Endometriosis historically has been defined as the presence of endometrial glands and stroma outside of the uterine cavity, usually within the pelvic cavity. 1 A large proportion of women with this condition may be asymptomatic, but symptoms can be substantial and may include heavy menstrual bleeding, painful menstrual cramps, dyspareunia, chronic and severe pelvic pain throughout the menstrual cycle, and perhaps infertility. Because physical, ultrasound, and magnetic resonance imaging examinations have relatively low sensitivity in ascertaining the presence of ectopic endometriotic tissue, diagnosis typically is made by direct visualization of the pelvis during a laparoscopic surgical examination, with confirmation through microscopic examination of excised tissue. In the United States, the annual incidence of surgically diagnosed disease has been estimated at 1.6 per 1,000 women ages 15–49 years. 2
Studies of factors that may predispose to endometriosis primarily have been clinic based, and widely differing subject source populations and inclusion criteria have been used. Several case-control studies have selected as cases patients at infertility clinics who were found upon laparoscopy to have endometriotic lesions (with or without another basis for their infertility), comparing them with control groups of patients without such lesions at infertility clinics, women undergoing tubal ligation, or women delivering a liveborn child. Other studies have used as cases women found to have endometriosis after presenting to gynecology clinics or hospitals for evaluation of pelvic pain or infertility, in comparison with patients at the same clinic or hospital with other diagnoses, or friends of cases. Additionally, in a few studies women with endometriosis diagnosed at the time of tubal ligation have been compared with women undergoing the same procedure who did not have visible evidence of endometriosis.
The disparate results obtained in these studies may be due in part to differing case definitions and subject-selection strategies. The purpose of this paper is to propose a standard disease definition for epidemiologic studies of endometriosis and suggest subject-selection strategies to increase the validity of clinic- or population-based studies.
The most widely accepted hypothesis for the etiology of endometriosis is that the condition results from the reflux of menstrual tissue containing endometrial stroma and glands from the uterus back through the fallopian tubes, followed by the adherence of this tissue to the mesothelium lining the peritoneum and pelvic organs. 3,4 Recent research indicates that menstrual reflux is a common occurrence, and small endometriotic implants occur frequently in asymptomatic women of reproductive age. 5,6 Consequently, there is a growing appreciation that some of what has been labeled pathologic is, instead, a temporary phase of a normal, ongoing process of cytolysis of recently implanted endometrial fragments. 7,8 Failure of local immunologic, hormonal, and other mechanisms to clear these fragments allows the lesions to vascularize, grow, and infiltrate, resulting in the pathologic condition defined in 1991 by a consensus panel of European gynecologists as “the presence of ectopic endometrium, in association with evidence of cellular activity in the lesions and of progression, such as the formation of adhesions, or by its interference with normal physiological processes.”7,9 (p. 432), 10 This perspective has led clinical researchers to suggest that the focus in etiologic studies now should be on what causes endometriosis to progress in some women to what has been termed “endometriotic disease.”11
The challenge for epidemiologists is to operationalize this disease definition. The presence of ectopic endometrium is determined by visual inspection of the pelvic cavity during laparoscopy. Increased levels of relevant biological markers in tissue, peritoneal fluid, or serum could potentially provide evidence of cellular activity in the ectopic implants; however, none of the numerous possible markers that have been investigated has yet shown sufficient sensitivity and specificity to be useful for diagnosis. In addition to pelvic adhesions, progressive disease has been proposed to include two types of lesions that may have differing etiologies: deeply infiltrating endometriotic lesions and cystic ovarian implants known as endometriomas. 7,12 The final criterion for endometriotic disease, interference with normal physiologic processes, may be the most challenging to operationalize. Positive correlations have been found fairly consistently between progressive disease and interference with menstrual, sexual, and reproductive functioning. Specifically, pelvic adhesions, endometriomas, and deep endometriotic lesions are all strongly associated with dysmenorrhea and dyspareunia, and endometrioma-related adhesions have been associated with infertility. 11,13–17 These symptoms of altered function, in the presence of progressive endometriosis, are highly predictive of endometriotic disease. Nevertheless, the underlying basis for similar symptoms in the presence of less severe endometriosis is less clear.
