ARTICLE
Auteur(s) : Claire
Deloche1, Philippe Bastien1, Stéphanie
Chadoutaud1, Pilar Galan2, Sandrine
Bertrais2, Serge Hercberg2, Olivier de
Lacharrière1
1L’Oréal Recherche, Clichy, FranceFax: (+33)1 47 56
82 21
2U557 Inserm (UMR INSERM/INRA/CNAM), 5 rue du Vertbois,
Paris, 75003 France
accepté le 6 Juin 2007
Hair loss in women is a common trait and the prevalence is age
dependent [1]. When a woman consults a dermatologist about hair
loss, her condition may or may not lead to alopecia, defined as a
decrease in hair density. Patterned androgenic alopecia occurs in
37% of post-menopausal women [2] but only in 10-13% of
non-menopausal women [2, 3]. Another hair trouble in non-menopausal
women is increased hair loss or hair shedding, also known as
chronic telogen effluvium (CTE) [3], affecting around 30% of women
in the USA, UK and Japan [4]. Both diffuse androgen-dependent
alopecia and chronic telogen effluvium are a major concern in
dermatology [5]. However, it is important to remember that in
clinical practice, women’s complaint of hair loss is still unclear
or controversial [6, 7]; other causes must be taken into
consideration.Iron deficiency has been suspected to represent one
of the possible causes of excessive hair loss in women. Iron
deficiency has been reported to be associated with chronic diffuse
hair loss [5, 6, 8, 9]. A total iron depletion was observed when
the serum ferritin was below 15 µg/L, and low iron stores with
serum ferritin was between 15 and 30 µg/L [10]; Rushton [11]
reported that the critical threshold of serum ferritin was 40 µg/L,
a level under which increased telogen hair shedding was observed.
So far, a direct relationship between ferritin levels and hair loss
has not been confirmed by other studies [12, 13] and is still a
matter of debate [14-16]. Some evidence was provided in a recent
report [17] that decreased serum ferritin is associated with hair
loss or alopecia in women; however, several parameters in this
study, such as sample sizes, the ferritin thresholds and the
clinical features [16, 17] need further explanation.To further
investigate the relationship between iron store and hair loss, we
decided to explore this relationship by assessing hair loss and
measuring serum ferritin concentration, which is closely related to
iron stores [10], in a very large sample of subjects by taking the
advantage of the SU.VI.MAX epidemiological study, which provided us
with the opportunity to study, for the first time, a large cohort
of 5110 women. The aim of the study was to evaluate a possible link
between iron stores based on the assessed ferritin level, and hair
loss in women.
Materials and methods
Volunteers
5110 women aged between 35 and 60 years were involved in the study.
This cohort of volunteers was part of the SU.VI.MAX trial, a large
French epidemiological study [18, 19] approved by the ethical
Committee of Paris-Cochin. SU.VI.MAX, which stood for
“SUpplementation en VItamines et Minéraux AntioXydants”, was a
randomized double blind, placebo-controlled, primary-prevention
trial designed to test the efficiency of daily supplementation with
anti-oxidant, vitamins and minerals at nutritional doses, in
reducing the incidence of cancers and ischemic vascular diseases in
a middle-age general population. Detailed description of the
SU.VI.MAX study-design, randomization and participant
characteristics have been previously reported [18, 19].
Hair loss assessment
Hair loss was assessed with the help of a set of descriptive
questions extracted from a self-assessment questionnaire which had
been sent to all volunteers at inclusion in the study (table 1). The quantification of hair loss was
estimated by quantification of the hairs lost during washing,
brushing, after drying with a towel, on the pillow after a night’s
sleep and on clothes. Hairs that shed were quantified as none or a
little, a few and a lot of hairs. Questions concerning the degree
of self-perceived hair loss were also provided to establish if the
volunteers consider themselves to be concerned by hair loss (non
existent, normal, abnormal) and were included in the data analysis.
Multiple correspondence analysis and hierarchical cluster analysis
(HCA) [20] were used to group the answers with similar expression
patterns. This allowed us to identify three distinct groups of
volunteers characterized by “absence of hair loss”, “moderate hair
loss” and “excessive hair loss” (table
2).
Table 1 Hair loss questionnaire
|
• Do you feel involved by hair loss?
|
□ Yes
|
□ No
|
|
- • If you feel involved by hair loss, do you estimate that your
hair loss corresponds to a transient hair loss?
