Serum phosphate levels are related to all-cause, cardiovascular and COPD mortality in men
European Journal of Epidemiology
Serum phosphate levels are related to all-cause, cardiovascular and COPD mortality in men
Natalia Campos-Obando 0 1 2 3
Lies Lahousse 0 1 2 3
Guy Brusselle 0 1 2 3
Bruno H. Stricker 0 1 2 3
Albert Hofman 0 1 2 3
Oscar H. Franco 0 1 2 3
Andre? G. Uitterlinden 0 1 2 3
M. Carola Zillikens 0 1 2 3
Mortality 0 1 2 3
0 Department of Respiratory Medicine, Erasmus MC , 3000 CA Rotterdam , The Netherlands
1 & M. Carola Zillikens
2 Department of Respiratory Medicine, Ghent University Hospital , 9000 Ghent , Belgium
3 Department of Epidemiology, Erasmus MC , 3000 CA Rotterdam , The Netherlands
Hyperphosphatemia has been associated with increased mortality in chronic kidney disease but the nature of such a relation in the general population is unclear. To investigate the association between phosphate (P) levels and all-cause and causespecific mortality, we assessed two cohorts from the Rotterdam Study, with follow-up of 14.5 (RS-I) and 10.9 (RS-II) years until January 2012 with availability of fasting phosphate levels. Deaths were classified according to International Classification of Diseases into 7 groups: cardiovascular, cancer, infections, external, dementia, chronic lung diseases and other causes. Sex-stratified Weibull and competing-risks models were adjusted for age, BMI and smoking. Hazard ratios are expressed per 1 mg/dL increase in phosphate levels. The total number of participants included 3731 (RS-I, 2154 women) and 2494 (RS-II, 1361 women) subjects. The main outcome measures were all-cause and cause-specific mortality. A significant positive association was found between phosphate and all-cause mortality in men (pooled HR (95% CI): 1.46 (1.26-1.69)) but not in women (0.90 (0.77-1.05)). In men, higher phosphate increased the risk for cardiovascular mortality (1.66 (1.29-2.14)), other causes (1.67 (1.16-2.40)) and chronic lung disease mortality (1.94 (1.02-3.72)), the latter driven by mortality due to chronic obstructive pulmonary disease (COPD) (4.44 (2.08-9.49)). No relations were found for mortality due to infections, cancer, dementia or external causes. In conclusion, serum P is associated with increased allcause, cardiovascular and COPD mortality in men but not women. The association with COPD mortality is novel and needs further research on underlying mechanisms.
Department of Internal Medicine, Erasmus MC,
PO Box 2040, 3000 CA Rotterdam, The Netherlands
Phosphorus is the sixth most common element in the
human body and the second mineral in abundance [
plays an important structural role in hard tissues, such as
bone, and exerts critical regulatory roles in metabolic and
signaling pathways [
The majority of phosphorus is stored in bone (85%)
where it is complexed with calcium in the form of
hydroxyapatite, whereas 15% of phosphorus is located in
the intracellular compartment while less than 1% is present
in extracellular fluids. In blood, phosphorus exists in two
main forms: a) an organic form bound to proteins (70%), b)
an ionized form (30%), known as inorganic phosphorus, or
phosphate, that circulates freely [
Traditionally, phosphate homeostatic mechanisms have
been ascribed to the actions of parathyroid hormone (PTH)
and 1,25-dihydroxyvitamin D3 (1,25(OH)2D3) [
Recently, an equally important new axis of phosphate
regulators was discovered [
], composed of the so-called
phosphatonins: fibroblast growth factor 23 (FGF23),
synthesized mainly in osteocytes, and its co-receptor a-Klotho
]. The FGF23/a-Klotho axis increases Purinary
Monogenic disorders causing extreme phosphate
concentrations are associated with rickets in severe
hypophosphatemia and calcinosis in severe
]. Recently, milder hyperphosphatemia was
shown to increase cardiovascular mortality in chronic
kidney disease (CKD) [
]. Subsequently, this association
was reported also in non-CKD population [
Interestingly, sex differences have been described with
associations found in men but not women for all-cause mortality
and subclinical atherosclerosis ; the underlying reasons
are not understood. In addition to serum phosphate levels
(P), high P intake has recently been found to increase
