Menopause. 2010 (Nov); 17 (6): 1201–1205 ~ FULL TEXT
Rafael Bolanos, MD, MSc, and Jose Francia, MD, MSc
San Marcos University,
OBJECTIVE: The aim of this study was to conduct an indirect comparison of the results from meta-analyses that evaluated the incidence of osteoporotic vertebral fracture in postmenopausal women exposed to hormone therapy (HT) or isoflavones.
METHODS: Systematic review and meta-analysis of HT and isoflavones related to the reduction of vertebral fracture risk in osteoporotic women versus the same control (placebo) were undertaken. Then, the combination of the overall results obtained from these two meta-analysis (indirect comparison) was adjusted to the common control (placebo).
RESULTS: The indirect odds ratio (OR), obtained from the combination of both individual meta-analyses, was calculated by using the following equation: OR(indirect) = OR(HT)/OR(isoflav), with a total indirect variance equivalent to the following equation: var(total) = var(HT) + var(isoflav). These calculations yielded a point estimate of 1.56 (95% CI, 0.39-6.19) for the indirect OR.
CONCLUSIONS: According to this indirect comparison, there is no statistically significant difference between HT or isoflavones in the reduction of vertebral fracture risk due to osteoporosis, and both interventions seem to be similar for this outcome.
From the FULL TEXT Article:
There is growing evidence about the positive effect of
isoflavones in the inhibition of bone resorption and in
the increase in bone mineral density (BMD).  As far as
we know, there are very few clinical trials that have evaluated
the relationship between isoflavone consumption and osteoporotic
fracture risk as the final variable of interest, which is a
critical point when pondering the actual usefulness of these
agents. Moreover, such trials have been conducted with relatively
small samples and the incidence of fractures was a
secondary objective of the study. Nevertheless, the Shanghai
Women’s Health Study  is the largest cohort study published
to date that evaluated the association between the routine
consumption of soy foods and the incidence of fracture in
24,403 postmenopausal women without fracture or cancer
background, between March 1997 and May 2000. This study
concluded that high consumption of soy foods may reduce the
risk of fracture in postmenopausal women, particularly in the
early postmenopausal years. The results from a previous meta-analysis 
concluded that there is a significant trend in favor of
isoflavones for the reduction of osteoporotic fracture risk.
On the other hand, intervention with hormone therapy (HT)
to reduce the risk of osteoporotic fracture has been evaluated
in some previous meta-analyses [4, 5]; however, none of them
included the results from the Women’s Health Initiative
(WHI)  study, one of the most important clinical trials
involving HT. Therefore, it would be important to update the
evidence by including its results.
Indirect treatment comparison (ITC) is the method by
which two independent meta-analyses are combined, each one
confronting a different intervention versus the same control.
This analysis has proven to be very useful when there are no
"head-to-head" comparisons, and its methodology has been
reviewed in several previous publications.  Thus, this review
intends to establish an indirect comparison of the efficacy
between both interventions (HT and isoflavones) by using the
results found in the corresponding meta-analyses that compared
the efficacy of these interventions versus that of placebo.
The aim of this study was to conduct an indirect comparison
of the results from systematic reviews that evaluated the
incidence of osteoporotic vertebral fracture in postmenopausal
women exposed to HT or isoflavones.
Study inclusion criteria
The systematic review for each intervention (HT or isoflavones)
was developed by using the following research
question: Does the use of isoflavones or HT, as compared with
placebo, reduce the risk of vertebral fracture in postmenopausal
The search was limited to clinical trials, reviews, and metaanalyses
in middle-aged adult women (>45 y). No language
restrictions were made.