- do you estimate that your hair loss corresponds to a persistent
hair loss?
|
|
|
|
|
• Currently, during hair washing, how much do you estimate your
hair loss?
|
□ Many hairs
|
□ Few hairs
|
□ Very few or None
|
|
• Currently, drying your hair with a bath towel, how much do you
estimate your hair loss?
|
□ Many hairs
|
□ Few hairs
|
□ Very few or None
|
|
• Currently, during hair brushing, how much do you estimate your
hair loss?
|
□ Many hairs
|
□ Few hairs
|
□ Very few or None
|
|
• Currently, after a night sleep, how much do you estimate your
hair loss on the pillow?
|
□ Many hairs
|
□ Few hairs
|
□ Very few or None
|
|
• Currently, during a day how much do you estimate your hair loss
on your clothes?
|
□ Many hairs
|
□ Few hairs
|
□ Very few or None
|
Table 2 Hair loss classification according to
descriptive questions on hair loss signs and hierarchical cluster
analysis
|
Cluster 1“Absence of Hair Loss”
|
Cluster 2 “Moderate Hair Loss”
|
Cluster 3 “Excessive Hair Loss
|
|
“not concerned by hair loss”
|
“hair loss self-perceived as normal hair loss”
|
“hair loss self-perceived as abnormal hair loss”
|
|
“no hair loss or a little during the washing”
|
“lose a few hairs during the washing”
|
“lose a lot of hairs during the washing”
|
|
“no hair loss or a little during hair brushing”
|
“lose a few hairs during hair brushing”
|
“lose a lot of hairs during hair brushing”
|
|
“no hair loss or a little on the bath towel”
|
“lose a few hairs on the bath towel”
|
“lose a lot of hairs on the bath towel”
|
|
“no hair loss or a little on the pillow”
|
|
“lose a few or a lot of hairs:
|
|
“no hair loss or a little on the clothes”
|
on the pillow”
|
|
“lose a few or a lot of hairs on the clothes”
|
Biochemical measurements
The iron status was assessed by measuring the serum ferritin level
and haemoglobin levels in each volunteer. At entry in the study, a
35 mL venous blood sample was collected. After collection (Becton
Dickinson tubes), haemoglobin was measured immediately
(cyanmethemoglobin method) and kept at + 4 °C in the dark
until centrifugation and preparation of the aliquots. Serum
ferritin levels were measured using a nephelometric assay (BNII
Berhing) [21].
Statistical analysis
Descriptive and inferential analyses were performed using SPSS
11.0, SPAD 5.1 and SAS 8.2 statistical softwares. The detection
limit of the serum ferritin assay caused some departure from
normality even for log-transformed data. Therefore analyses based
on ranks have been preferred. Analysis of variance (ANOVA) of serum
ferritin level based on ranks has been carried out independently in
post-menopausal and non-menopausal women, taking into account the
age factors, “hemoglobin” and “hair loss classes” (table 2). For non-menopausal women, the use of
intra-uterine devices or contraceptives was also taken into
consideration. Comparisons of mean ranked serum ferritin levels
relative to the hair loss classes were carried out using the
Tukey-Kramer procedure for pairwise comparisons [22]. The
relationship between the hair loss classes and serum ferritin
levels were fitted using generalized logits [23] controlled for
age, haemoglobin, and contraceptives or intra-uterine devices. The
“excessive hair loss” class was chosen as the reference class; the
significance level was set up at 0.05.
Results
Hair loss quantification
Based on the questionnaire, hierarchical cluster analysis (HCA)
using Ward’s criteria clearly identified three well-defined classes
of women [20]. The planar projection of individuals (figure 1) displays the
three clusters on the first factorial plan. The clusters are
stretched along the first component, corresponding to an intensity
hair loss axis. The U-shape representation of the clusters is
characteristic of a Guttman effect [20] which opposes the classes
“absence of hair loss” and “excessive hair loss” classes to
“moderate hair loss” class on the second factorial axis. The three
classes were characterized using modalities over-expressed in
respect to the whole population (table
2). The ellipsoids displays the classes “absence of hair
loss”, “moderate hair loss”, and “excessive hair loss” which
correspond to 43%, 48%, and 9% of the 5110 women, respectively.
Each point on this figure corresponds to a particular profile and
is sized proportionally to the frequency of the profile in the
studied population.
