Androgen testing for PCOS diagnosis
Polycystic ovary syndrome affects over 10% of all females and is now thought of as much more wide-ranging metabolic disorder than ‘just’ a disease of the female reproductive system. CLI chatted to James Hawley (Wythenshawe Hospital, Manchester University NHS Foundation Trust, Manchester, UK and Laboratory of Medical Science, Medical Research Council, London, UK) to find out more about the current thinking around the condition and how best to test for the androgens that drive it.
What is polycystic ovary syndrome (PCOS)?
Traditionally, PCOS – polycystic ovary syndrome – has been described as a common hormonal disorder affecting reproductive-age women, which is characterized by irregular/absent periods, high androgen levels and polycystic ovaries as well as other symptoms which include weight gain and patches of dark or oily skin. The pathophysiology is complicated. Historically, it has been thought of as condition resulting from ovarian malfunction, where higher levels of androgens are produced and follicles fail to mature (causing the polycystic appearance of the ovaries). Many patients also experience insulin resistance, this can contribute to other metabolic conditions such as diabetes mellitus. The genetic etiology remains unknown even though a family history of PCOS can be common.
As there is no cure, treatment involves managing the symptoms via lifestyle changes and medication to reduce insulin levels as well as medications to regulate the menstrual cycle and to induce ovulation to aid conception.
However, now, the way we think about PCOS is progressing and the terminology is evolving too. There’s a global drive to change the thinking behind PCOS and even rename the condition to better represent its etiology and complications. PCOS suggests that this is an ovarian-dominated condition, and that is simply not the case. The argument is that this does a disservice to the patient, as there are several complications associated with the PCOS that are either misrepresented or underrepresented with the name PCOS.
For example, patients have an adverse cardio–metabolic profile. They are more prone to cardiovascular disease and type 2 diabetes. Both of these significant health problems can be better managed with an earlier diagnosis, so there is an initiative to increase awareness of these complications. There is also the mental health aspect to consider – at present, it typically takes around two years from presentation to arriving at a diagnosis of PCOS; this can have a detrimental effect on patients’ mental health, causing increased anxiety and depression.
Accordingly, there is a push to standardize how we diagnose PCOS. There is increasing recognition that this is a significant condition affecting 10–12% of all females. As mentioned above, PCOS was traditionally always thought of as a condition that affects pre-menopausal females but that’s simply not the case. It’s also important to make a diagnosis in postmenopausal patients for the appropriate management and improvement of their cardio–metabolic health. We also have to be cognizant that the condition affects different ethnicities with different prevalences as well. A recent large study showed that whereas 8% of white females are affected that figure rises to approximately 16% of black females and then up to as high as 24% for Southeast Asians.
Healthcare professionals are trying to shift the paradigm to think of it not exclusively as an ovarian syndrome, but as a global metabolic condition that affects females of all ages that can be a significant health burden.
How is PCOS normally diagnosed?
There are three main criteria currently used to diagnose PCOS – evidence of ovarian dysfunction (irregular or no periods), ultrasound examination of ovarian morphology, and biochemical/clinical evidence of hyperandrogenemia. The presence of two out of these three is enough to satisfy the traditional, Rotterdam criteria for a diagnosis.
Whereas ovarian dysfunction (regular vs irregular menses) is reliant on a physician taking an accurate patient history, the assessment of ovarian morphology and the clinical/ biochemical assessment can be variable.
Ovarian morphology
Patients will often be referred for an ovarian ultrasound which will invariably detect cysts on the ovaries. Indeed, approximately 1 in 3 women will have cysts on their ovaries, a certain number of cysts and a certain diameter is required for them to be considered as consistent with PCOS.
