation among a subset of SUPeR
participants randomized to either (1) vaginal mesh hysteropexy with
sacrospinous fixation (hysteropexy) or (2) vaginal hysterectomy with
uterosacral ligament suspension (hysterectomy). DEMAND and SUPeR were
performed through the Pelvic Floor Disorders Network. Institutional
review board approval and participants’ written informed consent were
obtained at all study sites. Detailed study protocol and 3-year outcomes
for both studies are published.
Study Population:
The study population included a subset of 88 DEMAND participants who
underwent pelvic MRI between June 2014 and May 2018. A full list of
inclusion and exclusion criteria has been published. Additional
exclusion criteria were failure to capture the full vagina, poor
demarcation of the vaginal borders, incomplete MRIs, and MRIs taken
after reoperation. Patient characteristics—demographics, medical
history, and Pelvic Organ Prolapse Quantification (POP-Q)
measurements—from SUPeR were obtained at baseline (preoperative) and
follow-up (i.e., SUPeR study visit date closest to MRI examination
date).
MRI Protocol:
The MRI methods have been reported. Briefly, participants were trained
to perform maximal straining prior to MRI examination. Supine,
multi-planar, T2-weighted images were collected with a 3T system and a
pelvic phased array coil at rest (with prolapse fully reduced), maximal
strain, and recovery (rest period following strain without prolapse
reduced). The first rest scan provided a common reference configuration
of the vagina across all participants to obtain more reliable baseline
(rest) and dynamic (rest to strain) vaginal measurements. The final rest
(recovery) scan provided the physiological configuration of the vagina.
MRI scans were imported into 3D Slicer v4.10.0
(www.slicer.org) to (1) build a 3D pelvic
coordinate system (PCS) and (2) generate 3D vaginal models to compare
vaginal morphology and position across all participants while accounting
for differences in position in the MRI scanner.
Establishment of the 3D Pelvic Coordinate System:
A PCS was established to quantify vaginal angulation and position in 3D
space while accounting for differences in patient position in the MRI
scanner. The medio-lateral (X) axis was defined by the line connecting
the ischial spines, where its midpoint provided the origin of the PCS.
The anterior-posterior (Y) axis was given by the line orthogonal to the
X-axis starting from the origin to the point one-third along the
inferior-superior length of the pubic symphysis. The superior-inferior
(Z) axis was the cross product of the X- and Y-axes. The Y-Z plane
defined the midsagittal plane.
Vaginal Segmentation and 3D Reconstruction:
The vagina was manually segmented (with the lumen excluded) from axial
recovery MRI scans using 3D Slicer. Segmentations of multiple MRI slices
were stacked to reconstruct aliased (i.e., jagged edges of an object) 3D
surface models of the vagina with zero thickness. Vaginal models were
exported to Blender v2.83.2 (Blender Foundation, Amsterdam, The
Netherlands) to remove aliasing (sharp edges) by interpolating
segmentations between adjacent slices via a smoothing algorithm. As
described in a previous work, the smoothing algorithm allows unbiased,
global smoothing of aliased geometries while preserving their original
shape and volume. A diagram of the 3D reconstruction technique is shown
in Figure 1a .
3D Vaginal Position and Morphology Analysis:
Smoothed 3D vaginal models were imported into Mathematica v12.2.0
(Wolfram Research, Champaign, IL) to quantify vaginal morphology and
position via computational morphometry—an algorithmic technique that
performs automated measurements of anatomical structures by detecting
and extracting anatomical landmarks from 3D models (Figure 1b ).
All 3D-model-based vaginal position and morphology parameters were
static measures assessed at recovery (rest period following strain
without prolapse reduced).
Vaginal Position:
From the distal to proximal direction, each 3D vaginal surface model was
iteratively sliced along the axial plane at 1.5 mm intervals (i.e., half
of the MRI slice thickness), where for each vaginal slice (represented
as a thin ribbon), the centroid and lateral edges were calculated. The
centroidal and lateral edge points computed through this iterative
process defined the centerline and lateral margins of the vagina,
respectively. Using the centerline, the following vaginal position
parameters were measured with respect to the 3D PCS (Figure
2a): (1) vaginal apex position, the 3D coordinates of the most proximal
point on the centerline and (2) mid-vagina position, the 3D coordinates
of the midpoint of the centerline.
