A METHOD OF ASSESSING A FEMALE'S RISK OF HAVING PCOS AS WELL AS PRODUCTS AND USES RELATING THERETO

Abstract

The present invention relates to a method of assessing a female’s risk of having polycystic ovary syndrome (PCOS), a kit for use in assessing a female’s risk of having PCOS, the use of a marker combination in the assessment of a female’s risk of having PCOS, a computer system for use in a method according to the present invention as well as a computer program and a computer-readable storage medium comprising instructions, which when executed by a computer, cause the computer to carry out the method of the present invention.

Claims

1. A method of assessing a female’s risk of having polycystic ovary syndrome (PCOS), the method comprising a) providing a data set including an OA-value reflecting the length of the female’s menstrual cycle and/or the number of the female’s menstrual cycles per year, wherein an increased OA-value relative to the OA-value of a healthy reference population indicates an abnormal menstrual cycle length and/or number, a HA-value reflecting the female’s androgen status, wherein an increased HA-value relative to the HA-value of a healthy reference population indicates an increased androgen level in the female, and an AMH-value corresponding to the amount or concentration of anti-Mullerian hormone (AMH) in a sample obtained from the female; b) processing the data set provided in step a) with a processing unit, wherein the processing comprises combining values of the data set provided in step a) into one combined value; c) comparing the combined value obtained in step b) to the corresponding combined value as established in a reference population, wherein an increased combined value of the female relative to the combined value of a healthy reference population is indicative of an increased risk of PCOS; and d) indicating the female’s risk of having PCOS via an indication unit.

2. The method of claim 1, wherein the data set of step a) further includes a WEIGHT-value reflecting the female’s body weight, wherein an increased WEIGHT-value relative to the WEIGHT-value of a normal weight population indicates an increased body weight; and/or an AGE-value reflecting the female’s age.

3. The method of claim 1, wherein the HA-value corresponds to the amount or concentration of free testosterone (FT) in a sample obtained from the female, or the ratio of the amount or concentration of total testosterone (TT) and the amount or concentration of sex hormone-binding globulin (SHBG) in a sample obtained from the female (TT/SHBG), optionally multiplied by a constant a (a * TT/SHBG).

4. The method of claim 1, wherein in step b) the combined value is a weighted combined value obtained by weighted calculation of the values provided in step a) and in step c) the weighted combined value is compared to the corresponding weighted combined value of a reference population, wherein an increased weighted combined value of the female is indicative of an increased risk of having PCOS.

5. The method of claim 1, wherein the female has a high risk, if the combined value is above a threshold.sub.high; and the female has a moderate risk, if the combined value is above a threshold.sub.moderate and below threshold.sub.high; and the female has a low risk, if the combined value is below threshold.sub.moderate.

6. The method of claim 1, wherein one or more values of the reference population and/or the combined value of the reference population are retrieved from a database.

7. The method of claim 1, wherein the data set of step a) further includes a PHE-value reflecting one or more phenotypical characteristics known to be indicative of PCOS, wherein an increased PHE-value relative to the PHE-value of a healthy reference population indicates the presence of one or more phenotypical characteristics known to be indicative of PCOS is reflected by an increased PHE-value, and wherein the phenotypical characteristic is polycystic ovarian morphology (PCOM) and/or hyperandrogenemia hirsutism.

8. The method of claim 1, wherein the sample is a blood sample selected from the group consisting of serum, plasma, and whole blood.

9. The method of claim 1, wherein the female is a human.

10. The method of claim 1, wherein the method further includes determining one or more values of the data set provided in step a).

11. The method of claim 1, wherein the amount or concentration of one or more of the hormone(s) in the female’s sample is measured by an immunoassay and/or mass spectrometry.

12. A kit for assessing a female’s risk of having PCOS, the kit comprising reagents required to specifically measure in a sample obtained from the female (i) the amount or concentration of FT or (ii) the amount or concentration of TT and the amount or concentration SHBG; the amount or concentration of AMH; and optionally the amount or concentration of one or more further hormones indicative of PCOS.