Clinical/biochemical assessment
After taking the patient’s history, signs of hyperandrogenemia [which include hirsutism, acne or androgenic alopecia (male-pattern baldness), and, potentially, obesity] will be clinically assessed. Clinicians may also look to measure certain biochemical markers to help assess hyper-androgenemia, a hallmark of PCOS. Various guidelines, both internationally and nationally have been published throughout the years which aim help to standardize practice. However, we know this is not really the case and different laboratories offer different profiles and different investigations. Therefore, a lot of work still needs to be done to help to standardize and promote this in the UK as well as worldwide.
The concentrations of which androgens are routinely measured and what are the challenges?
In the biochemical work-up of PCOS, we are looking for evidence of hyperandrogenism. This can be demonstrated by elevated concentrations of total testosterone and androstenedione (the precursor to testosterone and itself androgenic). Androstenedione will interact with the androgen receptors and drive hyper-androgenism, but to a lesser extent than to testosterone.
Total testosterone, free androgen index and calculated free testosterone
We can also determine the free androgen index, which is a ratio of total testosterone to it’s binding globulin, sex hormone binding globulin (SHBG). Typically, around 2–3% of testosterone circulates free in the circulation, which is the physiologically active quotient of testosterone, and in an ideal world that is what we would measure. However, measurement of free hormone is very challenging and requires a specialist approach. Most assays are designed to measure total testosterone, which liberates testosterone from its binding globulins (SHBG) and any other proteins that it’s loosely bound to, like albumin. So, we measure the total serum or total plasma testosterone and then we can take the ratio of that to SHBG, which is the main binding globulin and binds 90–95% of testosterone; 5% of the rest is loosely bound to albumin. Then we can take a ratio of this which provides us with the free androgen index, which can be a useful marker to try to identify how much free testosterone is available.
Instead of free androgen index, you can also determine calculated free testosterone, which is probably a better indicator of the actual circulating free testosterone level. This takes into account measurements of SHBG, a measured (or standardized) albumin and total testosterone. These numbers then go through a calculation, for which several have been proposed, which then gives you an idea of how much free circulating testosterone is available. However, in general, fewer labs use this method than the free androgen index method as it’s more complicated to build the equations into the middleware systems. Additionally, there is still unfamiliarity in the medical community with reporting calculated free testosterone, but it is superior to the free androgen index as it takes into account the binding affinity of SHBG for testosterone.
Determining testosterone concentrations by mass spectrometry provides greater accuracy and specificity (Photo credit: J. Hawley)
Androstenedione measurement
At present, many labs do not routinely measure androstenedione despite guidance advocating that if total testosterone is normal, androstenedione should be measured in patients that are suspected of having PCOS. Measurement of androstenedione is important as approximately 1/3 of patients with PCOS will have a normal testosterone concentration but an elevated androstenedione concentration which drives the hyperandrogenemia. Hence, it is important to measure androstenedione concentrations, unfortunately this does not routinely happen in the UK. Thus, increasing the awareness and making sure people adhere to the guidance is important.
Assay methodology
It is also important to understand that not all laboratories offer the same profile of measurements. For example, not everywhere has the ability to measure androstenedione. Most will do total testosterone as a standard, and that is usually measured by immunoassay. However, some labs will use mass spectrometry (MS), which is a superior technology – it is more specific, it’s more sensitive and we can be more confident that our testosterone concentration results are not influenced by any interferences. Immunoassay is more prone to interferences, for example from cross-reactions with other steroid hormones or medication that the patient may be on, such as the oral contraceptive pill (OCP) which can contain norethisterone which can cross-react with testosterone assays and so give a falsely elevated testosterone concentrations. Some labs adopt a tiered approach whereby an initial testosterone measurement is completed by immunoassay followed by confirmatory testing using MS if the result is elevated.