Search strategy for identification of studies
The following terms were initially used for searching the articles:
Search 1: (flavonoids or isoflavones) and (fracture)
Search 2: (flav* or isoflav*) and (fracture)
Search 3: (flav* or isoflav* or phytoestrogen) and (fracture
Search 4: (hormone replacement or estrogen therapy or hormone
therapy) and (fracture)
Search 5: (HRT or HT or ERT) and (fracture or bone)
The search for studies ended on September 15, 2009. We
looked up in the trials register of the Cochrane Osteoporosis
Treatment Study Group and in the Cochrane Controlled Trials
Register. We also searched the following databases: Medline,
Embase, ProQuest, BIREME, Trip Database, LILACS, and
The bibliographic research for the first three searching
formulas (previously described) detected three clinical trials,
one cohort study, and one meta-analysis.
The reference lists of these publications were thoroughly
reviewed to complete the search. Upon the completion of
this process, we could not find another prospective study or
systematic review that could join the five studies we had
previously found. We did not search for unpublished trials
or summary books of congresses on this topic, and we did
not make direct contact with the researchers of these studies.
We contacted several national and international experts to
request a bibliographic search supplementary to ours and to
ask them if they knew about other studies additional to
those already found by our team. However, these consultations
did not contribute any additional information to
what we already had as a result of our bibliographic search.
The bibliographic search for the last two systematic
searches detected 17 studies, including one cohort study and
Three reviewers analyzed the published trials independently
to assess their quality. The same reviewers also extracted
data from each study independently. We used the Jadad
scale,8 which comprises five basic criteria:
(1) presence of randomization,
(2) description and adjustment of randomization,
(3) presence of blinding,
(4) description and adjustment of blinding, and
(5) description of study losses and withdrawals. According to Jadad’s original study, every
experimental clinical trial that gathers three or more criteria may be considered of sufficient quality.
The indirect odds ratio (OR), obtained from the combination
of the results from both individual meta-analyses, was
calculated by using the following equation: OR(indirect) =
OR(HT)/OR(isoflav), with a total indirect variance equivalent to
the following equation: var(total) = var(HT) + var(isoflav). Random
effects model was used in each meta-analysis.
STATA version 10 was the software used for the indirect
meta-analysis. The results for each study were expressed in
ORs and 95% CIs. The heterogeneity analysis was conducted
for each meta-analysis that synthesized the evidence of each
intervention. We also conducted the meta-regression of all the
studies (confronting the log OR vs the type of treatment) to
estimate the residual heterogeneity among the studies. The
level of significance to evaluate heterogeneity was defined as
P G 0.05.
The characteristics of the studies included for the metaanalyses
of HT and isoflavones are shown in Tables 1 and 2,
The systematic review that corresponds to the efficacy of
isoflavones versus a placebo in the reduction of vertebral
fracture risk identified three controlled, randomized clinical
trials [9–11] and one prospective cohort study,  whose results
were combined in a subsequent meta-analysis.  Then, we
conducted a new meta-analysis considering only the inclusion
of the three clinical trials, since the cohort study did not show
its results in accordance with the type of fracture. We finally
used the results from this new meta-analysis for the indirect
comparison with HT (Table 3).
On the other hand, the systematic review that corresponded
to HT efficacy versus placebo in the reduction of vertebral
fracture risk identified 14 clinical trials, [12–20, 22–26] one cohort
study,  and two previous meta-analyses. [4, 5] Then, we conducted
a new meta-analysis versus placebo for the subsequent
indirect comparison with isoflavones (Table 4). Nevertheless,
of the 15 studies we found in our review, we chose only 10 for
the new meta-analysis, because 5 of them did not comply with
our inclusion criteria. [22–26] The reason for the exclusion was the
absence of report of the incidence of vertebral fracture.
The combined OR of the meta-analysis that compared HT
versus placebo under the random effects model was 0.68 (95%
CI, 0.46–1.01), with a variance of 0.04. The combined OR of
the meta-analysis that compared isoflavones versus placebo
under the random effects model was 0.44 (95% CI, 0.12–
1.63), with a variance of 0.45. The indirect OR calculation
yielded an estimated value of 1.56 (95% CI, 0.39–6.19). To
estimate the residual heterogeneity among the studies, a
weighted regression model was then fitted with the log OR as
the outcome and the type of treatment as the predictor. The
model weighted each study according to the inverse variance.