The objectives of this study were to assess the
association of P with all-cause and, in detail, cause-specific
mortality within two cohorts of the population-based
Rotterdam Study, and to test for potential sex differences in
Materials and methods
The Rotterdam Study is a prospective study of men and
women designed to investigate the incidence and
determinants of chronic disabling diseases. Rationale and design
has been described elsewhere [
]. This research was
performed in two cohorts within the Rotterdam Study, the
Rotterdam Study I cohort (RS-I), initiated in 1990 in 7983
subjects, and the Rotterdam Study II cohort (RS-II)
initiated in 2000 in 3011 subjects. All participants were
55 years or more at recruitment and have been assessed at
baseline and through several follow-up visits. P was
measured in the non-fasting state at baseline visit of RS-I
(referred to as RS-I-1) and in the fasting state at the second
follow-up visit of RS-I (RS-I-3, referred to as RS-I) and the
baseline visit of RS-II (Fig. 1). The fasting state may
modify the association between P and mortality [
Therefore, our main analysis was based on data from
RS-I3 and RS-II because P was assessed in the fasting state;
subsequently we checked if the observed results followed
similar patterns in RS-I-1, where non-fasting samples are
available. A total of 3731 participants from RS-I and 2494
from RS-II were included for these analyses, all of them
with signed informed consent and available phosphate
levels. The Rotterdam Study was approved by the
institutional review board (Medical Ethics Committee) of
the Erasmus Medical Center and by the review board of the
Netherlands Ministry of Health, Welfare and Sports. The
approval has been renewed every 5 years.
The amount of phosphorus determined in blood
corresponds to the inorganic fraction, or phosphate (mg/dL),
assessed with a method based on the formation of
ammonium phosphomolybdate [
Total calcium determination (mg/dL) was done through
a colorimetric o-cresolphthalein complex one method
(Merck Diagnostica, Amsterdam, The Netherlands, for
RSI-1; and Roche, Mannheim, Germany, for RS-I and RS-II).
Levels of 25-hydroxyvitamin D (nmol/L) were determined
through an electrochemiluminescence immunoassay. We
applied cosinor regressions to adjust 25-hydroxyvitamin D
for season and year. After testing for seasonality applying
the dickey fuller test, we proceeded to perform a time
transformation on sine and cosine terms (sin(2*pi*time/
12)). Afterwards, we proceeded to regress the serum
vitamin D levels on those terms to get the mesor, that is, the
mean value of the cosinor regression. We then computed
the difference between the mean of each season and the
mesor, and adjusted every individual value accordingly
]. Levels of 1,25-dihydroxyvitamin D3 were
assessed in a subset of participants from RS-I-1 through
125Iradioimmunoassay (IDS, Boldon, UK). Creatinine was
determined through a sarcosine-based colorimetric assay
and standardized against isotope dilution mass
spectrometry (ID-MS). Cystatin C was assessed through particle
enhanced immunoturbidimetric assay. C-reactive protein
(CRP) levels were measured through an agglutination
method with antibodies. Magnesium (Mg) levels were
determined with a colorimetric method based on xylidyl
blue. Glucose and cholesterol levels were determined by
standard enzymatic methods .
We assessed the distribution of potential confounders
across P quintiles, such as age, body mass index (BMI),
smoking status, calcium, 25-hydroxyvitamin D levels,
creatinine, estimated glomerular filtration rate (eGFR),
C-reactive protein (CRP), glucose, magnesium, total
cholesterol to HDL cholesterol ratio and prevalent diabetes
mellitus and cardiovascular disease (CVD). BMI, smoking
status, prevalent diabetes mellitus and prevalent CVD were
assessed as previously described [
]. The diagnoses of
prevalent and incident chronic obstructive pulmonary
disease (COPD) cases was based on an obstructive
pre-bronchodilator spirometry (FEV1/FVC \ 0.7), according to
Fig. 1 Flowchart for time line,
design and sample size for the
GOLD guidelines [
]. P intake at baseline visit (RS-I-1)
was collected in a subsample of participants through a
validated semiquantitative food frequency questionnaire.
