Sage Journals: Discover world-class research

Abstract

Although it is generally accepted that endometriosis is responsible for dysmenorrhea, as well as other symptoms such as infertility, the precise relationship between the severity of dysmenorrhea and various characteristics of disease, such as stage and the type or site of endometriotic lesions, has been elusive and often a matter of heated debate, owing largely to conflicting reports. Here we review factors that are reported to be associated with dysmenorrhea in endometriosis. We also demonstrate, through a real example, that different statistical models employed in data analyses may yield somewhat different sets of risk factors, and the difference may not be always resolved within the same data set. In addition, we make the point that despite the best-fitting models, there may still be a sizable portion of variation in the severity of dysmenorrhea that can not be explained completely by the identified risk factors, suggesting that factors other than those of surgical findings or patient characteristics may also be responsible for dysmenorrhea severity. We review some possible factors that may also be responsible for the risk and severity of dysmenorrhea. Finally, we expose areas in need of further research.

Keywords

dysmenorrhea endometriosis risk factors severity statistical methodology

Dysmenorrhea refers to severe and disabling uterine pain during menstruation and reportedly affects 40–70% of women of reproductive age [1 –4]. Historically, dysmenorrhea has been categorized into two types: primary and secondary dysmenorrhea. The latter is diagnosed when symptoms are attributable to an underlying disease/disorder such as endometriosis, or a structural abnormality either within or outside the uterus, and the former is diagnosed when none of these are detected. While many women experience minor discomfort or pain during menstruation, dysmenorrhea is diagnosed when the pain is so severe as to interfere with normal activities, or to require medical attention. In women of reproductive age, dysmenorrhea appears to be the most prevalent pain among all chronic pelvic pains [5], which are primary and secondary indications for approximately half of all gynecologic laparoscopies [6]. A community-based survey of women in the UK found that 81% of women with chronic pelvic pain have dysmenorrhea [2]. Known risk factors for dysmenorrhea include: younger age (<30 years), being thin (BMI <20), smoking, early menarche (<12 years), longer cycles/duration of bleeding, irregular or heavy menstrual flow, presence of premenstrual symptoms, clinically suspected pelvic inflammatory disease, sterilization and history of sexual assault [7]. The most common cause of secondary dysmenorrhea is endometriosis [8], which is also the most prevalent diagnosis associated with pelvic pain symptoms [9 –11].

The link between dysmenorrhea and endometriosis also owes to the fact that the three common pharmacological modalities for treating endometriosis, specifically, gonadotropin-releasing hormone agonists, progesterones and androgenic agents, all suppress implant size [12] and alleviate, more or less, the pain associated with endometriosis [13,14]. Hence, it is logical to postulate that pelvic pain, including dysmenorrhea, may be attributed to the presence of endometrial implants.

However, even though it is generally accepted that endometriosis is causally associated with dysmenorrhea, as well as other symptoms such as infertility [15], the delineation of the exact relationship between the severity of dysmenorrhea and various characteristics of disease, such as stage and the type, site or histology of endometriotic lesions, has been tenuous and often a matter of conflicting reports [16,17]. Indeed, several studies failed to find any relationship between characteristics of endometriotic lesions and pelvic pain symptoms [18 –21]. Yet perhaps the most unsettling of all is a recent finding that in women receiving a second surgery for the return of pain, biopsy-proven endometriosis is not associated with the return of pain [22].

Controversies and conflicting reports are certainly not restricted to studies on the relationship between dysmenorrhea and various characteristics of endometriosis. Many factors could potentially contribute to the heterogeneous findings.

These include, but are not limited to, type of the study (prospective vs retrospective); indication for surgery (infertility, pain or pelvic mass); type of endometriosis; how the pain is evaluated; and genuine heterogeneity among different patient populations, which could be amplified in retrospective studies. Data analytic methodology or different statistical models may also be responsible. Even sample size and year of publication may have an influence [10].

In this paper, we provide an overview of the factors reported to be associated with dysmenorrhea in endometriosis. Through a real example, we show that different statistical models employed in data analyses may yield different sets of risk factors, and the difference is not always resolved within the same data set. In addition, despite the best-fitting models, there is still a sizable portion of variation in the severity of dysmenorrhea that can not be explained completely by the identified risk factors. This suggests that factors, other than those of surgical findings or patient characteristics, may also be responsible for the risk and severity of dysmenorrhea. Several possible factors that may also be responsible for determining the severity are discussed, in addition to stating areas that require further research.

Factors associated with the severity of dysmernorrhea in women with endometriosis

We performed a thorough search of the Pubmed database using the keywords dysmenorrhea, endometriosis, risk factors, severity and epidemiology, to retrieve papers on factors associated with the severity of dysmenorrhea in women with endometriosis. The results were summarized in Table 1 .

Table 1.

Factors reported to be associated with dysmenorrhea in women with endometriosis.