We propose that investigation of the pathologic condition termed endometriotic disease become the standard in future etiologic studies of endometriosis. Accordingly, we propose to define “definite” endometriotic disease as laparoscopically visible ovarian endometriomas of any size, endometriotic pelvic implants deeper than 5 mm, or any visible ectopic endometrial implants in the presence of ovarian or pelvic adhesions without another explanation (Table 1). We consider endometriosis that does not meet the above criteria, if accompanied by moderate to severe symptoms indicating interference with physiologic processes (infertility, dyspareunia, dysmenorrhea, or chronic pelvic pain), to be “possible” endometriotic disease. Infertility is defined here as unsuccessful attempted conception for 12 or more months in the absence of male, tubal, ovarian, or other basis for the infertility. Other symptoms are defined according to Brosens’ criteria, as noted on Table 1. 18 In our proposed definition, women would not be considered to have endometriotic disease if they have only ectopic endometrial implants without evidence of tissue damage or symptoms.
The presence of endometriotic disease fulfilling our proposed criteria does not correspond to the staging system developed by the American Fertility Society (now the American Society for Reproductive Medicine). This system classifies endometriosis using cutoff points of a global severity score derived from a weighted point system. Most women with endometriomas are classified as stage III or IV using this system, especially in the presence of the usually accompanying adhesions, but women with deeply infiltrating endometriosis often are classified as stage I or II. 11 Additionally, endometriosis stage may not correspond well with interference with physiologic processes. The staging system was designed to categorize women by probability of pregnancy, yet most research has found that stage is not correlated with infertility nor with the presence or extent of dysmenorrhea, dyspareunia, or chronic pelvic pain. 15,19–23
Use of our proposed endometriotic disease definition for potential study subjects requires the availability of operative notes and pathology reports from the diagnostic laparoscopic surgery. Routine operative notes usually contain documentation of the visualization of ectopic endometriotic implants, as well as the presence and extent of endometriomas and pelvic adhesions. Our recent experience indicates that lesion depth and morphology are less likely to be noted, however, either in operative notes or in pathology reports. Interference with physiologic processes may be inferred from chart review or determined by initial screening of potentially eligible subjects with ectopic endometrial tissue. Collaboration with clinicians may improve the documentation of surgical and histologic findings and symptom reporting and thus increase the accuracy of disease category assignment. If current theories regarding physiologic and pathologic endometriosis are correct, our proposed disease definition should increase the likelihood of observing true etiologic associations. Further separation of definite endometriotic disease into subcategories by possible etiologic pathways (endometriomas vs deep implants) and subcategorizing possible disease by the nature of symptoms (pain vs infertility) also may be beneficial.
Because of the relatively low incidence of endometriosis, the case-control design has been used most often in etiologic studies, and the balance of this paper discusses subject selection in this context. Case-selection issues discussed below, however, are equally applicable to case ascertainment in cohort studies, and control-selection issues apply to the identification of the cohort denominator population as well.
No registries exist to facilitate population-based case ascertainment of women with endometriosis. Thus, most etiologic studies of this condition have been clinic based, using women who present to various types of clinics or hospitals with menstrual pain, pelvic pain, or infertility and who are found upon surgical examination to have ectopic endometrial tissue. In such studies, determination of the population from which the cases arose and subsequent representative sampling from it for the control group are important validity issues. 24–27 Women who come to the study clinic or hospital for another reason often are used as controls in these studies and may offer the advantages of convenience, economy, and comparability with cases in terms of recall. Nevertheless, they constitute a nonrandom subset of the underlying population and as such may introduce bias. For the investigation of the impact of some potential risk factors, such as genetic characteristics, the degree of bias is likely to be small or absent. For exposures related to behaviors or to hormonal profiles, however, the results may be affected substantially by the control group selected. In a study of women diagnosed with endometriosis at a reproductive and endometriosis specialty clinic in New York, Darrow et al28 used two different comparison groups: other clinic patients (many of whom were infertile or who had other chronic gynecologic problems) and friends of cases. These investigators found that risks associated with menstrual cycle characteristics were consistently lower when clinic-based controls were used than when friend controls were used. For example, the odds ratio (OR) associated with a history of menstruating more than 20% of days was 1.5 using clinic-based controls vs 2.1 using friends. For heavy menstrual flow, the respective ORs were 1.3 and 1.8, and for severe cramps they were 1.5 and 2.0. Because a large proportion of the clinic-based controls were infertile or had other chronic gynecologic conditions that might be associated with abnormal hormonal profiles and menstrual symptoms, their use likely resulted in spuriously low effect estimates for these exposures. Conversely, in this study substantial differences were seen in the usual birth control method between cases and clinic-based controls but not friend controls (OR for usual use of barrier contraception = 2.5 using clinic-based controls vs OR = 0.9 using friend controls). 