Endometriosis is tissue resembling the lining of the uterus growing where it does not belong — on the ovaries, on the peritoneum lining the pelvis, on the bowel and bladder, occasionally further afield [1] [2]. It responds to the hormonal cycle much as the uterine lining does, and the result is inflammation, pain, adhesions and scarring, and often infertility.
It is common. A UK primary-care study of 5,540 diagnosed women matched to controls put the prevalence of diagnosed endometriosis at 1.5 percent [3]; population estimates that count undiagnosed disease are considerably higher, and Cochrane reviews open with the figure of about 10 percent of women of reproductive age [4] [5].
What sets it apart from most conditions in this collection is not the biology. It is the diagnostic delay. Across six studies in eight countries spanning twenty years, the interval between a woman first having symptoms and anyone naming the disease has run from about seven to nearly twelve years, and it has not measurably improved. That delay is the subject of the figure below, and it is the single most consequential fact about how this disease is actually experienced.
Two things make it worse. The first is that endometriosis is painful in a way that is easy to dismiss: the symptoms are period pain, pelvic pain, pain with sex — all of which can be, and routinely are, normalised. In the Austrian and German study, 74 percent of women had received at least one incorrect diagnosis before the right one [6]. Qualitative work has documented how menstrual concealment and the stigma around discussing periods contribute directly to this [7] [8].
The second is that pain severity does not track how much disease is present. Asymptomatic endometriosis exists, and the association between severe period pain and endometriosis appears to be independent of the macroscopic type of the lesions and of where they sit [9] — while the surgical staging system itself has been questioned precisely because it does not correlate well with symptoms, which creates problems for both diagnosis and choice of treatment [10]. This is not a subtlety; it undermines the intuition that a doctor can look and see how bad it is.
The cost is measurable. In 1,418 women across sixteen centres in ten countries, physical quality of life was significantly reduced compared with women who had similar symptoms but no endometriosis, and each affected woman lost an average of 10.8 hours of work per week, mostly through reduced effectiveness rather than absence [11]. The economic burden has been quantified repeatedly [12] [13] [14], and the psychological burden alongside it [15] [16] [17].
Start here: where the tissue comes from, and why that is not the whole answer
The oldest and still most-cited explanation is retrograde menstruation — menstrual tissue flowing backwards through the fallopian tubes into the pelvic cavity, where it implants and grows. John Sampson proposed it in 1927 [18] [19], and it remains the framework most descriptions use.
It has one large problem, and the study that exposed it is worth stating plainly. In 1984, blood was found in the peritoneal fluid of 90 percent of women with patent fallopian tubes examined by laparoscopy around the time of menstruation — and 90 percent of women with endometriosis, and eight of nine women taking oral contraceptives. When the tubes were blocked, only 15 percent had blood in the pelvis [20]. Retrograde menstruation is not a disease process. It is what menstruation normally does.
So the question is not why tissue arrives in the pelvis but why it implants and survives in a minority of women. Current accounts invoke local oestrogen production, immune escape and altered peritoneal environment [21] [22]. Two findings anchor the biology.
The lesion makes its own oestrogen. Aromatase, the enzyme converting androgens to oestrogen, is expressed in endometriotic tissue but not in normal endometrium [23] [24], its promoter is activated in endometriosis [25], and prostaglandin E₂ — an inflammatory mediator — stimulates that expression [26] [27]. That is a feed-forward loop: inflammation drives local oestrogen synthesis, oestrogen drives growth and more inflammation. It also explains why aromatase became a therapeutic target [28] [29].
The tissue resists progesterone. Gene-expression analysis of endometrium from women with endometriosis showed a progesterone-resistant signature [30] [31]; progesterone receptor isoform B is absent from endometriotic lesions [32] and its promoter is hypermethylated [33]; and the failure to metabolise oestradiol has been linked to the same defect [34] [35]. This matters clinically, because progestins are a mainstay of treatment and the tissue is partly deaf to them.
Centerpiece: the diagnostic delay
Endometriosis has no clean quantitative mechanism to simulate — there is no threshold, no dose-response, no reproducible curve linking a measurable quantity to the disease. Forcing one would be dishonest. What it does have is a well-documented, repeatedly measured number that matters enormously, so that is what the figure shows. Nothing here is fitted or invented; every value is read from a source.