13. (canceled)

14. A computer system comprising: a) a data set unit comprising computer instructions for providing a data set including an OA-value reflecting the length of the female’s menstrual cycle and/or the number of the female’s menstrual cycles per year, wherein an increased OA-value relative to the OA-value of a healthy reference population indicates an abnormal menstrual cycle length and/or number; a HA-value reflecting the female’s androgen status, wherein an increased HA-value relative to the HA-value of a healthy reference population indicates an increased androgen level in the female; and an AMH-value corresponding to the amount or concentration of AMH in a sample obtained from the female; b) a processing unit comprising computer instructions for processing the data sets of step a), wherein the processing comprises combining values of the data set provided in step a) into one combined value; c) a reference data unit comprising computer instructions for (i) storing and/or retrieving a reference data set including one or more of the reference values as established in a reference population and processing the reference data set into a combined value of the reference population; or (ii) storing and/or retrieving a combined value of the reference population; (d) comparing the combined value obtained in step b) to the corresponding combined value of step c), wherein an increased combined value of the female relative to the combined value of a healthy reference population is indicative of an increased risk of PCOS; and e) an indication unit indicating the female’s risk of having PCOS.

15. (canceled)

16. (canceled)

17. Computer-readable storage medium comprising instructions which, when executed by a computer, cause the computer to carry out steps a), b), c) and d) of the method of claim 1.

Description

FIGURES

[0369] FIG. 1 illustrates ROC curves resulting from weighted logistic regression evaluated using 200 Monte-Carlo cross-validation runs for 1955 cases and 1642 controls based on all variables of the PCOS risk score (age, BMI, OA, FAI, and AMH; Area under the curve (AUC):0.976) compared to single variables OA (AUC:0.898), FAI (AUC:0.765) or AMH (AUC:0.838) alone.

[0370] FIG. 2 shows ROC curves resulting from weighted logistic regression evaluated using 200 Monte-Carlo cross-validation runs for 1955 cases and 1642 controls based on all variables of the PCOS risk score (age, BMI, OA, FAI, and AMH; AUC:0.976) and combinations of AMH and FAI (AUC:0.877) or AMH and SHBG (AUC:0.873).

[0371] FIG. 3 illustrates a histogram of PCOS risk probabilities resulting from weighted logistic regression using age and BMI as well as OA, FAI and AMH using 1955 cases and 1642 controls. Crosses and circles indicate control and case subjects’ risks, respectively. Vertical lines denote PCOS risk classification derived using the predictiveness curve for 80% sensitivity and specificity.

[0372] FIG. 4 shows the Mean Regression coefficients (and mean SDs) of the weighted logistic regression model resulting from 200 Monte-Carlo cross-validation runs using 1955 cases and 1642 controls and the variables age and BMI (A) or age (B) as well as OA, FAI and AM H.

[0373] FIG. 5 depicts the Predictiveness curve of the PCOS risk score for 1955 cases and 1642 controls based on variables age, BMI, OA, FAI, and AMH (A) or variables age, OA, FAI, and AMH (B).

[0374] FIG. 6 shows ROC curves for prediction of case-control status of the independent second sample set consisting of 200 cases and 44 controls for the PCOS risk score and single variables of the PCOS risk alone. It visualizes the performance of the PCOS risk score based on 44 controls and 200 cases described in Example 3. The ROC curves as well as the AUC demonstrated that the combined variables of the PCOS risk score (age, BMI, OA, FAI and AMH) are superior as to using the single variables OA, FAI or AMH alone. The AUC of the PCOS risk score was 0.99 indicating a very good separation between cases and controls, followed by OA (AUC: 0.96), AMH (AUC: 0.76), and FAI (AUC:0.90). The performance of the PCOS risk score for the four-variable combinations (variables OA, FAI, AMH, and AGE of the PCOS risk) was also assessed (see Table 9).

[0375] FIG. 7 shows ROC curves for prediction of case-control status of the second sample set consisting of 200 cases and 44 controls for the PCOS risk score with the combined variables age, BMI, OA, FAI, AMH and combinations of AMH + FAI and AMH + SHBG. It illustrates the performance of the PCOS risk score including all variables (age, BMI, OA, FAI, and AMH) compared to a combination of AMH and FAI or a combination of AMH and SHBG (as suggested by Calzada et al.). The highest AUC was found for the PCOS risk score (AUC:0.99). An AUC of 0.90 was for the combination of AMH + FAI (AUC:0.90) and the combination of AMH +SHBG (AUC:0.90). The ROC curves as well as the AUC show that the combined variables of the PCOS Risk score (age, BMI, OA, FAI, and AMH; AUC: 0.96) are superior to using AMH combined with either FAI (AUC: 0.90) or AMH and SHBG (AUC:0.90)

[0376] FIG. 8 shows ROC curves for prediction of case-control status of the second sample set consisting of 200 cases and 44 controls for the biochemical detection of Hyperandrogenism (HA). FAI was compared to LH, FSH and the LH/FSH ratio. The highest AUC was found for FAI (AUC: 0.90) followed by the ratio of LH/FSH (AUC: 0.85).

