Link Search Menu Expand Document

An Informal Meta-Analysis of Estradiol Curves with Injectable Estradiol Preparations

By Aly W. | First published July 16, 2021 | Last modified July 24, 2021

Abstract / TL;DR

Injectable estradiol preparations such as estradiol valerate and estradiol cypionate in oil are frequently used as estrogens in transfeminine hormone therapy. However, there is little characterization of these preparations in transfeminine people and dosing recommendations by transgender health guidelines appear to be based on expert opinion rather than on clinical data. To help shed light on the properties of injectable estradiol and to better inform dosing considerations in transfeminine people, an informal meta-analysis of available clinical data on estradiol concentration–time curves with major injectable estradiol formulations was conducted. The included preparations were injectable estradiol benzoate in oil, estradiol valerate in oil, estradiol cypionate both in oil and as a suspension, estradiol enanthate in oil, estradiol undecylate in oil, and polyestradiol phosphate. The literature was searched for clinical concentration–time data with these injectable estradiol esters and these data were collected and analyzed. Meta-analysis consisted of data for each injectable estradiol preparation being processed and fit with pharmacokinetic models. Selected pharmacokinetic parameters were additionally determined and reported. The results of this work were discussed with regard to characteristics of injectable estradiol preparations like curve shapes, durations, estrogenic exposure, and variability between people and studies. Recommendations for injectable estradiol preparations by transgender health guidelines were also explored in light of the present results. Current guidelines recommend doses of these preparations that appear to be highly excessive with injection intervals that are too widely spaced. Based on the findings of the present meta-analysis, recommendations by guidelines should be reassessed. Finally, the fitted curves in this work were incorporated into an interactive web-based injectable estradiol simulator intended for use by transfeminine people and their medical providers to help guide therapeutic decisions.

Introduction

Estradiol is the main estrogen used in transfeminine hormone therapy and is available in a variety of different forms for use by different routes of administration. The most commonly employed forms are oral, sublingual, transdermal, and injectable preparations. Injectable estradiol preparations have been discontinued in many countries and hence are unavailable for use in transfeminine hormone therapy in many parts of the world, for instance in most of Europe (Glintborg et al., 2021). However, they are still used by many transfeminine people particularly in the United States and in the do-it-yourself (DIY) community. The most commonly used forms include estradiol valerate, estradiol cypionate, and estradiol enanthate all in oil. Injectable estradiol preparations have certain advantages over other estradiol forms that make them a popular choice for use in transfeminine hormone therapy. These include often lower cost, capacity to easily achieve higher estradiol levels that can be useful for testosterone suppression, less frequent administration, and theoretically reduced health risks relative to oral estradiol at equivalent doses due to the lack of the first pass with this route (Aly W., 2020). The higher estradiol levels with injections are particularly useful for estradiol monotherapy, in which an antiandrogen is not used.

Clinically used injectable estradiol preparations are formulated not as estradiol but as estradiol esters. When injected into muscle or fat in oil solutions or crystalline aqueous suspensions, these estradiol esters form depots at the injection site from which they are slowly released. Subsequent to release, estradiol esters are rapidly metabolized into estradiol and hence act as prodrugs. When estradiol itself is given by intramuscular injection in an aqueous solution or oil solution, it is rapidly absorbed and has a very short duration. Due to having lipophilic esters, most clinically used injectable estradiol esters are more fat-soluble than estradiol (as measured by oil–water partition coefficient (P)) (Table) . When these esters are administered as oil solutions by intramuscular or subcutaneous injection, their increased lipophilicity causes them to be released from the injection-site depot more slowly than estradiol and to therefore have longer durations. In the case of fatty acid esters, the longer the chain length of the ester—as in e.g. estradiol valerate (5 carbons) vs. estradiol enanthate (7 carbons) vs. estradiol undecylate (10 carbons)—the greater the fat solubility, the slower the rate of release from the depot, and the longer the time to peak levels and duration (Edkins, 1959; Sinkula, 1978; Chien, 1981; Kuhl, 2005; Kalicharan, 2017; Vhora et al., 2019). The durations of both injectable oil solutions and aqueous suspensions depend on the ester and its particular physicochemical properties, but the characteristics of these preparations are different and they work in distinct ways to produce their depot effects (Enever et al., 1983; Aly W., 2019). The durations of oil solutions are dependent on the lipophilicity of the ester as well as oil vehicle, whereas the durations of aqueous suspensions depend on the properties of the ester crystal lattice as well as crystal sizes (Chien, 1981; Enever et al., 1983; Aly W., 2019). The polymeric estradiol ester polyestradiol phosphate is more hydrophilic (water-soluble) than estradiol and works differently than other injectable estradiol preparations. Ιt is composed of many estradiol molecules linked together via phosphate esters (on average 13 molecules of estradiol per one molecule of polyestradiol phosphate) and has a prolonged duration due to slow cleavage into estradiol following injection. Estradiol esters are able to substantially prolong the duration of estradiol when used as injectables and these preparations have durations ranging from days to months depending on the ester and how it is formulated (Table).

There is very little in the way of research and review on the pharmacokinetics of injectable estradiol preparations in the transgender health literature. Transgender hormone therapy guidelines presently offer only brief descriptions and dosing recommendations that appear to be based mainly on expert opinion for this form of estradiol (e.g., Deutsch et al., 2016; Hembree et al., 2017). Many studies assessing the pharmacokinetics and concentration–time profiles of injectable estradiol preparations have been published but are largely confined to cisgender women and men rather than transgender people. These studies are scattered throughout the literature and have not been comprehensively reviewed or analyzed. Some review material exists on the pharmacokinetics of injectable estradiol preparations for use in hormonal birth control and menopausal hormone therapy in cisgender women (e.g., Düsterberg & Nishino, 1982; Kuhl, 1990; Garza-Flores, 1994; Kuhl, 2005) and androgen deprivation therapy for prostate cancer in cisgender men (e.g., Gunnarsson & Norlén, 1988). However, these publications discuss only small selections of the available research. Data on repeated administration of injectable estradiol preparations are more rare but have also been published (e.g., Gooren, 1984 [Graph]; various others). Multi-dose simulation has been done previously for polyestradiol phosphate (Henriksson et al., 1999; Johansson & Gunnarsson, 2000). However, it has not been explored for other injectable estradiol preparations to date. In contrast to injectable estradiol, excellent review literature and simulation exists for injectable testosterone preparations (e.g., Behre, Oberpenning, & Nieschlag, 1990; Behre & Nieschlag, 1998; Behre et al., 2004; Nieschlag & Behre, 2010; Nieschlag & Behre, 2012).

In order to aid understanding of concentration–time profiles with injectable estradiol preparations, I’ve developed an interactive web-based injectable estradiol simulator for transfeminine people and their medical providers. During work on this simulator, it became apparent that there is substantial variability in estradiol levels and curve shapes between different studies even with the same injectable estradiol ester. The injectable estradiol simulator was originally designed to simulate curves from only a single well-known pharmacokinetic study that directly compared estradiol benzoate, estradiol valerate, and estradiol cypionate in oil (Oriowo et al., 1980 [Graph]). However, due to the considerable differences in estradiol levels and curves across studies, it was decided that relying on only one study for such a project would be untenable. Instead, for the simulations to be reasonably accurate to the available data, many studies would need to be incorporated. Including additional studies would also allow for inclusion of other injectable estradiol esters in the simulator. As a result, the present work—an informal meta-analysis of estradiol curves with injectable estradiol formulations—was conducted for the simulator project.

Methods

A literature search was performed to identify studies reporting clinical estradiol concentration–time data with major injectable estradiol formulations (Table 1). All of these preparations have been used in transfeminine hormone therapy at one time or another in different parts of the world, although only estradiol valerate in oil and estradiol cypionate in oil are widely used today. Some of the injectable preparations included have notably been discontinued. Acceptable data for the search included mean and individual estradiol concentration data and Cmax estradiol levels (mean peak estradiol levels of individual subjects at time Tmax). Databases like PubMed, Google Scholar, and WorldCat were searched using relevant keywords (e.g., estradiol ester names and variations thereof as well as major brand names). Publications with relevant information were catalogued for data collection. Only single-dose data and multi-dose data that allowed estradiol levels to return to baseline between doses (as in e.g. repeated once-monthly combined injectable contraceptives) were included. Studies were included regardless of the hypothalamic–pituitary–gonadal axis (HPG axis) status of the participants. The study selection criteria aimed to maximize data inclusion due to scarcity of data for several preparations. If however there were many studies for a specific preparation, studies with only 1 or 2 subjects were generally skipped due to the limited additional value that they would provide. When data were in figures in papers—as was generally the case—they were extracted from the graphs using WebPlotDigitizer.