Differential diagnoses
Before making a diagnosis of PCOS, it is important to exclude other diagnoses and to exclude other endocrinopathies, such as hypo-thyroidism or hyperprolactinemia and Cushing’s syndrome if the presentation was suggestive of hypercortisolism. An important condition to rule out is late-onset congenital adrenal hyperplasia (CAH), which can present similarly to PCOS. Late-onset CAH is an attenuated form of classic CAH, where patients may struggle to conceive. They may have irregular periods and other symptoms of hyperandrogenemia, but this diagnosis relies on the measurement of 17-hydroxyprogesterone (17-OHP), a progestin that, again, we know is not commonly measured in the UK for the work-up of PCOS. The incidence of late-onset CAH is quite variable depending on the demographic – it ranges from one in 100 patients to one in 1000 patients, so it’s not uncommon when you think of the numbers involved with PCOS.
Additionally, it is important to realize that not all of these hormones are adrenally derived, testosterone and androstenedione are also synthesized in the ovaries. So, if there is ambiguity over the source of the hyperandrogenemia, we can also measure concentrations of dehydroepiandrosterone sulfate (DHEAS). This is a specifically produced by the adrenal glands so measuring it can provide a good idea of whether or not the testosterone is adrenal or ovarian in origin.
Why is MS useful for measuring androgen concentrations and what tips would you give?
Most guidance now advocates using MS as the first-line test for investigating testosterone levels and so we should really be adhering to that. Unfortunately, many labs either don’t have access to MS or the funding is such that it’s prohibitive to screen everybody using MS. Don’t forget these are large numbers – a lot of patients are investigated for this condition and it’s likely under investigated as well. So, in a publicly funded health service, we have to be pragmatic, which is why some places will employ multi-tiered approaches to diagnosing PCOS. Generally, the theory is that immunoassays – which rely on antibody–antigen reactions – are more prone to over-recovery, and so over-diagnose the condition because of cross-reactivity with other steroids. As mentioned, in the case of finding elevated testosterone on immunoassay, some labs will then usually send these samples away for confirmation testing using MS. MS is more specific than immunoassay, as it relies on the mass-to-charge ratio of the steroid hormone. We know the molecular weight of testosterone is 288 (289 after ionization) so we can set the quadrupole of the mass spectrometer to detect 289, which will filter out everything else. In the collision cell of the mass spectrometer, the molecules are bombarded with collision gas which will cause fragmentation, typically for testosterone, into fragments of molecular weight 97 and 109. We can program the mass spectrometer to scan for 289 for testosterone and then for the 97 and/or 109 fragments. Ideally, we will detect both fragments, so we’ve got an additional layer of specificity where we can look at the ratio of these results to make sure that there is no interference. Also, we test for an internal standard, which helps to control for losses either in the extraction of testosterone from the matrix itself or in the analysis. Internal standards can either be deuterated testosterone or carbon-13 labelled testosterone. 13C-labelled testosterone is readily available now and is not significantly more expensive than the deuterated forms, so 13C internal standards should be used where possible as these have the same physico-chemical properties as testosterone. This means it will elute at the same time on the chromatogram and we can then be sure that the chromatography has not been prone to any kind of deleterious effects, such as ion suppression, which is the Achilles heel of MS methods.
Another great advantage of MS is that we’re able to multiplex the methods. This means that we have the flexibility to look for multiple steroids in one analyses, whereas using immunoassay you have to use separate assays for each analyte – testosterone, androstenedione, etc. With MS, we can now build methods that measure testosterone, androstenedione and 17-HP into the same method so we can have one analysis sample for a targeted PCOS assessment, this makes using MS much more cost-effective.
Analytical range of testosterone measurement
When measuring testosterone, we have to cover a larger linear range of concentrations as we need to be able to detect the lower so-called ‘female’ end and the higher ‘male’ end of the concentration spectrum. We have assays that cover the analytical range from low concentrations for samples, 0.1–0.2 nanomoles per litre for females, and we’ll tend to go up to 40–50 nanomoles per litre for ‘male’ testosterone. There’s always great debate over what’s an appropriate reference range or normative range for female patients and a lot of work has been done over the years and to try and answer this question. Currently, we tend to think the normal range for most adult, pre-menopausal females will be around 0.3–1.9 nanomoles per litre.
