Validity of Self-reported Endometriosis

A Comparison Across Four Cohorts

A.L. Shafrir; L.A. Wise; J.R. Palmer; Z.O. Shuaib; L.M. Katuska; P. Vinayak; M. Kvaskoff; K.L. Terry; S.A. Missmer

Hum Reprod. 2021;36(5):1268-1278.

Abstract and Introduction

Abstract

Study Question: How accurately do women report a diagnosis of endometriosis on self-administered questionnaires?

Summary Answer: Based on the analysis of four international cohorts, women self-report endometriosis fairly accurately with a > 70% confirmation for clinical and surgical records.

What is Known Already: The study of complex diseases requires large, diverse population-based samples, and endometriosis is no exception. Due to the difficulty of obtaining medical records for a condition that may have been diagnosed years earlier and for which there is no standardized documentation, reliance on self-report is necessary. Only a few studies have assessed the validity of self-reported endometriosis compared with medical records, with the observed confirmation ranging from 32% to 89%.

Study Design, Size, Duration: We compared questionnaire-reported endometriosis with medical record notation among participants from the Black Women's Health Study (BWHS; 1995–2013), Etude Epidémiologique auprès de femmes de la Mutuelle Générale de l'Education Nationale (E3N; 1990–2006), Growing Up Today Study (GUTS; 2005–2016), and Nurses' Health Study II (NHSII; 1989–1993 first wave, 1995–2007 second wave).

Participants/Materials, Setting, Methods: Participants who had reported endometriosis on self-administered questionnaires gave permission to procure and review their clinical, surgical, and pathology medical records, yielding records for 827 women: 225 (BWHS), 168 (E3N), 85 (GUTS), 132 (NHSII first wave), and 217 (NHSII second wave). We abstracted diagnosis confirmation as well as American Fertility Society (AFS) or revised American Society of Reproductive Medicine (rASRM) stage and visualized macro-presentation (e.g. superficial peritoneal, deep endometriosis, endometrioma). For each cohort, we calculated clinical reference to endometriosis, and surgical- and pathologic-confirmation proportions.

Main Results and the Role of Chance: Confirmation was high—84% overall when combining clinical, surgical, and pathology records (ranging from 72% for BWHS to 95% for GUTS), suggesting that women accurately report if they are told by a physician that they have endometriosis. Among women with self-reported laparoscopic confirmation of their endometriosis diagnosis, confirmation of medical records was extremely high (97% overall, ranging from 95% for NHSII second wave to 100% for NHSII first wave). Importantly, only 42% of medical records included pathology reports, among which histologic confirmation ranged from 76% (GUTS) to 100% (NHSII first wave). Documentation of visualized endometriosis presentation was often absent, and details recorded were inconsistent. AFS or rASRM stage was documented in 44% of NHSII first wave, 13% of NHSII second wave, and 24% of GUTS surgical records. The presence/absence of deep endometriosis was rarely noted in the medical records.

Limitations, Reasons For Caution: Medical record abstraction was conducted separately by cohort-specific investigators, potentially introducing misclassification due to variation in abstraction protocols and interpretation. Additionally, information on the presence/absence of AFS/rASRM stage, deep endometriosis, and histologic findings were not available for all four cohort studies.

Wider Implications of the Findings: Variation in access to care and differences in disease phenotypes and risk factor distributions among patients with endometriosis necessitates the use of large, diverse population samples to subdivide patients for risk factor, treatment response and discovery of long-term outcomes. Women self-report endometriosis with reasonable accuracy (>70%) and with exceptional accuracy when women are restricted to those who report that their endometriosis had been confirmed by laparoscopic surgery (>94%). Thus, relying on self-reported endometriosis in order to use larger sample sizes of patients with endometriosis appears to be valid, particularly when self-report of laparoscopic confirmation is used as the case definition. However, the paucity of data on histologic findings, AFS/rASRM stage, and endometriosis phenotypic characteristics suggests that a universal requirement for harmonized clinical and surgical data documentation is needed if we hope to obtain the relevant details for subgrouping patients with endometriosis.

Study Funding/Competing Interest(S): This project was supported by Eunice Kennedy Shriver National Institute of Child Health and Development grants HD48544, HD52473, HD57210, and HD94842, National Cancer Institute grants CA50385, R01CA058420, UM1CA164974, and U01CA176726, and National Heart, Lung, and Blood Institute grant U01HL154386. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. AS, SM, and KT were additionally supported by the J. Willard and Alice S. Marriott Foundation. MK was supported by a Marie Curie International Outgoing Fellowship within the 7th European Community Framework Programme (#PIOF-GA-2011-302078) and is grateful to the Philippe Foundation and the Bettencourt-Schueller Foundation for their financial support. Funders had no role in the study design, conduct of the study or data analysis, writing of the report, or decision to submit the article for publication. LA Wise has served as a fibroid consultant for AbbVie, Inc for the last three years and has received in-kind donations (e.g. home pregnancy tests) from Swiss Precision Diagnostics, Sandstone Diagnostics, Kindara.com, and FertilityFriend.com for the PRESTO cohort. SA Missmer serves as an advisory board member for AbbVie and a single working group service for Roche; neither are related to this study. No other authors have a conflict of interest to report. Funders had no role in the study design, conduct of the study or data analysis, writing of the report, or decision to submit the article for publication.