Authors	Type of endometriosis	Sample size	Factors investigated	Pain assessment method	Multivariate analysis	Risk factors identified	Ref.
Koninckx et al. (1991)	Mixed	451	Type and histology of lesions	Medical chart	Yes	Depth to which the lesions penetrate	[34]
Marana et al. (1991)	Mixed	206	rAFS stage, age	Q	No	Not reported	[18]
Fedele et al. (1992)	Mixed	124	rAFS stage, type of endometriosis	Q	No	Ovarian endometriomas	[20]
Perper et al. (1995)	NS	47	Number, type and location of lesions; number, type and location of adhesions	Q	No	Number of endometrial implants	[43]
Vercellini et al. (1996)	Mixed	244	Stage and site of ectopic implants	VAS	No	None	[71]
Muzii et al. (1997)	NS	65	rAFS partial scores for superficial, deep and adhesion; number of superficial implants; number of typical, pigmented versus atypical, nonpigmented lesions	VAS	No	rAFS score	[21]
Porpora et al. (1999)	Mixed	72	Stage, site, and type	Q	Yes	Presence and extent of adhesion	[28]
Anaf et al. (2000)	Mixed, mostly with rectal–vaginal endometriosis	Pain score >7: n = 19 ≤7: n = 9	Histology and nerve density	VAS	No	Higher intrafibrotic and intraglandular nerve density, more endoneurial and perineurial invasion	[35]
Momoeda et al. (2002)	NS	Infertility group: n = 476 Pain group: n = 616	rAFS stage	Q	No	Disease stage	[23]
Fauconnier et al. (2002)	DIE	225	Location of DIE	Medical chart	Yes	Douglas pouch adhesions, low parity	[11]
Chapron et al. (2003)	DIE	209	Number of previous surgeries for endometriosis, rAFS, extensiveness of adnexal adhesion; Douglas obliteration; size of the posterior DIE implant; extent of the subperitoneal infiltration by the posterior DIE	VAS	Yes Ordinal logistic regression model with cumulative odds	Presence of a rectal or vaginal infiltration by the posterior DIE and extensiveness of adnexal adhesion	[27]
Tamburro et al. (2003)	Peritoneal endometriosis	35 patients 10 controls	TGF-β1 Immunohistochemistry	VAS	No	Expression of TGF-β1 in nerve fibers	[67]
Chopin et al. (2006)	Histologically confirmed OE of size ≥2 cm	239	Age, height, weight, BMI, parity, history of previous treatment, number, site and mean size of lesions, CA125, rAFS stage and histology	VAS during the month prior to surgery	Yes	rectal infiltration rAFS score >26	[24]
Milingos et al. (2006)	NS	95	rAFS stage	VAS	No	rAFS score >16	[30]
Leng et al. (2007)	OE and pertitoneal endometriosis	95	Various clinical factors	VAS	Yes	Nodules in uterosacral ligaments, invasive depth of uterosacral ligaments in patients with OE	[5]
Mehmud et al. (2007)	NS	50	rAFS stage Several clinical indicators	Four-point verbal rating	No	Presence of superficial implants	[29]
Vercellini et al. (2007)	All types	1054	rAFS stage, age	VAS	Yes	Younger age at surgery, rAFS stage	[72]
Liu et al. (2008)	OE	376	Size, rAFS stage, presence of adhesion and various clinical factors	Three-point verbal rating	Yes Various multivariate analytic methods	Co-occurrence of adenomyosis	[26]

DIE: Deep infiltrating endometriosis; NS: Not specified; OE: Ovarian endometriomas; Q: Questionnaire; rAFS: Revised American Fertility Society; VAS: Visual analog scale.

One factor that has been examined frequently in relation to the severity of dysmenorrhea is the revised American Fertility Society (rAFS) stage, yet the verdict is far from unanimous (Table 1). For example, while rAFS stage was found to be associated with the severity of dysmenorrhea by Muzii et al. [21], Momoeda et al. [23] and Chopin et al. [24], some other authors did not find such a correlation [18,20,25 –29]. Milingos et al. and Chapron et al. found that higher rAFS scores, instead of rAFS stage, are a risk factor [24,30].

Similarly, for specific characteristics of endometriotic lesions, it appears that there is no consensus as to which characteristics are closely related with the severity of dysmenorrhea (Table 1). Adhesion was reported by some to be associated with dysmenorrhea severity [11,27,28], but a larger study by Liu et al. did not find such an association [26]. This may appear a bit surprising since adhesion often entails the growth of sensory nerve fibers [31,32], leading to pain. It is possible that the discrepancy may be due to the subtype that Liu et al. investigated, specifically, ovarian endometriomas, which have been reported not to contain nerve fibers [33].

There is accumulating evidence to suggest that deep infiltrating endometriosis (DIE) is strongly associated with dysmenorrhea. Several studies reported that the depth of infiltration is associated with the dysmenorrhea severity [25,34], especially rectal infiltration [24,27]. This association appears to be corroborated by meticulous investigations of the distribution of nerve fibers and/or their marker, NGF, expression in endometriosis, demonstrating that the implants that penetrate the wall of adjacent organs have closer relationships with nerve fibers than those that do not [35,36]. It is unclear whether the presence of DIE can account for most of the variation in dysmenorrhea severity.

The recent finding by Liu et al. that the presence of adenomyosis is a strong and consistent risk factor for the severity of dysmenorrhea is worth commenting on [26]. Given that the prevalence of adenomyosis is as high as 90% in women with endometriosis under the age of 36 years [37,38] and that dysmenorrhea is one of chief complaints among women with adenomyosis [39], the finding appears to have uncovered something obvious that so far has been overlooked, and is in line with a recent report that patients with a history of prolonged dysmenorrhea are very likely to have adenomyosis [40]. Indeed, endometriotic lesions tend to be more severe in the posterior wall of the uterus, which would easily adhere to the rectum and infiltrate into the sacral nerve plexus, resulting in exacerbated dysmenorrhea [37]. Even after surgery, the persistence of dysmenorrhea and chronic pelvic pain has been suggested to be indicative of adenomyosis [41]. Alternatively, the well-documented uterine hyperperistalsis and dysperistalsis in women with endometriosis could easily result in increased uterine contractions and/or vasoconstriction, leading to dysmenorrhea [42].

There are some other isolated reports demonstrating factors that are associated with the severity of dysmenorrhea. For example, Perper et al. found that the total number of ectopic endometrial implants is associated directly with the intensity of dysmenorrhea experienced after examining the number, type and location of endometrial implants when these factors were evaluated during laparoscopy [43].

From Table 1 we can see that there is no apparent consensus regarding which factors are associated with the severity of dysmenorrhea.

There are many reasons for discrepant findings. Heterogeneity in patient population owing to the tertiary nature of the hospital, bias in referral, composition of the patient population (e.g., different proportion of disease severity), or type of endometriosis is an obvious factor. A less appreciated factor is the statistical methodology used in data analyses (see below).

