We here report a reversed-phase HPLC method for the determination of free cortisol in human urine, using methylprednisolone as the internal standard. Before chromatography, samples were extracted with a C18 solid-phase extraction column and the steroids were separated on a LiChrospher 100 C18 column with a mobile phase of methanol/acetonitrile/water (43/3/54 by vol). Linearity, precision, and accuracy of the method were established. The detection limit was 10 pmol of cortisol, and total CVs were <8%. With various solid-phase extraction columns the recovery of cortisol was 36–97%; recovery of the internal standard was 43–85%. Study of interference by 6 other steroids and metabolites and 24 drugs showed that carbamazepine and digoxin partly overlapped with cortisol, but this interference could be reduced by modification of the mobile phase. The HPLC method was compared with an RIA and an automated immunoassay method. The results obtained by HPLC averaged 40% of the RIA values.
- indexing terms: hydrocortisone
- sample preparation
- method comparison
- chromatography, reversed-phase
Measurement of urinary free cortisol excretion is clinically important, particularly in the diagnosis of Cushing syndrome (1)(2). As determined by mass spectrometry, the mean rate of cortisol secretion in healthy, unaffected adults is 27 ± 7.5 μmol/day (3), ∼53–93 nmol of which is daily excreted into urine as unmetabolized cortisol (4). Routine methods for determining urinary free cortisol concentrations are mostly based on competitive protein-binding (5), RIA (6), and HPLC (7)(8). RIAs are designed primarily for determinations of serum cortisol but are also applied to urine samples, either directly or after extraction of urine. Although the specificity of antisera used in most immunoassay systems is generally acceptable for serum cortisol measurements, urine contains many cross-reacting substances that interfere with immunological cortisol determinations (9). Some interfering compounds are known but most remain unidentified (9)(10)(11). The general performance of immunoassays is unsatisfactory (12). Because of cross-reacting substances, urinary free cortisol tends to be overestimated by RIA, the results obtained being generally two to three times greater than those obtained by HPLC (8)(11). HPLC methods, however, may be subject to interference from various coeluting substances. Before an HPLC analysis, separation of cortisol from interfering compounds, either by solvent or solid-phase extraction, can effectively improve the specificity of urinary free cortisol assays (11)(13)(14)(15).
When evaluating published HPLC methods for cortisol, we encountered problems from interfering drugs. We have therefore developed a new HPLC method for determining urinary free cortisol, optimizing the mobile-phase composition so that interference from drugs and other steroids is minimized. We also tested several types of solid-phase extraction columns for the prepurification of urine samples.
Materials and Methods
For HPLC analyses we used a chromatographic system consisting of two 2150 HPLC pumps combined with a high-pressure mixer and a 2152 HPLC controller (LKB-Produkter, Bromma, Sweden), a Waters 717plus Autosampler (Millipore, Milford, MA), a LaChrom column oven L-7350 (Merck, Darmstadt, Germany), and a 785A detector (Applied Biosystems, Foster City, CA) combined with a Hewlett-Packard ChemStation data system (Hewlett-Packard, Waldbronn, Germany). The 250 × 4.6 mm reversed-phase C18 column was a LiChrospher 100 (5 μm; Merck). Disposable filter units, Millex-HV 0.45-μm (average pore size), were from Millipore S.A., Molsheim, France.
We purchased hydrocortisone, cortisone, 6α-methylprednisolone, 11-deoxycortisol, and corticosterone from Sigma Chemical Co., St. Louis, MO. Methanol and acetonitrile, from Rathburn Chemicals (Walkerburn, Scotland, UK), were of HPLC grade. Solid-phase extraction columns were from J.T. Baker (Deventer, Holland), International Sorbent Technology (Mid Glamorgan, UK), and Millipore (Milford, MA). All other chemicals were of analytical reagent grade.
Stock calibration solutions (200–375 μmol/L) of steroids were prepared in methanol. The working internal standard solution (IS), prepared by diluting the stock solution of 328 μmol/L with methanol to 8 μmol/L, was stored at 4 °C.