Table 1: Major injectable estradiol formulations (ordered roughly from shortest- to longest-acting):

Estradiol esterAbbr.FormMajor brand names
Estradiol benzoateEBOil solutionProgynon-B
Estradiol valerateEVOil solutionDelestrogen, Mesigyna,a Progynon Depot
Estradiol cypionateECOil solutionDepo-Estradiol
  Aqueous suspensionbCyclofem,a Lunellea
Estradiol enanthateEEnOil solutionPerlutal,a Topasela
Estradiol undecylatecEUOil solutionDelestrec, Progynon Depot 100
Polyestradiol phosphatecPEPAqueous solutionEstradurin

a As combined injectable contraceptives also including a progestin. b Microcrystalline particle size. c No longer marketed.

Following their collection, data were processed, aggregated, and modeled. Data were adjusted for endogenous estradiol production and were normalized by dose. Adjustment for endogenous estradiol production was generally done via subtraction of baseline estradiol levels. In a number of cases however, subtraction of trough estradiol levels or of estradiol levels from a control group was required instead. Data were also weighted by sample size. In a handful of instances, certain missing information (e.g., time to peak levels, baseline levels, subject body weights) was filled in with reasonable assumptions to help maximize data inclusion. Data were processed in the form of mean estradiol curve data rather than individual-subject data (except for rare n=1 studies). The combined processed data from all studies for each injectable estradiol preparation were fit via least squares regression to one-, two-, and three-compartment pharmacokinetic models with first-order absorption and elimination that were obtained from the literature and other sources (e.g., Colburn, 1981; Wagner, 1993; Fisher & Shafer, 2007; Lixoft, 2008; Abuhelwa, Foster, & Upton, 2015; Certara, 2020). These models fit most curves from individual studies very well. Fitting the combined curve fits of all individual studies (as opposed to fitting all of the combined processed data directly) was additionally evaluated for each injectable estradiol preparation, and if it was feasible for the preparation and allowed for better fitting results, was employed instead. Fitting directly to the combined processed data has the effect of weighting individual studies by quantity of time points, whereas fitting the combined curve fits of studies eliminates this. The Akaike information criterion (AIC) was used to help guide model selection for fitting of the preparations. Curve fitting was performed using the Python library Lmfit with the Levenberg–Marquardt algorithm. Cmax concentrations are a different form of data than mean curve estradiol concentration–time data, and for this reason, were not included in the fitting unless data were very limited for a given injectable estradiol preparation. Outlying data were also excluded from fitting in a number of instances and this allowed for improved curve fits with more uniform area-under-the-curve levels. The main criterion used for excluding curves was fit area-under-the-curve levels that deviated considerably from what was typical for the injectable estradiol preparations (generally less than about 50% of the average or greater than about 150% of the average).

A selection of pharmacokinetic parameters were calculated for each injectable estradiol preparation using the single-dose fit curves and compartmental pharmacokinetic analyses. These parameters included maximal concentrations of estradiol after a single dose scaled to 5 mg (Cmax), time to maximal concentrations of estradiol after a single dose (Tmax), total area-under-the-curve concentrations of estradiol after a single dose (AUC0–∞), integrated mean concentrations of estradiol during repeated injections with a dose and dose interval of 5 mg once every 7 days (Cavg(5mg/7d)), terminal half-life after a single dose (t1/2), and the terminal 90% life after a single dose (t90%) (calculated as t1/2 × 3.322). Cmax and Tmax were defined and calculated as peak estradiol level and time to peak level of the fit mean curve as opposed to the mean peak level and mean time to peak level of individual subjects (which would not be possible to compute as most studies reported only estradiol mean curve data). Pharmacokinetic parameters were calculated using relevant pharmacokinetic equations and were compared against those computed by the PKSolver Microsoft Excel add-in application (Zhang et al., 2010) as a sanity check.

Results

The figures in the subsequent sections show the original data from studies adjusted for endogenous estradiol levels and normalized to a common dose as well as the curve fits to the data (or alternatively the curve fits of the fits of the data depending on the preparation) for the included injectable estradiol preparations. Estradiol benzoate, estradiol cypionate in oil, and estradiol cypionate suspension were fit to the fits of all individual studies for these preparations, whereas estradiol enanthate, estradiol undecylate, and polyestradiol phosphate were fit directly to the combined processed data for these esters. In the case of estradiol valerate, the two fitting approaches gave nearly identical curves, and so fitting the combined processed original data was done for simplicity for this preparation. Cmax studies were excluded in the fitting for all preparations except estradiol enanthate, for which available estradiol concentration–time data were otherwise very limited. The data for the injectable estradiol preparations were generally fit best by a three-compartment pharmacokinetic model (Desmos; V3C Fitter). As a result, and for consistency, this model was used in the fitting of all preparations.

Estradiol Benzoate

Injectable estradiol benzoate has been extensively used in the past in scientific research, most notably in studies elucidating the function and dynamics of the HPG axis. One such use of estradiol benzoate has been the estrogen provocation test, a diagnostic test of HPG axis function. Due to its use in research, a substantial amount of clinical estradiol concentration–time data with injectable estradiol benzoate exists. A total of 26 publications and concentration–time data for 453 individual injections were identified for estradiol benzoate (Geppert, 1975; Leyendecker et al., 1975; Keye & Jaffe, 1975; Shaw et al., 1975; Shaw, Butt, & London, 1975; Leyendecker et al., 1976; Canales et al., 1978; Shaw, 1978; Travaglini et al., 1978; Travaglini et al., 1979; Oriowo et al., 1980; Canales et al., 1981; White et al., 1981; Schweikert et al., 1982; Braun, Wildt, & Leyendecker, 1983; Kemeter et al., 1984; Välimäki et al., 1984; Goodman et al., 1985; Aisaka et al., 1986; Cano et al., 1986; Messinis & Templeton, 1987a; Messinis & Templeton, 1987b; Sumioki, 1987; Bider et al., 1989; Vizziello et al., 1993; Eriksson et al., 2006). A number of studies were excluded from fitting due to much higher or lower area-under-the-curve levels than average. A couple of studies were omitted from the meta-analysis as they only reported total estrogen levels rather than estradiol levels with estradiol benzoate (Akande, 1974; Weiss, Nachtigall, & Ganguly, 1976). Two studies were omitted due partly to being very old and using very early and inaccurate blood tests (Varangot & Cedard, 1957; Ittrich & Pots, 1965 [Graph]). A number of additional publications that might have estradiol concentration–time data with estradiol benzoate were not obtained (at least for now) and hence were not included (Selmi et al., 1976; Schaison et al., 1977; Chimbira et al., 1979; Kusuda & Onoue, 1982; Igarashi et al., 1983; Schaison, 1983; Punnonen, Multamäki, & Honkonen, 1984). The processed original data and fit of fits curve for estradiol benzoate are shown in Figure 1.

Figure 1: Published estradiol concentration–time curves and fit of fit curves (black line) with a single intramuscular injection of estradiol benzoate in oil solution over a period of 7 days. Curves were adjusted for endogenous estradiol levels, normalized to a dose of 5 mg, and fit with a compartmental pharmacokinetic model. Following this, the combined fit curves of the individual studies were fit using the same pharmacokinetic model. The original data from the studies for estradiol benzoate are also provided elsewhere (Sheets; Plotly).

