Smith JA, Vitale S, Reed GF, Grieshaber SA, Goodman LA, Vanderhoof VH, Calis KA, Nelson LM. Dry Eye Signs and Symptoms in Women With Premature Ovarian Failure. Arch Ophthalmol. 2004;122(2):151-156. doi:10.1001/archopht.122.2.151
To examine whether women with premature ovarian failure (POF) have abnormalfindings in ocular surface or tear parameters and whether they report symptomsof ocular discomfort compared with age-matched controls.
Sixty-five patients with POF and 36 age-matched healthy controls wereexamined for signs and symptoms of dry eye. The Ocular Surface Disease Indexquestionnaire and the 25-item National Eye Institute Visual Function Questionnaire(NEI-VFQ 25) were administered to the participants. Assessments of ocularsurface damage (Oxford and van Bijsterveld scores of vital dye staining) andtear status (Schirmer tests 1 [without anesthesia] and 2 [with anesthesia]and tear breakup time) were performed.
Women with POF scored significantly worse than controls on all ocularsurface damage parameters: Oxford score (3.2 vs 1.7; P =.001), conjunctival lissamine green (2.1 vs 1.3; P =.02), corneal fluorescein staining (1.2 vs 0.4; P =.005), and van Bijsterveld score (2.1 vs 1.3; P =.02). Further, the proportion of patients with POF meeting the dry eye diagnosticcriterion of a van Bijsterveld score greater than or equal to 4 was significantlygreater among women with POF than among controls (20% vs 3%; P = .02). The POF group also tended to have worse scores than controlson self-reported symptoms, as measured by the overall Ocular Surface DiseaseIndex (12.5 vs 2.1; P<.001) and the overall NEI-VFQ(94 vs 98; P = .001) afteradjustment for age and race. Schirmer test scores and tear breakup time didnot differ.
Women with POF were more likely to exhibit ocular surface damage andsymptoms of dry eye than age-matched controls. They were not, however, morelikely to have reduced tear production. To our knowledge, this associationbetween ocular surface disease and POF has not been previously reported. Thesedata provide further evidence of the multifaceted role of sex hormones inthe health and disease of the ocular surface.
Premature ovarian failure (POF) is defined as cessation of normal ovarianfunction in women younger than 40 years and affects 1% of women as determinedby a large cohort study of Minnesota women followed up for date and type ofmenopause.1 In addition to amenorrhea, womenwith POF exhibit hypoandrogenemia, hypoestrogenemia, and elevated gonadotropinlevels. Women with POF experience the same symptoms of estrogen deficiencyas do postmenopausal women, including hot flashes, night sweats, fatigue,and mood swings and have an increased risk of cardiovascular disease and osteoporosis.2 Although chemotherapy, irradiation, and chromosomalabnormalities can cause POF, an autoimmune origin is also well recognized.Women with autoimmune POF are at increased risk of potentially fatal autoimmuneadrenal insufficiency.3 In addition, womenwith POF demonstrate impaired immune regulation, including increased activationand total number of peripheral T cells, increased CD4/CD8 ratio, increasednumber of B cells,4 and reduced natural killercell activity.5 In most cases, the mechanismof POF is unknown.
In the United States, it is estimated that 15% of individuals aged 65to 84 years6 have keratoconjunctivitis sicca,defined as at least one symptom of dry eye often or all of the time. Thereare 2 major categories of keratoconjunctivitis sicca: aqueous tear deficiencyand evaporative tear deficiency.7 Aqueous teardeficiency is characterized by the decreased volume of tear production bythe lacrimal glands, chronic ocular surface inflammation, fluctuating visualdisturbance, decreased ability to perform activities of daily living, suchas reading and using a computer, and ocular discomfort.8 Evaporativetear deficiency can result from qualitative disturbance in the tear film withresultant instability, leading to increased evaporation and dryness of theocular surface. It is most often caused by meibomian gland disease. Althoughdry eye can occur in men or women of any racial group at any age, many studieshave found a higher risk of dry eye in women. Dry eye can be associated withautoimmune diseases, such as rheumatoid arthritis, systemic lupus erythematosus,or Sjögren syndrome.