29 The clinic-based controls (who may have been more likely than the general population to have had pelvic infections) may have had atypically low barrier contraceptive use, resulting in a spuriously elevated effect estimate. Friend controls, although convenient and inexpensive to ascertain, also are a nonrandom sample of the underlying population and may present problems as well, ranging from a loss of precision owing to overmatching to bias in effect estimates if the exposures of interest are related to gregariousness or to the basis for the friendship. 27,30,31
In one type of case-initiated study, women with endometriosis diagnosed in the process of an infertility evaluation constitute the entire case group. These studies often have used infertility clinic patients who do not have endometriosis as controls, presumably because these women are likely to have had laparoscopic surgery and therefore are known to be lesion-free. As with clinic-based controls in a reproductive specialty clinic, infertility clinic patients with ovulatory infertility are likely to have abnormal hormonal profiles that may bias the results of investigations of hormone-related factors such as menstrual characteristics, exercise, smoking, and body habitus, and all infertility clinic patients are atypical of the underlying population (which includes both fertile and infertile women) in terms of reproductive and contraceptive history. Evidence of this problem can be seen in a study done in a Boston infertility clinic using infertile women with endometriosis as cases and two types of controls: infertile women without endometriosis and women undergoing tubal ligation. 32 In this study, effect estimates associated with menstrual cycle characteristics were markedly lower in analyses using infertile controls compared with those using tubal ligation controls. For example, the OR associated with a history of irregular menstrual cycles was 1.0 using infertile controls, yet it was 4.7 using tubal ligation controls. For a history of menstrual pain, the corresponding ORs were 5.2 and 10.3. The validity of the use of women with documented fertility as controls, as in the above-cited study, depends on the exposures of interest. Menstrual cycle characteristics of fertile women are likely to be typical of the underlying population, but neither women undergoing tubal ligation nor those delivering a child are representative of the entire population of fertile women in terms of their recent reproductive and contraceptive histories.
Several strategies may decrease the likelihood of bias resulting from control group choices in clinic-based studies, and the importance of these choices and the interpretation of the results that stem from them depend on the exposures investigated. First, subjects can be restricted according to characteristics known to be associated with the likelihood of obtaining care at the study clinic (health plan or insurance coverage, local area of residence) to increase the likelihood that cases and controls arise from the same population. Second, controls who present to the study clinic for routine gynecologic care can be used in investigations of a broad range of exposures, as these women may be more representative of the underlying population than are women with infertility or other gynecologic problems. Third, although exclusive use of infertility clinic cases may be infeasible, given that only a small proportion of women diagnosed with infertility-related endometriosis are likely to meet our disease criteria, if infertility clinic patients with endometriotic disease are used in investigations of hormone-related factors (menstrual cycle characteristics, exercise, or diet), infertility clinic controls can be restricted to women who are likely to have typical hormonal profiles, such as women with tubal infertility or partners of infertile men.
Incomplete case ascertainment is likely in the study of diseases such as endometriosis that have expensive and invasive diagnostic procedures and may have low morbidity. Although underascertainment may occur in clinic-based studies of endometriosis, it is a larger potential problem in population-based studies, in which an important validity issue is the ascertainment of all or a representative sample of cases in the defined population. Only the subset of women who seek or allow surgical evaluation of their endometriosis-related symptoms will be available for study, and the exposure status of women with undiagnosed endometriotic disease will be unknown. If the likelihood of diagnosis is related in some way to the exposure of interest, bias will result. For instance, if women with endometriotic disease and severe symptoms present for medical evaluation more often than diseased women with moderate symptoms, and these symptoms are associated with a particular exposure such as high estrogen levels, then analyses of estrogen-related factors would yield spuriously elevated risks.