Estradiol Valerate

Studies with curve data on injectable estradiol valerate come from its use in menopausal hormone therapy and other therapeutic indications for estrogens, its use in combined injectable contraceptives, and use in scientific research. A total of 28 publications and concentration–time data for 309 individual injections were identified for estradiol valerate (Somerville, 1971; Somerville, 1972a; Somerville, 1972b; Somerville, 1972c; Somerville, 1975; Geppert, 1975; Leyendecker et al., 1975; Vermeulen, 1975; Oriowo et al., 1980; Rauramo et al., 1980; Rauramo, Punnonen, & Grönroos, 1981; Blackwell, Boots, & Potter, 1982; Düsterberg & Nishino, 1982; Düsterberg, & Wendt, 1983; Aedo et al., 1985; Düsterberg, Schmidt-Gollwitzer, & Hümpel, 1985; Reimann et al., 1987; Sang et al., 1987; Sherwin et al., 1987; Sherwin, 1988; Goh & Ratnam, 1988; Goh & Ratnam, 1990; Garza-Flores, 1994; Jilma et al., 1994; Goh & Lee, 1998; Göretzlehner et al., 2002; Kerdelhué et al., 2006; Valle Alvarez, 2011; Schug, Donath, & Blume, 2012). A few of these studies were excluded from fitting due generally to much higher or lower area-under-the-curve levels than average or due to being Cmax data. One study was omitted as it only reported estrone levels rather than estradiol levels (Ibrahim, 1996). Another study was not included due to being in pregnant women with concomitant pregnancy termination (Garner & Armstrong, 1977). One last study was omitted due partly to being very old and using very early and inaccurate blood tests (Ittrich & Pots, 1965 [Graph]). The processed original data and fit curve for estradiol valerate are shown in Figure 2.

Figure 2: Published estradiol concentration–time curves and fit curve (black line) with a single intramuscular injection of estradiol valerate in oil solution over a period of 30 days. Curves were adjusted for endogenous estradiol levels, normalized to a dose of 10 mg, and fit with a compartmental pharmacokinetic model. Subsequent fitting of the combined fits of individual studies for this preparation was explored but gave a nearly identical overall curve, so the overall fit curve for the combined processed original data was used for simplicity for this preparation. The original data from the studies for estradiol valerate are also provided elsewhere (Sheets; Plotly).

Estradiol Cypionate Oil

Estradiol cypionate in oil is used in menopausal hormone therapy and for other estrogen therapeutic indications. However, its use has been more limited relative to other injectable estradiol preparations like estradiol valerate. Only a handful of studies with relevant estradiol concentration–time data were identified for estradiol cypionate in oil. This included 4 publications and concentration–time data for 48 individual injections (Rosenfield et al., 1973; Rosenfield & Fang, 1974; Buckman et al., 1980; Oriowo et al., 1980; Lichten et al., 1996). No curves were excluded from fitting in the case of this preparation. The processed original data and fit of fit curves for estradiol cypionate in oil are shown in Figure 3.

Figure 3: Published estradiol concentration–time curves and fit of fit curves (black line) with a single intramuscular injection of estradiol cypionate in oil solution over a period of 30 days. Curves were adjusted for endogenous estradiol levels, normalized to a dose of 5 mg, and fit with a compartmental pharmacokinetic model. Following this, the combined fit curves of the individual studies were fit using the same pharmacokinetic model. The original data from the studies for estradiol cypionate in oil are also provided elsewhere (Sheets; Plotly).

Estradiol Cypionate Suspension

Estradiol cypionate suspension has been used exclusively in combined injectable contraceptives. For this reason, many relatively high quality pharmacokinetic studies have been conducted with this injectable preparation. A total of 9 publications and concentration–time data for 131 individual injections were identified for estradiol cypionate suspension (Fotherby et al., 1982; Aedo et al., 1985; Garza-Flores et al., 1987; Garza-Flores, 1994; Zhou et al., 1998; Rahimy & Ryan, 1999; Rahimy, Ryan, & Hopkins, 1999; Sierra-Ramírez et al., 2011; Thurman et al., 2013). One of these studies used subcutaneous injection instead of the usual intramuscular injection but the resulting curve was very similar to the curve for intramuscular injection in the same study (Sierra-Ramírez et al., 2011 [Graph]). Several Cmax studies were excluded from fitting for this preparation. One pharmacokinetic study only measured estradiol cypionate levels rather than estradiol levels and hence was not included (Martins et al., 2019 [Graph]). The processed original data and fit of fit curves for estradiol cypionate suspension are shown in Figure 4.

Figure 4: Published estradiol concentration–time curves and fit of fits curve (black line) with a single intramuscular (or in one case subcutaneous) injection of a microcrystalline aqueous suspension of estradiol cypionate over a period of 30 days. Curves were adjusted for endogenous estradiol levels, normalized to a dose of 5 mg, and fit with a compartmental pharmacokinetic model. Following this, the combined fit curves of the individual studies were fit using the same pharmacokinetic model. The original data from the studies for estradiol cypionate suspension are also provided elsewhere (Sheets; Plotly).

Estradiol Enanthate

Estradiol enanthate has been used exclusively in combined injectable contraceptives. Several pharmacokinetic studies have been conducted with it because of this. A total of 7 publications and concentration–time data for 270 individual injections were identified for estradiol enanthate (Recio et al., 1986; Wiemeyer et al., 1986; Wiemeyer et al., 1987; Schiavon et al., 1988; Garza-Flores et al., 1989; Garza-Flores, 1994; Martinez, 1995). However, 216 of the injections were from a single study and mainly included only Cmax levels. Wiemeyer et al. (1986) was excluded from fitting due to having unusually high area-under-the-curve levels with a small sample size (n=3). Because of the scarcity of estradiol concentration–time data available for estradiol enanthate, Cmax studies were included in the fitting for this preparation. The processed original data and fit curve for estradiol enanthate are shown in Figure 5.

Figure 5: Published estradiol concentration–time curves and fit curve (black line) with a single intramuscular injection of estradiol enanthate in oil solution over a period of 30 days. Curves were adjusted for endogenous estradiol levels, normalized to a dose of 10 mg, and fit with a compartmental pharmacokinetic model. The original data from the studies for estradiol enanthate are also provided elsewhere (Sheets; Plotly).

Estradiol Undecylate

Estradiol undecylate was formerly used in the treatment of prostate cancer and menopausal hormone therapy as well as other estrogen therapeutic indications. However, it was discontinued many years ago and is no longer used today. Nonetheless, estradol undecylate is of significant historical interest. A total of 3 publications and concentration–time data for 7 individual injections were identified for estradiol undecylate (Geppert, 1975/Leyendecker et al., 1975 [Graph]; Vermeulen, 1975 [Graph]). Unfortunately, these data were very low quality, with small sample sizes and considerable variations in estradiol levels. Moreover, estradiol undecylate is a very long-acting injectable estradiol ester with a duration measured in months, and the follow up in these studies only went to about 2 weeks post-injection. For these reasons, it was not possible to fit the data for estradiol undecylate in a reasonably accurate way—as suggested by area-under-the-curve estradiol levels that were only around one-third those of the other non-polymeric injectable estradiol esters. Limited multi-dose estradiol concentration–time data also exist for estradiol undecylate, but these data could not be incorporated (Jacobi & Altwein, 1979; Jacobi et al., 1980; Derra, 1981). The processed original data and fit curve for estradiol undecylate are shown in Figure 6.

Figure 6: Published estradiol concentration–time curves and fit curve (black line) with a single intramuscular injection of estradiol undecylate in oil solution over a period of 90 days. Curves were adjusted for endogenous estradiol levels, normalized to a dose of 50 mg, and fit with a compartmental pharmacokinetic model. The original data from the studies for estradiol undecylate are also provided elsewhere (Sheets; Plotly).

Polyestradiol Phosphate

Polyestradiol phosphate has been used primarily in the treatment of prostate cancer but has also been used for estrogen therapeutic indications like treatment of breast cancer and menopausal hormone therapy. While this injectable estradiol preparation has been used widely in the past, it appears to have recently been discontinued. All of the identified studies with estradiol concentration–time data on polyestradiol phosphate were in men with prostate cancer. A total of 11 publications and concentration–time data for 114 individual injections were identified for polyestradiol phosphate (Jönsson, 1976; Leinonen et al., 1979; Leinonen, 1980; Jacobi, 1982; Norlén, 1987; Gunnarsson & Norlén, 1988; Stege et al., 1988; Stege et al., 1989; Stege et al., 1996; Henriksson et al., 1999; Johansson & Gunnarsson, 2000). A few older and strongly outlying studies were excluded from the fitting. The processed original data and fit curve for polyestradiol phosphate are shown in Figure 7.