The role of sex hormones in dry eye has been investigated in severalstudies.9- 12 Mostrecently, Schaumberg et al13 found an increasedprevalence of dry eye in women who had received hormone therapy (9.1% in thosetreated with estrogen alone and 6.7% in those treated with estrogen plus progesterone/progestin);the prevalence was lowest in women who had never used hormone therapy (5.9%).It remains unclear, however, what role estrogen excess, androgen deficiency,and/or estrogen-androgen imbalance play in association with dry eye. Androgendeficiency, as seen in congenital androgen insensitivity syndrome and antiandrogentherapy, has been associated with dry eye.14,15 Androgendeficiency is also seen in Sjögren syndrome, and it has been proposedto lead to evaporative tear deficiency in affected women.16 Sincewomen with POF also suffer from androgen deficiency, we hypothesized thatthey would exhibit signs or symptoms of dry eye more frequently than age-matchedcontrols with normal ovarian status.
Sixty-five consecutive women with POF and 36 age-matched normal controlwomen were evaluated for subjective evidence of dry eye symptoms and objectivesigns of ocular surface disease. Premature ovarian failure was diagnosed asfollows: amenorrhea for at least 4 months and 2 consecutive instances of elevatedFSH concentrations (≥40 IU/dL) obtained at least 1 month apart. Women wereexcluded from the control group if they had any of the following: abnormalmenstruation, eye disease (except refractive error), contact lens use, oruse of prescription medications, including oral contraceptives. The protocolwas approved by the institutional review board for the National Institutesof Child Health and Human Development (Bethesda, Md), and all participantssigned an informed consent.
No controls had ever received hormone therapy at the time of the screeningexamination. Patients with POF who were currently using hormone therapy wereasked to discontinue it for at least 2 weeks prior to the visit. Control womenwere seen at any time during the menstrual cycle. All participants underwentETDRS (Early Treatment Diabetic Retinopathy Study) visual acuity assessment,masked slitlamp biomicroscopy, including a standardized grading of eyelidmargin thickness and hyperemia, conjunctival erythema, chemosis, tear filmdebris and mucus, and extent of meibomian gland plugging. Grading of externaleyelid disease was accomplished as follows. For meibomian gland disease, 5meibomian glands were selected in the central lower eyelid, and the numberof glands from which meibum could be readily expressed was graded as none(0), mild (+1) if 1 to 2 glands were plugged, moderate (+2) if 3 to 4 glandswere plugged, or severe (+3) if all 5 glands were plugged. Eyelid margin erythemawas graded as none (0) if there was no erythema, mild (+1) if redness waslocalized to a small region of the eyelid(s) margin, moderate (+2) if rednessaffected most or all of the eyelid(s) margin, severe (+3) if redness affectedmost or all of the eyelid(s) margin and skin, or very severe (+4) if therewas marked diffuse redness of the eyelid(s) margin and the skin. Eyelid marginswelling was graded as none (0), mild (+1) if localized to a small regionof the eyelid(s), moderate (+2) if it affected most or all of the eyelid(s)but was not prominent, severe (+3) if most or all of the eyelid(s) was affectedand prominent, or very severe (+4) if the swelling was prominent, with eversionof the eyelid(s).
In addition, we performed tests of tear production (Schirmer test 1[without anesthesia] and 2 [with anesthesia]) and an assessment of ocularsurface damage using vital dye staining with 5 µL of 2% sodium fluoresceinor 10 µL of 0.5% lissamine green instilled using capillary tubes. Thevan Bijsterveld grading method assessed lissamine green staining of the corneaand the temporal and nasal bulbar conjunctiva. The cornea and conjunctivawere graded separately (0-3 for each zone) and then combined for the totalvan Bijsterveld score.17 The Oxford methodassessed sodium fluorescein staining of the cornea and lissamine green stainingof the nasal and temporal bulbar conjunctiva, graded separately (0-5 for eachzone) and then combined for the total Oxford score.18 Determinationof tear film stability was assessed by fluorescein tear breakup time. If thetear breakup time was less than 10 seconds, the test was repeated for a totalof 3 values, and the average was determined. The tests were performed accordingto standard operating procedures and in the following order: Schirmer 1, slitlampexamination, vital dye staining, tear breakup time, and Schirmer 2.
For statistical analysis, the maximum (worse) score for the 2 eyes ofeach individual was used for Oxford, lissamine green, and van Bijsterveld,and the minimum (worse) score for the 2 eyes was used for Schirmer 1, Schirmer2, and tear breakup time. Tear breakup time values greater than or equal to10 seconds19 were coded as 10 (normal), andthose less than 10 seconds were defined as abnormal. A score of 5 mm or lesson Schirmer 1 or a van Bijsterveld score greater than or equal to 4 were usedas objective evidence of dry eye, following the past and current EuropeanCommunity Study Group on Classification Criteria for the diagnosis of dryeye for Sjögren Syndrome.20 Grading ofthe extent of external disease was performed using a standardized scheme.