[0377] FIG. 9 illustrates a histogram of PCOS risk probabilities for the second sample set consisting of 200 cases and 44 controls The weights of the PCOS risk were derived using weighted logistic regression with 200 MCCV runs and 100 repetitions using age. BMI, OA. FAI and AMH in the sample set of 1866 cases and 1675 controls. The vertical lines indicate thresholds at low=0.2 and high=0.8 resulting in 96.0% sensitivity and 90.9% specificity. Crosses and circles indicate control and case subjects’ risks, respectively.

EXAMPLES

Example 1: Derivation of a PCOS Risk Score

[0378] In total, N = 1642 controls and N = 1955 cases were used for the derivation of the PCOS risk score based on combination of the numeric variables for age, BMI, FAI, and AMH as well as the categorical variable for Oligo Anovulation (OA) (yes, no). Additionally, the PCOS risk score was derived using a four-variable combination namely OA, age, FAI, and AMH but excluding BMI as well as using a three-variable combination namely OA, FAI, and AMH but excluding BMI and age.

Cases

[0379] 1955 women diagnosed with PCOS aged 20 to 45 years and not using contraceptives were considered OA was based on information of irregular menstrual cycles and/ or cycle length, Hyperandrogenism (HA) was derived as the free androgen index (FAI) based on levels of serum testosterone (nmol/l) and serum sex hormone-binding globulin (SHBG) (nmol/l):

[00005]$\text{FAI}\text{=}\text{testosterone}\text{/}\text{SHBG * 100}$

Patients were evaluated for PCOM by an ovarian volume ≥10 ml, and/or an antral follicle count (AFC) above threshold.

[0380] In addition, serum Anti-Müllerian Hormone (AMH) was measured using Elecsys AMH Plus immunoassay.

[0381] The PCOS cases covered PCOS patients representing the four phenotypes according to the Rotterdam criteria (PCOS Consensus Workshop Group, Fertil Steril 2004; 81:19-25). [0382] Phenotype A (Oligo Anovulation (OA) +, Hyper-Androgenism (HA) +, polycystic ovarian morphology (PCOM) +) [0383] Phenotype B (OA +, HA +, PCOM -) [0384] Phenotype C (OA -, HA +, PCOM +) [0385] Phenotype D (OA +, HA -, PCOM +)

Controls

[0386] In the derivation of the PCOS risk score, the control group consisted of 1642 healthy women between 20 and 45 years without having PCOS.

[0387] Age and body mass index (BMI), antral follicle count (AFC) and serum AMH values were available, Testosterone and SHBG levels to depict the FAI were simulated based on information gained from the cases as well as the reference range values for healthy women to reflect the expected distribution of these variables in healthy subjects based on reference range studies. The simulation was done by sampling from the expected SHBG and Testosterone distributions in healthy women.

[0388] The following table lists the statistics of the variables for cases and controls:

TABLE-US-00002 Descriptive statistics of the variables of the PCOS risk score overall and for patients (cases) and controls. For the controls, the FAI was derived based on simulated values for Testosterone and SHBG. OA of controls was based on information of the cycle length or simulated. The control group represented 45.6% of the overall number of patients and controls. All (N=3597) Controls (N=1642) Cases (N=1955) OA* Regular Cycle N=1394 (38.8%) N=1345 (81.9%) N=49 (2.5%) OA N=2203 (61.2%) N=297 (18.1%) N=1906 (97.5%) FAI* Mean 4.6 2.7 6.1 SD 4.5 2.2 5.3 Median 3.2 2.1 4.9 Q1 ... Q3 1.8 ... 5.8 1.3 ... 3.4 2.6 ... 7.6 Min... Max 0.20 ... 67.1 0.20 ... 27.2 0.28 ... 67.1 AMH (ng/ml) Mean 4.9 2.7 6.7 SD 4.0 2.3 4.2 Median 3.9 2.2 5.6 Q1 ... Q3 2.1 ... 6.5 1.1 ... 3.6 3.8 ... 8.7 Min... Max 0.010 ... 23.0 0.010 ... 22.1 0.010 ... 23.0 Age (years) Mean 30.8 33.9 28.2 SD 5.6 5.1 4.5 Median 30.0 34.0 28.0 Q1 ... Q3 27.0 ... 35.0 30.0 ... 38.0 25.0 ... 31.0 Min ... Max 20.0... 44.0 20.0 ... 44.0 20.0 ... 44.0 BMI Mean 26.1 25.1 26.9 SD 5.7 5.0 6.2 Median 24.7 23.8 25.9 Q1 ... Q3 21.8 ... 29.5 21.6 ... 27.7 22.1 ... 30.7 Min ... Max 13.9... 52.7 14.8 ... 52.7 13.9 ... 51.2 Testosterone (nmol/L)* Mean 1.7 1.4 2.0 SD 0.92 0.78 0.97 Median 1.5 1.2 1.7 Q1 ... Q3 1.0 ... 2.2 0.89 ... 1.8 1.2 ... 2.5 Min ... Max 0.17 ... 7.3 0.23 ... 6.2 0.17 ... 7.3 SHBG (nmol/L)* Mean 56.0 69.5 44.6 SD 35.8 40.7 26.3 Median 47.8 59.7 38.3 Q1 ... Q3 31.5 ... 70.5 42.2 ... 86.6 26.2 ... 57.2 Min ... Max 7.1 ... 370 8.8 ... 37.0 17.1 ... 342 *partially based on simulations. OA: Oligomenorrhea and/or anovulation FAI: Free Androgen Index; FAI = Testosterone / SHBG *100 BMI: Body Mass Index; BMI = weight (kg) / size (m)2. AMH: Anti-Mullerian Hormone SHBG: Sex Hormone-Binding Globulin

[0389] Of note, there was a difference in age between cases and controls due to the design of the study.

PCOS Risk Score

[0390] The proposed PCOS risk score calculates the patient’s risk of having PCOS ranging from 0 to 1, with higher values meaning a higher risk of having PCOS.

[00006]$\begin{array}{c}\text{risk}\text{=}\log \left(\frac{p}{1-p}\right)=\text{log}\left(\frac{P\left(y=1\right)}{1-P\left(y=1\right)}\right)\\ \begin{array}{l}=W_0+X_{age}\ast W_{age}\\ +X_{BMI}\ast W_{BMI}\left(optional\right)\\ +X_{OA}\ast W_{OA}\\ +X_{PAI}\ast W_{PAI}\\ +X_{AMH}\ast W_{AMH}\in \left[0;1\right]\end{array}\end{array}$

[00007]$\begin{array}{l}\text{with}\\ P\left(y=1\right)=\text{probablitity of being a case,}\\ {\text{W}}_0=\text{intercept of the weighted logistic regression model,}\\ \text{variables}{X}_{age}=\text{age in years,}\\ X_{BMI}=\text{BMI}\left(\text{optional}\right).\\ X_{OA}=\text{Oligo Anovulation}\left(\text{Regular/OA}\right)\\ X_{FAI}=\text{FAI}\text{.}\\ X_{AMH}=\text{AMH}\text{in ug/mL and}\\ \text{weights}{W}_{age}\text{for age}\text{.}\\ W_{BMI}\text{for BMI}\left(\text{optional}\right).\\ W_{OA}\text{for OA}\text{.}\\ W_{FAI}\text{for FAI and}\\ W_{AMH}\text{for AMH}\text{.}\end{array}$

[0391] Weighted logistic regression model was established with case-control status as endpoint within a Monte-Carlo cross-validation (MCCV) with 200 runs (Xu & Liang 2001).

[0392] The variables Age, BMI (optional), and OA as well as FAI were included for the derivation of the PCOS risk, whereas the variables AMH, testosterone and SHBG were included as log-transformed variables. To account for imbalance between the higher numbers of cases versus controls a weighted logistic regression model was applied to derive the PCOS risk. This means each subject was assigned a weight, which was considered within the model estimation of the logistic regression (Hastie et al. (2009). The weights were chosen according to Elkan (2001) by applying costs for the false classification of cases and controls each.

[0393] MCCV. For each of the MCCV runs the data set was randomly split into training and test set (80% and 20%, respectively), while maintaining the ratio of cases and controls. On the current training set, the model was built and the performance was evaluated by means of area under the ROC curve (AUC) using the respective test set. The estimated overall performance of the logistic regression model was given as mean AUC. Mean sensitivity and specificity was calculated to estimate the model performance for cases and controls, separately. The stability of the regression model was evaluated by providing the mean of the regression coefficients together with the standard deviation (SD) and coefficient of variation (CV).

[0394] The mean regression coefficients for each variable from MCCV were considered as the weights W the PCOS risk.