Figure 7: Published estradiol concentration–time curves and fit curve (black line) with a single intramuscular injection of an aqueous solution of polyestradiol phosphate over a period of 90 days. The graph was clipped to maximum estradiol levels of 600 pg/mL (~2,200 pmol/L) for better viewability. Curves were adjusted for endogenous estradiol levels, normalized to a dose of 160 mg, and fit with a compartmental pharmacokinetic model. The original data from the studies for polyestradiol phosphate are also provided elsewhere (Sheets; Plotly).

Other Injectable Estradiol Preparations

A number of clinical studies with estradiol concentration–time data for other injectable estradiol preparations were also identified. These formulations included unesterified estradiol in oil (Ittrich & Pots, 1965 [Graph]) or an “aqueous” form (type unspecified but probably suspension) (Jones et al., 1978 [Graph]), estradiol/progesterone as a macrocrystalline aqueous suspension (Garza-Flores et al., 1991) or in microspheres (Garza-Flores, 1994; Espino y Sosa et al., 2019 [Graph]), estradiol/megestrol acetate as a microcrystalline aqueous suspension (Yan et al., 1987 [Graphs]), estradiol dipropionate in oil (Presl et al., 1976 [Graph]), estradiol benzoate/estradiol phenylpropionate in oil (Rauramo et al., 1980; Rauramo, Punnonen, & Grönroos, 1981), estradiol benzoate/estradiol phenylpropionate/testosterone propionate/testosterone phenylpropionate/testosterone isocaproate in oil (Vermeulen, 1975), and estradiol benzoate/estradiol dienanthate/testosterone enanthate benzilic acid hydrazone in oil (Sherwin & Gelfand, 1987; Sherwin et al., 1987; Sherwin, 1988 [Graphs]). These preparations were not included in the present meta-analysis due to their relative obscurity and the limited data available for them. In addition, there were concerns about fitting the employed pharmacokinetic models to the multi-component and microsphere formulations. No estradiol concentration–time data were identified for certain other injectable estradiol forms of interest like unesterified estradiol in aqueous solution, estradiol benzoate as a microcrystalline aqueous suspension, or estradiol benzoate butyrate/dihydroxyprogesterone acetophenide in oil (a less widely used combined injectable contraceptive).

All Injectable Estradiol Preparations Together

Figure 8 shows the curve fits for all of the injectable estradiol preparations scaled to a single dose of 5 mg (or equivalent) together in the same figure. The dose for polyestradiol phosphate was scaled to be about 6.5 times higher than the other injectable estradiol preparations in order to make it roughly equivalent to them in terms of total estradiol exposure. This was because polyestradiol phosphate was found to produce much lower area-under-the-curve estradiol levels than the other injectable estradiol preparations (see the Discussion section). Estradiol undecylate was not included in Figure 8 as a decent fit curve could not be obtained for it due to the very limited data available for this preparation.

Figure 8: Curve fits of published estradiol concentration–time data with different injectable estradiol preparations by intramuscular injection scaled to equivalent doses and plotted over a period of 20 days in a single combined graph. Polyestradiol phosphate is scaled to a 6.5-fold higher dose that is roughly equivalent to that for the other esters as it gave total estradiol levels that were around 6 or 7 times lower than the other esters at the same dose.

For simulated estradiol concentration–time curves with repeated injections of these injectable estradiol preparations, please see the accompanying interactive web simulator.

Selected Pharmacokinetic Parameters

The table below shows selected pharmacokinetic parameters for the fit curves of the included injectable estradiol preparations (Table 2). Estradiol undecylate was not included in the table due to the lack of data needed to achieve a decent curve fit for this preparation and the uncertainty of its parameters.

Table 2: Selected pharmacokinetic parameters for estradiol with injectable estradiol preparations:

Estradiol PreparationCmax
(pg/mL) a
Tmax
(d)
t1/2
(d)
t90%
(d)
AUC0–∞
(pg•d/mL) a
Cavg
(pg/mL) b
Estradiol benzoate in oil9350.781.03.42327332
Estradiol valerate in oil2952.13.09.91886269
Estradiol cypionate oil1554.36.722.32150307
Estradiol cypionate suspension2411.25.116.92096299
Estradiol enanthate in oil1606.54.615.12183312
Polyestradiol phosphate c3418.028.494.22117302

a Scaled to a single 5 mg injection. b Scaled to repeated injections of 5 mg every 7 days. c Scaled instead to a single 32.5 mg injection or to repeated injections of 32.5 mg every 7 days (6.5 times higher dose than with the other esters).

Terminal half-life (t1/2) is the time for the concentration of estradiol to decrease by 50% after pseudo-equilibrium of distribution has been reached—not the time required for half of an administered dose of the estradiol ester to be eliminated (Toutain & Bousquet-Mélou, 2004). It is calculated using only the terminal portion of a concentration–time curve, without the absorption or distribution phases influencing it (Toutain & Bousquet-Mélou, 2004). Due to flip–flop kinetics with depot injectables and the very short blood half-life of estradiol (~0.5–2 hours), what is being described by the terminal half-life in the case of depot estradiol injectables is not actually elimination of estradiol from blood but is rather the absorption of estradiol from the injection-site depot (Toutain & Bousquet-Mélou, 2004; Yáñez et al., 2011).

Discussion

Data Quality, Limitations, and Variability Between Studies

The accuracies of the curve fits for the different included injectable estradiol preparation are limited by the available data for these preparations. The quantity and quality of data are variable among these preparations. In some cases, such as with estradiol valerate in oil and estradiol cypionate in suspension, the data are overall quite good. In other instances, such as with estradiol cypionate in oil and estradiol enanthate in oil, the available data are more limited. There was undersampling of certain parts of the concentration–time curve with some preparations, for instance estradiol benzoate in oil (the early curve), estradiol enanthate in oil (much of the curve), and polyestradiol phosphate (the late curve). In the case of estradiol undecylate in oil, the available data for this preparation weren’t adequate to achieve a decent curve fit at all. The fit curves and calculated pharmacokinetic parameters of the included injectable estradiol preparations should be interpreted with the imperfect data in mind. For example, the curve shapes and pharmacokinetic parameters for the different preparations should not be taken as precise determinations in most cases but instead as rough estimates that would no doubt change with more and better data. Indeed, the fits and pharmacokinetic parameters were often noticeably sensitive to the influences of individual studies. Modeling decisions, such as the choice of pharmacokinetic model, or whether to fit directly to the combined processed data versus to the fits of individual studies, also yielded significantly different curve fits as well as calculated pharmacokinetic parameters.

Due to scarcity of data for several injectable estradiol preparations, the study selection criteria maximized data inclusion in order to allow for better curve fits at the risk of including potentially less reliable data. As examples, studies were included regardless of the status of the HPG axis of the participants, and Cmax data were included in the fitting if data were very limited. In the case of HPG axis state, studies with cycling women may result in greater error due to more variable levels of endogenous estradiol. Moreover, acute high levels of estradiol can induce a surge in luteinizing hormone levels after several days in gonadally intact women, and this may cause a delayed bump in estradiol levels (Wiki). One of the more overt instances of this can be seen in a study of estradiol benzoate in such women (Shaw, 1978 [Graph]). Many if not most of the included studies with estradiol benzoate involved women with intact HPG axes, whereas studies of this sort were uncommon with the other preparations. In the case of Cmax data, these data when Cmax corresponds to the mean of individual peaks are a different type of data than the peak of the mean curve of all individuals. Cmax levels can differ in both magnitude and timing compared to the mean curve peak (e.g., Oriowo et al., 1980 [Graph]; Rahimy, Ryan, & Hopkins, 1999). This is because for instance not all individuals peak at the same time and this variability in time to peak normally serves to dilute peak levels for the mean curve when compared to individual maximal concentrations. However, Cmax levels are in any case generally in the vicinity of the mean curve peak. While Cmax levels were excluded in the fitting for most injectable estradiol preparations, they were included in the case of estradiol enanthate. This was because the available mean and individual estradiol curve data were very limited for this specific preparation, and inclusion of Cmax data allowed for improved fitting in spite of its limitations. Lastly, some of the included data was once-monthly multi-dose, and research with once-monthly estradiol enanthate-containing combined injectable contraceptives has found that the time to peak levels may shift with repeated long-term use (Schiavon et al., 1988; Garza-Flores, 1994).