The Chronic Kidney Disease Epidemiology Collaboration
equations based on creatinine [
] were applied to
estimate eGFR (mL/min). Additionally, cystatin C-based
eGFR was estimated for subjects with creatinine-based
eGFR less than 60 mL/min, as previously recommended
Assessment of all-cause and cause-specific mortality
respective field reviewed and confirmed the diagnosis.
Information on cause-specific mortality was available until
Different causes of mortality were recorded according to
ICD-10 codes and firstly grouped into cardiovascular
diseases (CVD), cancer and other causes. To perform
comprehensive analyses, the group of other causes was further
categorized into external causes, dementia, infections,
chronic lung disease and other causes in the strict sense, as
previously described [
Information on vital status is obtained continuously from
the municipal authorities in Rotterdam. The cohorts are
monitored for mortality through computerized linkage of
the study database to medical files of general practitioners.
Two research physicians independently coded the mortality
events according to ICD-10. Medical specialists in the
Subjects with fasting P measurements from RS-I and RS-II
were analyzed separately and in a meta-analysis.
Additionally, we analyzed subjects with non-fasting P from
Due to sex differences in P [
] and in its association
with health outcomes [
], we built sex-stratified analyses.
We compared the distribution of potential confounding
factors applying age-adjusted tests of trend across P
quintiles. We estimated P levels across smoking categories
applying ANOVA and post hoc (Tukey?s) tests. Initially,
the association of P with mortality was assessed through
Cox models, testing the proportionality assumption of the
hazards via the Schoenfeld residuals test. All significant
HRs from Cox? models were found to be constant over
follow-up time; therefore, we found no evidence for a
timedependent effect of P levels on mortality. In a second step,
we compared the semi-parametric Cox model with
parametric models, and found that Weibull regression
models?albeit with highly similar results to Cox regressions?
provide better statistical fit to the data than Cox models.
Weibull models provided also better fit than the rest of
parametric models. We applied Cox-Snell residuals graphs
and Akaike (AIC) and Bayesian information criteria (BIC)
to compare among models, as previously recommended
]. Models with lower AIC and BIC correspond to a
better fit. Therefore, the results reported in this manuscript
correspond to Weibull regression models. Finally, we also
performed competing-risks regressions models which
allow for informative censoring due to the multiple
possible causes of death [
]; these models provide an estimate
of the effect of the exposure on the probability of
developing the outcome over time [
Hazard ratios (HRs) are expressed per increase in 1 mg/
dL (0.32 mmol/L) of P or in quintiles; the latter were built
to explore a potential dose?effect relationship between
phosphate levels and mortality.
The analysis time was set at the date of blood drawing.
Subjects were followed until the first of the following
events happened: death, lost to follow-up, or censoring by
1st January, 2012.
Adjustments were made firstly for age, BMI and
smoking because they are related to mortality and P;
subsequently other covariates that have been associated with
mortality were added to the model and retained if they
changed the beta estimate more than 10%, including eGFR,
glucose, hsCRP, Mg levels, cholesterol to HDL cholesterol
ratio, calcium, 25-hydroxyvitamin D and prevalent
Results from RS-I and RS-II were meta-analyzed using
Primary analyses were done with subjects with complete
information on covariates. Subsequently, missing values
were imputed via multiple imputation with chained
equations, allowing missingness at random. We followed
specific guidelines for imputation for survival analysis.
We repeated analyses including only subjects with normal
P (2.5?4.5 mg/dL; 0.81?1.45 mmol/L). We further
adjusted the analyses for phosphate dietary intake and
1,25dihydroxyvitamin D3 levels in a subset of participants from
RS-I-1 (n = 4046).
Additionally, we performed stratified analyses according
to smoking categories.
We used SPSS (version 21.0, Armonk, NY: IBM Corp),
Stata (version 13, College Station TX: Stata Corp LP) and
Comprehensive Meta-Analysis (version 2.2, Biostat,
Englewood, NJ). A two-sided p \0.05 was considered
Serum phosphate correlates
A general descriptive summary of main continuous
covariates is depicted in Table 1. The distribution of
relevant covariates and risk factors across even quintiles of P
for RS-I and RS-II is depicted in Table 2. P was higher in
women than men in both cohorts (pdifference \ 0.001). P
levels were different across smoking categories in both
sexes and cohorts (ANOVA p \ 0.001); this difference
was due to higher P in current smokers (Tukey?s tests [
0.05 between former and never smokers).