The random effects meta-regression was used because
it allows for unexplained variability between treatment
groups. The meta-regression model estimates a single T2 to
describe the residual heterogeneity among all trials having
accounted for the difference in estimates between trials
comparing placebo versus HT and trials comparing placebo
versus isoflavones (τ2 = 0.17). For this τ2 value, the percentage
residual variation due to heterogeneity was 35.87%
(I2 = 0.3587).
The previous meta-analysis from Bolanos et al,  which
evaluates the efficacy of isoflavones in the reduction of fracture
risk, is composed of three randomized clinical trials and one
prospective cohort study. The combination of results from
clinical trials (experimental designs), together with prospective
cohort studies (observational designs), is a valid model when
there is compatibility in the objectives of the study and when
both designs have sufficient quality to be included into the
combined analysis. It is very important to highlight this current
epidemiological trend because the combination of both designs
allows for more information when there are common objectives.  Nevertheless, such a cohort study was excluded from
our new indirect meta-analysis because it did not show results
regarding the type of fracture.
Because of the significant heterogeneity obtained when
we combined these three studies, it was pertinent to use the
random effects model to obtain the estimation for the overall
effect.  According to the final results, the overall effect of
isoflavones (combined OR) obtained an OR value of 0.44
(95% CI, 0.12–1.63), with a variance of 0.45, and these
results were used for this adjusted indirect comparison
The meta-analysis of Guyatt et al,  which compares HT
with placebo, combines the results from six clinical trials that
evaluated vertebral fracture risk among its objectives, but its
analysis did not include the results from the WHI trial,  which
could be considered a weakness, taking into account that the
WHI trial is a primary prevention study in the assessment of
osteoporotic fracture risk with the use of HT. The results from
this meta-analysis show an overall effect favorable to HT,
reaching an estimated OR value of 0.68 (95% CI, 0.46–
1.01). Although the previous meta-analysis of Torgerson and
Bell-Syer  considered a greater number of studies, its results
were similar to those of Guyatt et al ; however, it did not
include WHI trial findings either. This is why we decided to
conduct a new systematic review under the inclusion criteria
considered by our work team. We finally conducted the new
analysis including 10 studies that were valid enough to be
combined (Table 4).
The indirect OR, obtained from combining the results
from both individual meta-analyses, was 1.56 (95% CI,
0.39–6.19); thus, no significant difference was found between
Based on these reviews, the indirect comparison between
HT and isoflavones to reduce osteoporotic vertebral
fracture risk suggests that there is still no sufficient evidence
to ensure that there is a statistically significant and robust
difference between these two interventions. That is, both
could be equally efficacious versus placebo. Nevertheless, it
is important to point out some weaknesses of this study.
First, clinical trials with isoflavones that gather information
about vertebral fracture risk are still scarce. The combination
of the results shows an important heterogeneity, which must
be evaluated in future studies with similar designs. At this
point, there may be potential explanations for the lack of
efficacy of ipriflavone observed in the study of Alexandersen
et al ; for example, statistical power was not sufficient to
detect statistically significant differences between the ipriflavone
group and the placebo group because fracture incidence
was considered as a secondary endpoint, or it could be
questioned if the population studied was too old or had too
little bone mass. Second, in our analysis, we considered
vertebral fracture to be the most important outcome because
it occurs at a metabolically very active anatomic location and
it is usually an early complication of osteoporosis, although
its clinical assessment is more prone to error in subclinical
cases that have only radiological interpretation. Third, HT
was heterogeneous with respect to the route of administration,
the hormones used, and the dose, which could imply
important differences in the long-term efficacy and safety.
Fourth, in the studies that evaluated the intervention with
isoflavones, ipriflavone (synthetic isoflavone that is metabolized
to genistein [10%] by the action of the intestinal flora)
was used, which could generate differences in the magnitude
of the efficacy, considering that the main sources of isoflavone
(eg, soy) are composed of a mixture of these compounds
(genistein, daidzein, and glycitein).