There was considerable variability between studies in terms of estradiol levels and concentration–time curve shapes with the same injectable estradiol preparation. The reasons for the large variability across studies are not fully clear. In any case, there are many potential factors that may contribute to this variability. These include preparation- and injection-related factors like formulation (e.g., oil vehicle, other components and excipients, concentration, particle size), injection volume, site of injection (e.g., buttocks, thigh, upper arm), injection technique (e.g., force of injection—and resulting depot droplet dimensions), and syringe dead space. They additionally include various subject- and research-related variables like differing blood-testing methodology, differing sample characteristics (e.g., age, weight, gender, ethnicity, physical activity, HPG axis state), and sampling error (Sinkula, 1978; Chien, 1981; Minto et al., 1997; Larsen & Larsen, 2009; Larsen et al., 2009; Florence, 2010; Larsen, Thing, & Larsen, 2012; Kalicharan, 2017). Older studies, which used potentially less accurate blood tests and tended to have smaller numbers of subjects, seemed to particularly add to the variability between studies. These studies may represent less reliable data than more recent research with larger sample sizes. The exclusion criteria helped to remove outliers for the different injectable estradiol preparations however. This meta-analysis does not take into account the potential factors underlying the variability between studies. To do so would be difficult, as in many cases information on these variables is not provided in individual studies and research quantifying their precise influences and relative importances is limited.

It is in any case known from other studies that different oil vehicles are absorbed at different rates from the injection site (Svendsen & Aaes‐Jørgensen, 1979; Schultz et al., 1998; Larsen et al., 2001) and can result in different concentration–time curve shapes (Ballard, 1978 [Excerpt]; Knudsen, Hansen, & Larsen, 1985). This is thought to be due to differences in oil lipophilicity and depot release rates. Viscosity of oils has also been hypothesized to potentially influence rate of depot escape (Schug, Donath, & Blume, 2012). However, research so far has not supported this hypothesis (Larsen & Larsen, 2009; Larsen, Thing, & Larsen, 2012). Oil vehicles can vary with injectable estradiol preparations even for the same estradiol ester. For instance, pharmaceutical estradiol valerate is formulated in sesame oil, castor oil, or sunflower oil depending on the preparation (Table). It is notable however that these three oils have similar lipophilicities (Table). On the other hand, homebrewed injectable estradiol preparations used by DIY transfeminine people often employ medium-chain triglyceride (MCT) oil as the oil vehicle. This oil (in the proprietary form of Viscoleo) has notably been found to be much more rapidly absorbed than conventional oils like sesame oil and castor oil in animals (Svendsen & Aaes‐Jørgensen, 1979; Schultz et al., 1998; Larsen et al., 2001). In addition, although based on very limited data, MCT oil has been found to give spikier and shorter-lasting depot injectable curves in humans (Knudsen, Hansen, & Larsen, 1985). As such, injectable estradiol preparations using MCT oil as the vehicle may have differing and less favorable concentration–time curve shapes than pharmaceutical injectable estradiol products. Other excipients, like benzyl alcohol, as well as factors like injection site and volume, have additionally been found to influence pharmacokinetic properties with depot injectables (Minto et al., 1997; Kalicharan, Schot, & Vromans, 2016). Excipients besides oil vehicle also vary by formulation (Table).

An implication of the variability between studies is that there is not a single estradiol concentration–time curve for a given injectable estradiol preparation but rather there are many, with these curves determined by variables such as formulation, dose/administration, and subject characteristics, among others. Hence, the curve fits determined in this meta-analysis represent only an estimation of the most typical and hence likely case, but the true curve for a preparation in a given context may be quite different.

Fitting all studies for a given injectable estradiol preparation individually first, and then fitting the fits of these studies, allowed for improved curve fits relative to directly fitting all of the combined processed original data for the preparation. The latter approach has limitations in that it has the effect of inherently weighting individual studies by quantity of time points (resulting in studies with greater time sampling having greater influence on the fit). Additionally, and more problematically, this approach can lead to distortions in curve shape due to different studies sampling different portions of the curve to differing extents in conjunction with systematic differences in curves between these studies. These are problems that fitting the fits of individual studies instead can solve. However, it is not possible to fit all individual studies, as some studies have limited time sampling and curve characterization which precludes fitting them appropriately. Cmax data are an example of this, which on their own cannot be fit properly. As such, it was not possible to fit the fits of the individual studies for all injectable estradiol preparations. Consequently, the fitting approach in this regard was not the same across esters, with some fit instead directly to the combined processed original data (e.g., estradiol enanthate, polyestradiol phosphate).

In spite of the various limitations of this work, aggregated analysis and modeling with injectable estradiol preparations has not previously been done. This informal meta-analysis provides among the most detailed insight into estradiol levels and curve shapes with these preparations available to date.

Durations and Curve Shapes

The curve shapes of non-polymeric injectable estradiol esters in oil relate strongly to lipophilicity. The more lipophilic the ester, the lower the peak levels and the more protracted the estradiol concentration–time curve. Accordingly, estradiol benzoate, one of the least lipophilic estradiol esters, has one of the spikiest curves and shortest durations, whereas more lipophilic estradiol esters, like estradiol cypionate in oil and estradiol enanthate, have comparatively flatter curves with delayed peaks and longer durations.

Duration of Estradiol Valerate

The estradiol concentration–time curve for injectable estradiol valerate in the well-known Oriowo et al. (1980) [Graph] study is notably spikier and shorter-lasting than the overall curve for estradiol valerate in this meta-analysis. On the other hand, the overall curve for injectable estradiol valerate in this meta-analysis was similar to (and considerably influenced by) the curves from several relatively recent and presumably better-quality studies of this injectable estradiol ester (e.g., Göretzlehner et al., 2002; Valle Alvarez, 2011; Schug, Donath, & Blume, 2012). It’s noteworthy that Oriowo et al. (1980) used a peanut oil-based formulation of estradiol valerate that differed from pharmaceutical injectable estradiol valerate preparations, which generally use sesame oil or castor oil as the carrier (as well as other excipients) (Table). This may have influenced the curve shape of estradiol valerate in Oriowo et al. (1980). The study also had a small sample size relative to the more recent studies (n=9 versus n=17, n=32, and n=24×2, respectively). Based on the newer and overall data, estradiol valerate appears to have a curve that is noticeably flatter and more prolonged than that suggested by Oriowo et al. (1980).

Duration of Estradiol Cypionate in Oil versus Estradiol Enanthate

Available estradiol concentration–time data for injectable estradiol cypionate in oil and estradiol enanthate in oil are more limited than with several of the other injectable estradiol preparations, and no direct comparisons of these two preparations exist at present. Based on some of the available literature on these injectable estradiol esters, most notably discussion by Oriowo et al. (1980) and a review of the pharmacokinetics of combined injectable contraceptives (Garza-Flores, 1994 [Graph]), it seemed that the duration of estradiol enanthate in oil was longer than that of estradiol cypionate in oil. However, this was based on limited research from separate and hence indirectly comparative studies of these esters. The estradiol cypionate in oil data from the relevant Garza-Flores (1994) figure was based on Oriowo et al. (1980) [Graph], and there are reasons to be cautious about relying on these data alone. The main concern is that curve shapes with the same injectable estradiol preparation can vary considerably across studies, as the present meta-analysis has shown. The reasons for this have yet to be fully clarified as already discussed, but among other factors may include varying formulations across studies of the same injectable estradiol ester. It is notable in this regard that Oriowo et al. (1980) used a formulation of estradiol cypionate that differs from conventional pharmaceutical estradiol cypionate in oil preparations—specifically, the study used a peanut oil-based formulation (with few other specifics) rather than the cottonseed oil-based preparation employed in marketed pharmaceutical formulations (Table). The study also had a somewhat small sample size (n=10) and may have had significant sampling error. Hence, single studies, perhaps particularly Oriowo et al. (1980), should be interpreted cautiously.