Trial Registration Number: N/A.

Introduction

Endometriosis, in which endometrial-like tissue is found outside of the uterus, affects approximately 10% of reproductive-aged women and can lead to debilitating pelvic pain, diminished quality of life, infertility, and significant health care costs (Missmer et al., 2004; Shafrir et al., 2018; Zondervan et al., 2018; Ghiasi et al., 2020). It is a heterogeneous disease in terms of lesion characteristics, symptom presentation, and response to treatment (Zondervan et al., 2020). As such, studies with large sample sizes that include diverse populations are needed to identify and adequately analyze informative endometriosis subgroups. Grouping all endometriosis participants into one category may bias estimates towards the null if some subgroups but not others are associated with certain risk factors or with predicting treatment responses.

Currently, definitive diagnosis of endometriosis requires surgical visualization, with some arguing for the additional requirement of histologic confirmation of endometrial glands and/or stroma in at least one excised lesion (Agarwal et al., 2019). Thus, many studies rely on medical record abstraction. However, conducting medical record abstraction in large epidemiologic studies can be costly and time-consuming while still yielding inconsistent information (Becker et al., 2014; Vitonis et al., 2014). Consequently, many studies include small samples from single clinical sites that may provide care to biased subsets of the general population of women at risk for endometriosis. Indeed, the prevalence of endometriosis alone may differ by an order of magnitude when comparing specialty clinic populations to general populations (Ghiasi et al., 2020). Additionally, analyses of insurance claims data (e.g. International Classification of Diseases (ICD)-10 codes), which capture a more comprehensive population, are limited to patients who received an ICD-10 code for billing purposes and have been shown to inadequately capture the full extent of endometriosis disease (Whitaker et al., 2019). For example, endometriosis remains coded within 'non-inflammatory disorders of female genital tract,' which is entirely inconsistent with current knowledge, and thus may be overlooked. In addition, there is no coding option to define endometriosis subgroups such as superficial or deep disease (Whitaker et al., 2019).

Currently, overcoming these limitations requires reliance on self-reported endometriosis via self-administered questionnaires. To date, only a few studies have assessed the validity of self-reported endometriosis, finding a wide range of validation proportions—32–89%––across different populations and using different data sources (e.g. medical records and inpatient hospital registries) as the gold standard for confirmation (Treloar et al., 2000; Missmer et al., 2004; Saha et al., 2015, 2017). No study has compared the validity of self-reported endometriosis against medical records across populations that differ with respect to age and race/ethnicity.

Therefore, we compared endometriosis diagnoses reported via self-administered questionnaires to medical record notation across four diverse cohorts. We also assessed the availability of endometriosis phenotypic details, including American Fertility Society (AFS) or revised American Society of Reproductive Medicine (rASRM) stage and visualized macro-presentation of endometriosis lesions (e.g. superficial peritoneal, endometrioma, deep endometriosis).

1 2 3 4 5

Next Section

Table I. Population characteristics and validation study eligibility of the Black Women's Health Study, Etude Epidémiologique aupres de femmes de la Mutuelle Générale d l'Education Nationale, Growing Up Today Study, and Nurses' Health Study II.
	BWHS	E3N	GUTS	NHSII first wave	NHSII second wave
Enrollment years	1995	1989–1991	1996 & 2004	1989	1989
Age range at baseline (yrs)	21–69	39–66	9–17	25–42	25–42
Baseline population (N)	59,001	98,995	15,044	116,429	116,429
Post-high school education (%)	83%	87%	98%	100%	100%
Questionnaire cycles of endometriosis self-report	1995–2013	1990–2005	2005–2016	1989–1993	1989–2007
Total self-reported endometriosis cases (N)	1,605	2,104	491	1,766	9,452
Endometriosis case type for validation¹	Incident	Incident	Incident & prevalent	Incident	Incident & prevalent
Eligible endometriosis cases (N)²	1,605	558	491	1,766	711
Endometriosis cases selected for validation study (N)	1,605	200	380	200	711

BWHS, Black Women's Health Study; E3N, Etude Epidémiologique aupres de femmes de la Mutuelle Générale d l'Education Nationale; GUTS, Growing Up Today Study; NHSII, Nurses' Health Study II.
¹Participants were considered to have prevalent endometriosis if they reported their endometriosis had been diagnosed before baseline. Incident endometriosis cases were diagnosed after baseline.
²Number of self-reported endometriosis cases eligible for inclusion in the cohort's case validation study.