The Ocular Surface Disease Index (OSDI)21 (AllerganInc, Irvine, Calif) was used to quantify the effect of dry eye on qualityof life, including gritty or painful eyes, limitation in performance of commonactivities, such as reading and working on a computer, and the effect of environmentaltriggers, such as wind, on dry eye symptoms during a 1-week recall period.The effect of dry eye on activities of daily living and visual functioningwas determined using the 25-item National Eye Institute Visual Function Questionnaire(NEI-VFQ 25).22,23 Responses tothe questionnaires were scored using the methods described by the authorsof these instruments.21,24 Forthe NEI-VFQ, subscale scores for general vision, ocular pain, near vision,distance vision, social functioning, mental functioning, role functioning,dependency, driving, color vision, and peripheral vision, as well as an overallscore, were computed. The NEI-VFQ scores can range from 0 to 100, with lowerscores indicating more problems or symptoms. For the OSDI, 3 subscales forocular discomfort, visual functioning, and environmental triggers were computed,as well as an overall score. The OSDI scores can range from 0 to 100, withhigher scores indicating more problems or symptoms.
Preliminary analyses were performed, adjusting for age and race. Becauseof a racial imbalance between the controls and the other 2 groups, we retainedrace (grouped as black/Hispanic/other vs white) as a potential confounderin all subsequent analyses. Logistic regression models (SAS version 8.02;SAS Institute, Cary, NC) were used to perform comparisons between controlsand patients. χ2 or Fisher exact tests were used to compareproportions in 2 × κ contingency tables. Spearman correlationcoefficients were computed to assess the strength of the linear relationshipbetween pairs of variables. All participants answered that they had "no difficultyat all" with the single color vision item on the NEI-VFQ, so this subscalewas not included in the analyses.
The basic demographic and visual acuity characteristics of participantsin the 2 groups are presented in Table 1. As expected, the age distributions did not differ between the2 groups. However, the controls had markedly more black (39%) and other (22%)participants compared with the POF group. Ninety-one percent of women withPOF had a visual acuity of 20/20 or better in the better eye compared with94% of controls. Visual acuity in the worse eye was somewhat worse in patientsthan in controls: 22% had a visual acuity worse than 20/20 vs 14% of controls(P = .41).
Mean values for ocular surface and tear film parameters are presentedin Table 2. Women with POF hadworse mean scores than controls on all measures of ocular surface damage:total Oxford score (P = .001),conjunctival lissamine green (P = .02), corneal fluorescein staining (P<.001),and total van Bijsterveld score (P = .02). The percentages of patients with abnormal ocular surface andtear parameters are presented in Table 3. Thirteen (20%) of 65 women with POF had a van Bijsterveld scoregreater than or equal to 4 compared with 1 (3%) of 36 controls (P = .02). Furthermore, significantly morePOF patients (38%) showed abnormal corneal fluorescein staining (score >1)than did controls (8%) (P<.001 after adjustmentfor race and age). In addition, significantly more women with POF (30/45 [67%])demonstrated eyelid margin erythema than controls (11 [31%] of 36; P = .03). The proportion of women with anymeibomian gland plugging was similar in each group (controls: 36%, POF: 40%).The mean tear breakup time, Schirmer 1, and Schirmer 2 did not differ betweenwomen with POF and controls. Although women with POF were roughly twice aslikely to have a Schirmer 1 score less than or equal to 5 mm (23% vs 11%),this difference was not statistically significant. The proportion of womenwith POF that met the past and current European Community Study Group on ClassificationCriteria for Sjögren syndrome dry eye25 (vanBijsterveld ≥4 and/or Schirmer 1 score ≤5 mm) was greater than thatof controls (37% vs 22%), although this difference was not statistically significant(P = .19). No significant differences in the percentagewith tear film debris were observed.