A small but interesting pharmacokinetic study which directly compared injectable testosterone cypionate (n=6) and testosterone enanthate (n=6) both in oil is relevant to the topic in question. This study found that equivalent doses of these testosterone esters using otherwise identical formulations produced virtually identical testosterone concentration–time curves (Schulte-Beerbühl & Nieschlag, 1980 [Graph]). The findings of this study are consistent with the fact that the lipophilicities of testosterone cypionate and testosterone enanthate (as measured by predicted log P) are very similar when directly compared (e.g., 5.1 vs. 5.11 with ALOGPS, 6.29 vs. 6.11 with ChemAxon, and 6.4 vs. 6.3 with XLogP3, respectively) (Table). This of course is of importance as lipophilicity is thought to be the key factor determining the release kinetics of oil-based depot injectables (Sinkula, 1978; Shah, 2007; Larsen & Larsen, 2009; Larsen, Thing, & Larsen, 2012; Shahiwala, Mehta, & Momin, 2018). Analogously similar lipophilicities can be seen when comparing estradiol cypionate and estradiol enanthate, which employ the same ester moieties (e.g., predicted log P values of 6.47 vs. 6.45 with ALOGPS and 7.1 vs. 7.0 with XLogP3, respectively) (Table). Hence, on a theoretical level, injectable estradiol cypionate and estradiol enanthate, like injectable testosterone cypionate and testosterone enanthate, might be expected to produce very similar curves—at least provided all other variables, such as formulation, are held constant.

The present meta-analysis found that the overall estradiol curve for estradiol cypionate in oil was significantly less spikey and more prolonged than that observed in Oriowo et al. (1980). It is noteworthy in this regard that all of the other studies included for estradiol cypionate in oil specifically employed pharmaceutical Depo-Estradiol and that the overall curve for this preparation appears to be more consistent with its licensed injection interval for use in menopausal hormone therapy (1–5 mg once every 3–4 weeks) (Depo-Estradiol Label). Moreover, this meta-analysis found that injectable estradiol cypionate in oil and estradiol enanthate in oil had fairly similar and comparably flat and prolonged estradiol concentration–time curves. However, estradiol cypionate in oil appeared to peak earlier than estradiol enanthate, while estradiol enanthate was eliminated more rapidly than estradiol cypionate in oil in the terminal portion of the curve. In any case, the available concentration–time data for these preparations are limited, and the present work is not able to determine whether these estradiol esters have truly differing pharmacokinetic properties, as the apparent differences between the curves for these preparations may simply be due to statistical error. Taken together, estradiol cypionate in oil may have a less spikey and longer-lasting curve than that implied by Oriowo et al. (1980), and estradiol cypionate in oil and estradiol enanthate may have more similar curves than has been previously assumed.

Curve Shape of Estradiol Cypionate Suspension

While estradiol cypionate as an aqueous suspension is a relatively long-lasting injectable estradiol preparation similarly to estradiol cypionate in oil and estradiol enanthate in oil, it seems to differ in the shape of its estradiol concentration–time curve from these preparations. Estradiol cypionate as a suspension has a curve that appears to peak significantly earlier than estradiol cypionate in oil and other longer-acting oil-based injectable estradiol preparations. This might relate to the differing mechanisms of depot action and unique properties of injectable aqueous suspensions (Aly W., 2019). In line with this notion, injectable medroxyprogesterone acetate suspension (Depo-Provera) also appears to peak rapidly despite having a very long duration (longer durations tending to be associated with delayed peaks in the case of oil-based depot injectables) (Graphs). Although aqueous suspensions generally last longer than oil solutions as injectables (Enever et al., 1983; Aly W., 2019), this is not always the case, and estradiol cypionate suspension interestingly seems to be shorter-acting than estradiol cypionate in oil.

Estradiol Exposure and Potency

The average estradiol levels with the non-polymeric injectable estradiol esters when scaled to a dose and dosing interval of 5 mg every 7 days were around 300 pg/mL (~1,100 pmol/L). For comparison, in premenopausal cisgender women, estradiol production is on average about 200 μg/day (or 6 mg per month/cycle) and mean estradiol levels are around 100 pg/mL (~370 pmol/L) (Aly W., 2019). After adjusting for the molecular weight of the ester, the estradiol levels for a given dose of non-polymeric injectable estradiol esters are in fairly close agreement with the estradiol levels for an equal quantity of estradiol produced endogenously by the ovaries in premenopausal cisgender women (very roughly around 1.2 mg estradiol per 7 days for injectable estradiol esters and 1.4 mg estradiol per 7 days for ovarian production to achieve average integrated estradiol levels of around 100 pg/mL). The preceding is in accordance with the fact that injectable estradiol valerate has been reported to have approximately 100% bioavailability (with this being less characterized but likely also the case for the other non-polymeric injectable estradiol esters) (Düsterberg & Nishino, 1982; Seibert & Günzel, 1994).

Although non-polymeric injectable estradiol esters have differing estradiol concentration–time curve shapes, they all appear to achieve fairly similar area-under-the-curve levels of estradiol when compared to one another. This is in accordance with the fact that differences in molecular weight and hence estradiol content with the different estradiol esters are fairly minor (all of the assessed non-polymeric esters range from 62 to 76% of that of estradiol in terms of estradiol content, and all but estradiol undecylate are in the range of 69 to 76%) (Table). The appearance of differences in area-under-the-curve levels of estradiol in the present meta-analysis is probably just due to statistical error, and true differences cannot be established by this meta-analysis. An implication of the similar area-under-the-curve estradiol levels with the different non-polymeric injectable estradiol esters is that these preparations can all be expected to deliver a roughly comparable amount of estradiol for the same dose.

On the other hand, the polymeric ester polyestradiol phosphate appears to produce around 6- to 7-fold lower area-under-the-curve and average estradiol levels than non-polymeric estradiol esters. This suggests that polyestradiol phosphate does not have 100% bioavailability, and is supported by the fact that this ester is used clinically at substantially higher dosages than other injectable estradiol esters (40–320 mg/month), even for the same indications such as menopausal hormone therapy and treatment of prostate cancer (Wiki; Estradurin Labels). This does not seem to have been previously described in the literature, and the reasons for it are unknown. It seems possible that polyestradiol phosphate may be partially excreted before it can be cleaved into estradiol and thereby rendered partly inactive, in turn necessitating the use of higher doses to achieve the same estradiol levels and therapeutic effect.

Although two given injectable estradiol preparations may produce equivalent total estradiol levels, this does not necessarily mean that they will always have the same estrogenic potency (i.e., strength of effect at a given dose). It is plausible that spikier estradiol concentration–time curves, like with estradiol benzoate, may have overall lower estrogenic potency than more steady curves, like with estradiol enanthate. This is because estrogen receptors for a given tissue may become saturated at a certain point due to the finite quantity of available receptors in the tissue. As a result, high peak estradiol levels with spikier curves might effectively be “wasted” to varying extents in different tissues. On the other hand, more spikey estradiol curves, due to higher peak estradiol levels, might have greater influence on tissues that require high estradiol levels for effect such as the liver (and by extension on coagulation and associated health risks) (Aly W., 2020). However, these possibilities are speculative, and although some literature exists that is relevant to this issue (e.g., Parkes, 1937Bradbury, Long, & Durham, 1953), there is very little research in this area. Consequently, it is not currently possible to take into account timewise variations in estradiol levels or differing estradiol curve shapes when assessing the comparative estrogenic potency between injectable estradiol preparations (or between other estradiol forms/routes). It is also noteworthy that these variations depend on injection interval and may be reduced with shorter injection intervals that maintain steadier estradiol levels, and this must also be considered.

Variability Between Individuals

There is substantial variation in total estradiol levels and curve shapes between people with the same injectable estradiol preparation. Indicators of interindividual variability such as standard deviation or 95% range have not been included in this meta-analysis at this time due to the large amount of additional time and work this would require (e.g., additional extraction of error bars from all studies and analysis). In any case, individual studies that were included show this marked interindividual variation (e.g., Oriowo et al., 1980; Derra, 1981 [Graph]; Aedo et al., 1985 [Graphs]; Sang et al., 1987 [Graphs]; Rahimy & Ryan, 1999 [Graph]; Valle Alvarez, 2011 [Graph]; Schug, Donath, & Blume, 2012 [Graphs]). Highly variable estradiol levels are already well-established with oral and transdermal estradiol (Kuhl, 2005; Wiki). Less variability might be expected with non-polymeric injectable estradiol esters since these preparations appear to have approximately complete bioavailability. However, it seems that even with injectable forms of estradiol, the variability between people is still quite substantial. An implication of this is that the appropriate dose and dosing interval of an injectable estradiol formulation for a given person will vary considerably. This emphasizes the importance of blood work to ensure that injectable estradiol preparations are neither overdosed—which can increase health risks such as blood clots (Aly W., 2020)—nor underdosed—which may result in suboptimal testosterone suppression and therapeutic efficacy.