We are now seeing more testosterone prescriptions for perimenopausal and menopausal female patients and there are undoubtedly benefits of this, such as improved bone density and estradiol production. As a consequence, we are now seeing higher concentrations of testosterone in these patients and it’s important to remember that patients need to have not only their testosterone levels monitored but also, checks on their full blood count. If patients are over-prescribed or over-administer testosterone, there is a risk of polycythemia, which makes the blood more viscous, there are complications associated with it, such as increased risk of clots, stroke or heart attack.
Mass spectrometer set-up
In terms of the mass-spectrometer set-up, any MS method can be broadly separated into three different stages: the sample preparation, the chromatography and detection in the mass spectrometer.
Testosterone is one of several androgens that can be tested for in PCOS diagnosis (Adobe Stock)
Sample preparation
Sample preparation – usually the purification of testosterone or similar steroids from the matrix or the sample of interest is a crucial stage to get right. The most common samples that we see here are serum and plasma and they have their own challenges. They are highly proteinaceous and contain salt, and trying to extract something that’s in nanomolar concentrations from a small volume of sample can be challenging. There’s always a balance to be struck – the more sample that you have, in general, the ‘dirtier’ the extraction will be. So you can have various options with the extraction. One option, used in other analyte scenarios is colloquially termed ‘dilute and shoot’ where the sample is massively diluted and injected into the mass spectrometer. This is not usually a great idea for analytes like testosterone, as even if the sample looks clear to the eye, it will still be salty and proteinaceous, which would not be good for the mass spectrometer.
We have a number of options for cleaning up the sample. The fastest one is to precipitate (and so remove) the larger proteins using zinc sulfate. Then, we’ve got an organic additive which can contain the internal standard and will precipitate more proteins leaving an extract that looks quite clear by eye but there is still a lot present that we can’t see. It is an improvement on dilute and shoot and it’s the quickest way that we can prepare a sample for analysis.
Further to that, another option is a liquid–liquid extraction, where we mix the sample with a solvent that is immiscible with the serum or the plasma. As testosterone is hydrophobic, it will migrate into the solvent rom the remaining serum components giving a pretty clean extract. There’s equipment now that can be used to semi-automate these extractions, which is advantageous as it reduces inter-operator variability and helps to prepare the sample.
Moving up the scale, there are some of the more expensive options, such as a supported liquid extraction (SLE), which can be a synthetic or an organic diatomaceous earth bed that the sample is loaded on to. This, again, will remove a lot of the salts, proteins and phospholipids. When the testosterone is eluted from the bed with solvent it is very ‘clean’, which in this context means that the sample is usually free from ion suppression.
After that there is solid phase extraction. Here the sample is loaded onto a bed of solid (stationary) phase material which can be subjected to more rigorous wash steps (such as 30–40% methanol) to try to remove contaminants without removing the testosterone, so we can clean the matrix up sufficiently so that it’s conducive to being injected onto the mass spectrometer.
Chromatography
There are various columns available now where we can look at different retentive mechanisms. We’ve got the classic C18 column, which is a hydrocarbon chain that provides hydrophobicity, which testosterone (itself hydrophobic) will stick to in the right conditions. Therefore, we can develop a method to be specific for testosterone, where we wash off anything we need to remove before we elute the testosterone. We need to be careful about any isobaric interferences, like epitestosterone (used in sports doping) and to make sure we’ve got separation from this and other isobars. The structure of testosterone is based on various aromatic rings, so we can also use pentafluorophenol (PFP) HPLC columns or biphenyl columns, which give a different type of selectivity and allow the separation of challenging steroids from similar isobaric steroids.
