Variability and quantification of serum medroxyprogesterone acetate levels

Abstract

Quantification of serum progestin levels in clinical contraceptive studies is now routinely performed to understand progestin pharmacokinetics and to correct for unreliable self-reporting of contraceptive use by study participants. Many such studies are focussed on the three-monthly progestin-only intramuscular (IM) injectable contraceptive depot medroxyprogesterone acetate (DMPA-IM). Methods commonly used to measure serum MPA levels include liquid chromatography coupled to mass spectrometry (LC/MS) and radioimmunoassay (RIA); however, RIA methods have not been used in recent years. We review the available literature and find that these methods vary widely in terms of use of organic solvent extraction, use of derivitization and choice of organic solvent and chromatography columns. There is a lack of standardization of LC/MS methodology, including a lack of detailed extraction protocols. Limited evidence suggests that RIA, without organic solvent extraction, likely over-estimates progestin levels. Maximum MPA concentrations in the first two weeks post-injection show wide inter-individual and inter-study variation, regardless of quantification method used. Standardization of quantification methods and sampling time post-injection is required to improve interpretation of clinical data, in particular the side effects arising at different times depending on the pharmacokinetic profile unique to injectable contraceptives.

1. Introduction

The three-monthly progestin-only intramuscular (IM) injectable contraceptive, depot-medroxyprogesterone acetate-IM (DMPA-IM), administered as 150 mg MPA in a one mL suspension, is the most commonly used hormonal contraceptive in sub-Saharan Africa [1], the region with the highest burden of HIV/AIDS [2, 3]. Clinical observational studies, animal and in vitro studies together suggest that DMPA-IM use may have side-effects including increased risk of HIV-1 and other sexually transmitted infections, altered vaginal microbiome and inflammation in the female genital tract (FGT) and thinning of FGT epithelium [4–10]. Several plausible biological mechanisms may explain these effects on HIV-1 acquisition [5]. Importantly, the effects of DMPA-IM are concentration-dependent [11–15]. The pharmacokinetic profile of DMPA-IM and other injectable contraceptives is characterized by a sharp increase in concentration in the first few weeks, followed by a slow decline to a steady low plateau concentration maintained over the 3 months prior to the following injection [16]. Thus, the effects on HIV-1 acquisition risk may be greater during the initial peak concentration (C_max) than during the second and third months when MPA concentrations are lower. Given the widespread use of DMPA-IM as well its association with several side-effects, much of the literature on HIV-1 and contraceptives has focussed on serum concentrations of MPA.

Recent clinical studies have shown that there is often discordance between the self-reported use of a particular hormonal contraceptive and its detection in blood serum or plasma, suggesting that self-reporting of contraceptive use is unreliable [17–24]. Therefore, the objective quantification of circulating levels of MPA and other progestins is becoming a routine procedure in clinical studies of contraceptives. While LC/MS is the current method of choice for quantifying MPA levels in serum, LC/MS methodology varies widely in the literature. Here we discuss the commonly-used chromatographic, mass spectrometry and RIA methods for quantification of MPA, and potential sources of variability related to methodology with a view to inform on the importance of standardizing choice of methodology, extraction protocols and sampling times. We also discuss inter-individual and inter-study variation in reported MPA C_max values.

2. Quantification methods for measuring serum MPA levels

Traditionally, serum MPA and other progestin levels were measured by RIA, either with (conventional RIA) or without (direct RIA) prior organic solvent extraction [25–30]. Since RIA is labour-intensive, chromatographic and spectrometry methods, which employ a prior organic solvent extraction step, have become more commonly used in recent years. Progestin levels can be measured by various liquid chromatographic (LC) methods such as high-performance liquid chromatography (HPLC) [31, 32] or ultra-performance liquid chromatography (UPLC) [17, 33], either alone or coupled to a mass spectrometry instrument with a single or two mass analysers (LC/MS or LC-MS/MS). Gas chromatography (GC) or GC coupled to mass spectrometry (GC/MS) are also used spectrometry methods [34–37]. The sensitivity and specificity of these methods vary due to variations in methodological approaches, as discussed below, as well as in the specific compound being measured. It is therefore difficult and beyond the scope of this review to give a general assessment of the sensitivity and specificity of the methods. However, based on the very limited available literature for MPA, UPLC/MS and HPLC/MS appear to be the most sensitive (~20–25 pg/mL) [18, 33, 38, 39], followed by conventional RIA (25 – 200 pg/mL) [40, 41] and GC/MS (0.5 ng/mL – 0.001 μg/mL) [34, 36].