Link between hair loss and serum ferritin levels
In non-menopausal women, 41.5%, 48.6% and 10% displayed absence,
moderate and excessive hair loss, respectively (table 3A). Among them, 23% (881/3759) presented
iron depleted stores (serum ferritin levels < 15 µg/L). In
addition, 57% [(881+1269)/3759] of non-menopausal women presented
low iron stores (serum ferritin < 40 µg/L) (table 3A, figure 2).
In post-menopausal women, the iron loss caused by bleeding
during menstruation does not occur anymore. Among them, 47.4%, 46%
and 6.6% display absence, moderate and excessive hair loss
respectively (table 3B). At that time,
only 5% (67/1351) had depleted iron stores (serum ferritin <
15 µg/L) (table 3B). Similarly,
the low iron stores (serum ferritin < 40 µg/L) affected
only 23% [(67+243)/1351] of post-menopausal women versus 57% of
non-menopausal women.
The detection limit of the serum ferritin assay caused some
deviation from normality, even for log-transformed data. Since the
parametric analyses could thus be biased, we decided to perform
analyse based on ranks, The ANOVA based on ranks showed a
significant effect of the mean serum ferritin level on hair loss
(p = 0.01) only in non-menopausal women. The Tukey-Kramer test
for pairwise comparisons showed significant differences between
excessive hair loss and both absence (p = 0.01) and moderate
(p = 0.02) hair loss (figure 3).
In non-menopausal women, with serum ferritin levels below
40 µg/L or below 15 µg/L, excessive hair loss was
significantly more frequent (11.4% [(90+156)/(881+1269)] and 10.2%,
respectively), compared to women with optimal levels of ferritin,
above 70 µg/L (6.8%) (table 3A). Anemia
(defined as hemoglobin < 120 g/L) concerned 10%
[(173+43)/(881+1269)] and 19.6% of women presenting a serum
ferritin level lower than 40 µg/L or 15 µg/L, respectively. With a
serum ferritin level above 40 µg/L, only 1.8% [(23+7)/(937+672)] of
the women showed anemia (table 4). Thus,
a relationship between hair loss and anemia was noticed only when
the body’s iron stores was below 40 µg/L.
Adjusted for age, haemoglobin, and the use of contraceptives or
intra-uterine devices [21], generalized logits based on ranks for
the absence versus excessive hair loss and moderate versus
excessive hair loss were performed. A generalized logits model was
preferred over a conventional polytomic logistic regression, since
the assumption of proportional odds was rejected (score test for
the proportional assumption, p = 0.009). The results showed
that low serum ferritin levels appeared to be highly significantly
linked to the presence of excessive hair loss in non-menopausal
women (p = 0.005 for excessive hair loss versus absence of
hair loss; p = 0.005 for excessive hair loss versus moderate
hair loss). Model adequacy was supported by Hosmer and Lemeshow
Goodness-of-Fit test [23]. In addition, the model based on the log
transformed ferritin data (figure 4) shows similar
results.
Table 3A Impact of iron store levels on hair loss
status in non-menopausal women (n = 3759)
|
Ferritin (µg L-1)
|
|
Hair loss
|
|
Total
|
|
Absence of hair loss
|
Moderate hair loss
|
Excessive hair loss
|
|
< 15
|
Count
|
365
|
426
|
90
|
881
|
|
% within ferritin
|
41.4%
|
48.4%
|
10.2%
|
100.0%
|
|
[15-40]
|
Count
|
512
|
601
|
156
|
1269
|
|
% within ferritin
|
40.3%
|
47.4%
|
12.3%
|
100.0%
|
|
[40-70]
|
Count
|
385
|
469
|
83
|
937
|
|
% within ferritin
|
41.1%
|
50.1%
|
8.9%
|
100.0%
|
|
≥ 70
|
Count
|
297
|
329
|
46
|
672
|
|
% within ferritin
|
44.2%
|
49.0%
|
6.8%
|
100.0%
|
|
Total
|
Count
|
1559
|
1825
|
375
|
3759
|
|
% within ferritin
|
41.5%
|
48.6%
|
10.0%
|
100.0%
|
Table 3B Impact of iron store levels on hair loss
status in post-menopausal women (n = 1351)
|
Ferritin (µg L-1)
|
Hair loss
|
Total
|
|
Absence of hair loss
|
Moderate hair loss
|
Excessive hair loss
|
|
< 15
|
Count
|
30
|
30
|
7
|
67
|
|
% within ferritin
|
44.8%
|
44.8%
|
10.4%
|
100.0%
|
|
[15-40]
|
Count
|
106
|
117
|
20
|
243
|
|
% within ferritin
|
43.6%
|
48.1%
|
8.2%
|
100.0%
|
|
[40-70]
|
Count
|
175
|
142
|
21
|
338
|
|
% within ferritin
|
51.8%
|
42.0%
|
6.2%
|
100.0%
|
|
≥ 70
|
Count
|
330
|
332
|
41
|
703
|
|