The six estimates. A 1996 postal survey of 218 women with surgically confirmed disease found a mean delay of 11.73 ± 9.05 years in the USA and 7.96 ± 7.92 years in the UK [36]. A Norwegian questionnaire study of 261 women found a mean of 6.7 ± 6.2 years, median 5.0 [37]. The ten-country multicentre study of 1,418 women found 6.7 years [11]. An Austrian and German study of 171 women found a median of 10.4 years (SD 7.9) [6]. A Dutch study of 139 women found a median of 89 months — 7.4 years [38].
Four things the arithmetic shows.
The distribution is heavily right-skewed. Every study reporting a spread reports one comparable to its own centre — the standard deviation exceeds three-quarters of the mean in all four cases, and is essentially equal to it for the UK. Where both statistics are given, Norway's mean of 6.7 exceeds its median of 5.0. The average understates the tail badly: a mean of seven years is compatible with a substantial minority waiting twenty.
Independent designs converge. The Norwegian questionnaire study in 2003 and the sixteen-centre prospective study across ten countries in 2011 arrive at the same mean to one decimal place — 6.7 years — sharing no data, no population and no method.
Twenty years produced no clear improvement. A line through the six central estimates falls by less than two years across the whole span and still sits above seven years at the end of it. The Norwegian authors said the same thing directly of their own data: mean delay "had not changed significantly with time" [37].
And the Dutch decomposition does not add up. That study reports its 89-month median as 7 months of patient delay, 35 months of general-practitioner delay and 5 months of gynaecologist delay [38]. Those sum to 47 months — a little over half the total — leaving 42 months outside the three named segments. The figure shows the gap rather than hiding it, because it is a real feature of the published numbers and this review is not going to explain it away. What the accounted-for portion does show is unambiguous: 74 percent of it sits with the first-line clinician. And that is the quantitative echo of a conclusion the ten-country study reached independently, finding the delay "principally in primary care" and 8.3 years in predominantly state-funded centres against 5.5 years elsewhere [11].
The teaching point is that this is a recognition problem, not a technology problem. The delay does not sit where a new test would fix it. It sits in the years before anyone considers the diagnosis — and the same study that measured the delay also found it longer in women with more pelvic symptoms [11], which is the opposite of what a rational triage system would produce.
Three honest limits. Most of these studies recruited through self-help groups, referral centres or specialist clinics, which selects for women who eventually got diagnosed and probably for those with worse experiences — the true population delay is not what these numbers measure. All six report a delay to surgical diagnosis, so as imaging displaces laparoscopy the quantity itself is changing definition. And the series is not a time trend: six different countries with six different health systems and six different recruitment methods cannot be read as one population followed forward, which is why the figure draws no trend line through them.
Pillar 1: measurement and diagnosis
It starts as clinical suspicion
There is no blood test — Cochrane's assessment is that no blood biomarker can currently replace diagnostic surgery [5]. The diagnosis begins with a pattern of symptoms — cyclical or chronic pelvic pain, painful periods, pain with intercourse, subfertility — and the pattern is genuinely informative. In the UK primary-care case-control study, women later diagnosed with endometriosis had substantially raised odds of dysmenorrhoea (OR 8.1), subfertility (OR 8.2), ovarian cysts (OR 7.3), dyspareunia or postcoital bleeding (OR 6.8), abdominopelvic pain (OR 5.2) and menorrhagia (OR 4.0), and 73 percent had abdominopelvic pain, dysmenorrhoea or menorrhagia against 20 percent of controls [3].
Those are large effects. The reason they do not produce a prompt diagnosis is that each symptom is individually common, and the same study found raised odds of being diagnosed with irritable bowel syndrome instead (OR 1.6) [3] — the misdiagnosis pathway made visible in the data.
Imaging has changed what surgery is for
Transvaginal ultrasound and MRI now detect ovarian endometriomas and deep infiltrating lesions well enough to make the diagnosis in many women without an operation. MRI has been evaluated for diagnosing deep pelvic endometriosis and predicting its extent [39]; transvaginal sonography has been assessed for deep pelvic disease [40] and compared directly with clinical examination and MRI [41] [42]; and the IDEA consensus standardised how the pelvis should be examined and described sonographically, so that findings mean the same thing between centres [43]. A Cochrane review has assessed imaging both as a replacement for diagnostic surgery and as a triage test [4].