In some population-based studies, the defined population is all women who present to a specific hospital for tubal ligation; cases are women who are discovered to have endometriotic lesions, whereas controls are a random sample of the population in whom no visual evidence of lesions is noted. Two strengths of this type of study are the elimination of the possibility of case underascertainment (as all members of the population undergo surgical examination) and the ability to explore the entire spectrum of endometriosis. It may be difficult and costly, however, to find an adequate sample of women in this population meeting our disease definition. On average, approximately 5% of women undergoing tubal ligation have been found to have endometriosis, only 10% of which is classified as moderate or severe, so as many as 40,000 women undergoing tubal ligation may be needed to find 200 with definite endometriotic disease. 6,33–39
Selection of controls who are representative of the population in terms of exposures of interest is also critical in population-based studies. 30,40 Population-based controls may be obtained from available rosters (health maintenance organizations, the military, driver’s licenses, or voting lists) or through household survey techniques, including random digit dialing. 27 Use of a population-based control group is not without potential problems: obtaining subjects often requires an expensive and logistically difficult enumeration of a substantial fraction of the underlying population, recall bias may result from querying ill cases and well controls, and well controls may have little motivation to participate. 30 Although the first issue may be an insurmountable problem in some situations, the latter two issues may not have as much impact on study results as is sometimes assumed. Recent studies have found little difference between cases and population-based controls, or between well and ill controls, in the accuracy of self-reported exposures. 41–44 The effect of a low level of participation on representativeness was evaluated by Olson et al, 40 who compared census information obtained from random digit dialing survey respondents (65% of those approached) with that obtained from the entire census area. They observed that the respondents were similar to the entire population on demographic and residential characteristics and most health measures. 40
A final issue in population-based studies of endometriosis is the potential presence of undiagnosed disease (both symptomatic and asymptomatic) in control subjects. No data exist on the prevalence of undiagnosed symptomatic endometriotic disease, but estimation is possible. In one general survey of nearly 18,000 U.S. women ages 18–50 years in 1994, 9% reported undiagnosed chronic pelvic pain, and investigators in another study found that 18% of women with chronic pelvic pain had stage III or stage IV endometriosis. 45,46 It is likely, therefore, that less than 2% of a population-based control group would have undiagnosed endometriotic disease with this symptom. The prevalence of asymptomatic endometriotic disease has been directly assessed. In a recent large study of 3,384 asymptomatic women who had a tubal ligation, Sangi-Haghpeykar and Poindexter 35 found that 16 (0.5%) had ovarian endometriosis and 28 (0.8%) had deep endometriosis. Thus, only a small percentage of a population-based control group is likely to have undiagnosed symptomatic or asymptomatic endometriotic disease, and so the effect estimates of a population-based study would be biased to only a very small degree.
The primary strategy to decrease the likelihood of bias in population-based studies of endometriosis is reducing case underascertainment. Complete ascertainment can be achieved through laparoscopic examination of the entire population, a protocol that is feasible only in studies of women obtaining tubal ligation. In more general population-based studies of endometriotic disease, utilization can be made of populations in which there is little or no financial disincentive for diagnosis, such as military or health maintenance organizations in the United States, or general populations in countries with universal health care access. Case underascertainment due to nonfinancial disincentives for diagnosis is more difficult to remedy. Our proposed endometriotic disease definition may help decrease the likelihood of incomplete ascertainment, as women meeting our disease criteria are likely to be symptomatic and may therefore seek medical care. In circumstances in which the inclination to seek medical care is an important aspect of disease diagnosis, Savitz and Pearce 47 advise that restriction of controls to that segment of the underlying population with a similar inclination may help prevent bias. These authors suggest one of two strategies to achieve this: (1) restrict controls to those women who have a demonstrated tendency to seek medical care for other conditions of low morbidity or (2) select controls more broadly and then restrict them on the basis of how they would respond if confronted with the case’s problem. Either of these strategies can be used in studies of endometriotic disease. Controls can be restricted to women who have made recent medical visits for minor unrelated illnesses or for routine gynecologic care, or restriction can be made in interview-based studies to women who state they would seek medical care and allow laparoscopic evaluation if they developed symptoms of endometriosis.
Because of the variable nature of the manifestations of endometriosis and the difficulties of diagnosis, neither the ideal design nor the ideal case and control groups are likely to be achieved in etiologic studies of this disease. Case-definition and subject-selection decisions should be made, therefore, with consideration both of the available study populations and of the specific exposures of interest. Careful attention to these decisions may lead to more consistency in results of future etiologic investigations of endometriosis, and specification of them should help researchers interpret etiologic studies in relation to each other and estimate the direction of the bias that may have resulted from a study design’s deviation from the ideal.
We thank Sascha Dublin for her invaluable comments on an earlier draft of this manuscript.