Table II. Physician, surgical, and pathology record receipt and confirmation proportions by cohort.
Cohort	BWHS	E3N	GUTS	NHSII first wave	NHSII second wave
Baseline population	59 001	98 995	15 044	116 429	116 429
Years in which endometriosis was self-reported	1995–2013	1990–2005	2005–2016	1989–1993	1989–2007
Years when participants were contacted for validation	2008–2016	2007	2011–2017	1996	2009–2011
Eligible women who self-reported endometriosis (N)	1605	558	491	1766	711
Contacted for medical record review (N)	1605	200	380	200	711
Unable to contact/never responded¹	1128	—	139	16	169
Denied diagnosis on consent form	200	—	64	3	77
Confirmed diagnosis but not access to medical records²	—	—	61	32	15
Permission obtained to review medical records³	277 (19.7%)	183 (91.5%)	116 (36.7%)	149 (75.6%)	360 (56.8%)
Medical records never received	52	0	27	17	132
Inadequate information received⁴	0	15	4	0	11
Relevant medical records received and reviewed	225/277 (81.2%)	168/183 (91.8%)	85/116 (73.3%)	132/149 (88.6%)	217/360 (60.3%)
Found no clinical diagnosis and no surgery performed	18	—	4	12	3
Clinical diagnosis noted but no surgery performed⁵	28	—	16	14	4
Surgically confirmed	134	137	63	98	185
Clinical diagnosis noted but disconfirmed at surgery⁶	0	—	2	8	2
Found no clinical diagnosis and disconfirmed at surgery	45	—	0	0	8
Pathology-confirmed but not visualized at surgery	0	—	0	0	9
Pathology-confirmed but no surgical report obtained	0	—	0	0	6
Overall confirmation proportion ⁷	162/225 (72.0%)	137/168 (81.5%)	81/85 (95.3%)	112/132 (84.8%)	206/217 (94.9%)
Surgically confirmed proportion⁸	134/179 (74.9%)	137/168 (81.5%)	63/65 (96.9%)	98/106 (92.5%)	185/204 (90.7%)
Clinically confirmed proportion⁹	18/28 (64.3%)	—	16/20 (80.0%)	14/26 (53.8%)	4/7 (57.1%)

¹Number of participants we were either unable to contact or never responded to repeated mailings.
²These participants confirmed their endometriosis diagnosis but either denied access to their medical records or did not respond to requests for access.
³The percentage providing permission to obtain medical records was calculated out of the total contacted minus those who denied their endometriosis diagnosis.
⁴Reasons inadequate information was received for medical records included no information on disease history and symptoms or medical records not relating to primary care physician, surgeon or gynecologic visits.
⁵Evidence found in the medical records that the participant had been told by her clinician that she most likely had endometriosis but no evidence that she ever had surgery to confirm her diagnosis.
⁶The eight surgically disconfirmed endometriosis cases in the NHSII first wave were all clinically confirmed (found documentation that the woman was told by her clinician that she most likely had endometriosis) and included in the overall and the clinically confirmed proportion.
⁷The numerator of this measure includes those who were surgically confirmed and clinically confirmed.
⁸The denominator of this measure includes those who had evidence of surgery in their medical records.
⁹The denominator of this measure includes those who did not have evidence of surgery in their medical records.

Table III. Differences in surgical and physician confirmation proportions by self-reported laparoscopic confirmation. ¹
Cohort	GUTS	NHSII first wave	NHSII second wave
	Reported Laparoscopic Confirmation
Medical records received and reviewed (N)	64	105	210
Found no clinical diagnosis and no surgery performed	1	0	3
Clinical diagnosis noted but no surgery performed²	3	0	3
Surgically confirmed	59	97	180
Clinical diagnosis noted but disconfirmed at surgery³	1	8	2
Found no clinical diagnosis and disconfirmed at surgery	0	0	8
Pathology-confirmed but not visualized at surgery	0	0	9
Pathology-confirmed but no surgical report obtained	0	0	5
Overall confirmation proportion ⁴	63/64 (98.4%)	105/105 (100%)	199/210 (94.8%)
Surgically confirmed proportion⁵	59/60 (98.3%)	97/105 (92.4%)	180/199 (90.5%)
Clinically confirmed proportion⁶	3/4 (75.0%)	—	3/6 (50.0%)
	Did Not Report Laparoscopic Confirmation ⁷
Medical records received and reviewed (N)	20	27	7
Found no clinical diagnosis and no surgery performed	3	12	0
Clinical diagnosis noted but no surgery performed²	13	14	1
Surgically confirmed	3	1	5
Clinical diagnosis noted but disconfirmed at surgery	1	0	0
Found no clinical diagnosis and disconfirmed at surgery	0	0	0
Pathology-confirmed but not visualized at surgery	0	0	0
Pathology-confirmed but no surgical report obtained	0	0	1
Overall confirmation proportion	17/20 (85.0%)	15/27 (55.6%)	7/7 (100%)
Surgically confirmed proportion^5,8	3/4 (75.0%)	1/1 (100%)	5/5 (100%)
Clinically confirmed proportion⁶	13/16 (81.3%)	14/26 (53.8%)	1/1 (100%)