It will be important to carry out new comparative (head-to-head)
clinical trials that assess these two available alternatives
(HT and isoflavones) in the fragility fracture risk,
considering that there are still very few studies that have
focused on this point and take BMD (and not the fracture
incidence) as the primary outcome of the analysis. For
example, in a comparative head-to-head design, Morabito
et al  evaluated the efficacy and safety of HT versus a
genistein concentrate in the bone mass. Both interventions
(HT and isoflavones) differed significantly from the placebo
with regard to the changes in BMD; however, the authors did
not find significant differences in the magnitude of the effect
between both treatments.
ITC can provide useful information, but this analysis may
have some limitation. ITC requires extrapolation from known
results to situations in which a study has not been done, and
therefore, the validity of ITC estimates may be questionable.
Significant differences may exist between trials that compare
one treatment to a control and trials that compare another
treatment to the same control. Thus, the two sets of trials may
be characterized by differences in patient characteristics, and
such heterogeneity between patients may result in a different
effect linking the treatment of interest. Differences in the
length of follow-up, measurement of outcomes, and diagnostic
criteria may also yield very weak results. 
Keeping in mind these observations, if a comparison
between two treatments is of relevance and randomized controlled
clinical trial (direct evidence) cannot be conducted,
investigators may proceed to an indirect comparison and
interpret, with caution, results based on such analyses. Making
reasonable ITCs is useful for healthcare decision makers who
face the option of performing an ITC or who rely on information
generated through indirect evidence. Moreover, combining
both direct and indirect evidence in the evaluation of
two interventions may provide more precise estimates, as
indicated by narrower CIs, than do results based on direct
evidence alone. 
According to this ITC, there is no statistically significant
difference between HT and isoflavones in the reduction of
vertebral fracture risk due to osteoporosis, and both interventions
seem to be similar for this outcome.
We thank Paola Laverde for her valuable and
professional support in the systematic review.
Ma D-F, Qin L-Q, Wang P-Y, Katoh R.
Soy isoflavone intake inhibits bone resorption and stimulates bone formation in menopausal women:
meta-analysis of randomized controlled trials.
Eur J Clin Nutr 2008; 62:155-161.
Xianglan Z, Xiao-Ou S, Honglan L, et al.
Prospective cohort study of soy food consumption and risk of bone fracture among postmenopausal
women (Shanghai Women’s Health Study).
Arch Intern Med 2005; 165:1890-1895.
Isoflavones and risk of fracture in postmenopausal women: systematic review and metaanalysis [abstract].
Ann Epidemiol 2009; 19:651.
Guyatt G, Wells G, Tugwell P, et al.
Meta-analysis of the efficacy of hormone replacement therapy in treating and preventing osteoporosis in
Endocr Rev 2002;23:524-539.
Torgerson D, Bell-Syer S.
Hormone replacement therapy and prevention of vertebral fractures: a meta-analysis of randomised trials.
BMC Musculoskeletal Disorders 2001;2:7.
Rossouw J, Anderson GL, Prentice RL, et al.
Risks and benefits of estrogen plus progestin in healthy postmenopausal women.
JAMA 2002;288: 321-333.
Glenny A, Altman D, Song F, et al.
Indirect comparisons of competing interventions.
Health Technol Assess 2005;9:1-134: iii-iv.
Jadad A, Moher D, Nichol G, Penman M, Tugwell P, Walsh S.
Assessing the quality of randomized controlled trials: an annotated bibliography of scales and checklists.
Control Clin Trials 1995;16:62-73.
Maugeri M, Panebianco P, Russo MS, et al.
Ipriflavone-treatment of senile osteoporosis: results of a multicenter, double-blind clinical trial of
Arch Gerontol Geriatr 1994;19:253-263.
Passeri M, Biondi D, Costi E, et al.
Effects of 2-year therapy with ipriflavone in elderly women with established osteoporosis.
Ital J Mineral Electrolyte Metab 1995;9:137-144.
Alexandersen P, Toussaint A, Christiansen C, et al.