1. Mishell DR, Stenchever MA, Droegemueller W, Herbst AL. Comprehensive Gynecology. 3rd ed. St. Louis: Mosby-Year Book, 1997.
2. Houston DE, Noller KL, Melton LJ, Selwyn BJ, Hardy RJ. Incidence of pelvic endometriosis in Rochester, Minnesota, 1970–1979. Am J Epidemiol 1987; 125: 959–969.
3. Sampson JA. The development of the implantation theory for the origin of peritoneal endometriosis. Am J Obstet Gynecol 1940; 40: 549–557.
4. Nisolle M, Donnez J. Peritoneal endometriosis, ovarian endometriosis, and adenomyotic nodules of the rectovaginal septum are three different entities. Fertil Steril 1997; 68: 585–596.
5. Halme J, Hammond MG, Hulka JF, Raj SG, Talbert LM. Retrograde menstruation in healthy women and in patients with endometriosis. Obstet Gynecol 1984; 64: 151–154.
6. Liu DT, Hitchcock A. Endometriosis: its association with retrograde menstruation, dysmenorrhoea and tubal pathology. Br J Obstet Gynaecol 1986; 93: 859–862.
7. Vercellini P, Bocciolone L, Crosignani PG. Is mild endometriosis always a disease? Hum Reprod 1992; 7: 627–629.
8. Thomas EJ. The relevance of asymptomatic endometriosis. Hum Reprod 1996; 11 (suppl 3): 103–109.
9. Audebert A, Backstrom T, Barlow DH, Benagiano G, Brosens I, Buhler K, Donnez J, Evers JLH, Pellicer A, Mettler L, Rönneberg L, Smith SK, Thomas EJ. Endometriosis 1991: a discussion document. Hum Reprod 1992; 7: 432–435.
10. Wardle PG, Hull MG. Is endometriosis a disease? Baillieres Clin Obstet Gynaecol 1993; 7: 673–685.
11. Koninckx PR, Oosterlynck D, D’Hooghe T, Muelemen C. Deeply infiltrating endometriosis is a disease whereas mild endometriosis could be considered a non-disease. Ann NY Acad Sci 1994; 734: 333–341.
12. Evers JLH. Endometriosis does not exist; all women have endometriosis. Hum Reprod 1994; 9: 2206–2209.
13. Cornillie FJ, Oosterlynck D, Lauweryns JM. Deeply infiltrating pelvic endometriosis: histology and clinical significance. Fertil Steril 1990; 53: 978–983.
14. Koninckx PR, Lesaffre E, Meuleman C, Cornillie FJ, Demeyere S. Suggestive evidence that pelvic endometriosis is a progressive disease, whereas deeply infiltrating endometriosis is associated with pelvic pain. Fertil Steril 1991; 55: 759–765.
15. Fedele L, Bianchi S, Bocciolone L, Di Nola G, Parazzini F. Pain symptoms associated with endometriosis. Obstet Gynecol 1992; 79: 767–769.
16. Zreik TG, Olive DL. Pathophysiology: the biologic principles of disease. Obstet Gynecol Clin North Am 1997; 24: 259–268.
17. Porpora MG, Koninckx PR, Piazze J, Natili M, Colagrande S, Cosmi EV. Correlation between endometriosis and pelvic pain. J Am Assoc Gynecol Laparosc 1999; 6: 429–434.
18. Brosens IA. Endometriosis: a disease because it is characterized by bleeding. Am J Obstet Gynecol 1997; 176: 263–267.
19. Guzick DS, Silliman NP, Adamson GD, Buttram VC, Canis M, Malinak LR, Schenken RS. Prediction of pregnancy in infertile women based on the American Society for Reproductive Medicine’s revised classification of endometriosis. Fertil Steril 1997; 67: 822–829.
20. Buttram VC. Conservative surgery for endometriosis in the infertile female: a study of 206 patients with implications for both medical and surgical therapy. Fertil Steril 1979; 31: 117–121.
21. Fedele L, Arcaini L, Parazzini F, Candiani GB, Bianchi S. Stage and localization of pelvic endometriosis and pain. Fertil Steril 1990; 53: 155–158.
22. Marana R, Muzii L, Caruana P, Sell-Acqua S, Mancuso S. Evaluation of the correlation between endometriosis extent, age of the patients and associated symptomatology. Acta Eur Fertil 1991; 22: 209–212.