¹Participants who reported laparoscopic confirmation of their endometriosis diagnosis on self-administered questionnaires.
²Evidence in the medical records that the participant had been told by her clinician that she most likely had endometriosis but no evidence that she ever had surgery to confirm her diagnosis.
³The eight surgically disconfirmed cases in the NHSII first wave were all clinically confirmed (found documentation that the woman was told by her clinician that she most likely had endometriosis) and included in the physician confirmation proportion.
⁴The numerator of this measure included those who were surgically confirmed and clinically confirmed.
⁵The denominator of this measure includes those who had evidence of surgery in their medical records.
⁶The denominator of this measure includes those who did not have evidence of surgery in their medical records.
⁷Total N does not add up to 21 for GUTS as one participant was missing laparoscopic confirmation status self-report and therefore could not be stratified.
⁸Evidence of surgery was found for 11 of the participants who did not self-report having had a laparoscopic confirmation. For the one participant from NHSII first wave and five participants from NHSII second wave, all had hysterectomies at which the presence of endometriosis was documented. For the four GUTS participants, three of them had self-reported a clinical diagnosis that preceded the data of the medical record documentation of laparoscopic surgery and one had a tubal ligation during which endometriosis was visualized.

Authors and Disclosures

A.L. Shafrir^1,2,*, L.A. Wise³, J.R. Palmer^4,5, Z.O. Shuaib⁶, L.M. Katuska⁷, P. Vinayak⁸, M. Kvaskoff^9,10, K.L. Terry^2,8,11 and S.A. Missmer^1,2,8,12

¹Division of Adolescent and Young Adult Medicine, Department of Pediatrics, Boston Children's Hospital and Harvard Medical School, Boston, MA, USA ²Boston Center for Endometriosis, Brigham and Women's Hospital and Boston Children's Hospital, Boston, MA, USA ³Department of Epidemiology, Boston University School of Public Health, Boston, MA, USA ⁴Slone Epidemiology Center, Boston University, Boston, MA, USA ⁵Section of Hematology-Oncology, Boston University School of Medicine, Boston, MA, USA ⁶Massachusetts Department of Public Health, Boston, MA, USA ⁷Department of Nutrition, Harvard T.H. Chan School of Public Health, Boston, MA, USA ⁸Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, MA, USA ⁹CESP, Fac de médecine—Univ. Paris-Sud, Fac. de médecine—UVSQ, INSERM, Université Paris-Saclay, Villejuif Cedex, France ¹⁰Gustave Roussy, Espace Maurice Tubiana, Villejuif Cedex, France ¹¹Department of Obstetrics and Gynecology, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, USA ¹²Department of Obstetrics, Gynecology, and Reproductive Biology, College of Human Medicine, Michigan State University, Grand Rapids, MI, USA

*Correspondence address
Boston Children's Hospital, 1 Autumn Street, 5th Floor, Boston, MA 02215, USA. E-mail: amy.shafrir@childrens.harvard.edu

Authors' roles
S.A.M. designed the study with critical discussions from L.A.W., J.R.P., and M.K., A.L.S., L.A.W., J.R.P., Z.O.S., L.M.K., P.V., M.K., and S.A.M. contributed to the data collection and execution of the study with A.L.S., Z.O.S., L.M.K., P.V., and M.K. contributing to the statistical analyses. A.L.S., L.A.W., J.R.P., M.K., K.L.T., and S.A.M. contributed to the interpretation of the data. A.L.S. drafted the original manuscript version and all authors contributed to reviewing and editing the final manuscript draft.

Conflict of interest
L.A. Wise has served as a fibroid consultant for AbbVie, Inc for the last 3 years and has received in-kind donations (e.g. home pregnancy tests) from Swiss Precision Diagnostics, Sandstone Diagnostics, Kindara.com, and FertilityFriend.com for the PRESTO cohort. SA Missmer serves as an advisory board member for AbbVie and a single working group service for Roche; neither are related to this study. No other authors have a conflict of interest to report. Funders had no role in the study design, conduct of the study or data analysis, writing of the report, or decision to submit the article for publication.