% within ferritin
|
46.9%
|
47.2%
|
5.8%
|
100.0%
|
|
Total
|
Count
|
641
|
621
|
89
|
1351
|
|
% within ferritin
|
47.4%
|
46.0%
|
6.6%
|
100.0%
|
Table 4 Impact of iron store levels on hemoglobin level
in non-menopausal women (n = 3759)
|
Ferritin (µg /L)
|
Hemoglobin
|
|
|
|
< 120 g/L
|
≥120 g/L
|
Total
|
|
< 15
|
number
|
173
|
708
|
881
|
|
%
|
19.6%
|
80.4%
|
100.0%
|
|
[15-40]
|
number
|
43
|
1226
|
1269
|
|
%
|
3.4%
|
96.6%
|
100.0%
|
|
[40-70]
|
number
|
23
|
914
|
937
|
|
%
|
2.5%
|
97.5%
|
100.0%
|
|
≥ 70
|
number
|
7
|
665
|
672
|
|
%
|
1.0%
|
99.0%
|
100.0%
|
|
Total
|
number
|
246
|
3513
|
3759
|
|
%
|
6.5%
|
93.5%
|
100.0%
|
Discussion
This study, carried out for the first time on a very large cohort
of women, provides strong arguments in favor of an association
between depleted iron stores and hair loss, particularly excessive
hair loss in women before menopause. It is likely that women
classified in the present study under “excessive hair loss” were
mainly women with androgenetic alopecia [17], however considering
their large number, women with CTE [3] might also be included.
Since serum ferritin measurement has been reported to be the most
sensitive assay for estimating the iron status in an adult
population [24], we used this assay to verify a possible link
between iron status and hair loss.
Our results fully agreed with previous reports [5, 6, 8, 9] and
provide further evidence that the iron status has to be taken into
consideration when studying hair loss in women [5] and contrasts
with some previous studies where no link between iron deficiency
and hair loss was detected. This discrepancy could be explained by
differences in the design of the studies. None of the previous
studies was performed on an important sample of the general
population nor did they take into consideration parameters such as
age and haemoglobin concentration in post-menopausal women, and
also neglecting the adjustment of the results with the use of
contraceptives or intra-uterine devices in non-menopausal women. It
is known that menstruating women using intra-uterine devices have
significantly lower serum ferritin levels than those without
contraception, or using oral contraception [21]. A potential
weakness of our study, based on the way the data on hair loss were
collected, was compensated by the successful logistic regression
model, establishing a link between serum ferritin level and
excessive hair loss in non-menopausal women.
According to these results, a decrease in ferritin levels might
be considered as a potential risk factor for excessive hair loss.
For example, a decrease in 30 units of serum ferritin level in
non-menopausal women presenting an initial serum ferritin
concentration of 70 µg/L (reference mean value of serum ferritin)
would lead to a 28% increase in the odds of excessive hair
loss.
Our results support the “threshold hypothesis” [17], which
states that decreased iron stores lower the threshold for
developing different types of alopecia. Nevertheless, additional
studies are required to better understand the biological
significance of the critical iron status level of 40 µg/L in the
etiology of hair loss.
In hair follicles, iron is implicated as a metabolic factor.
Iron is also a main constituent for hemoglobin, and iron depletion
leads to anemia. The present study indicates that anemia appears
for a ferritin threshold much lower than for hair loss.
Consequently in clinical practice, an excessive hair loss supports
the need to check serum ferritin levels. The optimum serum ferritin
levels to reach for effectiveness on hair loss are higher than
those usually recommended for treatment of anemia.
Acknowledgements
We would like to thank Mrs M. Cartron for her help, Mrs A. Bielicki
for her technical assistance, Pr A.M. Roussel (Joseph-Fourier
University), Drs P. Preziosi (U557 Inserm), B.A. Bernard and C.
Bouillon for critical comments of the manuscript.
Disclosure: The authors attest that they have no
conflicts of interest to disclose.
References
1 Birch MP, Messenger JF, Messenger AG. Hair
density, hair diameter and the prevalence of female pattern hair
loss. Br J Dermatol 2001; 144: 297-304.