Sample preparation and chromatography are the big things that you have to get right. If you develop what looks like a very good MS method using clean samples and/or non-patient samples, it can seem to be working well until you introduce an actual patient sample and then you see ion suppression effects and low counts where before you may have observed high counts in standards. This can compromise the whole assay with either your recovery or your limit of quantitation. Ion suppression can be caused by certain medication that the patient might be taking, or plasticizers in the sample collection tube. Trying to make robust methods that are not prone to ion suppression can be challenging.
Mass spectrometer
The mass spectrometer is a detector really, so at this point we are trying to optimize the detection. As long as the sample preparation and the chromatography has been done well, we should be able to optimize detection.
What are the future perspectives on this topic?
There are probably two major areas of development: saliva testing and detection of 11-oxygenated androgens.
Saliva testing
As we’ve mentioned earlier, looking at free (rather than bound) hormone is always attractive because that gives you an idea of what’s actually physiologically active rather than just the amount of total hormone present, and it’s just not complicated by things like binding proteins. In saliva, therefore, we can determine the concentration of free testosterone, this has shown good promise. However, there’s a paucity of data available on this. In addition, saliva testing is non-invasive as well, which is usually appealing to patients. One of the great advantages of MS analysis is that we can do multiplex testing, where in one analysis we can determine the concentrations of testosterone and the other steroid hormones. However, this is challenging as we are trying to detect concentrations that are another order of magnitude lower – moving from nanomoles per litre into picomoles per litre to try to get down to that sensitivity. It’s certainly helpful that saliva is ‘cleaner’ than serum; it’s not as proteinaceous and it’s not salty, but it is still challenging. Really you need a high-end mass spectrometer to achieve the sensitivities required and you need to pay attention to detail with the sample preparation and the chromatography, which we briefly touched on earlier..
Detection of 11-oxygenated androgens
Another emerging area of interest is the 11-oxygenated androgens, particularly, for example, 11-ketotestosterone. This has been shown to be as potent as testosterone and it’s got the same avidity for the androgen receptor as testosterone. Currently, there are several studies looking at the utility of 11-ketotestosterone (11-KT) in the diagnosis of hyperandrogenemia conditions, such as PCOS. Wider availability of 11-KT testing could be very interesting, and it can also be tested for in saliva, so it’s a very interesting avenue to pursue. 11-KT is usually an adrenally derived steroid that is made from 11β-hydroxyandrostenedione which is peripherally metabolized to 11-KT, so this could also help guide whether the androgens are ovarian or adrenal in origin.
These are exciting future developments. Also, further down the line, there may be some machine learning algorithms that could help provide a more objective interpretation of data to help reduce variability in result interpretation, but I think this is a way off.
The interviewee
James Hawley MSc FRCPath
Principal Clinical Scientist
Wythenshawe Hospital, Manchester University NHS Foundation Trust, Manchester, UK and Laboratory of Medical Science, Medical Research Council, London, UK
Email: james.hawley@mft.nhs.uk
Bibliography
1. Elhassan YS, Hawley JM, Cussen L, Abbara A, Clarke SA et al. Society for Endocrinology Clinical Practice Guideline for the Evaluation of Androgen Excess in Women. Clin Endocrinol (Oxf). 2025;103(4):540–566 (https://doi.org/10.1111/cen.15265).
2. Teede HJ, Tay CT, Laven JJE, Dokras A, Moran LJ et al; International PCOS Network. Recommendations from the 2023 international evidence-based guideline for the assessment and management of polycystic ovary syndrome. Eur J Endocrinol. 2023;189(2):G43–G64 (https://doi.org/10.1093/ejendo/lvad096).
3. Bizuneh AD, Joham AE, Teede H, Mousa A, Earnest A, Hawley JM et al. Evaluating the diagnostic accuracy of androgen measurement in polycystic ovary syndrome: a systematic review and diagnostic meta-analysis to inform evidence-based guidelines. Hum Reprod Update. 2025;31(1):48–63 (https://doi.org/10.1093/humupd/dmae028).