[0395] The PCOS risk, i.e. the probability of being with PCOS (being a case), was than estimated by:

[00008]$\text{risk}\text{=}\widehat{p}=\widehat{P}\left(y=1\right)=\frac{\exp \left({\widehat{W}}_0+X\widehat{W}\right)}{1+\exp \left({\widehat{W}}_0+X\widehat{W}\right)},$

[00009]$\begin{array}{l}\text{with}\\ {\widehat{W}}_0=\text{estimated intercept of logistic regression model,}\\ \widehat{W}\text{estimated weights for age, BMI}\left(\text{optional}\right),\text{OA, FAI and AMH and}\end{array}$

[00010]$\begin{array}{l}X\widehat{W}=X_{age}\ast {\widehat{W}}_{age}+X_{BMI}\ast {\widehat{W}}_{BMI}+X_{OA}\ast {\widehat{W}}_{OA}+X_{PAI}\ast {\widehat{W}}_{FAI}\\ +X_{AMH}\ast {\widehat{W}}_{AMH}.\end{array}$

[0396] The PCOS risk classifies as low, moderate and high as follows: The risk thresholds were derived using the predictiveness curve as proposed by Pepe et al. (2007) given that at least 80% sensitivity and 80% specificity are achieved.

Results

[0397] FIG. 1 visualizes the performance of the PCOS risk score based on 1642 controls and 1955 cases. The ROC curves as well as the AUC show superior performance for the PCOS risk score with the variables (age, BMI, OA, FAI, and AMH) as compared to single variables OA, FAI, or AMH alone. The AUC of the PCOS risk score was 0.98 indicating a very good separation between cases and controls, followed by OA (AUC: 0.90). AMH (AUC: 0.84), and FAI (AUC: 0.77). The performance of the three- and four-variable combinations (variables OA, FAI, AMH, and AGE of the PCOS risk or OA, FAI or AMH) was also assessed (see Table 7). The ROC curves as well as the AUC demonstrated that the combined variables of the PCOS risk (age, OA, FAI and AMH) are superior as to using OA, FAI or AMH alone. The AUC of the PCOS risk reaches 98%, indicating that a very good separation between cases and controls can be achieved, followed by OA with 90% and AMH with 84%, The AUC of FAI reaches only about 77%. The three-variable model (combination of OA, AMH and FAI vs OA, FAI or AMH alone) resulted in a PCOS Risk Score AUC of 0.970.

[0398] FIG. 2 shows the performance of the PCOS risk score including all variables (age, BMI, OA, FAI, and AMH) compared to a combination of AMH and FAI or a combination of AMH and SHBG (as suggested by Calzada et al.). The ROC curves as well as the AUC show that the variables of the PCOS Risk score (age, BMI, OA, FAI, and AMH) are superior to using AMH combined with either FAI or SHBG. The performance of the model when either all variables of the PCOS risk (except for BMI) or OA, FAI or AMH alone was also assessed (data not shown). The ROC curves as well as the AUC proved that the variables of the PCOS risk (age, OA, FAI and AMH) are superior as to using AMH binarized using a cutoff of 5.03 ng/ml together with either FAI or SHBG. The AUC of the PCOS risk reaches 98%, indicating that a very good separation between PCOS cases and controls without PCOS can be achieved, followed by AMH (binarized using cutoff 5.03 ng/ml) plus SHBG with 82%. The AUC of AMH (binarized using cutoff 5.03 ng/ml) plus FAI reaches only about 76%.

[0399] The stability of the weighted logistic regression model was evaluated using 200 MCCV runs and displayed in FIG. 4 as well as Tables 3 and 4. Results indicate that OA has the largest influence on the PCOS risk, followed by AMH and FAI. Small standard deviations indicate quite stable regression coefficients throughout the MCCV runs.

TABLE-US-00003 Table of weights of the PCOS risk derived based on weighted logistic regression using age and BMI as well as OA, FAI and AMH by case-control status based on 1955 cases and 1642 controls. Variable Mean SD FAI 1.02 0.06 AMH 1.21 0.07 Age -0.19 0.01 BMI 0.04 0.01 OA 5.09 0.10

TABLE-US-00004 Table of weights of the PCOS risk derived based on weighted logistic regression using age as well as OA, FAI and AMH by case-control status based on 1955 cases and 1642 controls. Variable Mean SD FAI 1.10 0.05 AMH 1.17 0.07 Age -0.19 0.01 OA 5.11 0.09

[0400] The estimated PCOS risk probabilities resulting from weighted logistic regression using age and BMI as well as OA, FAI and AMH are displayed in FIG. 3 by case-control status. The histogram shows a clear separation of cases and controls with a high estimated risk for the cases and a low risk for the controls. Few subjects are considered as having a moderate risk of PCOS of around 50%. Dashed lines indicate thresholds to group females as low risk, moderate risk and high risk (from left to right) indicating 80% sensitivity and 80% specificity. Similar results were obtained weighted logistic regression using the three-and four-variable models (see Table 7).