We collected 24-h urine specimens with no preservatives. If analysis was delayed, they were stored at −20 °C. Before analysis, the samples were filtered and 25 μL of 8 μmol/L IS was added to 2 mL of urine. The steroids were then extracted with 3-mL (500 mg) “Bakerbond C18” cartridges (J.T. Baker), which had been activated with 2 mL of methanol followed by 2 mL of water. After application of the samples, the cartridges were washed with two 2-mL aliquots of 25 mmol/L borate buffer, followed by acetone, 200 mL/L, in water. One milliliter of hexane was added and the cartridges were air-dried under reduced pressure for 4 min. The steroids were eluted with two 1-mL aliquots of ethyl acetate. The eluate was dried under nitrogen and dissolved in 75 μL of 400 mL/L methanol. We then injected 25 μL of the reconstituted sample into the HPLC system.
Chromatographic conditions and calculations.
The mobile phase was methanol, acetonitrile, and water (43:3:54 by vol). The system was run isocratically at 40 °C with a flow rate of 1 mL/min. The detection wavelength was 242 nm. Integration was performed by the valley-to-valley method. Urinary free cortisol concentrations (UFC) were calculated from peak areas of internal standard (IS) and cortisol (C) as follows: UFC = (C peak area/IS peak area) × urine IS (amount per liter) × f, where f is a correction factor [f = relative response (IS/C) × relative recovery (IS/C)]. Routinely, the amount of IS per liter of urine is 100 nmol/L, and f = 0.87. The relative response obtained from the ratio of IS to C peak areas averaged 0.99, and the relative recovery (the ratio of absolute recoveries of IS and C added to urine), 0.88 (see Results).
. Cortisol RIA kits were from Orion Diagnostica (Espoo, Finland). Urinary cortisol was also measured, without extraction, by the Technicon Immuno 1 analyzer (Bayer, Tarrytown, NY), which uses a competitive immunoassay format with two incubations. First, the cortisol–antibody conjugate and enzyme conjugate are reacted with the patient’s sample; monoclonal immunomagnetic particles are added in a second incubation. The particles are washed and the enzymatic activity of the complex is determined with p-nitrophenyl phosphate as substrate.
The linearity of the method was tested for cortisol and IS. Increasing amounts of analytes, from 10 to 1000 pmol in 400 mL/L methanol, were injected.
Analytical recovery of the various solid-phase extraction columns was evaluated by adding from 50 to 1000 pmol of cortisol and IS to a 2-mL urine sample and processing them through the whole assay procedure.
Various drugs (see Table 3⇓ ) commonly administered to patients were obtained from the hospital pharmacy. The drugs were dissolved in methanol and further diluted with 400 mL/L methanol before injection.
The regression equations were calculated by the method of standardized principal component (16), and the coefficient of correlation was determined by linear regression.
Linearity and sensitivity.
Calibration curves for peak areas (y) vs quantity of cortisol (x) were linear from 10 to 1000 pmol (Fig. 1⇓ ). The minimum detectable amount of cortisol and IS was 10 pmol. Using 2 mL of urine and injecting 25 μL of the reconstituted extract corresponds to 15 nmol/L in urine. A relative response factor of IS compared with that of cortisol was calculated and found to be 0.99 for the entire range. This factor was incorporated into the formula for internal standard quantification (see above).
Absolute recoveries of cortisol and IS were slightly different. With Bakerbond C18 columns the mean (n = 5) recovery of 50–800 pmol of added cortisol and IS was 97% and 85%, respectively. Thus we incorporated a relative recovery factor (IS/cortisol) of 0.88 in the equation for calculation of the final results.
As shown in Table 1⇓ , the within-assay CV calculated from values for two samples (321 and 56.6 nmol/L, 20 replicates each) was 2.1% and 4.4%; the total CVs were 6.2% and 7.7% (387 and 97 nmol/L, 20 replicates). These determinations were conducted on urine pools stored in frozen aliquots and thus reflect the entire process, including solid-phase extraction.
Chromatographic separation of cortisol.
Typical chromatograms of a steroid calibrator and urine samples are shown in Fig. 2⇓ . Cortisol elutes as a sharp symmetrical peak at 21.3–21.7 min and the IS at 39.3–40.1 min. The retention-time instabilities (CVs) over a run of 20 samples for cortisol and IS were 2.8% and 3.0%, respectively; the retention-time ratio, however, remained unchanged: 0.54. The majority of urine samples gave chromatograms similar to that shown in Fig. 2B⇓ .