Vaginal Angulation:
The vaginal centerline points were divided into proximal and distal
halves to demarcate the upper and lower vagina. A line of best fit was
computed for each set of points to define the upper and lower vaginal
axes. From these axes, the following vaginal angulation parameters were
calculated with respect to the 3D PCS (Figure 2b): (1) upper
vaginal sagittal angle, the angle between the upper vaginal and
anterior-posterior axes; (2) lower vaginal sagittal angle, the angle
between the lower vaginal and anterior-posterior axes; (3) upper-lower
vaginal sagittal angle, the angle between the upper and lower vaginal
axes in the sagittal plane; (4) upper vaginal coronal angle, the angle
between the upper vaginal and superior-inferior axes; (5) lower vaginal
coronal angle, the angle between the lower vaginal and superior-inferior
axes; and (6) upper-lower vaginal coronal angle, the difference between
the upper and lower vagina coronal angles (i.e., a measure of the
overall lateral orientation of the vagina). A more horizontal sagittal
angle (towards the sacrum) is given by smaller values and a more
vertical sagittal angle is given by larger values. A more medial coronal
angle (aligned with the midline) is given by values closer to zero and a
more lateral coronal angle (tilted to the left or right) is given by
values further away from zero.
Vaginal Dimensions:
Using the vaginal centerline, lateral margins, and surface model, the
following vaginal dimension parameters were quantified (Figure
2c): (1) vaginal length, the length of the vaginal centerline; (2)
maximum transverse vaginal width, the largest pairwise straight-line
distance between the vaginal lateral margins across all vaginal slices;
(3) vaginal surface area, the surface area of the vaginal model; and (4)
vaginal volume, the volume enclosed by the vaginal surface.
Vaginal Mobility/Distensibility and Genital Hiatus Measures:
Measurements of vaginal mobility/distensibility and GH size were
obtained from the primary DEMAND study by methods previously described.
In short, rest (with prolapse fully reduced) and strain sagittal MRIs
were co-registered in 3D Slicer using the 3D PCS. The superior-inferior
and anterior-posterior axes defined the midsagittal plane. In this
plane, the anterior and posterior vaginal walls were outlined. A line
was drawn between the posterior margin of the external urethral meatus
and anterior margin of the perineal body (PB) to approximate the vaginal
introitus length, GH size, and the level of the hymen (i.e., hymenal
line). The position of the vaginal apex, AVW (i.e., point along the AVW
wall corresponding to its half-length), and PB (i.e., the posterior
margin of the GH) were identified. Using these landmarks, the following
measurements commonly associated with prolapse recurrence were obtained:
vaginal mobility (displacement of the vaginal apex, AVW, and PB from
rest to strain); vaginal distensibility (elongation of the AVW and
vaginal introitus from rest to strain); and GH size (at recovery and
strain). Vaginal mobility/distensibility parameters are dynamic
measures, where displacement is the straight-line distance between the
rest and strain position and elongation is the difference in length
between rest and strain. The GH size parameters are static measures
assessed at recovery and strain.
Definition of Prolapse Recurrence:
Prolapse recurrence was defined as prolapse beyond the level of the
hymen at strain on MRI. Using the midsagittal trace of the vaginal wall
and hymenal line in the strain MRI, vaginal protrusion past the hymenal
line indicated prolapse recurrence.
Statistical Analysis:
Descriptive statistics of demographic and medical history information
were calculated and stratified by recurrence or success within each
surgery group. Linear models including surgery group,
recurrence/success, and their interaction were fit to each vaginal
morphology and position measure. Model-estimated means and standard
errors were calculated for each combination of surgery group and
recurrence/success. Model-estimated differences and 95% confidence
intervals were calculated for the difference between success and
recurrence within each surgery group and for the difference between
surgery groups within recurrence and success. The association between
(1) vaginal morphology/position parameters and (2) vaginal
mobility/distensibility and GH measures was analyzed using Pearson
correlation. All statistical tests were two-sided and evaluated at a
significance level of 0.05. Due to the exploratory nature of the
analysis, no adjustment for multiple comparisons was performed. All
statistical analyses were performed using SAS software v9.4 (SAS
Institute Inc, Cary, NC).
RESULTS:
Study Population:
Of the 88 women from the primary DEMAND study, 82 met analysis inclusion
criteria (41 hysteropexy, 41 hysterectomy). Of those, 75 (38
hysteropexy, 37 hysterectomy) were imaged at 30-42 months, six (2
hysteropexy, 4 hysterectomy) were imaged prior to 30 months, and one
(hysteropexy) was imaged at 48 months. Full baseline and 30–42-month
follow-up characteristics of the study cohort are given in Table
1 . The population primarily consisted of white (81.7%), older (65±8
years), postmenopausal (97.6%) women. Based on MRI criteria, 34 (41%)
women had recurrence, with 12/41 (29%) in the hysteropexy group and
22/41 (54%) in the hysterectomy group.
Physiological Vaginal Morphology and Position
Model-estimated group differences in physiological vaginal
characteristics related to morphology (angulation, dimensions) and
position are provided in Table 2 .
