[10 – 12], a lipoprotein newly recognised as indepen- dently associated with the development of atherosclero-
sis both at coronary and cerebrovascular levels [13 – 15]. Nevertheless, the established effects of exogenous
androgens on HDL and Lpa, have not been defini- tively confirmed for endogenous sex steroids. Indeed,
although an increase of HDL-C and ApoAI has been shown in response to androgen suppression induced by
both orchidectomy or chemical castration [16 – 20], male hypogonadism is not associated with low HDL-C levels
[21]. Also, studies on the correlations between androgen and lipoprotein blood concentrations gave rather confl-
icting results. Indeed, HDL-C has been reported as positively related to testosterone T and dihydrotestos-
terone DHT serum concentrations in man by several studies [22 – 25], although others [26,27] have found no
significant associations. Similarly, no significant associ- ations have been reported between Lpa and androgen
concentrations [28,29], with the exception of a positive correlation
to dehydroepiandrosterone-sulfate
[29]. Therefore, in spite of the abundance of data for the
importance of endogenous testosterone in normal HDL metabolism, its precise role has not yet been clarified.
Besides, it is not clear whether in humans androgen action on lipids is exerted by testosterone itself or, as it
occurs at the level of other androgen targets, by some specific end-products or intermediates in the pathway
of testosterone liver metabolism, such as DHT or estra- diol [30].
The aim of this study was to assess whether in males the inhibition of liver 5 alpha-reductase and, as a
consequence, the decrease in DHT can modify lipo- protein circulating concentrations.
To this purpose, the effects on lipids of finasteride, a 5 alpha-reductase inhibitor, recently introduced as
treatment of benign prostatic hyperplasia BPH were studied [31].
2. Materials and methods
Twenty eight patients, aged 64.4 9 6.3 M 9 S.D. range: 55 – 74, consecutively referred to the outpatient
care centre of the department of Urology of the Univer- sity of Parma Italy for BPH, were studied. They all
gave informed written consent before participating in the study. The study protocol was approved by the
ethical committee of the University of Parma.
BPH was diagnosed on the presence of symptoms of lower urinary tract obstruction and of enlarged
prostates, documented by digital rectal examination and transrectal ultrasonography. However, because of
the need for an untreated control group, only patients in an early phase of the disease and with symptoms of
mild obstruction were enrolled into the study. A com- plete clinical and laboratory evaluation was performed
before randomisation, to exclude the presence of dia- betes mellitus, thyroid disease, severe renal or liver
failure; among blood tests, also Protein C assay was performed, to exclude any possible intercurrent inflam-
mation that could per se induce changes in lipoproteins, namely Lpa concentrations. No one of the patients
enrolled was taking medications known to influence lipoprotein metabolism or alcohol in excessive daily
intake.
They were randomly allocated to one of two groups: finasteride group 13 patients treated with Finasteride
at the dosage of 5 mg o.i.d. PROSCAR, Merck Sharp and Dome for 6 months and control group 15 un-
treated patients. All patients were asked not to change their diet habits and their physical activity throughout
the experimental period.
Before starting therapy and after 3 and 6 months, blood samples were collected to assay DHT, Estradiol
E2, free testosterone fT, as well as to assess lipid profile. Blood samples were drawn at 08:00 – 09:00 h,
after an overnight fast; they were centrifuged at 3000 rpm for 10 min and serum was stored at − 80°C until
assay. Each series of assays was performed in a single batch.
Total cholesterol TC and triglycerides, TG were assayed by an enzymatic-colorimetric reaction, employ-
ing reagents from Boerhinger Mannheim, FRG. HDL-C was assayed by the same procedure as TC,
after precipitation of VLDL and LDL by phospho- tungstic acid and magnesium ions. HDL3-C was as-
sessed on supernatant, after precipitation of LDL, VLDL and HDL2 with polyethylene glycol PEG at
concentration of 7 mmoll Immuno AG, Vienna, AU. HDL2-C was obtained by calculating the difference
between total HDL-C and the HDL3-C subclass. LDL- C was estimated by Friedewald formula [32].