P was within normal range in 95.5 and 94.9% of
participants in the fasting state (RS-I and RS-II, respectively)
and in 89.7% of participants in the non-fasting state
Serum phosphate and all-cause mortality
During 14.5 year (median) and 10.9 year (mean) follow-up
a total of 1631 and 469 fatal events occurred in RS-I and
RS-II, respectively. We found a significant interaction
between P and sex for all-cause mortality in RS-I
(pinteraction \ 0.001) and performed sex-stratified analyses. The
results for the comparison of goodness-of-fit between
parametric models and the semiparametric Cox model are
displayed in Supplementary Table 1 (AIC and BIC criteria)
and in Fig. 2 (Cox-Snell residuals plot). Both methods
showed that Weibull models provide a better fit to our data
among the parametric and semiparametric models.
The associations between P and all-cause mortality are
depicted in Table 3. Results from RS-I and RS-II were
meta-analyzed (pooled HR (95% CI)). A significant
association between P and all-cause mortality was found in men
(1.46 (1.26?1.69)) but not in women (0.90 (0.77?1.05)).
Table 1 General characteristics
of subjects in RS-I and RS-II
with serum phosphate levels,
BMI and smoking information
available, stratified by sex
Chol to HDL ratio
Chol to HDL ratio
BMI body mass index, 25(OH)D 25-hydroxyvitamin D levels, CRP C-reactive protein, eGFR estimated
glomerular filtration rate, Mg magnesium, Chol to HDL ratio total cholesterol to HDL cholesterol ratio
Conversion to SI Units: to convert 25-hydroxyvitamin D levels to ng/mL multiply by 0.4; to convert
glucose to mg/dL multiply by 18.02; to convert creatinine to mg/dL multiply by 0.011; to convert
magnesium to mg/dL multiply by 2.43
Adjustments in a full model composed of age, BMI,
smoking, prevalent cardiovascular disease and levels of
calcium, 25-hydroxyvitamin D, eGFR, CRP, Mg, glucose
and total cholesterol to HDL cholesterol ratio levels did not
substantially modify results (men: 1.49 (1.27?1.74);
women: 0.92 (0.79?1.07)).
Similarly, results from RS-I-1 with non-fasting
phosphate showed a significant association of phosphate with
all-cause mortality in men (1.12 (1.02?1.23); no events =
1389), but not in women (0.99 (0.91?1.08); no
events = 1779).
To explore whether there was a dose?response pattern in
the association we found in men, we analyzed P in even
quintiles and all-cause mortality in RS-I, the cohort with
most events, (Table 4) and set the first quintile (lowest) as
reference. We observed a significant trend for increasing P
and mortality (ptrend \ 0.001) with significant HRs for the
fourth (1.35 (1.08?1.69)) and fifth quintile (1.49
(1.19?1.86)) compared with the first quintile.
Results after excluding subjects with abnormal P were
similar to the unrestricted analyses (men: 1.44 (1.21?1.70);
women: 0.87 (0.74?1.03)). Adjustments for phosphate and
energy intake in men from RS-I-1 did not modify the
results between non-fasting phosphate and all-cause
mortality (1.13 (1.02?1.24); no events = 1117). Further
adjustments for 1,25 dihydroxyvitamin D3 levels in a
subset from RS-I-1 did not modify results (data not shown).
Statistically significant p-values (\0.05) are highlighted in bold font
*P values corresponds to age-adjusted significance of trend across quintiles
BMI body mass index, 25(OH)D 25-hydroxyvitamin D levels, CRP C-reactive protein, prevalent DM prevalent diabetes mellitus, eGFR
estimated glomerular filtration rate, Mg magnesium, Chol to HDL ratio total cholesterol to HDL cholesterol ratio, prevalent CVD prevalent
Conversion to SI Units: to convert 25-hydroxyvitamin D levels to ng/mL multiply by 0.4; to convert glucose to mg/dL multiply by 18.02; to
convert creatinine to mg/dL multiply by 0.011; to convert magnesium to mg/dL multiply by 2.43
Serum phosphate and cause-specific mortality in men
We did not observe associations between P and
causespecific mortality in women (data not shown). In contrast,
the pooled results in men (Table 5) showed a significant
positive relation between P and CVD mortality (1.66
(1.29?2.14)). Exclusion of male subjects with prevalent
CVD disease yielded similar results (1.69 (1.28?2.23)).