Laparoscopy with histology remains the reference standard — every diagnostic-accuracy study above is measured against it, and both Cochrane reviews state plainly that no non-invasive test yet diagnoses endometriosis accurately in clinical practice [4] [5]. But the role has shifted: imaging increasingly answers the question, and surgery is reserved for cases imaging cannot resolve or where treatment is being done anyway.
Staging, and the caveat that matters most
The revised American Society for Reproductive Medicine classification assigns a stage from I to IV based on lesion type, size, location and adhesions [44]. It is universally used and it does not measure what patients care about: the system has been challenged on exactly the grounds that it does not correlate well with underlying symptoms [10], the relationship between stage, lesion location and pain has been examined directly [45] [9], and the 1996 delay survey found that stage of disease did not affect the time to diagnosis either [36].
Part of the explanation may be that pain in endometriosis is not purely a local phenomenon: a controlled psychophysical study found evidence of central sensitisation in affected women [46]. If the nervous system's pain processing has itself changed, counting lesions will not predict the experience — and that reframes endometriosis pain as a chronic pain condition requiring chronic pain management, not merely a plumbing problem awaiting an operation.
Adenomyosis overlaps
Adenomyosis — endometrial tissue within the muscular wall of the uterus itself — frequently coexists with endometriosis, shares features of its molecular abnormality [47] [48] [49] [50], and is now diagnosable by transvaginal ultrasound and MRI [51]. The boundary is not always crisp: bladder lesions have been argued to sit on it [52]. The two together carry obstetric and postpartum consequences [53], and a woman assessed for one should be assessed for the other.
Pillar 2: treatment
There is no cure. Treatment is symptom- and goal-directed, and the goals — pain control, fertility, avoiding recurrence — pull in different directions. The ESHRE guidelines are the reference [54] [55] [56].
Suppressing the cycle
Because the lesions are oestrogen-responsive, most medical treatment works by flattening the hormonal cycle. NSAIDs address the prostaglandin component that also drives primary dysmenorrhoea [57].
Combined oral contraceptives and progestins are first-line, generally taken continuously to eliminate cyclical stimulation. A randomised trial found a low-dose oral contraceptive comparable to a GnRH agonist for pelvic pain [58]; Cochrane has reviewed progestagens and anti-progestagens for endometriosis-associated pain [59]; and a levonorgestrel-releasing intrauterine system was randomised against a depot GnRH analogue in 82 women, with chronic pelvic pain falling significantly under both and no difference between them over six months [60]. Progestin implants have been compared with depot medroxyprogesterone acetate on pain scores [61]. The caveat is the one from the biology above: endometriotic tissue is partly progesterone-resistant [30] [35], which is a mechanistic reason why progestins fail in some women rather than a failure of adherence.
GnRH agonists induce a temporary menopause. They work, and the cost is hypo-oestrogenic side effects including bone loss — which is why add-back therapy exists. In a 12-month randomised trial of 201 women, leuprolide with placebo add-back produced a 6.3 ± 2.3 percent loss of bone density at 52 weeks, while the add-back groups were protected, and all four groups improved on pain by week 8 [62]. Add-back does not compromise the analgesic benefit; omitting it compromises the skeleton. Depot leuprolide has also been trialled in women with chronic pelvic pain and clinically suspected endometriosis [63], and GnRH agonists have been studied after conservative surgery [64] [65] and before IVF [66].
Oral GnRH antagonists are the newer option, and they do not require an injection. Two 6-month phase 3 trials randomised 872 and 817 women with surgically diagnosed endometriosis and moderate-to-severe pain to two doses of elagolix or placebo, with co-primary endpoints of clinical response for dysmenorrhoea and for non-menstrual pelvic pain at three months [67]. Because the degree of oestrogen suppression is dose-dependent, the class offers a dial rather than a switch — the same trade-off between efficacy and hypo-oestrogenic harm, made adjustable.
Aromatase inhibitors follow directly from the local-oestrogen finding, and have been combined with a GnRH agonist after surgery [68].
Surgery
Surgical removal of lesions is effective for pain. A randomised, placebo-controlled trial of laparoscopic excision was conducted [69], as was a double-blind randomised trial of laser laparoscopy [70], and excisional versus ablative technique for ovarian endometriomas has been compared [71]. Recurrence after conservative surgery is common, which is why post-operative medical suppression has been studied repeatedly [64] [65].