2 Venning VA, Dawber RP. Patterned androgenic alopecia
in women. J Am Acad Dermatol 1998; 18: 1073-7.
3 Rushton DH, Norris MJ, Dover R,
Busuttil N. Causes of hair loss and the developments in hair
rejunevation. Int J cosm Sci 2002; 24: 17-23.
4 Rushton DH. Nutritional factors and hair loss. Clin Exp
Dermatol 2002; 27: 396-404.
5 Rushton DH. Management of hair loss in women. Dermatol
Clin 1993; 11: 47-53.
6 Rushton DH, Ramsay ID, James KC,
Norris MJ, Gilkes JJH. Biochemical and trichological
characterization of diffuse alopecia in women. Br J Dermatol 1990;
123: 187-97.
7 Futterweit MD, Dunaif A, Yeh H-C, et al.
The prevalence of hyperandrogenism in 109 consecutive female
patients with diffuse alopecia. J Am Acad Dermatol 1988; 19:
831-6.
8 Roberts JL. Examining the etiology of telogen effluvium
in pre-and postmenaupausal women: a chart review study. Tokyo,
Japan: Proceeding, Tri-continental hair research Meeting, 2001;
(Poster 157).
9 Haycox C. The incidence of depleted iron stores in North
american females presenting with hair loss. Tokyo, Japan:
Proceedings, Tri-Continental hair research Meeting, 2001; (Poster
159).
10 Milman N, Kirchhoff M. Iron stores in 1359, 30-to
60-year old Danish women: evaluation by serum ferritin and
hemoglobin. Ann Hematol 1992; 64: 22-7.
11 Rushton DH, Ramsay ID. The importance of adequate
serum ferritin levels during oral cyproterone acetate and ethinyl
oestradiol treatment of diffuse androgen-dependent alopecia in
women. Clin Endocrinol (Oxf) 1992; 36: 421-7.
12 Aydingöz IE, Ferhanoglu B, Guney O. Does
tissue iron status have a role in female alopecia? J Eur Acad
Dermatol Venereol 1999; 13: 65-7.
13 Auerbach R. Low iron levels. Arch Dermatol 1968; 98:
681.
14 Sinclair R. There is no clear association between low
serum ferritin and chronic diffuse telogen hair loss. Br J Dermatol
2002; 147: 982-4.
15 Chamberlain AJ, Dawber RPR. Significance of iron
status in hair loss in women. Br J Dermatol 2003; 149: 428.
16 Rushton DH, Dover R, Norris MJ. Is there
really no clear association between low serum feritin and chronic
diffuse telogen hair loss? Br J Dermatol 2003; 148: 1282-4.
17 Kantor J, Kessler LJ, Brooks DG,
Cotsarelis G. Decreased serum ferritin is associated with
alopecia in women. J Invest Dermatol 2003; 121: 985-8.
18 Hercberg S, Preziosi P, Briançon S,
et al. A primary prevention trial using nutritional doses of
antioxidant vitamins and minerals in cardiovascular diseases and
cancers in a general population: the SU. VI. MAX study-design,
methods, and participant characteristics. Control Clin Trials 1998;
19: 336-51.
19 Hercberg S, Galan P, Preziosi P,
Bertrais S, Mennen L, Malvy D, Roussel AM,
Favier A, Briançon S. The SU.VI.MAX study: a randomized,
placebo controlled trial of the health effects of antioxidant
vitamins and minerals. Arch Intern Med 2004; 164: 2335-42.
20 Lebart L, Morineau A, Piron M. In: Statistique
exploratoire multidimensionnelle. Paris: Dunod, 1995: 93-176.
21 Galan P, Yoon HC, Preziosi P, et al.
Determining factors in the iron status of adult women in the
Suvimax women. Eur J Clin Nutr 1998; 52: 383-8.
22 Westfall PH, Tobias RD, Rom D, et al. In:
Multiple comparisons and multiple tests using the SAS system. Cary,
NC: SAS Institute Inc, 1999: 73-6.
23 Stokes ME, Davis CS, Koch GG. In: Categorical
data analysis using the SAS system. Cary, NC: SAS Institute Inc,
2000: 211-70.
24 Hercberg S, Galan P. Nutritional anaemias. In:
Flemming AF, ed. Clinical Haematology_International Practice
and Research. Bailllière, London: Tindall, 1992: 145-68.
|
Version imprimable
Version PDF