Methodological issues

Conceivably, there are many potential factors that may determine the risk or severity of dysmenorrhea. Evaluation of just a single factor at a time often entails the risk of bias and confounding and, thus, spurious findings, particularly when data are collected retrospectively. Despite the obvious advantage of controlling for confounding, not all published studies used multivariate statistical methodology, as can be seen from Table 1.

For many studies that did use multivariate statistical methods, such as logistic regression [27,44], a subtle, less appreciated, yet no less important point is that, for the data at hand, it is often the case that several statistical models, which often entail different assumptions and could lead to not-so-congruent results, fit the data equally well and in many cases it may be challenging to determine which is the best-fitting model.

To see this point, we use the data published recently by Liu et al. [26]. Liu et al. included 850 consecutive patients with ovarian endometrioma undergoing surgery at a tertiary hospital [26]. Among the 850 patients, 110 were lost to follow-up, leaving 710 who were available for analysis. Of these 710 patients, 376 (53%) complained of dysmenorrhea, with 245 (65.2%), 107 (28.5%) and 24 (6.4%) being mild, moderate and severe, respectively.

For each patient, information was collected, with the use of medical charts and interviews, on demographics; age at surgery; age at menarche; BMI at surgery; pelvic exams and results; type of surgery; mode of surgery (conservative or semi-radical); complaint of dysmenorrhea or not; duration and severity of dysmenorrhea; number of induced abortions prior to surgery if applicable; presence of adenomyosis or not; previous use of endometriosis-related medication or not prior to the surgery; previous endometriosis-related surgeries; laterality of endometrioma; size of the largest endometrioma; presence of adhesion or not; rAFS scores and stage; postoperative use of medication or not; and improvement in symptoms.

For the data set of 376 women with dysmenorrhea, Liu et al. first performed a multiple logistic regression for binary or dichotomous response by lumping women with moderate and severe dysmenorrhea into one group. In this case, the severity data were collapsed into binary data: mild, or moderate and severe (Figure 1A). This approach yielded two factors that are positively associated with the severity of dysmenorrhea: the presence of adenomyosis and shorter duration of dysmenorrhea (Table 2). Obviously, the data could alternatively be collapsed into mild and moderate versus severe, which may or may not yield identical results. Liu et al. chose the first way of collapsing owing to the fact that there were few in the ‘severe’ group [26].

Figure 1.

Severity data structure.

Table 2.

Parameter estimates of the dichotomous logistic regression model on risk factors for severity of dysmenorrhea when moderate and severe are catergorized together.

Covariate	Estimate	Standard error	Odds ratio (95% CI)	p-value
Presence of adenomyosis	0.1880	0.0502	1.207 (1.093, 1.332)	0.0002
Duration (in 12 months)	-5.62 × 10⁻³	2.75 × 10⁻³	0.9944 (0.9891, 0.9998)	0.0414

Data taken with permission from Liu et al. [26].

The second approach was based on a proportional odds model using the logistic link. This model assumes, implicitly, that the data were ordered, categorical data, with an implicit underlying order (scale of severity) in the data (Figure 1B). With this model, the presence of adenomyosis, the rAFS stage and previous use of medication were identified as risk factors, and having previous induced abortions was identified as a protective factor associated with severity (Table 3). The same class of the model with different links, such as extreme value, probit and Cauchit, all identified the presence of adenomyosis as one of risk factors, but they also resulted in a slightly different set of risk factors (data not shown). The score test indicated that the assumption of proportional odds is acceptable. The analysis based on partial proportional odds models also led to the same proportional odds model [26].

Table 3.

Parameter estimates of the proportional odds regression model on factors associated with the severity of dysmenorrhea.

Covariate	Estimate	Standard error	Odds ratio (95% CI)	p-value
Presence of adenomyosis	0.7602	0.2285	2.139 (1.367, 3.347)	0.0004
rAFS stage II	1.4591	0.7141	4.302 (1.061, 17.439)	0.021
rAFS stage III	0.8848	0.6721	2.423 (0.649, 9.044)	0.094
rAFS stage IV	1.5027	0.6821	4.494 (1.180, 17.109)	0.014
Previous medical treatment	0.6356	0.2914	1.888 (1.067, 3.343)	0.015
Previous abortions	-0.2883	0.1256	0.750 (0.586, 0.959)	0.011

rAFS: Revised American Fertility Society.

Data taken with permission from Liu et al. [26].

The third approach is a polychotomous logistic regression model [45]. Similar to the fourth model as detailed below, here it is assumed that a woman with moderate dysmenorrhea does not necessarily have to go through the stage of mild to become ‘moderate’; instead, different severity levels are merely different categories or labels (of patients) (Figure 1C). Therefore, the severity of dysmenorrhea can be viewed as nominal polychotomous data. Under this model, the presence of adenomyosis, rAFS stage and the size of the largest cyst were identified to be risk factors (Table 4).

Table 4.

Parameter estimates of the final polytomous logistic regression model on factors associated with the severity of dysmenorrhea.

Covariate	Severity	Estimate	Standard error
Presence of adenomyosis	Moderate	0.5513	0.2416
	Severe	1.3013	0.4534
rAFS stage	Moderate	0.2124	0.1599
	Severe	0.7437	0.3459
Size of the largest cyst	Moderate	-0.5155	0.2218
	Severe	0.3884	0.3836

rAFS: Revised American Fertility Society.

Data taken from Liu et al. [26].