Example 2: Investigation of Risk Thresholds

[0401] The PCOS risk derived by means of weighted logistic regression assigns the majority of cases with a high risk of having PCOS, whereas the controls are estimated to have a low risk (see Tables 5 and 6).

TABLE-US-00005 Table of PCOS risk score classification derived based on weighted logistic regression using age as well as OA, FAI and AMH by case-control status based on 1955 cases and 1642 controls. Status PCOS risk low moderate high Total control 1315 276 51 1642 case 27 365 1563 1955 Total 1342 641 1614 3597

TABLE-US-00006 Table of PCOS risk score classification derived based on weighted logistic regression using age and BMI as well as OA, FAI and AMH by case-control status based on 1955 cases and 1642 controls. Status PCOS risk low moderate high Total control 1315 279 48 1642 case 26 366 1563 1955 Total 1341 645 1611 3597

[0402] All in all the PCOS risk without BMI leads to comparable risk classification as the when including BMI into the PCOS risk.

[0403] The classification into risk groups based on the predictiveness curve results in risk thresholds of 7% for low risk (specificity ≥80%) and of 78% for high risk (sensitivity ≥ 80%). Based on these thresholds, the low risk group contains approximately 39%, the moderate group 16% and the high risk group 45% of the women based on 1955 cases and 1642 controls (FIGS. 5A and B).

TABLE-US-00007 Table of AUC (area under the curve) resulting from weighted logistic regression evaluated using 200 Monte-Carlo cross-validation runs for 1955 cases and 1642 controls based on different combinations of variables. *: applying a cut-off of 5.03 for AMH as suggested by Mahajan & Kaur 2019 AUC PCOS risk score (age, BMI, OA, FAI, and AMH) 0.98 PCOS risk score (age, OA, FAI, and AMH) 0.98 PCOS risk score (OA, FAI, and AMH) 0.97 OA 0.90 FAI 0.76 AMH 0.84 AMH + FAI 0.88 AMH + SHBG 0.87 AMH* 0.72 AMH* + FAI 0.76 AMH* + SHBG 0.82

Example 3: Evaluation of the PCOS Risk Score

[0404] The performance of the PCOS risk score was evaluated on a second independent sample set of 200 cases and 44 controls.

Controls

[0405] The controls consisted of 44 healthy women between aged 18 - 38 years without having PCOS. The median age was 25.5 years (standard deviation=5.02) and the majority had a normal body mass index (BMI, median=21.9 kg/m.sup.2, standard deviation=1.88). All women included in this control group had regular cycles based on information of menstrual cycles and/or cycle length. Serum Anti-Mullerian Hormone (AMH) was measured using Elecsys AMH Plus immunoassay. Hyperandrogenism (HA) was derived as the free androgen index (FAI) based on levels of serum testosterone (nmol/L) and serum sex hormone-binding globulin (SHBG) (nmol/L).

[00011]$\text{FAI}\text{=}\text{testosterone}\text{/}\text{SHBG * 100}$

Testosterone and SHBG levels to depict the FAI were determined by Elecsys Testosterone II (nmol/L) and Elecsys SHBG immunoassays (nmol/L) .

[0406] The three serum assays were measured from a serum sample taken at days 1-3 of the menstrual cycle.

Cases

[0407] 200 women diagnosed with PCOS aged 20 to 41 years and not using contraceptives were considered. Oligomenorrhea and/or anovulation (OA) was based on information of irregular menstrual cycles and/ or cycle length. Hyperandrogenism (HA) was derived as the free androgen index (FAI) based on levels of serum testosterone (nmol/L) and serum sex hormone-binding globulin (SHBG) (nmol/L):

[00012]$\text{FAI}\text{=}\text{testosterone}\text{/}\text{SHBG * 100}$

Patients were evaluated for PCOM by an ovarian volume ≥10 mL and/or an antral follicle count (AFC) above threshold based on transvaginal ultrasound examination. Serum Anti-Mullerian hormone (AMH) was measured using the Elecsys AMH Plus immunoassay.