Variations in methodological approaches within and between these techniques can result in differences in determined concentrations across studies. These include but are not limited to use of different antisera, with variable degrees of cross-reactivity with progestin metabolites, for RIA, and use of different chromatographic columns, instruments and protocols for chromatography and spectrometry methods. Whether inter-study variations in reported serum progestin levels are real or due to different methods of quantification and/or steps prior to quantification is unclear. These issues will be discussed below.

2.1. Organic solvent extraction in RIA

Organic solvents are used to extract or remove steroids bound to proteins, including corticosteroid binding globulin (CBG), sex hormone-binding globulin (SHBG) and albumen. Extraction is also used to remove other interfering proteins, less hydrophobic steroid metabolites and conjugates, [42] and to concentrate the steroid of interest. Deconjugation or hydrolysis of conjugated steroids are important for rendering the steroid of interest volatile and thermally stable for accurate measurement [43]. Although MPA does not bind CBG or SHBG, it does bind albumin [44, 45]. Whether organic solvent extraction is performed, and the organic solvent of choice, may therefore result in inter-study variability [46]. While some studies have reported that values obtained for MPA using conventional RIA (extracted) are 5–10 times lower than in other studies using direct (non-extracted) RIA [25, 40, 41, 47], these studies did not compare these methods in parallel on the same samples, making it difficult to draw conclusions. There is to our knowledge only one study that directly compared the impact of extraction on the same samples. In this report, non-extracted serum MPA levels were 2–3 times higher by direct RIA than when extracted using diethyl ether before RIA [28]. For conventional RIA, it is difficult to deduce the importance of different organic solvents based on limited evidence from the literature. The abovementioned study showed no difference in MPA levels when using two solvents with different polarity (benzene:iso-octane and diethyl ether) for extraction prior to RIA [28]. However, the chosen organic solvent may improve selectivity; for example, petroleum ether selectively extracts progesterone (P₄) rather than corticosteroids [48]. Taken together, these studies suggest that for MPA, the addition of an organic solvent extraction step makes a greater difference than choice of organic solvent when using RIA.

2.2. RIA versus chromatographic methods (GC, GC/MS, LC/MS, UPLC-MS/MS)

Only a few studies directly compare levels obtained by RIA to those obtained by GC, GC/MS or LC/MS, on the same samples. In one such study, MPA levels from the same samples were 5–10 times greater using direct RIA (without extraction) as compared to GC with extraction [36]. Another study estimated that mean MPA C_max from the same samples measured by direct RIA (without extraction) was about 5 times greater than those measured by LC-MS with hexane extraction [49]. The authors suggest the discrepancy was caused by cross-reacting metabolites not removed due to no extraction before RIA, thereby over-estimating MPA levels, although they did not measure the cross-reaction of MPA metabolites in their study. Thus in the above studies the reason for the discrepancies between values are unclear. In a recent study, the cross-reacting metabolites of estradiol (E₂) did not account for the higher serum E₂ levels observed with conventional RIA compared to LC-MS/MS; however, this metabolite effect may be steroid-specific [50].