The fourth model used was the semiparametric polychotomous logistic regression model [46]. Define Y as the random variable of dysmenorrhea severity, which takes value of 1 (mild), 2 (moderate) or 3 (severe). Let X be a vector of covariates of length M, and x, a realization of X. The polychotomous logistic regression model is defined as:

\begin{array}{l} P (Y = k Ɩ x) = \\ \frac{exp Θ (k Ɩ x)}{exp Θ (1 Ɩ x) + exp Θ (2 Ɩ x) + exp Θ (3 Ɩ x)} \end{array}

\begin{array}{l} where k = 1, 2 or 3, \\ exp Θ (k Ɩ x) = exp [Θ (k Ɩ x; β_{k})] = \end{array}

exp [\sum_{j = 1}^{J} β_{j k} B_{j} (x)]

J is a positive integer and B₁…B_j are basis in the J-dimensional linear space of function spanned by x, with each B being piecewise linear splines and their selected tensor products, and β_k =(β_k1…β_kj)^T is a vector of parameters to be estimated [46]. This model differs from the third model in that it takes a semiparametric form instead of a parametric one; thus, it is more robust. Under this model, the presence of adenomyosis was identified to be the only risk factor.

The last approach was the Cox regression model for the discrete or grouped survival time data [47]. Here, we assumed that the severity of dysmenorrhea can progress through various stages, hence, patients who start with ‘none’ or ‘mild’ can deteriorate to ‘moderate’ or even ‘severe’ and are unlikely to reverse this progression (Figure 1D). With this model, we can treat the severity as if they were discrete ‘survival time’ without censoring. However, we note that longer survival time actually means a more severe dysmenorrhea; hence, the ‘risk factors’ identified through this approach are actually ‘protective factors’ for dysmenorrhea severity. This model yielded three factors, the presence of adenomyosis, rAFS stage and age at menarche (Table 5).

Table 5.

Parameter estimates of the Cox model for discrete survival data on factors associated with the severity of dysmenorrhea.

Covariate	Estimate	Standard error	Hazard ratio	p-value
Presence of adenomyosis	-0.504	0.125	0.604	0.000
rAFS stage	-0.208	0.077	0.812	0.007
Age at menarche	0.078	0.040	1.081	0.048

rAFS: Revised American Fertility Society.

Data taken from Liu et al. [26].

It is interesting to note that all five statistical models consistently identified the presence of adenomyosis as a risk factor for the severity of dysmonorrhea, but the set of risk factors are somewhat different under different models. While it is reassuring that the presence of adenomyosis is consistently and robustly identified by different models with different assumptions, indicating that this factor is indeed a risk factor for severity, it is somewhat disappointing that no single model stood out as the best-fitting one, since all five models appeared to fit the data equally well as judged by some diagnostic tests.

This example serves to remind us that:

Different statistical models may yield different results

Several models may fit the data equally well

It may not possible to determine which model fits the data the best

If the results can be different within the same study, one can imagine how discrepant it could be when different studies are published by different teams using different data and methodologies. Of course, if a factor is consistently identified by different groups, then the likelihood that it is a genuine one would be higher.

One final point, perhaps not appreciated as well, is that there are factors other than host or specific characteristics of the endometriotic lesions that may also be responsible for dysmenorrhea and its severity. This can be seen from the variation of severity that can be explained by the statistical model. For the above example, the semiparametric polychotomous logistic regression model only explained approximately 20% of variation in severity when presence of adenomyosis is included in the model. Another 80% is currently unaccounted for. For the binary logistic model, the percentage of correct classification under the fitted model is 65.2% for the same data, far from perfect.

Other possible factors

If there are other factors, besides the lesion characteristics and host factors, that are also responsible for the severity of dysmenorrhea, what could they be?

To understand this, we point out that dysmenorrhea is essentially an intense visceral pain during menstruation and is one of many forms of pain, which is ‘an unpleasant sensory and emotional experience associated with actual or potential tissue damage, or described in terms of such damage’ [48]. As such, the perception of dysmenorrhea may involve mind as well as body. In other words, affective and cognitive as well as physiological factors may be involved.

Several studies suggest that somatization and poor coping are also positively associated with menstrual pain intensity, suggesting that central factors should also be considered. Somatization refers to a tendency to experience and communicate somatic distress in response to psychosocial stress and to seek medical help for it [49]. The broad category of somatization includes the somatoform disorders that encompass pain disorders and is found to be strongly associated with chronic pain and dysmenorrhea [50,51].

Some investigators reported that women with dysmenorrhea tend to be more preoccupied with bodily sensations, tend to express greater negative attitudes toward illness and have a more negative effect toward menstruation than women without dysmenorrhea [52,53]. Goldstein-Ferber and Granot reported that high levels of somatization predict pain intensity, and cultural variables and personality traits also play a role in the perception of dysmenorrhea [54].

Granot et al. reported that, compared with women without dysmenorrhea, women with dysmenorrhea have an enhanced perception, as measured by visual analog rating of pain intensity, and response, as measured by pain-evoked potential latencies, when the pain stimuli (laser) was applied to locations unrelated with menstruation or even the reproductive tract (hand), and these changes were seen across the entire menstrual cycle, not just during menstruation [55]. These changes appear to suggest a systemic phenomenon since the pain stimuli were applied to a location seemingly unrelated to the reproductive system.

Bajaj et al. reported that women with dysmenorrhea show altered somatosensory perception in several modalities [56]. Specifically, dysmenorrheic women show reduced pain thresholds during menstruation to heat and pressure stimulation, both within and outside areas of referred menstrual pain, with a lower heat-pain threshold at sites unrelated to reproductive tract (arm and thigh) and a lower pressure-pain threshold at all sites, compared with women without dysmenorrhea. The decrease is more pronounced within the referral pain areas (low back and abdominal sites) [56]. It is speculated that these alterations may depend on a spinal mechanism of central hyperexcitability, induced by recurring dysmenorrhea [56].