Comparisons of the OSDI and NEI-VFQ subscales were done for patientsvs controls (Table 4). After adjustingfor age and race, OSDI scores for all 3 subscales and the overall scale weresignificantly higher (worse) for patients than for controls. These significantdifferences in OSDI scores did not change when adjustment was made for anyof the ocular surface or tear parameters (Oxford, lissamine green, cornealfluorescein staining, van Bijsterveld, Schirmer 1 or 2, and tear breakup time)(data not shown). For the NEI-VFQ overall scale and subscales, general vision,ocular pain, near vision, distance vision, mental functioning, and driving,POF patients had significantly lower (worse) scores than did controls. Scoreson the NEI-VFQ subscales remained significantly worse for patients than forcontrols after adjustment for the ocular surface and tear parameters thatdiffered significantly between patients and controls (total Oxford, conjunctivallissamine green, corneal fluorescein staining, and total van Bijsterveld scores)(data not shown). Compared with controls, patients (n = 22) had significantlyworse mean responses to both of the visual analog questions, "how dry do youreyes feel most of the time?" (29 vs 3; P = .004),with responses ranging from 0 = not dry at all to 100 = very dry, and "howoften do you experience burning, stinging, or grittiness of your eyes?" (18vs 2; P = .003), with responses ranging from from100 = most of the time to 0 = never, even after adjustment for age and race.
Of those patients who ever used contact lenses, the proportion who discontinueduse did not statistically significantly differ between the groups (control:3/7 [43%]; POF: 12/31 [39%]). In addition, the proportion who discontinuedcontact lens use because of dryness did not differ between the groups (1 [33%]of 3 controls vs 5 [42%] of 12 patients). Use of tear substitutes or lubricatingointments was, however, statistically significantly more common among patients(21/65 [32%]) than among controls (1/36 [3%]; P =.001). This association did not change after adjustment for age and race (P = .007). The women with evidence of ocular surface damage(van Bijsterveld ≥4) compared with those without were significantly morelikely to demonstrate positive antinuclear (6/13 vs 9/51; P = .04) and antiadrenal (2/13 vs 0/50; P = .04) autoantibodies but notthyroid peroxidase or parietal autoantibodies. The frequency of clinicallyapparent autoimmune disease was low and did not differ between the 2 groups.
We examined the associations between the objectively measured ocularsurface and tear parameters and the subjective OSDI and NEI-VFQ scores usingvarious analytic methods, including Spearman correlation coefficients forthe continuous variables and Kruskal-Wallis tests for differences in the medianOSDI (or NEI-VFQ) score among groups defined according to severity of ocularsurface disease (eg, none, mild, severe). Within the POF group, the followingshowed significant association: worse near vision with lower (more abnormal)tear breakup time (r = 0.33; P = .01) and worse peripheral vision with lower Schirmer 1 score (r = 0.31; P = .02). Within thecontrol group, there were no associations between any of the objective parameters(Oxford, Lissamine green, corneal fluorescein staining, van Bijsterveld, Schirmer1 or 2, or tear breakup time) and OSDI scores (nearly all r values were <0.2). Although there was a consistently significantassociation of lower (more normal) Oxford score with better driving scoresamong the controls, it should be noted that only 2 controls had abnormal Oxfordscores.
The women with POF showed significantly more severe ocular surface damagethan controls on all grading scales and scored significantly worse on allof the dry eye symptom assessments and the visual function questionnaire.Further, a larger proportion of POF patients met a diagnostic criterion fordry eye by the severity of ocular surface vital dye staining (van Bijsterveldscore ≥4) than did controls; however, aqueous tear production, as measuredby the Schirmer test, was not significantly different between the groups.
Sex hormone messenger RNAs, proteins, and receptors have been foundin ocular tissues, including the cornea, conjunctiva, meibomian glands, lacrimalgland acinar cells, and retinal pigment epithelium.26- 33 Estrogensin general are immune response stimulators, and androgens act as immunosuppressors.34 Sullivan et al35 haveproposed that androgen insufficiency contributes to meibomian gland dysfunction,tear film instability, and evaporative dry eye in menopause, aging, Sjögrensyndrome, complete androgen insensitivity syndrome, and antiandrogen use.Hykin and Bron36 have documented increasedmeibomian gland disease and evaporative tear deficiency in postmenopause andwith aging. In addition, androgens and estrogens can have profound effectson both the cellular and humoral immune systems. For example, androgens canenhance T-cell suppressor activity and decrease autoreactive antibody formation.37 This may be one of the reasons that autoimmune diseasesare more common in women than in men.38 Theexact pathologic mechanisms of POF remain unclear; however, there is evidencethat autoimmunity may play a role in some cases. For example, POF can be seenin autoimmune polyglandular syndrome type I, a disorder characterized by thedestruction of endocrine glands, impaired cellular immunity, and ectodermaldystrophy.39 In contrast, a mutation in FOXL2, a forkhead transcription factor gene that causesblepharophimosis-ptosis-epicanthus-inversus syndrome may be associated withPOF.40 None of the patients in our study demonstratedany eyelid abnormalities. Local ocular disruption of sex hormone homeostasisor androgen deficiency alone may be responsible for the ocular surface diseasethat we found in women with POF. A common genetic factor could be responsiblefor both the ovarian and ocular disease; perhaps a shared structural proteinor other factor is required to maintain both developing ovarian folliclesand a stable tear film and healthy ocular surface. It is possible that thedry eye phenotype may signal a particular cause of POF since not all patientshad dry eye.