Two studies have directly compared conventional RIA (with extraction) to chromatographic methods on the same samples. In one study, MPA serum levels measured by conventional RIA with petroleum ether extraction were 16–29% higher than GC/MS after petroleum ether extraction [51], while in the second study MPA values measured by GC/MS were 13–92% lower than conventional RIA with petroleum ether extraction [52]. While comparisons with MPA are limited, there are some studies comparing RIA and LC/MS to quantify other steroids from the same samples. In one study, levels of the endogenous androgens testosterone, androstenedione and dehydroepiandrosterone were comparable between a diethyl ether extraction before RIA and LC/MS with a hexane:dichloromethane extraction on the same samples, done in parallel [53]. In another study, serum levels of the progestin levonorgestrel were 20–25% higher when measured by conventional RIA than LC/MS, despite the use of a chromatographic separation step and highly specific antiserum for RIA [54–56].

There are also some examples of similar serum MPA levels reported between conventional RIA and chromatographic methods in different studies, using different samples. For instance, in breast cancer patients, similar plasma MPA levels were reported between one study after 500 mg orally administered MPA when measured using hexane extraction and UPLC [32] and in another study after 400 mg oral MPA using conventional RIA with petroleum ether extraction [57]. However, differences in study design such as oral dose, sampling time and cancer origin make it difficult to compare findings between different studies.

In summary, if both samples are extracted prior to analysis, there appears to be a 0.25- or 0.9-fold difference in steroid serum concentrations determined by RIA versus LC/MS or RIA vs GC/MS, respectively. Taken together, these studies suggest that differences in specificity between RIA and chromatographic methods reside mostly in whether or not prior extraction is performed, provided a highly specific antibody is used for RIA.

2.3. Differences between chromatographic methods (GC/MS, LC/MS, HPLC)

GC/MS for steroid hormones was first introduced in the early 1960s, followed by RIA methods in the late 1960s [46]. GC/MS has the advantage of improved sensitivity to measure only the particular steroid of interest, even at low levels, and in smaller sample volumes [46]. However, GC/MS is a costly and low-throughput method, due to the long sample process time, extraction and derivatization that is required due to the non-polarity of steroids when using GC [58]. More recently developed LC/MS methods have the benefit of shorter sample preparation protocols, shorter run times and the ability to multiplex and measure levels of several progestins simultaneously [46, 59–61]. If one compares chromatographic methods between different studies, using different samples, similar serum levels after 25 mg IM MPA injection were reported using LC/MS [62] and GC/MS [63].

Differences in extraction protocols between methods involving chromatographic procedures might contribute to variation across studies of MPA. Detailed extraction protocols, including organic solvent choice and columns used, are generally lacking. Differences in extraction methods include varying starting volumes of serum, use of different extraction solvents and different extraction phases (liquid-liquid, solid-phase or solid-liquid extraction), as well as the volume and solvent used to reconstitute samples prior to detection [21, 64]. Studies with inefficient extraction would not detect accurate levels of MPA. Inefficient extraction not only refers to sub-optimal extraction of the steroid, i.e. not extracting all the steroid out of the serum or losing steroid during the extraction process, but also to matrix effects, where other compounds from the serum are co-extracted and in turn influence the ionisation of the measured steroid. Ion suppression or ion enhancement would result in lower or higher observed concentrations, respectively. For more detailed discussion on extraction method challenges and matrix effects, please refer to a specialized review on the matter [65]. Efficient extraction and any potential procedural losses are calculated and corrected for making use of internal standards. For LC-MS, GC-MS, and HPLC-MS, deuterated analogues of the specific analytes are typically used. These deuterated analogues should have similar extraction recovery and matrix effects. For LC-, GC- and HPLC-MS assays, they should also have similar ionization response in electrospray ionization and co-elute with the analyte [66]. However, not all deuterated analogues are equally as well suited and choice of deuterated analogue can have a significant effect on the results obtained [67, 68]. All of the above would influence the percentage bias/accuracy between the real concentration and the observed concentration. The mobile phases as well as instrument parameters and the chromatography column used are additional possible sources of variation in specificity. The available data are insufficient to draw firm conclusions about the differences in accuracy among methods. This highlights the need for standardized methodologies, analysis and specifications for extraction solvent choice for MPA as well as for other steroids [46, 69].