[0408] The PCOS cases covered PCOS patients representing the four phenotypes according to the Rotterdam criteria (PCOS Consensus Workshop Group, Fertil Steril 2004;81:19-25).

[0409] The following table lists the baseline characteristics for the cases and controls.

TABLE-US-00008 Descriptive statistics of the variables of the PCOS risk score overall and for patients (cases) and controls of the independent second sample set All (N=244) Controls (N=44) Cases (N=200) PCOS Phenotype A (OA+, HA+, PCOM+) N=121 (49.6%) N=0 N=121 (60.5%) B (OA+, HA+, PCOM-) N=6 (2.5%) N=0 N=6 (3.0%) C (OA-, HA+, PCOM+) N=5 (2.0%) N=0 N=5 (2.5%) D (OA+, HA-, PCOM+) N=67 (27.5%) N=0 N=67 (33.5%) None N=1 (0.41%) N=0 N=1 (0.50%) OA* Regular Cycle N=48 (19.7%) N=42 (85.5%) N=6 (3.0%) OA N=196 (80.3%) N=2 (4.5%) N=194 (97.0%) FAI* Mean 5.16 1.30 6.01 SD 5.34 0.931 5.54 Median 3.52 0.966 4.64 Q1 ... Q3 1.67 ... 6.51 0.594 ... 1.65 2.40 ... 7.87 Min ... Max 0.128 ... 39.7 0.129 ... 3.58 0.318 ... 39.7 AMH (ng/mL) Mean 5.88 3.37 6.41 SD 3.71 1.74 3.81 Median 5.13 3.21 5.70 Q1 ... Q3 3.16... 7.56 2.02 ... 4.47 3.54 ... 8.41 Min ... Max 0.0455 ... 23.0 0.308 ... 7.48 0.0455 ... 23.0 Age (years) Mean 27.8 26.5 28.0 SD 4.84 5.02 4.78 Median 27.0 25.5 28.0 Q1 .. Q3 24.0 ... 31 0 23.0 ... 30.5 24.0...31.0 Min ... Max 18.0 ... 41.0 18.0 ... 36.0 20.0...41.0 BMI Mean 25.7 22.2 26.5 i SD 5.71 1.88 5.97 i Median 24.2 21.9 25.5 Q1 ... Q3 21.5 ... 29.0 21.1 ... 23.6 22.2... 30.4 Min ... Max 16.7 ... 45.3 18.2 ... 26.0 16.7 ... 45.3 Testosterone (nmol/L)* Mean 1.78 0.914 1.97 SD 0.931 0.393 0.907 Median 1.56 0.913 1.74 Q1 ...Q3 1.09 ... 2.37 0.598 ... 1.15 1.32 ... 2.58 Min ... Max 0.0870 ... 4.79 0.0870 ... 1.79 0.350 ... 4.79 SHBG (nmol/L)* Mean 52.3 68.6 44.3 SD 32.3 39.1 24.3 Median 45.7 60.8 38.2 Q1 ...Q3 29.4 ... 66.6 56.6 ... 111 27.0 ... 58.9 Min ... Max 8.30 ... 192 36.0 ... 192 8.30... 152 OA: Oligomenorrhea and/or anovulation FAI: Free Androgen Index; FAI = Testosterone / SHBG *100 BMI. Body Mass Index; BMI = weight (kg) / size (m).sup.2 AMH: Anti-Mullerian Hormone SHBG: Sex Hormone-Binding Globulin HA: HyperAndrogenism PCOM: PolyCystic Ovary Morphology

Sensitivity and Specificity of the PCOS Risk Score

[0410] Different thresholds were applied to the independent second sample set of 200 cases and 44 controls. Best results were achieved at fixed risk probability thresholds of 0.2 and 0.8 resulting in 96.0% sensitivity and 90.9% specificity (see FIG. 9).