Lpa was assayed by one step ELISA, employing reagents purchased from Immuno A.G. Vienna, Aus-
tria. Monospecific, polyclonal anti-apo a antibodies in solid phase were the capture antibodies, whereas the
detection antibody was a monospecific and monovalent anti-apo a coupled with peroxidase. Reference materi-
als were a mixture of 4 – 6 Lpa isoforms. Values have been expressed in terms of total lipoprotein mass, with
a minimum detection limit of approximately 10 mgl. Inter- and intra-assay coefficients of variation for three
concentrations of Lpa 110 mgdl, 250 and 420 mgl were respectively: 7.6, 6.9, 5.3 and 5.2, 5.5, 4.2.
FT, E2 and SHBG were assayed by a radioimmuno- logical procedure RIA, employing reagents from Di-
agnostic Products Corporation Los Angeles, CA. Minimum detection limits were: 0.52 nmoll for fT,
29.36 pmoll for E2 and 0.04 nmoll for SHBG. Inter- assay and intra-assay coefficients of variation for three
concentrations were: for fT 16, 67, 140 pmoll, 5.5, 3.7, 3.4; and 4.3, 4.0, 3.2; for E2 74, 166, 248
pmoll 14.0, 7.5, 9.0; and 12.0, 5.2, 7.5; for SHBG 10.8, 64, 116 nmoll, 3.1, 5.3, 6.9; and 2.8, 3.0, 3.6.
DHT also was determined by RIA, after extraction of samples with a mixture of n-exane and ethanol 98
exane; 2 ethanol, employing a commercial kit from Diagnostic Products Corporation Los Angeles, CA.
Minimum detection limit was 0.01 nmoll. Inter-assay and intra-assay coefficients of variation for three con-
centrations 0.33, 1.00, 2.36 nmoll were respectively: 8.5, 2.3, 8.4 and 6.2, 4.5, 3.1.
On the same occasion of blood sample collection, a standardised interview was performed to exclude any
possible substantial modification of lifestyle, in particu- lar smoking habits and exercise, and dietary habits,
assessed by self-reported alcohol intake and 24-h recall. Nutrient intake was estimated by 24-h recall, using
computerised food tables Dieta 2000, Computek s.r.l. Verona, Italy.
In addition, an assessment of body composition was performed by both anthropometry and bio-impedance
analysis BIA. The following anthropometric measure- ments were taken: weight and height; skinfold thick-
ness, measured in triplicate to the nearest 1 mm at triceps, biceps, subscapular and suprailiac sites with a
Harpender skinfold caliper; circumferences, measured at waist, hip and thigh. From anthropometry the fol-
lowing indexes were calculated: body mass index BMI, kgm
2
as the ratio between weight and height; fat mass percent FMSKF, from four skinfold thickness, by
Durnin and Womersley [33] and Siri equations [34]; the ratio of waist to hip circumferences WHR. In addi-
tion, whole body bioelectric resistance R was mea- sured at 800 mA and 50 kHz with a portable impedance
analyser type BIA, Akern, Firenze, Italy. Resistance R, reactance X and the angular transformation of
the ratio of R to X, the phase angle PA were regis- tered. From bioelectric parameters, fat mass percent
was estimated by the equations developed by Segal et al. [35] or Deurenberg et al. [36], according to age
under or over 62 years FMBIA.
Statistical analysis was performed by a computer employing the SPSS-X program Statistical Package for
the Social Sciences, University of Pittsburgh. For each parameter, the normality of distribution by
Lilliefors test was assessed, which showed a normal distribution for all variables with the exception of TG
and Lpa. For these variables, statistical analysis was performed by both non-parametric tests and parametric
tests on log-transformed values. So, the differences from baseline during treatment were tested for signifi-
cance by ANOVA for repeated measures, as well as Friedman test when appropriate. However, results were
represented by non-transformed values.
Bivariate correlation analysis and stepwise multiple regression analysis were used to study the association
among the absolute changes in lipid, hormone and body composition.
Finally, agglomerative hierarchical cluster analysis was employed to identify patient subsets according to
their different responses in HDL subclasses. Also, the distance between cases was measured as the squared
Euclidean distances and clusters were combined by the average linkage method.
3. Results