Insights for Clinical Guidelines and Dosing Recommendations

Clinical guidelines for transgender health (see also Aly W. (2020)) provide recommendations on doses and dosing intervals of injectable estradiol valerate in oil and estradiol cypionate in oil (Table 3). Dosing recommendations are not given for other injectable estradiol preparations, which are much less commonly used in transgender medicine. The recommended doses for estradiol valerate and estradiol cypionate vary widely depending on the guidelines, whereas the recommended intervals are consistently once every 1 to 2 weeks. The doses for estradiol valerate range from 2 to 20 mg/week or 5 to 80 mg/2 weeks and the doses for estradiol cypionate range from <1 to 10 mg/week or <2 to 80 mg/2 weeks. For reference, the Endocrine Society guidelines and the University of California, San Francisco (UCSF) guidelines are the most major clinical guidelines for transgender hormone therapy at present (Aly W., 2020). The Endocrine Society guidelines recommend 5 to 30 mg/2 weeks or 2 to 10 mg/week for either estradiol valerate or estradiol cypionate (Hembree et al., 2017). Conversely, the UCSF guidelines recommend <20 to 40 mg/2 weeks for estradiol valerate and <2 to 5 mg/2 weeks for estradiol cypionate (with the option to divide dose into weekly injections if cyclical side effects occur) (Deutsch et al., 2016).

Table 3: Recommended doses and injection intervals of injectable estradiol preparations (specifically estradiol valerate and estradiol cypionate) in transgender medicine clinical guidelinesa:

GuidelinesEster(s)Dose Ranges and Intervals
Endocrine Society / Hembree et al. (2017)Estradiol valerate or cypionate5–30 mg/2 weeks or 2–10 mg/week i.m.
UCSF / Deutsch (2016)Estradiol valerateInitial–low: <20 mg/2 weeks i.m.
Initial: 20 mg/2 weeks i.m.
Maximum: 40 mg/2 weeks i.m.
Note: “May divide dose into weekly injections for cyclical symptoms”
Note: Specifically for transfeminine adults
 Estradiol cypionateInitial–low: <2 mg/2 weeks i.m.
Initial: 2 mg/2 weeks i.m.
Maximum: 5 mg/2 weeks i.m.
Note: “May divide dose into weekly injections for cyclical symptoms”
Note: Specifically for transfeminine adults
UCSF / Olson-Kennedy et al. (2016)Estradiol valerate5–20 mg/2 weeks
Maximum: 30–40 mg/2 weeks
Note: Specifically for transfeminine youth
 Estradiol cypionate2–10 mg/week
Note: Specifically for transfeminine youth
Fenway Health (2015)Estradiol valerateInitial: 5–10 mg/week i.m.
Usual: 20 mg/2 weeks i.m.
Maximum: 40 mg/2 weeks i.m.
 Estradiol cypionateInitial: 2.5 mg/2 weeks i.m.
Usual: 5 mg/2 weeks i.m.
Maximum: 10 mg/2 weeks i.m.
Callen-Lorde (2018)Estradiol valerateInitial: 10–20 mg/2 weeks
Maximum: 20–40 mg/2 weeks
 Estradiol cypionateInitial: 2.5 mg/2 weeks
Maximum: 5 mg/2 weeks
Davidson et al. / Tom Waddell Health Center (2013)Estradiol valerate or cypionateInitial: 20–40 mg/2 weeks i.m.
Average: 40 mg mg/2 weeks i.m.
Maximum: 40–80 mg/2 weeks i.m.
Bourns / Sherbourne Health / Rainbow Health Ontario (2019)Estradiol valerateInitial: 3–4 mg/week or 6–8 mg/2 weeks
Usual: Variable
Maximum: 10 mg/week
Trans Care BC (2019)Estradiol valerateInitial: 5 mg/week i.m. or s.c.
Usual: 10–20 mg/week i.m. or s.c.
Every 2 weeks at 2x dose may be tolerated in some
Dahl et al. / Vancouver Coastal Health (2015)Estradiol valerate20–40 mg/2 weeks i.m.
Note: “Alternative estrogen therapy for 3–6 months only”
European Society for Sexual Medicine / T’Sjoen et al. (2020)Estradiol valerate5–30 mg/1–2 weeks i.m.
 Estradiol cypionate2–10 mg/week i.m.
TransLine (2019)Estradiol valerateInitial/Usual: 5–10 mg/week
Maximum: 20 mg/week
 Estradiol cypionateInitial/Usual: 1.25–2.5 mg/week
Maximum: 5 mg/week

a Several other guidelines recommend doses and intervals that appear to be taken directly from the Endocrine Society or UCSF guidelines and thus are not listed here but can be found elsewhere (Aly W., 2020).

A number of concerns arise when the doses and intervals of injectable estradiol valerate and estradiol cypionate recommended by the major transgender clinical guidelines are considered in the context of the present informal meta-analysis and when they are compared between guidelines. Based on the present work, dosages of injectable preparations recommended by the major transgender clinical guidelines appear to result in estradiol exposure that is markedly higher than that with the recommended dosages for other routes and forms of estradiol (e.g., oral or transdermal). Whereas a dosage of 5 mg/week of any non-polymeric injectable estradiol ester appears to give average estradiol levels of around 300 pg/mL (~1,100 pmol/L), which are already supraphysiological, doses of injectable estradiol valerate or estradiol cypionate recommended by guidelines are as high as 15 to 20 mg per week. The average estradiol concentrations that would be expected to result from such doses per this meta-analysis (e.g., ~600–1,200 pg/mL or 2,200–4,400 pmol/L at 10–20 mg/week) (Figure 9) would vastly exceed the ranges for estradiol levels in transfeminine people advised by the same guidelines (generally about 50–200 pg/mL or ~180–730 pmol/L) (Table). These highly supraphysiological estradiol levels are not unexpected when normal production of estradiol in premenopausal cisgender women is considered (~1.4 mg per week or 6 mg per month/cycle giving mean estradiol levels of ~100 pg/mL or 370 pmol/L) (Aly W., 2019). Clinical safety data on high doses of injectable estradiol esters like estradiol valerate and estradiol cypionate are lacking at present, but excessive estrogenic exposure is known to increase the risk of health complications such as blood clots (Aly W., 2020). The very high doses of these preparations that are recommended by guidelines should raise considerable reservations about their safety.

Figure 9: Simulated estradiol levels with injectable estradiol valerate at the doses and interval (5–40 mg/2 weeks) preferentially recommended by current major transgender care guidelines. Steady-state estradiol levels are reached by about the third injection with this injection interval and levels do not further accumulate.

The present author elsewhere has listed doses of injectable estradiol preparations that are roughly comparable in terms of total estradiol exposure to doses for other estradiol forms and routes used in transfeminine people (Aly W., 2020). These doses range from about 1 to 6 mg per week for “low dose” to “very high dose” therapy with non-polymeric injectable estradiol esters (Graph). This dose range for injectable estradiol is likely to be more appropriate for use in transfeminine people than current recommendations by many guidelines. Although high estradiol levels can be useful in transfeminine hormone therapy when antiandrogens are not used due to their greater efficacy in terms of testosterone suppression than physiological levels, only modestly supraphysiological estradiol levels (e.g., ~200–300 pg/mL or 730–1,100 pmol/L) appear to be required for strong testosterone suppression (Aly W., 2019; Langley et al., 2021). In relation to this, doses of injectable estradiol need not be excessive.