3. Variability in serum MPA levels

Within studies measuring the same progestin by the same method of quantification, substantial inter-individual differences occur in reported serum concentration ranges. For example, there is 4-fold variability for MPA by LC/MS [70], 12-fold for MPA by GC/MS [51] and 38-fold for MPA by conventional RIA [71]. This suggests that the variation could be explained by inter-individual biological differences and may not only depend on the method of quantification. We therefore carried out an extensive PubMed search of reported C_max MPA serum concentrations, for different doses and methods of MPA administration (IM injection, subcutaneous injection, oral). This table is available at Mendeley Data (https://data.mendeley.com/datasets/5sck77c9b9/3) [72]. Taken together, these publications show that there is wide inter-individual and inter-study variability in reported MPA serum levels. For DMPA-IM, the majority of publications show a range of C_max values from 2.6–30 nM [28, 47, 73–77], while three publications show higher upper peak levels ranging from 60–100 nM [75, 76, 78]. A plateau concentration of about 2.6 nM (1 ng/mL) is maintained for 2–3 months [16, 30, 79, 80].

There is also a wide range of reported t_max, the time at which C_max is reached. While several studies report a C_max at 1 week post-injection [28, 78, 81–83], this time point was the first sampling time post-injection. Only a few studies measured frequently within the first few weeks following injection and reported C_max earlier than 7 days [30, 43, 69, 71, 80, 85, 87]: for example, in one study, serum was sampled daily for 2 weeks and a C_max was observed at 2–4 days post-injection [30]; in another study, serum was sampled every 2 days for 2.5 weeks and a C_max was observed at 3–5 days [75]. Taken together this suggests that C_max is reached by 2–7 days, and that some studies may underreport C_max levels due to later sampling times.

In order to depict differences and similarities in reported DMPA-IM C_max (after 150 mg IM dose), we compiled a graph of the published peak serum concentrations from 18 studies (in ng/mL and nM) over the first 30 days post-injection (Figure 1). This information was taken from studies where C_max was reported at any time point during the first month, even if the first time point was at 30 days. The graph shows that the highest values are obtained around 2–7 days post-injection. However, since some studies did not measure levels at 2–7 days, reported C_max depended on the sampling time and whether MPA levels were measured at frequent time points within the first month post-injection. Studies where the first sample is only taken at 2 weeks or 1 month are likely to have missed measurement of C_max. These data suggest that appropriate sampling time is essential for determining C_max. The wide ranges of C_max in the graph indicate that there is high inter- and intra-study variation, which cannot be explained by different quantification methods (i.e. RIA versus LC/MS or GC/MS), since overlapping ranges of C_max are shown for all quantification methods at Day 7. There is also far less variation in plateau ranges of MPA after 1–3 months, despite methodological differences [16, 17, 21, 28, 30, 71, 74, 81, 82, 84–91].

Figure 1. Reported MPA serum levels in the first 30 days following 150 mg of DMPA-IM.

The serum concentrations and time points at which they were measured across 30 days are summarized from 18 published studies, each represented by a different colour. Where numerical values were not reported, these were read off the graph for several studies [30, 41, 73–75, 77, 79, 81]. For these studies, the points in this figure are representative of concentrations that could be clearly read off the graph, and not all points are plotted. Some studies measured serum MPA levels at time points later than 30 days (not shown). Points are either: mean (single dot), mean + standard deviation (SD) and mean – SD (2 dots) or range of C_max (2 dots), as indicated in the legend table. Method of quantification, sampling times and number of women in the study are indicated in the legend table.