Similar evidence is also provided by Brinkert et al. who found that, despite the absence of overt gastrointestinal symptoms or viscero-somatic sensitization, women with dysmenorrhea have increased sensitivity in the sigmoid colon and rectum [57]. This hypersensitivity may result from centrally mediated viscero-visceral hyperalgesia owing to recurrent intense menstrual pain [57]. Alternatively, this may suggest that dysmenorrhea may be a manifestation of central hyperexcitability [58], which is known to play an important role in neuropathic pain [59].

The notion that altered somatosensory perception in women with dysmenorrhea may be centrally mediated appears to be supported by the observation that induced peritoneal endometriosis in rats results in hyperalgesia in the vaginal area, an area remote to the lesions [60]. Thus, there exists a potent cross-system, viscero–visceral interaction in which pathophysiology in one organ influenced the physiology and response to pathophysiology of another organ [61]. If this is also true in humans, then the imperfect correlation between stage of endometriosis and severity of dysmenorrhea may be something to be expected.

Coping strategy also appears to have an effect on the severity of dysmenorrhea. Some earlier studies noted personality differences between women with and without dysmenorrhea [62,63]. Based on a multivariate analysis involving stress, neuroticism and other variables, Whitehead et al. reported that childhood reinforcement of menstrual illness behavior significantly predicted adult menstrual symptoms and disability days, and that childhood reinforcement scales were useful to predict which functional disorders (dysmenorrhea or irritable bowel syndrome) these subjects had even after stress and neuroticism were controlled for [64]. Indeed, pain catastrophizing, that is, a disproportionally negative view of the pain experience, influences perception of menstrual pain intensity: high pain catastrophizers, in comparison with low pain catastrophizers, reported greater menstrual pain intensities, greater affective menstrual pain intensity, greater variability in the use of pain-coping strategies, lower perceived effectiveness of over-the-counter medications and nonmedical pain-coping strategies and greater disability [65]. Hence, a psychological dimension is also involved in the perception of dysmenorrhea. Not surprisingly, a recent systemic review found some evidence that behavioral interventions may be effective for the management of dysmenorrhea [66].

Even within the body, the histology and/or morphology of endometriotic lesions may not provide sufficient information for the severity of dysmenorrhea. Tamburro et al. report that higher expression of TGF-β1 is associated with the physical appearance of endometriotic implants and the severity of dysmenorrhea appears to be related to the expression of TGF-β1 in nerve fibers found in endometriotic lesions [67]. TGF-β1 is a multifunctional proinflammatory cytokine. Given the well-documented link between proinflammatory cytokines and pain (see, for example, [68]), it is perhaps not surprising to see the elevated expression of TGF-β1 in nerve fibers found in endometriotic lesions.

Along the same line, it was recently reported that Il-1 receptor antagonist (Il-1ra) concentrations in peritoneal fluid were significantly lower in women with endometriosis than those without, especially in women with disease-related dysmenorrhea [69]. Il-1ra is an endogenous Il-1β inhibitory bioactivity factor. In endometriosis, Il-1β is known to be elevated [70].

It is not clear as to whether the increased TGF-β1 expression or decreased Il-1ra production is a major culprit, an unwitting accomplice or merely an innocent bystander, in causing dysmenorrhea or general pelvic pain in women with endometriosis. Nonetheless, it is possible that certain biochemical or genomic factors may have a greater correlation with the risk and severity of dysmenorrhea than the histology/morphology of endometriosis. Future research is needed to identify biomarkers for the risk and severity of dysmenorrhea in women with endometriosis, so that a better understanding of the underlying mechanisms can be gained through this research. In addition, the identified biomarkers may potentially be targets for intervention.

Conclusion

While it is generally considered that endometriosis is responsible for dysmenorrhea, it is somewhat disheartening to see that the correlation between the severity of endometriosis and the severity of dysmenorrhea is far from perfect. A reappraisal of this imperfect correlation should prompt us to find its causes and identify other factors that are responsible. Once we have a clear understanding of which factors determine the risk and severity of dysmenorrhea, we should have a better insight into the possible mechanisms of dysmenorrhea in women with endometriosis, and may devise better therapeutics and appropriate interventional procedures.

Through a real example, we demonstrate that:

Different statistical models may yield different results;

Even within the same data set, it is possible that several statistical models with different assumptions can fit the data equally well;

It is often the case that the choice of the best-fitting model can be difficult, if not impossible.

These may explain as why different studies, which yield different data in addition to different patient populations, often report incongruent findings. In addition, we demonstrate that even the best-fitting model may not completely account for all variations in the severity of dysmenorrhea, indicating that there are other factors, not included in the analysis, that may also be involved.

While many clinical investigators may, when it comes to analyzing the dysmenorrhea severity data, view statistical methodology with glazed eyes, the truth of the matter is that a rigorous statistical analysis is one of the hallmarks of a good study and is crucial to delineate the relationship between characteristics of endometriotic lesions and severity. A rigorous analysis may unveil things that can otherwise be easily missed, and may provide some much-needed insight into the relationship between endometriosis and dysmenorrhea. Some knowledge and appreciation of the importance of proper statistical analysis would go a long way.

There are ways to reduce or minimize the heterogeneity in the identification of risk factors for dysmenorrhea severity. For example, standardization of pain evaluation, increasing the sample size and the proper use of multivariate statistical methodology can, in principle, help to find common ground. In addition, focusing on a specific type of endometriosis, for example, ovarian endometrioma, may also help.

Yet we also should keep an open mind that psychological factors may contribute to the perceived severity of dysmenorrhea. In addition, it is possible and quite plausible that biochemical, genomic and other markers may provide a better correlation with the severity than the histology or morphology of endometriotic lesions. So far little work has been done in either of the two areas.

It is generally agreed that the most effective treatment of endometriosis-associated dysmenorrhea would come from the identification and treatment of its underlying cause(s). Based on the rather short list of references in this article, it is evident we still have a long way to go before we can confidently and effectively treat endometriosis-associated dysmenorrhea, let alone tackle prevention. Given the prevalence of endometriosis and dysmenorrhea, their debilitating nature and their heavy economic burdens, more research in this area is undoubtedly warranted.