Solid-phase extraction columns.
We tested eight different solid-phase extraction columns by adding 50–1000 pmol of cortisol and IS to urine samples and taking them through the procedure described in Materials and Methods. The best recoveries were obtained with Bakerbond C18 columns (Table 2⇓ ).
Correlation between methods.
The correlation between the RIA and our HPLC assay was determined with 88 patients’ samples (Fig. 3⇓ , A and B). The correlation by the standardized principal component method was: RIA = 1.98 HPLC − 22.5, r = 0.78. The correlation between the RIA and Immuno 1 analyzer (Fig. 3C⇓ ) was Immuno 1 = 1.29 RIA + 20.2 (r = 0.96, n = 68) and that between Immuno 1 and our HPLC (Fig. 3D⇓ ) was Immuno 1 = 1.49 HPLC + 91.8 (r = 0.81, n = 70). RIA and Immuno 1 analyzer gave values ∼1.5–2-fold higher than HPLC; moreover, the Immuno 1 analyzer correlated poorly, and the intercept on the y-axis was unacceptably high.
We analyzed 24-h urine samples (n = 28) from apparently healthy individuals who, to our knowledge, did not have any adrenal-related disease. The results ranged from 30 to 145 nmol (mean = 83, median = 76, and SD = 32 nmol). A provisional upper reference limit based on the 95th percentile was 144 nmol (52 μg); the lower limit (5th percentile) was 30 nmol (11 μg).
Some urine samples contained substances that partly overlapped with cortisol. Interference studies were conducted with some steroids and several common drugs. Table 3⇓ shows their relative retention times and shows that most of the commonly administered medications do not interfere. Only prednisolone completely overlaps with cortisol. Carbamazepine, with a relative retention time of 0.51, and digoxin (0.52) can ordinarily be separated from cortisol (0.54). If their concentrations are very high compared with that of cortisol, the peaks tend to overlap. Reducing the percentage of methanol in the mobile phase completely separates them from cortisol but the analysis times are longer.
Serum and urinary free cortisol concentrations are mostly measured by RIA. Results of quality-assessment schemes show that performance of serum cortisol RIAs is relatively satisfactory, but not RIA measurements of urinary free cortisol (12). The major problem of measuring urinary free cortisol is the large number of interfering substances in urine. Many steroids are metabolized and excreted as conjugates into urine, in concentrations possibly several orders of magnitude higher than those of the intact hormone. Because they may cross-react in immunoassays, conjugates form a particular problem in direct assays. HPLC methods that largely eliminate the problem of conjugate cross-reactivity have been described (11)(14)(15), but our initial studies indicated that certain commonly used drugs may interfere in these.
Our HPLC cortisol method was optimized to eliminate interferences by drugs commonly encountered in hospital patients. If the sample contains high concentrations of carbamazepine or digoxin, the percentage of methanol in the mobile phase has to be slightly reduced; the only drawback of this change is a fairly long analysis time of ∼1 h per sample. Prednisolone also interferes, as appears also to be the case in other HPLC methods. Possible interference from other drugs not encountered in the present study should be considered. However, the provisional upper reference limit established for urinary free cortisol, 144 nmol/day, is in agreement with previously reported values determined by HPLC (10)(13)(17).
Solid-phase extraction is a convenient way to extract steroids from urine, but there are large differences in extraction recovery of various columns (Table 2⇑ ). The two most efficient columns are Bakerbond C18 and Isolute C18MF. In any case, the slight difference in recovery between cortisol and the IS must be taken into account in calculation of final results. Our method, with solid-phase extraction before HPLC, should further be useful for determination of other urinary glucocorticoids such as cortisone and 11-deoxycortisol (Fig. 2A⇑ ). The IS used, 6α-methylprednisolone, elutes quite late and does not overlap with other peaks.
In conclusion, the results obtained by this HPLC are about half the quantities obtained by immunoassay, and especially in the low range of values the correlation is poor. This is in agreement with findings from earlier studies (8)(13). Although our HPLC method has not been validated against a Reference Method, the similarity of our results and those obtained by GC-MS (4) suggests that HPLC results are relatively accurate. In contrast, the direct immunoassay of urinary cortisol does not fulfill generally accepted quality requirements.
- © 1997 The American Association for Clinical Chemistry