We also found an association between higher P and
chronic lung disease mortality (1.94 (1.02?3.72)). Most of
these cases clustered within COPD mortality. Therefore,
we further investigated such a relation (Table 6), and found
a significant association (4.44 (2.08?9.49)). Most likely due
to power constraints, this association was not significant in
Statistically significant p-values (\0.05) are highlighted in bold font
*Phosphate levels in mg/dL
Hazard ratios from Weibull models; first quintile was set as reference
Statistically significant p-values (\0.05) are highlighted in bold font
*Hazard ratios from Weibull models, expressed per 1 mg/dL (0.32 mmol/L) increase in phosphate levels
Studies combined from meta-analyses using fixed-effect models
RS-II (05 cases in contrast to 28 cases in RS-I) but there
was no evidence for statistical difference between both
estimates (pheterogeneity = 0.780). Further adjustments for
glomerular filtration rate did not abolish the association
between P and COPD mortality (4.16 (2.05?8.43)).
Furthermore, the association was found to be consistent in
subjects without chronic kidney disease (CKD) (6.58
(2.59?16.7)); whereas we found no association in subjects
with CKD (1.14 (0.20?6.63)), although the latter analysis is
constrained due to low number of events and driven only
by RS-I. Non-fasting phosphate levels and COPD mortality
Table 6 Serum phosphate levels and chronic obstructive pulmonary
disease (COPD) mortality in men from RS-I and RS-II, adjusted for
age, BMI and smoking, follow-up until year 2012
Statistically significant p-values (\0.05) are highlighted in bold font
*Hazard ratios from Weibull models, expressed per 1 mg/dL
(0.32 mmol/L) increase in phosphate levels
Studies combined from meta-analyses using fixed-effect models
in men from RS-I-1 also displayed a significant association
(1.54 (1.05?2.27), no events = 69).
P was also found to be positively associated with
mortality from other causes (1.67 (1.16?2.40)).
We found no significant associations between P and
death due to cancer, infections, dementia or external
Results from competing-risks regression models were
similar to Weibull models and showed a significant
association between P and mortality due to CVD (1.50
(1.12?2.02)), other causes (1.40 (1.01?1.93)) and COPD
(2.42 (1.62?3.63)); no other significant associations were
found (Supplementary Tables 2 and 3).
Analyses after applying multiple imputation yielded
significant associations for P and all-cause, CVD, COPD
and other causes of mortality in men (data not shown).
Missingness of covariates of interest was less than 6%.
Results after excluding male subjects with abnormal P
were similar to the unrestricted analyses (Supplementary
Tables 4 and 5). Likewise, our findings remained
essentially unaltered after adjustments for calcium and
25-hydroxyvitamin D levels; and were only slightly attenuated
after further adjustments for levels of calcium,
25-hydroxyvitamin D and eGFR (CVD 1.65 (1.27?2.14), COPD
3.79 (1.87?7.69), other causes 1.76 (1.21?2.56)). Similar
results were obtained after adjustments for cystatin-based
eGFR. Additionally, the analyses after exclusion of male
subjects with eGFR \ 60 mL/min showed a positive
association between P and mortality due to other causes
(1.72 (1.13?2.61)) and COPD (6.58 (2.59?16.7)) - as
previously mentioned- and a borderline association between P
and CVD mortality (1.36 (1.00?1.85)).
Smoking adjustment did not attenuate the association
between P and CVD or COPD mortality (data not shown).
The results from the stratified analyses according to
smoking categories (Supplementary Tables 6 and 7)
showed that in studies combined the associations between
P and all-cause and CVD mortality were in the same
direction and did not show statistical evidence for a
difference across categories (pheterogeneity = 0.752 for all-cause
mortality and pheterogeneity = 0.796 for CVD mortality). The
relation between P and COPD mortality in men from RS-I
(RS-II excluded due to few events) was not statistically
different among former and current smokers
(pheterogeneity = 0.494).