Future perspective

Given the imperfect correlation between the severity of dysmenorrhea and various clinical and histological characteristics of endometriosis, perhaps a more profitable way to delineate the causal relationship between endometriosis and endometriosis-related dysmenorrhea may be through the identification of biomarkers for dysmenorrhea in women with endometriosis. In addition, other factors, such as somatization and coping strategies, may also play a role and warrent further investigation. Once we have a much better handle on factors and biomarkers that are predictive of the dysmenorrhea severity, we should be in a better position for understanding the mechanisms of endometriosis-related dysmenorrhea and will have a better chance to develop more rational and efficacious treatment strategy for managing dysmenorrhea.

Executive summary

While it is generally regarded that endometriosis is responsible for dysmenorrhea, the correlation between the severity of dysmenorrhea and various clinical and histological characteristics of endometriosis is far from perfect.

Aside from the presence of deep infiltrating endometriosis and, to a lesser extent, of adenomyosis, the roles of other factors in determining the severity of dysmenorrhea have been contentious and controversial.

The discrepancy may arise from a myriad of factors, including the use of different statistical methodology. Yet even the best-fitting models still can not account for all variations in dysmenorrhea severity.

One way to delineate the causal relationship between the severity of endometriosis and the severity of dysmenorrhea may be through the identification of biomarkers for dysmenorrhea in women with endometriosis.

Other factors, such as somatization and coping strategies, may also play a role in determining the dysmenorrhea severity and are worth investigation in the future.

A reappraisal of this imperfect correlation should prompt us to find its causes and identify other factors that are responsible.

Once we have a clear understanding of which factors determine the risk and severity of dysmenorrhea, we should have a better insight into the possible mechanisms of dysmenorrhea in women with endometriosis, and may devise better therapeutics and appropriate interventional procedures.

Footnotes

The authors received funding from the Shanghai Science and Technology Commission (Grant 074l19517). The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.

References

Pullon

Reinken

Sparrow

: Prevalence of dysmenorrhoea in Wellington women. N. Z. Med. J. 101(839), 52–54 (1988).

Zondervan

Yudkin

Vessey

: The community prevalence of chronic pelvic pain in women and associated illness behaviour. Br. J. Gen. Pract. 51(468), 541–547 (2001).

Weissman

Hartz

Hansen

Johnson

: The natural history of primary dysmenorrhoea: a longitudinal study. BJOG 111(4), 345–352 (2004).

Patel

Tanksale

Sahasrabhojanee

Gupte

Nevrekar

: The burden and determinants of dysmenorrhoea: a population-based survey of 2262 women in Goa, India. BJOG 113(4), 453–463 (2006).

Latthe

Say

Gulmezoglu

Khan

: WHO systematic review of prevalence of chronic pelvic pain: a neglected reproductive health morbidity. BMC Public Health 6, 177 (2006).

Howard

: Laparoscopic evaluation and treatment of women with chronic pelvic pain. J. Am. Assoc. Gynecol. Laparosc. 1(4 Pt 1), 325–331 (1994).

Latthe

Mignini

Gray

Hills

Khan

: Factors predisposing women to chronic pelvic pain: systematic review. Br. Med. J. 332(7544), 749–755 (2006).

French

: Dysmenorrhea in adolescents: diagnosis and treatment. Paediatr. Drugs 10(1), 1–7 (2008).

Mathias

Kuppermann

Liberman

Lipschutz

Steege

: Chronic pelvic pain: prevalence, health-related quality of life, and economic correlates. Obstet. Gynecol. 87(3), 321–327 (1996).

10.

Guo

Wang

: The prevalence of endometriosis in women with chronic pelvic pain. Gynecol. Obstet. Invest. 62(3), 121–130 (2006).

11.

Fauconnier

Chapron

Dubuisson

Vieira

Dousset

Breart

: Relation between pain symptoms and the anatomic location of deep infiltrating endometriosis. Fertil. Steril. 78(4), 719–726 (2002).

12.

Olive

: Optimizing gonadotropin-releasing hormone agonist therapy in women with endometriosis. Treat. Endocrinol. 3(2), 83–89 (2004).

13.

Giudice

Kao

: Endometriosis. Lancet 364(9447), 1789–1799 (2004).

14.

Olive

Pritts

: The treatment of endometriosis: a review of the evidence. Ann. NY Acad. Sci. 955, 360–372; discussion 389–393, 396–406 (2002).

15.

Farquhar

: Extracts from the ‘clinical evidence’. Endometriosis. Br. Med. J. 320(7247), 1449–1452 (2000).

16.

Whiteside

Falcone

: Endometriosisrelated pelvic pain: what is the evidence? Clin. Obstet. Gynecol. 46(4), 824–830 (2003).

17.

Vercellini

: Endometriosis: what a pain it is. Semin. Reprod. Endocrinol. 15(3), 251–261 (1997).

18.

Marana

Muzii

Caruana

Dell'Acqua

Mancuso

: Evaluation of the correlation between endometriosis extent, age of the patients and associated symptomatology. Acta Eur. Fertil. 22(4), 209–212 (1991).

19.

Fedele

Parazzini

Bianchi

Arcaini

Candiani

: Stage and localization of pelvic endometriosis and pain. Fertil. Steril. 53(1), 155–158 (1990).

20.

Fedele

Bianchi

Bocciolone

Di Nola

Parazzini

: Pain symptoms associated with endometriosis. Obstet. Gynecol. 79(5 Pt 1), 767–769 (1992).

21.

Muzii

Marana

Pedulla

Catalano

Mancuso

: Correlation between endometriosis-associated dysmenorrhea and the presence of typical or atypical lesions. Fertil. Steril. 68(1), 19–22 (1997).

22.