Some guidelines such as the Endocrine Society guidelines recommend the same doses and intervals for both estradiol valerate and estradiol cypionate, whereas other guidelines such as the UCSF guidelines recommend different doses for these two injectable estradiol esters. Concerningly, the doses for estradiol valerate and estradiol cypionate recommended by the UCSF guidelines differ by roughly an order of magnitude (<20 to 40 mg/2 weeks for estradiol valerate and <2 to 5 mg/2 weeks for estradiol cypionate). These estradiol esters appear to produce similar average estradiol levels (e.g., around 300 pg/mL or 1,100 pmol/L at a dosage of 5 mg/week) and have concentration–time curve shapes that are not extremely different, with estradiol cypionate being only somewhat flatter and more prolonged than estradiol valerate. As such, it would appear that similar doses should be appropriate for these esters. This is supported by the fact that the same doses of estradiol valerate and estradiol cypionate are used in combined injectable contraceptives in cisgender women (both 5 mg once per month) and that these doses were carefully determined during an intensive clinical development programme for these preparations (Garza-Flores, 1994; Newton, d’Arcangues, & Hall, 1994; Sang, 1994; Toppozada, 1994). This programme notably included dose-ranging and direct-comparison studies. Based on the present analysis, the current recommendations by the UCSF guidelines may result in marked overdosage in the case of estradiol valerate and potential underdosage in the case of estradiol cypionate.

Transgender health guidelines recommend an injection interval for estradiol valerate and estradiol cypionate in oil of once every 1 to 2 weeks. Although an injection interval of 2 weeks seems technically feasible in the case of both of these preparations, such an interval would appear to result in substantial fluctuations in estradiol levels, with high peak levels and low troughs. This is particularly true in the case of the shorter-acting estradiol valerate (Figures 9, 10). Considering the wide fluctuations and unknown effects of this variability, as well as the fact that testosterone suppression when applicable may depend on sustained higher estradiol levels, it may be advisable that a once-weekly interval be preferentially recommended for these preparations. This would achieve steadier estradiol levels and would reduce potential problems due to high or low estradiol levels (Figure 10). Alternatively, a shorter interval of once every 5 days may be used with estradiol valerate to further reduce the variability in estradiol levels that occurs with this preparation (Figure 10). On the other hand, an injection interval of once every 10 days to 2 weeks may be practical and allowable in the case of the longer-acting estradiol cypionate in oil (as well as estradiol enanthate) (Figure 10)—provided that the injection cycles are well-tolerated and testosterone suppression remains adequate. When selecting different injection intervals, doses should be scaled by the interval to maintain equivalent total estradiol exposure (e.g., 3.5 mg/5 days, 5 mg/7 days, 7 mg/10 days, or 10 mg/14 days for high-dose non-polymeric injectable estradiol esters).

Figure 10: Simulated estradiol levels with a high dosage of injectable estradiol valerate or estradiol cypionate in oil at different injection intervals (doses scaled by interval to be equivalent in total estradiol exposure).

Considering the concerns about the doses and intervals of injectable estradiol preparations recommended by transgender care guidelines, the question of how these recommendations were determined arises. Unfortunately, current guidelines do not generally describe how they arrived at their recommendations nor do they usually cite sources to support them. It is notable that the UCSF guidelines recommend doses and intervals for injectable estradiol preparations that are nearly identical to those advised by Christian Hamburger and Harry Benjamin in 1969 in one of the first textbooks on transgender health (Hamburger & Benjamin, 1969 [PDF]). These authors recommended a dose of 10–40 mg/2 weeks for estradiol valerate and of 2–5 mg/2 weeks for estradiol cypionate (although Benjamin additionally stated that after 4–8 months, the same doses could be used at a longer injection interval of once every 4 weeks). These recommendations were notably made before estradiol blood tests became practicably available and were prior to the advent of modern pharmacokinetic studies. Hence, the recommendations for at least these guidelines appear to be based mainly on past expert opinion and long-standing historical precedent rather than on pharmacokinetic or clinical data. The same is likely to also be true for most other guidelines. High doses with certain injectable estradiol preparations (namely estradiol valerate) were probably originally employed for the purpose of achieving longer and more convenient injection intervals. This was notably prior to the risks of excessive estrogenic exposure like blood clots becoming known, and these doses may simply have never been revised. The reasons that dose recommendations for injectable estradiol in transfeminine people have remained as they have for so long may be because injections are not as commonly used as other forms of estradiol and because many clinicians may only test estradiol levels at trough with these preparations. This is noteworthy as trough levels only describe the lowest point of the estradiol curve with injectable estradiol preparations and can give a misleading impression of average or total estradiol exposure.

Among the surveyed guidelines for transgender hormone therapy, only the UCSF guidelines (Deutsch, 2016) and the Sherbourne Health/Rainbow Health Ontario guidelines (Bourns, 2019) referenced pharmacokinetic literature in their discussion of injectable estradiol. The specific publications cited by these guidelines were Düsterberg & Nishino (1982), Sierra-Ramírez et al. (2011), and Thurman et al. (2013). Although it is favorable to see guidelines considering published pharmacokinetic data for informing use of these preparations, there are a few concerns about the studies that were cited. Düsterberg & Nishino (1982) in its study of injectable estradiol valerate had a very small sample size (n=2), and this study was excluded as an outlier in the present meta-analysis due to unusually high estradiol levels. The findings of Düsterberg & Nishino (1982) also do not seem to have actually been used to guide dosing recommendations in the case of the UCSF guidelines, since if this were the case the recommended doses should have been much lower. On the other hand, Bourns et al. (2019) cited the same study and recommended injectable estradiol valerate at doses of 3–4 mg/week or 6–8 mg/2 weeks. These doses are well below those recommended by other transgender care guidelines and appear to be more appropriate for use in transfeminine people in light of the present meta-analysis. Sierra-Ramírez et al. (2011) and Thurman et al. (2013), although better-quality studies than Düsterberg & Nishino (1982), described injectable estradiol cypionate suspension rather than estradiol cypionate in oil. The oil-based version of estradiol cypionate is the form normally used in transfeminine hormone therapy, and there are important differences between these estradiol cypionate preparations such that pharmacokinetic studies for the suspension can’t necessarily be generalized to the oil solution. These preparations do in any case produce similar total estradiol levels however and hence doses should be similar for them.

This meta-analysis is only informal and unpublished research. Nonetheless, based on its findings and the preceding discussion, current dosing recommendations for injectable estradiol preparations by most transgender clinical guidelines appear to be highly excessive and likely unsafe, with injection intervals that may additionally be too widely spaced, and guidelines should consider reassessing these recommendations. In addition, the transgender medical community should make an effort to better characterize the pharmacokinetics and optimal dosing schemes of injectable estradiol preparations in transfeminine people in the future. Since clinical data on these preparations are scarce and will probably remain so in the near-term, use of published pharmacokinetic data should be further considered for guiding dosing recommendations for injectable estradiol. As identified and catalogued by this meta-analysis, there is a great deal of data in this area that could be used to better inform care guidelines with respect to the use of injectable estradiol preparations in transfeminine people.

Interactive Web Simulator

This informal meta-analysis of estradiol concentration–time data with injectable estradiol preparations was conducted for the purpose of deriving accurate and representative estradiol curves for incorporation into a web-based injectable estradiol simulator intended for use by transfeminine people and their clinicians. This web app is able to simulate both single-injection curves and repeated-injection curves with these preparations. An informational page for this simulator can be found at the following location:

And the injectable estradiol simulator itself can be found at the following page:

Future Possibilities

There are various possibilities for further work on this project in the future. For example, assessment of interindividual variability for estradiol levels with injectable estradiol preparations could be included in the meta-analysis. As another example, it would be fairly straightforward and valuable to expand the meta-analysis as well as simulator to other hormonal preparations such as injectable testosterone preparations and other estradiol routes and forms like oral estradiol, sublingual estradiol, and estradiol pellets. Pharmacokinetic literature for some of these preparations has already been collected by this author. However, these future possibilities would require much additional time and effort to complete.

Special Thanks

A special thank you to Violet and Lila for their indispensable input and guidance on modeling topics during the work on this project. An additional thanks to Violet for deriving a special three-compartment pharmacokinetic model that was used in this work. Please also check out Violet’s own projects Tilia—an effort to empower trans people with tools to manage their hormonal transitions—and TransKit—a work-in-progress pharmacokinetic simulation library specifically tailored for transgender hormone therapy. Lastly, thank you to all the peer reviewers who carefully reviewed this article prior to it being posted.

Supplementary Material