As previously mentioned, analyses in men from RS-I-1
showed that non-fasting phosphate levels were also
associated with chronic lung disease mortality and COPD
mortality, and these associations were not abolished after
further adjustments for phosphate and energy intake:
chronic lung disease mortality: 1.79 (1.19?2.68); no
events = 59; COPD mortality: 1.87 (1.20?2.91), no
events = 49.
This prospective population-based cohort study among
elderly demonstrated that P was positively associated with
all-cause mortality in men but not in women, supporting an
effect modification by sex previously described [
analyzing in detail cause-specific mortality in men, we
found that this association was driven by mortality due to
CVD, COPD and other causes. The association between
increasing P and the composite endpoint of fatal and
nonfatal CVD incidence in non-CKD population in
sex-combined analyses has been reported before but is still scarce
]. Our results provide evidence of an association
between higher P?even within normal range?and death
due to CVD in men. On the other hand, to the best of our
knowledge the association we found with COPD mortality
is novel. These results remained significant after
adjustments for several potential confounders, were observed
also after restricting the analyses to subjects with normal P
and showed no heterogeneity between cohorts.
Several mechanisms have to be considered when
analyzing P and mortality, including phosphate being a marker
of another risk factor or through direct pathogenic
First, P levels are regulated by a complex interplay of
factors that have been linked to mortality, such as
1,25dihydroxyvitamin D3, PTH and FGF23. Low levels of
25-hydroxyvitamin D and 1,25-dihydroxyvitamin D3 have
been found to be associated with increased mortality [
Nevertheless, the vitamin D adjustments did not modify
PTH abnormalities have also been associated with
mortality. Primary excess of PTH is associated with
increased cardiovascular mortality [
], but in this context
P tends to be low. Secondary elevations of PTH in impaired
kidney function have been inconsistently associated with
mortality. This compensatory mechanism in CKD is
triggered when eGFR falls below 47 mL/min [
PTH levels were not available, the proportion of patients in
our cohorts with eGFR below that threshold was
considerably low (4% in RS-I and 2% in RS-II) suggesting that
secondary hyperparathyroidism is unlikely to explain our
findings. Nevertheless, PTH values seem to rise within
normal range in the general population without CKD [
at higher thresholds of decreasing eGFR (\ 120 mL/min);
whether increasing PTH values within normal range are
associated with all-cause mortality in the long term is
Other important players in P homeostasis that might
underlie its associations with mortality are the
phosphatonins FGF23 and a-Klotho [
]. FGF23 is synthesized
mainly in osteocytes  and requires the presence of
aKlotho to bind to its receptor with high affinity and for
]. FGF23/a-Klotho axis decreases P through
increased urinary phosphate excretion and both molecules
are anti-ageing factors [
]. Primary causes of excess
FGF23, such as in hereditary hypophosphatemic rickets,
have been associated with cardiovascular calcification in
cases of excessive phosphate treatment. Secondary FGF23
elevation occurs in CKD at earlier stages than PTH [
in response to P retention, and it has been linked to
increased mortality [
]. Similar to PTH, FGF23
elevations within normal range have been described at high
thresholds of eGFR in population without CKD [
FGF23 levels have also been associated with mortality in
this setting [
]. Nevertheless, FGF23 seems not to induce
vascular calcification in most studies [
4, 32, 33
Recently, soluble klotho has been linked to increased
mortality in CKD patients [
] although the lack of a
validated assay for its measurement might be a concern for
Another potential confounder could be smoking. Similar
to previous reports [
], P was found to be higher in current
smokers. Although adjustments for smoking did not alter
our analyses, due to heavy current and former smoking in
men it is difficult to fully dissect its effects. Nevertheless,
in studies combined the stratified analyses by smoking
status showed that the associations between P and all-cause
and CVD mortality appeared to be of the same direction
and similar magnitude across smoking categories. The
group of former smokers?who had similar P as never
smokers?displayed the most statistically significant
associations possibly due to larger number of subjects in this
category. Specifically, P was related to COPD mortality
comparably in current and former smokers men from RS-I
but only significant in the latter group; a relation in
nonsmokers could not be tested due to low numbers in this
subgroup. Therefore we do not anticipate that current
smoking explains the association between P and COPD
Regarding direct effects, P itself is able to induce
vascular calcification, a process with high resemblance to bone
ossification and that increases mortality [
pathways are known such as (a) differential gene
expression in vascular smooth muscle cells with up-regulation of
markers critical for mineralization ; and (b) elastin
degradation, thought to be mediated by P induction of
matrix metalloproteinase (MMP)-9.