Our study had several limitations. There were more white women in thePOF group than in the control group; however, to date, there is no evidencefor a racial difference in the prevalence of dry eye. Furthermore, adjustmentfor race did not modify any of the differences we found between the groups.The average tear breakup time in 129 normal women aged 20 to 24 years hasbeen reported as 18.9 seconds.41 While themean tear breakup time of 6.3 seconds in our POF group is less than thesepreviously published values, this difference may have been due to our methodof coding values greater than or equal to 10 seconds as 10. There are fewpublications that document sex- and age-specific data for normal tear production,and most studies have shown that Schirmer test results are highly variable.Normal scores on the Schirmer test with anesthesia have been reported to rangefrom 23.71 mm (range, 14-30 mm) in normal men and women42 to33.3 mm in a sample of 48 normal Indian women.43 Feldmanand Wood44 found wide variability in scoreson the Schirmer test with anesthesia in 10 healthy women (age range, 19-38years) tested daily for 30 days; mean values ranged from 8.5 to 21 mm, andSDs ranged from 2.7 to 7.6. The mean tear production in our controls (Schirmer1 = 13.2 mm, Schirmer 2 = 9.6 mm) was lower than we expected for healthy youngwomen with no history of ocular disease. These findings may indicate thatthe published normal values for these measures are not applicable to womenin the 18- to 40-year-old age group, or that our controls or procedures weredifferent from those in published studies. However, this difference wouldtend to make the controls appear more like dry eye cases, thus tending todiminish any associations observed (ie, any bias would be in a conservativedirection).
The women with POF and dry eye in this study had somewhat better scoreson the overall NEI-VFQ and OSDI than those published for dry eye cases. Ina previous study of 75 patients with dry eye evaluated with the NEI-VFQ,45 the mean ± SD overall score was 87.9 ±8.4, and the mean ocular pain score was 69.5 ± 18.7. In our study,the women with POF who met the more stringent diagnostic criteria for Sjögrensyndrome–related dry eye (Schirmer 1 ≤5 mm and/or van Bijsterveld≥4) had a somewhat higher (better) mean NEI-VFQ overall score of 93 ±7. However, their mean ± SD ocular pain subscale score was 82 ±15, lower (worse) than reported scores for patients with cataract (86 ±19)46 and low vision (97.3 ± 8.3).47 For women with POF meeting the same criteria, theoverall mean ± SD OSDI score was 12.2 ± 13.3, lower (better)than the published overall OSDI scores for 83 patients with mild to moderatedry eye (18.1 ± 17.1).21
We found an increased prevalence and severity of both signs and symptomsof ocular surface disease in patients with karyotypically normal spontaneousPOF compared with age-matched normal women. The dysregulation of sex hormonesor immunologic dysfunction seen in POF may play a role in the pathogenesisof the ocular surface disease. Epithelial dysfunction or lacrimal gland diseasemay play a role. An abnormality in tear composition likewise could resultin the corneal and conjunctival damage seen. The key role of sex hormonesin ocular surface homeostasis is further supported by these findings; however,the specific pathophysiologic mechanisms involved in dry eye in POF remainto be determined. Additional studies are needed to further characterize thepathologic mechanisms of dry eye in POF.
Corresponding author and reprints: Janine A. Smith, MD, NationalEye Institute, National Institutes of Health, 10 Center Dr, MSC 1863, Bethesda,MD 20892-1863 (e-mail: firstname.lastname@example.org).
Submitted for publication March 5, 2003; final revision received July22, 2003; accepted August 15, 2003.
Results from this study were presented at the Association for Researchin Vision and Ophthalmology annual meeting; May 9, 2002; Fort Lauderdale,Fla.
We are grateful for the support of the POF support group in the referralof women to the National Institutes of Health Clinical Center, Bethesda, forparticipating in these studies.