Recurrence versus Success
After hysteropexy, women with recurrence had larger upper-lower and
upper vaginal coronal angles oriented -9.5º (p=0.009) and -7.4º (p=0.02)
farther towards the left side of the body (more laterally deviated)
whereas women with success had smaller upper-lower and upper vaginal
coronal angles nearly aligned with the midline of the body (closer to
0º). Following hysterectomy, the vaginal apex and mid-vagina were -8.7
mm (p=0.01) and -6.1 mm (p=0.02) more inferiorly positioned (lower) in
the recurrence group than in the success group.
Hysteropexy versus Hysterectomy
Within the recurrence group, women treated with hysteropexy had an
upper-lower vaginal coronal angle oriented -7.7º (p=0.04) farther to the
left side of the body while women treated with hysterectomy had an
upper-lower vaginal coronal angle aligned with the midline of the body
(0º). In addition, with recurrence, the hysteropexy group had a 5.8 mm
(p=0.02) wider transverse vaginal width and 12.1 cm2(p=0.04) larger vaginal surface area than the success group. Within the
success group, women treated with hysteropexy had a smaller upper
vaginal sagittal angle oriented -6.4º (p=0.04) more horizontally towards
the sacrum compared to women treated with hysterectomy. Additionally,
within successes, the vaginal apex and mid-vagina were -8.8 mm (p=0.005)
and -5.2 mm (p=0.03) more posteriorly positioned (closer to the sacrum)
in the hysteropexy group than in the hysterectomy group.
Relationship with Vaginal Mobility/Distensibility and GH
Size
None of the correlations exceeded ±0.4 (moderate correlation)
(Table S1). Of the vaginal mobility measures, greater AVW
displacement correlated with a smaller transverse vaginal width
(r=-0.24, p=0.03). Regarding vaginal distensibility, greater vaginal
introitus elongation correlated with a more posterior vaginal apex
(r=-0.26, p=0.02) and greater AVW elongation correlated with a lower
vaginal apex (r=-0.26, p=0.02) and mid-vagina (r=-0.34, p=0.002), larger
(more horizontal) lower vaginal sagittal angle (r=0.27, p=0.01), and
more obtuse (straighter) upper-lower vaginal sagittal angle (r=0.29,
p=0.008). In addition, elongation of the vaginal introitus correlated
with a more posterior vaginal apex (r=-0.26, p=0.02). Lastly, a larger
GH at rest correlated with a lower mid-vagina (r=-0.28, p=0.01), larger
(more vertical) upper vaginal sagittal angle (r=0.24, p=0.03), and more
obtuse upper-lower vaginal sagittal angle (r=0.23, p=0.04).
DISCUSSION:
Main Findings:
After hysteropexy, a more laterally oriented vagina, particularly the
upper vagina, was associated with prolapse recurrence. After
hysterectomy, a lower vaginal apex and mid-vagina were associated with
prolapse recurrence. Among recurrences, women treated with hysteropexy
had a more laterally deviated vagina, wider transverse vaginal width,
and larger vaginal surface area compared to women treated with
hysterectomy. Among successes, the hysteropexy group had a more
horizontal upper vagina and more posteriorly positioned vaginal apex and
mid-vagina (pulled towards the sacrum) compared to the hysterectomy
group.
Weak correlations were observed between (1) physiological vaginal
morphology and position factors and (2) vaginal mobility/distensibility
and GH size measures, where only the superior-inferior position of the
mid-vagina explained >10% of the variation in AVW
elongation. Most notably, a lower (more inferiorly positioned)
mid-vagina and straighter (more obtuse) upper-lower vaginal sagittal
angle were both correlated with greater AVW elongation and larger
physiological rest GH.
In the primary DEMAND study, assessment of vaginal morphology with
respect to prolapse recurrence and vaginal surgery was limited to
midsagittal (2D) MRI measurements of vaginal length. By expanding the
vaginal analysis to 3D, this study was able to distinguish differences
in lateral orientation and position of the vagina in recurrences versus
successes within the hysteropexy and hysterectomy group, respectively.
Furthermore, the current study identified vaginal angulation and
position factors correlated with vaginal mobility/distensibility and GH
measures found to be associated with prolapse recurrence in the primary
DEMAND study.
Interpretation:
Regardless of prolapse surgery type or outcome, the mean postoperative
upper-lower vaginal sagittal angle across groups
(~171°-177°) was larger than its normal value (160°) for
women of similar age and vaginal parity. Thus, normal vaginal anatomy
was not maintained postoperatively—a key goal of many pelvic
reconstructive procedures for POP. Studies have shown that prolapse
surgeries often straighten the vagina, resulting in an irregular vaginal
angulation and position that may predispose to prolapse recurrence.
It is speculated that there are two key vaginal support structures for
maintaining vaginal angulation and position, the uterosacral ligaments
(USLs) and levator ani muscles (LAM). The USLs pull the upper vagina
horizontally and posteriorly, positioning it above the levator plate and
towards the sacral hollow.