Stratton

Sinaii

Segars

: Return of chronic pelvic pain from endometriosis after raloxifene treatment: a randomized controlled trial. Obstet. Gynecol. 111(1), 88–96 (2008).

23.

Reports some quite unexpected findings and finds no correlation between the return of pain and the return of surgically confirmed endometriosis.

24.

Momoeda

Taketani

Terakawa

: Is endometriosis really associated with pain? Gynecol. Obstet. Invest. 54(Suppl. 1), 18–21; discussion 21–23 (2002).

25.

Chopin

Ballester

Borghese

: Relation between severity of dysmenorrhea and endometrioma. Acta Obstet. Gynecol. Scand. 85(11), 1375–1380 (2006).

26.

Leng

Lang

Dai

: [Relationship between pain symptoms and clinico-pathological features of pelvic endometriosis]. Zhonghua Fu Chan Ke Za Zhi 42(3), 165–168 (2007).

27.

Liu

Yuan

Wang

Shen

Guo

: Risk factors for dysmenorrhea and its severity in women with ovarian endometriomas. Gynecol. Obstet. Invest. (2008) (In Press).

28.

Chapron

Fauconnier

Dubuisson

Barakat

Vieira

Breart

: Deep infiltrating endometriosis: relation between severity of dysmenorrhoea and extent of disease. Hum. Reprod. 18(4), 760–766 (2003).

29.

Porpora

Koninckx

Piazze

Natili

Colagrande

Cosmi

: Correlation between endometriosis and pelvic pain. J. Am. Assoc. Gynecol. Laparosc. 6(4), 429–434 (1999).

30.

Mehmud

Akhtar

Sadia

: Endometriosis: frequency and correlation between symptomatology and disease stage. J. Coll. Physicians Surg. Pak. 17(4), 199–202 (2007).

31.

Milingos

Protopapas

Kallipolitis

: Endometriosis in patients with chronic pelvic pain: is staging predictive of the efficacy of laparoscopic surgery in pain relief? Gynecol. Obstet. Invest. 62(1), 48–54 (2006).

32.

Tulandi

Chen

Al-Took

Watkin

: A study of nerve fibers and histopathology of postsurgical, postinfectious, and endometriosis-related adhesions. Obstet. Gynecol. 92(5), 766–768 (1998).

33.

Sulaiman

Gabella

Davis

: Presence and distribution of sensory nerve fibers in human peritoneal adhesions. Ann. Surg. 234(2), 256–261 (2001).

34.

Al-Fozan

Bakare

Chen

Tulandi

: Nerve fibers in ovarian dermoid cysts and endometriomas. Fertil. Steril. 82(1), 230–231 (2004).

35.

Koninckx

Meuleman

Demeyere

Lesaffre

Cornillie

: Suggestive evidence that pelvic endometriosis is a progressive disease, whereas deeply infiltrating endometriosis is associated with pelvic pain. Fertil. Steril. 55(4), 759–765 (1991).

36.

Anaf

Simon

El Nakadi

: Relationship between endometriotic foci and nerves in rectovaginal endometriotic nodules. Hum. Reprod. 15(8), 1744–1750 (2000).

37.

First report on the relationship between endometriotic foci and nerves in endometriosis.

38.

Anaf

Simon

El Nakadi

: Hyperalgesia, nerve infiltration and nerve growth factor expression in deep adenomyotic nodules, peritoneal and ovarian endometriosis. Hum. Reprod. 17(7), 1895–1900 (2002).

39.

Reports strong expression of the NGF-TrkA pathway in deep adenomyotic nodules, which could explain why this type of lesion infiltrates in richly innervated anatomical sites and, consequently, is associated closely with pain. However, care should be exercised since the paper actually does not show any hyperalgesic behavior per se (i.e., extreme sensitivity to pain).

40.

Kunz

Beil

Huppert

Noe

Kissler

Leyendecker

: Adenomyosis in endometriosis – prevalence and impact on fertility. Evidence from magnetic resonance imaging. Hum. Reprod. 20(8), 2309–1236 (2005).

41.

Kunz

Herbertz

Beil

Huppert

Leyendecker

: Adenomyosis as a disorder of the early and late human reproductive period. Reprod. Biomed. Online 15(6), 681–685 (2007).

42.

Benson

Sneeden

: Adenomyosis: a reappraisal of symptomatology. Am. J. Obstet. Gynecol. 76(5), 1044–1057; discussion 1057–1061 (1958).

43.

Kissler

Zangos

Kohl

: Duration of dysmenorrhoea and extent of adenomyosis visualised by magnetic resonance imaging. Eur. J. Obstet. Gynecol. Reprod. Biol. 137(2), 204–209 (2008).

44.

Parker

Leondires

Sinaii

Premkumar

Nieman

Stratton

: Persistence of dysmenorrhea and nonmenstrual pain after optimal endometriosis surgery may indicate adenomyosis. Fertil. Steril. 86(3), 711–715 (2006).

45.

Rapkin

Mayer

: Gastroenterologic causes of chronic pelvic pain. Obstet. Gynecol. Clin. North Am. 20(4), 663–683 (1993).

46.

Perper

Nezhat

Goldstein

Nezhat

: Dysmenorrhea is related to the number of implants in endometriosis patients. Fertil. Steril. 63(3), 500–503 (1995).

47.

Vercellini

Fedele

Aimi

Pietropaolo

Consonni

Crosignani

: Association between endometriosis stage, lesion type, patient characteristics and severity of pelvic pain symptoms: a multivariate analysis of over 1000 patients. Hum. Reprod. 22(1), 266–271 (2007).

48.

Analyzes data involving 1054 patients with endometriosis and shows that the association between endometriosis stage and severity of pelvic symptoms was marginal and inconsistent.

49.

Ananth

Kleinbaum

: Regression models for ordinal responses: a review of methods and applications. Int. J. Epidemiol. 26(6), 1323–1333 (1997).

50.