Surgery for endometriomas carries a specific cost that should be weighed before it is done. A meta-analysis of eight prospective cohorts, 237 patients, found a significant fall in serum anti-Müllerian hormone after ovarian cystectomy — a weighted mean difference of −1.13 ng/mL, and −1.52 ng/mL in women whose pre-operative level was 3.1 or above [72]. Removing an endometrioma removes ovarian reserve with it, which matters most in exactly the women who came in wanting to conceive.
Fertility
Endometriosis is found at high rates in infertile women, including those ovulating normally with normospermic partners [73] [74], and the association is well characterised [75] [76] [77].
The cleanest evidence is a trial of 341 infertile women with minimal or mild endometriosis, randomised during diagnostic laparoscopy either to have the visible lesions resected or ablated, or to have the diagnostic procedure alone. Cumulative pregnancy continuing to 20 weeks was 30.7 percent with treatment against 17.7 percent without (P = 0.006), with fecundity rates of 4.7 versus 2.4 per 100 person-months and a rate ratio of 1.9 (95% CI 1.2–3.1) [78]. The two ways of reporting the result agree: 4.7 divided by 2.4 is 1.96, which is the rate ratio, so the fecundity and cumulative-probability analyses are describing the same effect.
That is a real benefit and a modest one: the difference of 13.0 percentage points works out, on those two figures, at about eight women treated per additional continuing pregnancy. For more advanced disease, or when surgery has not worked, assisted reproduction is the route, and the effect of endometriosis on IVF outcomes has been studied specifically [79] [80] [66].
The part that is not a drug or an operation
Because pain in endometriosis involves central sensitisation [46], persists after adequate surgery, and coexists with depression and anxiety [16] [15], the condition is properly managed as a chronic pain disorder with a multidisciplinary team rather than as a series of operations. Cochrane has even assessed hormonal suppression against dietary therapy and placebo for painful symptoms [81]. The women who do worst are typically those who receive repeated surgery for pain that was never going to be fixed by removing more tissue.
Pillar 3: what is unresolved
A non-invasive diagnostic test. This is the field's central goal and it has not been achieved. Peripheral biomarkers have been systematically reviewed [82] [83], a panel of 28 evaluated at once [84], serum cytokines assessed [85], circulating microRNAs identified by genome-wide analysis [86], and endometrial alterations proposed as markers [87]. CA-125 is elevated in advanced disease [88] but is neither sensitive nor specific enough. The Cochrane verdict on blood biomarkers is that none can currently replace surgery [5], and the same holds for imaging in the general case [4] — though imaging is much closer, and for endometriomas and deep disease it is already there.
Closing the delay. Nothing in the six studies suggests the delay is falling. Given where it sits, the interventions that would work are recognition-side — taking severe period pain seriously in adolescents, not normalising it, and referring on suspicion rather than on proof [11] [6] [7].
Non-hormonal treatment. Every effective medical therapy currently works by suppressing oestrogen, which means contraception is an unavoidable side effect and long-term use has costs. The aromatase and prostaglandin biology [26] [28] and progesterone-resistance work [35] point at targets that would not require shutting down the cycle, but none has yet displaced hormonal suppression.
Better classification. The r-ASRM stage does not predict pain and does not predict fertility well [44] [45]. What a classification should predict — and how to build one that does — remains open [89] [10].
Malignant potential. Endometriosis is benign, but ARID1A mutations have been found in endometriosis-associated ovarian carcinomas [90], and the relationship deserves stating carefully: the association is real, the absolute risk to an individual woman is low, and the finding is more interesting as biology than as a reason for alarm.
Dig deeper in lmmol
The natural pair for this review is polycystic ovary syndrome — the other very common, very under-recognised reproductive condition, and an instructive contrast: PCOS has a clean quantitative mechanism and a treatable number, while endometriosis has neither and is defined instead by a recognition failure. Both share infertility as a presenting complaint [80]. The chronic-pain and mental-health dimension connects to depression, which coexists with pelvic endometriosis and worsens with pain rather than with lesion burden [16] [15], and to migraine as another cyclically patterned pain condition where central sensitisation is central to the mechanism [46]. The bone-density cost of GnRH agonist therapy is the same problem treated in osteoporosis, which is why add-back exists [62]. And inflammatory bowel disease is worth reading alongside this one for the diagnostic-confusion reason: endometriosis is frequently labelled irritable bowel syndrome first [3]. The full collection is at health.