23. Stovall DW, Bowser LM, Archer DF, Guzick DS. Endometriosis-associated pelvic pain: evidence for an association between the age of disease and a history of chronic pelvic pain. Fertil Steril 1997; 68: 13–18.
24. Miettinen OS. The “case-control” study: valid selection of subjects. J Chron Dis 1985; 38: 543–548.
25. Greenland S. Control-initiated case-control studies. Int J Epidemiol 1985; 14: 130–134.
26. Wacholder S, McLaughlin JK, Silverman DT, Mandel JS. Selection of controls in case-control studies. I. Principles. Am J Epidemiol 1992; 135: 1019–1028.
27. Wacholder S, Silverman DT, McLaughlin JK, Mandel JS. Selection of controls in case-control studies. II. Types of controls. Am J Epidemiol 1992; 135: 1029–1041.
28. Darrow SL, Vena JE, Batt RE, Zielezny MA, Michalek AM, Selman S. Menstrual cycle characteristics and the risk of endometriosis. Epidemiology 1993; 4: 135–142.
29. Darrow SL, Selman S, Batt RE, Zielezny MA, Vena JE. Sexual activity, contraception, and reproductive factors in predicting endometriosis. Am J Epidemiol 1994; 140: 500–509.
30. Robins J, Pike M. The validity of case-control studies with nonrandom selection of controls. Epidemiology 1990; 1: 273–284.
31. Flanders WD, Austin H. Possibility of selection bias in matched case-control studies using friend controls. Am J Epidemiol 1986; 124: 150–153.
32. Signorello LB, Harlow BL, Cramer DW, Spiefelman D, Hill JA. Epidemiologic determinants of endometriosis: a hospital-based case-control study. Ann Epidemiol 1997; 7: 267–274.
33. Mahmood TA, Templeton A. Prevalence and genesis of endometriosis. Hum Reprod 1991; 6: 544–549.
34. Kirshon B, Poindexter AN, Fast J. Endometriosis in multiparous women. J Reprod Med 1989; 34: 215–217.
35. Sangi-Haghpeykar H, Poindexter AN. Epidemiology of endometriosis among parous women. Obstet Gynecol 1995; 85: 983–992.
36. Moen MH. Endometriosis in women at interval sterilization. Acta Obstet Gynecol Scand 1987; 66: 451–454.
37. Strathy JH, Molgaard CA, Coulam CB, Melton LJ. Endometriosis and infertility: a laparoscopic study of endometriosis among fertile and infertile women. Fertil Steril 1982; 38: 667–672.
38. Verkauf BS. The incidence, symptoms and signs of endometriosis in fertile and infertile women. J Fla Med Assoc 1987; 74: 671–674.
39. Trimbos JB, Trimbos-Kemper GC, van der Does CD, van Hall EV. Findings in 200 consecutive asymptomatic women having a laparoscopic sterilization. Arch Gynecol Obstet 1990; 247: 121–124.
40. Olson SH, Kelsey JL, Pearson TA, Levin B. Evaluation of random digit dialing as a method of control selection in case-control studies. Am J Epidemiol 1992; 135: 210–222.
41. Kerber RA, Slattery ML. Comparison of self-reported and database-linked family history of cancer data in a case-control study. Am J Epidemiol 1997; 146: 244–248.
42. Berwick M, Chen YT. Reliability of reported sunburn history in a case-control study of cutaneous malignant melanoma. Am J Epidemiol 1995; 141: 1033–1037.
43. Friedenreich CM, Howe GR, Miller AB. The effect of recall bias on the association of calorie-providing nutrients and breast cancer. Epidemiology 1991; 2: 424–429.
44. Khoury MJ, James LM, Erickson JD. On the use of affected controls to address recall bias in case-control studies of birth defects. Teratology 1994; 49: 273–281.
45. Mathias SD, Kuppermann M, Liberman RF, Lipschutz RC, Steege JF. Chronic pelvic pain: prevalence, health-related quality of life, and economic correlates. Obstet Gynecol 1996; 87: 321–327.
46. Gruppo Italiano per lo Studio dell’Endometriosi. Prevalence and anatomical distribution of endometriosis in women with selected gynaecological conditions: results from a multicentric Italian study. Hum Reprod 1994; 9: 1158–1162.
47. Savitz DA, Pearce N. Control selection with incomplete case ascertainment. Am J Epidemiol 1988; 6: 1109–1117.