Kooperberg

Bose

Stone

: Polychotomous regression. J. Am. Stat. Assoc. 92, 117–127 (1997).

51.

Kalbfleisch

Prentice

: The Statistical Analysis of Failure Time Data. Wiley, NY, USA (1980).

52.

Merskey

Bogduk

: Classification of Chronic Pain: Descriptions of Chronic Pain Syndromes and Definitions of Pain Terms (2nd Edition). IASP Press, WA, USA (1994).

53.

Lipowski

: Somatization: the concept and its clinical application. Am. J. Psychiatry 145(11), 1358–1368 (1988).

54.

Hardt

Gerbershagen

Franke

: The symptom check-list, SCL-90-R: its use and characteristics in chronic pain patients. Eur. J. Pain 4(2), 137–148 (2000).

55.

Kroenke

Swindle

: Cognitive-behavioral therapy for somatization and symptom syndromes: a critical review of controlled clinical trials. Psychother. Psychosom. 69(4), 205–215 (2000).

56.

Rierdan

Koff

: Premenarcheal predictors of the experience of menarche: a prospective study. J. Adolesc. Health Care 11(5), 404–407 (1990).

57.

Sigmon

Dorhofer

Rohan

Boulard

: The impact of anxiety sensitivity, bodily expectations, and cultural beliefs on menstrual symptom reporting: a test of the menstrual reactivity hypothesis. J. Anxiety Disord. 14(6), 615–633 (2000).

58.

Goldstein-Ferber

Granot

: The association between somatization and perceived ability: roles in dysmenorrhea among Israeli Arab adolescents. Psychosom. Med. 68(1), 136–142 (2006).

59.

Granot

Yarnitsky

Itskovitz-Eldor

Granovsky

Peer

Zimmer

: Pain perception in women with dysmenorrhea. Obstet. Gynecol. 98(3), 407–411 (2001).

60.

Demonstrates that women with dysmenorrhea appear to have enhanced pain perception as compared with women without dysmenorrhea.

61.

Bajaj

Madsen

Arendt-Nielsen

: A comparison of modality-specific somatosensory changes during menstruation in dysmenorrheic and nondysmenorrheic women. Clin. J. Pain 18(3), 180–190 (2002).

62.

Brinkert

Dimcevski

Arendt-Nielsen

Drewes

Wilder-Smith

: Dysmenorrhoea is associated with hypersensitivity in the sigmoid colon and rectum. Pain 132(Suppl. 1), S46–S51 (2007).

63.

Interesting paper showing that women with dysmenorrhea demonstrate intestinal hypersensitivity even without overt gastrointestinal symptoms or viscero-somatic sensitization, suggesting the possibility of viscerovisceral convergence and cross-sensitization.

64.

Bajaj

Madsen

Arendt-Nielsen

: Endometriosis is associated with central sensitization: a psychophysical controlled study. J. Pain 4(7), 372–380 (2003).

65.

Campbell

Meyer

: Mechanisms of neuropathic pain. Neuron 52(1), 77–92 (2006).

66.

Cason

Samuelsen

Berkley

: Estrous changes in vaginal nociception in a rat model of endometriosis. Horm. Behav. 44(2), 123–131 (2003).

67.

Berkley

Rapkin

Papka

: The pains of endometriosis. Science 308(5728), 1587–1589 (2005).

68.

Thought-provoking paper summarizing some recent findings on mechanisms underlying endometriosis-associated pain and their sensitivity to estrogen.

69.

Holmlund

: The experience of dysmenorrhea and its relationship to personality variables. Acta Psychiatr. Scand. 82(2), 182–187 (1990).

70.

Gomibuchi

Taketani

Doi

: Is personality involved in the expression of dysmenorrhea in patients with endometriosis? Am. J. Obstet. Gynecol. 169(3), 723–725 (1993).

71.

Whitehead

Crowell

Heller

Robinson

Schuster

Horn

: Modeling and reinforcement of the sick role during childhood predicts adult illness behavior. Psychosom. Med. 56(6), 541–550 (1994).

72.

Walsh

LeBlanc

McGrath

: Menstrual pain intensity, coping, and disability: the role of pain catastrophizing. Pain Med. 4(4), 352–361 (2003).

73.

Proctor

Murphy

Pattison

Suckling

Farquhar

: Behavioural interventions for primary and secondary dysmenorrhoea. Cochrane Database Syst. Rev. Issue 3. Art. No.: CD002248. DOI: 10.1002/14651858.CD002248.pub3 (2007).

74.

Tamburro

Canis

Albuisson

Dechelotte

Darcha

Mage

: Expression of transforming growth factor β1 in nerve fibers is related to dysmenorrhea and laparoscopic appearance of endometriotic implants. Fertil. Steril. 80(5), 1131–1136 (2003).

75.

Uceyler

Sommer

: Cytokine regulation in animal models of neuropathic pain and in human diseases. Neurosci. Lett. 437(3), 194–198 (2008).

76.

Zhang

Wen

Deng

Lin

: Decreased levels of peritoneal interleukin-1 receptor antagonist in patients with endometriosis and disease-related dysmenorrhea. Fertil. Steril. 88(3), 594–599 (2007).

77.

Yang

: Peritoneal cellular immunity and endometriosis. Am. J. Reprod. Immunol. 38(6), 400–412 (1997).

78.

Vercellini

Trespidi

De Giorgi

Cortesi

Parazzini

Crosignani

: Endometriosis and pelvic pain: relation to disease stage and localization. Fertil. Steril. 65(2), 299–304 (1996).

79.

Vercellini

Fedele

Aimi

Pietropaolo

Consonni

Crosignani

Dysmenorrhea: Risk Factors in Women with Endometriosis

Abstract

Keywords

Factors associated with the severity of dysmernorrhea in women with endometriosis

Methodological issues

Other possible factors

Conclusion

Future perspective

Footnotes

References