[0411] Moreover, the following table lists the ROC area under the curve values (AUC) for different variable combinations applied to the independent second data set showing a superior performance of the PCOS risk score combination:

TABLE-US-00009 Table of AUC (area under the curve) resulting from weighted logistic regression evaluated using 200 MCCV runs and 100 repetitions for 200 cases and 44 controls using different variable combinations. *: applying a cut-off of 5.03 for AMH as suggested by Mahajan & Kaur 2019 Variable combination AUC PCOS risk score (age, BMI, OA, FAI, AMH) 0.99 OA only 0.96 FAI only 0.90 LH/FSH ratio only 0.85 LH only 0.76 FSH only 0.56 AMH only 0.76 AFC only 0.82 AMH + FAI 0.90 AMH + SHBG 0.90 AMH* 0.72 AMH*+ FAI 0.92 AMH* + SHBG 0.92 PCOS risk score w/o BMI 0.97

REFERENCES

[0412] Calzada et al. “AMH in combination with SHBG for the diagnosis of polycystic ovary syndrome”; J Obstet Gynaecol. 2019, 17:1-7.

[0413] Cho et al. “The biological variation of the LH/FSH ratio in normal women and those with Polycystic Ovarian Syndrome”; 2005 Endocrine Abstracts 2005; 9 P80.

[0414] Escobar-Morreale H. F., “Polycystic ovary syndrome: definition, aetiology, diagnosis and treatment”; Nature Reviews Endocrinology 2018; Vol 14, 270-284.

[0415] Indian et al. “Simplified 4-item criteria for polycystic ovary syndrome: A bridge too far?”; Clin. Endocrinol. (Oxf). 2018; doi: 10.1111/cen.13755.

[0416] International evidence-based guideline for the assessment and management of polycystic ovary syndrome 2018.

[0417] Mahajan, Nalini, & Kaur, Jasneet. 2019, Establishing an Anti-Muelledan hormone cut-off for diagnosis of polycystic, ovarian syndrome in women of reproductive age-bearing Indian ethnicity using the automated Anti-Muellerian hormone assay. Journal of Human Reproductive Sciences, 12(2), 104-113.

[0418] Malini and George “Evaluation of different ranges of LH:FSH ratios in polycystic ovarian syndrome (PCOS) - Clinical based case control study’ General and Comparative Endocrinology 2018; 260: 51-57

[0419] Mireya Calzada, Natividad Lopez, Jose A. Noguera, Jaime Mendiola, Ana I, Hernàndez, Shiana Corbalan, Maria Sanchez a Alberto M. Torres (2019): AMH in combination with SHBG for the diagnosis of polycystic ovary syndrome, Journal of Obstetrics and Gynaecology, JOURNAL OF OBSTETRICS AND GYNAECOLOGY

[0420] Nicholas et al. ,The utility of Anti-Müllerian Hormone in diagnosing Polycystic Ovary Syndrome amongst women presenting to an infertility clinic”; Hum Reprod. Abstract ESHRE 2014.

[0421] Nordenstrom A. and Falhammar H., ,,MANAGEMENT OF ENDOCRINE DISEASE: Diagnosis and management of the patient with non-classic CAH due to 21-hydroxylase deficiency.”; Eur J Endocrinol. 2018; pil: EJE-18-0712.R2. doi:10.1530/EJE-18-0712.

[0422] Pepe et al., “integrating the Predictiveness of a Marker with Its Performance as a Classifier”; Am J Epidemiol. 2008, 167(3): 362-368.

[0423] Pigny et al. ,,Comparative assessment of five serum antimüllerian hormone assays for the diagnosis of polycystic ovary syndrome.”: Fertil Steril. 2016; 105(4):1063-1069.e3.

[0424] R Core Team. 2015. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria (https://www.r-project.org/)

[0425] Sahmay et al. “Diagnosis of Polycystic Ovary Syndrome: AMH in combination with clinical symptoms”; J Assist Reprod Genet. 2014; 31(2): 213-220.

[0426] Scouten, W. H ., “A survey of enzyme coupling techniques.”; Methods in Enzymology 135:30-65. 1987.

[0427] Tijssen, P., Practice and theory of enzyme immunoassays, Elsevier Science Publishers B.V., Amsterdam (1990), the whole book, especially pages 43-78 and pages 108-115

[0428] Vermeulen, A., L. Verdonck, and J. Kaufman, A critical evaluation of simple methods for the estimation of free testosterone in serum. Journal of Clinical Endocrinology & Metabolism, 1999. 84(10): p. 3666-3672.

[0429] Xu. Qing-Song, & Liang, Yi-Zeng. 2001. Monte Carlo cross validation. Chemometrics and Intelligent Laboratory Systems, 56(1), 1-11.

[0430] The Rotterdam ESHRE/ASRM-Sponsored PCOS Consensus Workshop Group. Revised 2003 consensus on diagnostic criteria and long-term health risks related to polycystic ovary syndrome Fertil Steril 2004; 61:19-25.