Detachment,
elongation, or injury of the USLs can lead to anterior tilt and loss of
horizontal-posterior orientation of the upper vagina. This allows the
upper vagina to displace anteriorly past the levator plate and
inferiorly towards the GH, increasing the risk of vaginal descent and
POP. This anterior and inferior deviation of the upper vagina has been
observed in prolapse surgeries like hysterectomy which involve
disruption of the USLs and was also seen in the hysterectomy group of
this study. With mesh-augmented prolapse surgery, these vaginal
angulation and position changes would involve loosening or lengthening
of the mesh arms.
Interestingly, this study showed that the upper vagina was displaced
laterally in women with prolapse recurrence after hysteropexy,
particularly towards the patients’ left side. In these cases, the right
mesh arm appeared loose and wave-like compared to the left mesh arm on
MRI, suggesting consistent asymmetrical loss of support. Clinically, the
right sacrospinous ligament is believed to be more easily identifiable
and accessible during sacrospinous fixation, mainly by right-handed
surgeons. Thus, it is plausible that this preference may account for the
disproportionate “failure” of the right mesh arm (potentially due to
asymmetrical tensioning and subsequent gradual lengthening) that leads
to a laterally displaced upper vagina towards the patient’s left side.
This study also demonstrated that among successes, hysteropexy was
better able to preserve a more horizontally and posteriorly oriented
upper vagina than hysterectomy.
The effects of apical support loss are amplified with the presence of
LAM defects, where the levator plate is responsible for upper vaginal
support. When defective, the LAM is less able to pull the lower vagina
towards the pubic bone and reflexively contract to stabilize the upper
vagina in response to increased intrabdominal pressure. As a result, the
levator plate is straighter and more dorsally oriented, which places
abnormal loading and strain on the USLs and accounts for their elongated
and hypertrophic appearance seen clinically. Loss of levator plate
angulation is often indicative of LAM defects and accompanied with an
enlarged GH, both of which have been associated with prolapse
recurrence.
Though it is unclear which is the result or cause of the other,
defective apical vaginal support (ligaments, mesh) and distal vaginal
support (pelvic floor muscles) can result in a straighter vagina that
leaves the AVW more susceptible to elongate and descend with
intraabdominal pressure.
Similar
observations were noted in this study, where a larger upper-lower
vaginal sagittal angle and lower mid-vagina correlated with greater AVW
elongation and larger recovery GH, and another study that reported that
a lower mid-vagina correlated with a larger strain GH and prolapse
recurrence. Longitudinal randomized clinical trials that
radiographically assess vaginal anatomy before and after prolapse
surgery are needed to distinguish the causative mechanisms of abnormal
postoperative vaginal angulation and position—surgical technique
versus anatomical defects—and their individual role in prolapse
recurrence.
Strengths and Limitations:
This study prospectively evaluated vaginal anatomy after vaginal surgery
in a well-characterized cohort of women from a randomized clinical
trial. A major strength of this work was that the vagina was analyzed in
3D using computational morphometry. Previous studies have been limited
to 2D (planar) and manual vaginal measurements. The computational
methods of this study allowed evaluation of the full vaginal geometry
relative to a coordinate system which minimized subjectivity and
variability of vaginal measurements through 3D modeling and automation.
A major limitation of this study was that the MRI sequences used for the
analysis were obtained in the supine position. Thus, the ability to
capture the full extent of prolapse was dependent on patient effort and
proper performance of straining. To address this, participants underwent
training on how to properly maximally strain prior to imaging and were
required to perform multiple attempts to achieve maximal strain during
the MRI examination. Only about half of SUPeR participants enrolled in
DEMAND. Thus, the DEMAND cohort was not a random sample of SUPeR
patients. However, when comparing women enrolled versus not enrolled in
DEMAND, baseline patient characteristics were similar. Another
limitation of this study was the absence of controls (preoperative MRI,
immediate/short-term postoperative MRI) which made it difficult to
isolate the individual impact of surgery versus POP on vaginal anatomy.
CONCLUSION:
Following
hysterectomy, women with prolapse recurrence have a more laterally
deviated vagina compared to women with success. After hysterectomy,
women with prolapse recurrence have a lower vaginal apex and mid-vagina
than women with success. A lower mid-vagina and straighter upper-lower
vaginal sagittal angle correlated with greater AVW elongation and larger
recovery GH. Findings provide considerations for prolapse surgeries
suggesting that postoperative changes in vaginal anatomy that result in
a straighter, more inferiorly positioned vagina may predispose to
prolapse recurrence, particularly with an enlarged GH. Future studies
will assess 3D shape variation of the vagina and pelvic floor muscles
after hysteropexy versus hysterectomy to investigate the relationship
between pelvic floor muscle and vaginal morphology with respect to
prolapse surgery, prolapse recurrence, and LAM defects.