Ipriflavone in the treatment of postmenopausal osteoporosis: a randomized controlled trial.
Cauley J, Robbins J, Chen Z, et al.
Effects of estrogen plus progestin on risk of fracture and bone mineral density.
Lufkin E, Wahner H, O’Fallon W, et al.
Treatment of postmenopausal osteoporosis with transdermal estrogen.
Arch Intern Med 1992;117:1-9.
A four-year randomized controlled trial of hormone replacement and bisphosphonate, alone or in combination, in women with post-menopausal osteoporosis.
Am J Med 1998;104:219-226.
Alexandersen P, Riis BJ, Christiansen C.
Monofluorophosphate combined with hormone replacement therapy induces a synergistic effect on bone mass by dissociating bone formation and resorption in postmenopausal women: a randomized study.
J Clin Endocrinol Metab 1999; 84:3013-3020.
Delmas P, Confavreux E, Garnero P, et al.
A combination of low doses of 17A-estradiol and norethisterone acetate prevents bone loss and normalizes
bone turnover in post-menopausal women.
Osteoporos Int 2000; 11:177-187.
Gallagher J, Fowler S, Detter J, Sherman S.
Combination treatment with estrogen and calcitriol in the prevention of age-related bone loss.
J Clin Endocrinol Metab 2001;86:3618-3628.
Cauley J, Black D, Barrett-Connor E, et al.
Effects of hormone replacement therapy on clinical fractures and height loss: the Heart and Estrogen/Progestin Replacement Study (HERS).
Am J Med 2001;110: 442-450.
Mosekilde L, Beck-Nielsen H, Sorensen O, et al.
Hormonal replacement therapy reduces forearm fracture incidence in recent post-menopausal women: results of the Danish Osteoporosis Prevention Study.
Recker R, Davies M, Dowd R, Heaney R.
The effect of low-dose continuous estrogen and progesterone therapy with calcium and vitamin D on bone in elderly women.
Arch Intern Med 1999;130:897-904.
Lindsay R, Hart D, Forrest C, Baird C.
Prevention of spinal osteoporosis in oophorectomised women.
Herrington D, Reboussin D, Broshnihan K, Sharp P, Shumaker S, Snyder
Effects of estrogen replacement on the progression of coronary artery atherosclerosis.
N Engl J Med 2000;343:522-529.
Rvan P, Bidstrup M, Wasnich R, Davis J, McClung M, Balske A.
Alendronate and estrogen-progestin in the long term prevention of bone loss: four-year results from the early post-menopausal intervention cohort study.
Arch Intern Med 1999;131:935-942.
Ishida Y, Soh H, Tsuchida M, Kawahara S, Murata H.
Comparison of the effectiveness of hormone replacement therapy, bisphosphonate, calcitonin, vitamin D and vitamin K in post-menopausal osteoporosis: a one year prospective randomized controlled trial.
The Writing Group for the PEPI Trial.
Effect of hormone therapy on bone mineral density.
Greenspan S, Bankhurst A, Bell N, et al.
Effects of alendronate and estrogen, alone or in combination on bone mass and turnover in postmenopausal
Bone 1998;23(Suppl 1):S174.
Shrier I, Boivin J-F, Steele R, et al.
Should meta-analysis of interventions include observational studies in addition to randomized controlled trials? A critical examination of underlying principles.
Am J Epidemiol 2007; 166:1203-1209.
Borenstein M, Hedges L, Higgins J, Rothstein H.
Fixed-effect versus random-effects models.
In: Introduction to Meta-Analysis, West Sussex, UK:
John Wiley & Sons, Ltd. 2009:77-86.
Morabito N, Crisafulli A, Vergara C, et al.
Effects of genistein and hormone- replacement therapy on bone loss in early postmenopausal women:
a randomized double-blinded placebo-controlled study.
J Bone Miner Res 2002;17:1904-1912.
Canadian Agency for Drugs and Technologies in Health.
Indirect treatment comparisons in meta-analysis. March 2009.
Return to the SOY PROTEIN Section