The association we found between P and COPD
mortality has never been described in humans before;
interestingly there is additional evidence for the pathogenicity
of high P stemming from rodent models with fgf23 or
klotho knockout. These animals display similar phenotypes
characterized by severe hyperphosphatemia and features of
premature aging, such as osteoporosis, ectopic
calcifications, pulmonary emphysema and short life span [
Heterozygous klotho mice also display emphysematous
lungs. Remarkably, a low phosphate diet is able to alleviate
or rescue the phenotype -including the lung emphysema;
and a high phosphate diet worsens it [
suggesting that phosphate itself accelerates ageing [
induces alveoli destruction, and that this process can be
modified by diet manipulation [
A new concept of phosphotoxicity as a risk factor for
mammalian ageing has emerged lately [
] and there are
concerns that increasing phosphate intake through food
additives may negatively influence multiple aspects of
health . Indeed it has been shown that high absolute P
intake was positively related to all-cause mortality -not
explained by CVD mortality [
]. Recently, a healthy
diet?according to the Alternate Healthy Eating Index
(2010) score?was associated with lower risk of COPD in
]; interestingly in men but not women this
beneficial association was driven mostly by a drastic
reduction in red and processed meat consumption, expected
to contain high phosphate [
]. A positive relation between
cured meat intake and COPD risk has previously been
reported in cross-sectional (NHANES III) and prospective
]. Importantly, when spirometric definitions
for lung volumes and COPD have been applied, cured meat
intake has been shown to be negatively associated with
lung function, and positively related with COPD risk
]; the latter study showed that these associations
were found predominantly in men. Cured meat
consumption has also been shown prospectively to increase the
hospital readmission rate in COPD patients [
From a mechanistic point of view, previous research
] has shown that phosphate is able to directly induce
injury in mice and human lung epithelial cells through
increased DNA oxidative stress and apoptosis; indeed
phosphate medium is used experimentally to induce
oxidative lung injury. Interestingly, a-Klotho exerts
protective antioxidant effects against lung injury induced by P
], hyperoxia, and acute a-Klotho deficiency [
data show that lung tissue is a target for phosphotoxic
insult. Remarkably, increased P intake down-regulates
aKlotho expression in rodents [
]; therefore low P diet may
be a therapeutic strategy to increase Klotho [
A genetic variant associated with low FGF23 was found
to be associated with emphysema in smokers with COPD.
More studies are needed to elucidate further the underlying
mechanisms, especially considering that COPD ranks high
in the most common causes of death worldwide.
The reasons for the sex difference between P and
mortality are not clear. Interestingly, the vascular calcification
induction by P is attenuated by 17b-estradiol, suggesting a
potential hormonal reason for this difference [
the fact that menopause is characterized by low estradiol
levels, hormone replacement therapy-na??ve
postmenopausal women with higher 17b-estradiol levels
display lower coronary calcification scores than those with
lower 17b-estradiol [
]. Additionally, coronary infusion
of 17b-estradiol exerts vasodilation in postmenopausal
women, but not men [
]. Testosterone and estradiol play
important roles as P regulators [
Although men had a less healthy profile at baseline than
women, multiple adjustments did not abolish our results.
Moreover, a previous study showed that P is associated
with subclinical atherosclerosis in men (but not women)
without prevalent cardiovascular and cerebrovascular
disease at baseline [
This study has several limitations.
1,25-dihydroxyvitamin D3 levels were available only in a subgroup. PTH and
FGF23 measurements were not available and it is known
that kidney function in elderly can be misclassified even by
eGFR. Our findings cannot be generalized to other
ethnicities other than European Caucasians. Nevertheless,
there are several strengths, such as the availability of two
well-characterized cohorts with long follow-up, the
detailed information on cause-specific mortality and the
availability of multiple potential confounders. The
completeness of follow-up was high (94 and 92% in RS-I and
RS-II) indicating that obtained estimates are valid.