Symbol	Frequency of sampling post-injection	Number of women measured	Reference
Conventional RIA
	Sampled every 2–3 days#	2	[41]
	Sampled every day for 2 weeks and later time points#	3	[30]
	Sampling times not mentioned#	10	[28]
	Sampled twice weekly (first sample at Day 2) *	5	[95]
	Sampled weekly (first sample at Week 1)#	20	[74]
	Sampled prior to dose, at Day 14 and prior to next dose*	9	[86]
	Sampled twice weekly$	5	[79]
Direct RIA
	Sampled every 2–3 days (first sample at Day 3)#	3	[75]
	Sampled weekly (first sample at Week 1)#	8	[73]
	Sampled at Day 3, Week 4, Week 13, Week 26*	9	[91]
RIA, extraction unknown
	Sampled weekly for 2 weeks, then every 4 weeks#	6	[76]
	Sampled at 1 month and 3 months*	16	[88]
	Sampled at Week 1, 2, 4, 6#	10	[78]
LC/MS
	Sampled every 2 weeks (first sample at Week 2)#	30	[96]
	Sampled weekly (first sample at Week 1)#	29	[81]
	Sampled every 4 weeks#	61	[89]
GC/MS
	Sampled at Day 2, 5, 32 and later time points#	6	[77]
	Sampled every 2–3 days (first sample at Day 3) – measured after 1 injection*	9	[83]
	Sampled every 2–3 days (first sample at Day 3) – measured after 8 injections*	9	[83]

SD: standard deviation

^*

values on the graph represent mean – SD, mean + SD

^#

values on the graph represent range (min-max)

^$

values on the graph represent mean

Where organic solvent extraction is unknown, access to the publication was restricted and information was obtained from the abstract only

4. Conclusion

The wide variation in quantification methods used for serum progestins contributes to the large inter-study variation reported for serum DMPA-IM C_max levels in the literature. While there is some evidence that direct RIA, i.e. without prior organic solvent extraction, may over-estimate MPA levels, there are too few studies directly comparing direct RIA versus LC/MS on the same samples, making it difficult to draw the conclusion that RIA methods are inaccurate. It should also not be assumed that concentrations measured by conventional RIA are incorrect or artificially high unless determined on the same samples in parallel by both conventional RIA and either GC/MS or LC/MS. Current studies would benefit from reporting clear and detailed extraction methods. These issues highlight the need for standardization of quantification methods [46, 69] similar to the testosterone standardization project implemented by the Centers for Disease Control and Prevention, National Center for Environmental Health, Division of Laboratory Sciences (CDC/NCEH/DLS) [92]. These international standardization guidelines should require that specific test sensitivity, accuracy and percentage bias levels, as well as established reference ranges for the measurement of progestins should be established. Additionally, there should be standardized sampling times and defined sampling frequencies for future clinical studies involving pharmacokinetics of DMPA-IM. Standardized methodology and sampling times will facilitate comparisons between studies.

Our analysis of published MPA serum C_max levels measured by different quantification methods indicates that inter-study variability does not only depend on quantification method. Variability is likely to depend on a number of additional factors including sample size, whether prior organic solvent extraction occurred during quantification, and organic solvent choice. Furthermore, differences in reported C_max clearly depend on when the sample was taken. With these limitations in mind, taken together the data suggest that C_max serum levels appear to be reached within the first week post injection, with day 7 showing the highest concentrations with a range of 2.6–99.6 nM. These concentrations then decline to a less variable plateau concentration of around 2.6 nM which is maintained until time of next injection. Differences in quantification method appear to have a greater effect on variability in serum MPA levels for C_max than for the plateau levels. However, large inter-individual variation is observed not only for peak levels of MPA but for all contraceptive progestins [93]. Importantly, these studies all measure MPA in serum or plasma but there is limited or no data on the progestin levels reached in target tissues of the FGT [94], which are also likely to show large inter-individual variation. More research on MPA levels in FGT tissues is urgently needed, since the concentration-dependent effects of MPA are likely to be relevant at the site of heterosexual HIV-1 infection.

Abbreviations

MPA

medroxyprogesterone acetate

DMPA-IM

intramuscular depot medroxyprogesterone acetate

C_max

maximum or peak concentration

LC/MS

liquid chromatography/mass spectrometry

GC/MS

gas chromatography/mass spectrometry

RIA

radioimmunoassay

FGT

female genital tract

HPLC

high performance liquid chromatography

UPLC

ultra-performance liquid chromatography

CBG

corticosteroid binding globulin

SHBG

sex hormone-binding globulin

E₂

estradiol

P₄

progesterone