In conclusion, we found that higher P is associated with
increased all-cause mortality and cause-specific mortality
due to CVD, COPD and other causes in elderly men but not
in women, adding more evidence for a modification of
these associations by sex. We hereby provide evidence to
support that the concept of phosphotoxicity also among
non-CKD general population deserves further attention
and, if causally related, it occurs independently of vitamin
D levels and kidney function. Our study suggests that
moderation of phosphate intake might be relevant also in
non-CKD population for healthy ageing. Finally, we
consider that the available evidence calls for a review of the
currently accepted normal range of P. Further research is
needed to clarify the underlying mechanisms, especially for
COPD mortality, and to elucidate the reasons for the sex
difference in the association of P with mortality.
Acknowledgments The authors thank the participants and staff of the
research center of the Rotterdam Study. The authors also
acknowledge Dr. F. Rivadeneira, for his critical review of the manuscript. The
Rotterdam Study has been approved by the Medical Ethics
Committee of the Erasmus MC (Registration Number MEC 02.1015) and
by the Dutch Ministry of Health, Welfare and Sport (Population
Screening Act WBO, License Number 1071272-159521-PG). The
Rotterdam Study has been entered into the Netherlands National Trial
Register (NTR; www.trialregister.nl) and into the WHO International
Clinical Trials Registry Platform (ICTRP; www.who.int/ictrp/net
work/primary/en/) under shared catalogue number NTR6831. All
participants provided written informed consent to participate in the
study and to have their information obtained from treating physicians.
Author contribution Dr. Zillikens and N. Campos-Obando are the
study guarantors and take responsibility for the integrity of the data
and the accuracy of the data analysis. Study concept and design: N.
Campos-Obando and Dr. Zillikens. Acquisition of data: Prof.
Hofman, Dr. Zillikens, Prof. Uitterlinden, Prof. Stricker, Prof. Brusselle,
Prof. Franco, Dr. Lahousse. Analysis and interpretation of data: N.
Campos-Obando and Dr. Zillikens. Drafting of the manuscript: N.
Campos-Obando and Dr. Zillikens. Critical review of the manuscript
for important intellectual content: all authors. Statistical analyses: N.
Campos-Obando. Obtained funding: Prof. Hofman, Prof. Uitterlinden.
Administrative, technical and material support: Dr. Zillikens, Prof.
Uitterlinden. Study supervision: Dr. Zillikens.
Funding The Rotterdam Study is funded by Erasmus Medical Center
and Erasmus University, Rotterdam, Netherlands Organization for the
Health Research and Development (ZonMw), the Research Institute
for Diseases in the Elderly (RIDE), the Ministry of Education, Culture
and Science, the Ministry for Health, Welfare and Sports, the
European Commission (DG XII), and the Municipality of Rotterdam. The
funding sources had no influence in the study design, collection,
analysis, interpretation of data, writing of the report and in the
decision to submit the article.
Compliance with ethical standards
Conflict of interest Authors declare that they have no conflict of
Ethical approval The Rotterdam Study was approved by the Medical
Ethics Committee of Erasmus Medical Center.
Open Access This article is distributed under the terms of the Creative
Commons Attribution 4.0 International License (http://creative
commons.org/licenses/by/4.0/), which permits unrestricted use,
distribution, and reproduction in any medium, provided you give
appropriate credit to the original author(s) and the source, provide a
link to the Creative Commons license, and indicate if changes were
1. Koeppen BM , Stanton BA . The renal system . In: Koeppen BM , Stanton BA editors. Berne & Levy physiology, 6th edn. Philadelphia , PA; 2010 . p. 557 - 636 .
2. Shimada T , Kakitani M , Yamazaki Y , et al. Targeted ablation of Fgf23 demonstrates an essential physiological role of FGF23 in phosphate and vitamin D metabolism . J Clin Invest . 2004 ; 113 ( 4 ): 561 - 8 .
3. Hu MC , Shiizaki K , Kuro-o M , Moe OW . Fibroblast growth factor 23 and Klotho: physiology and pathophysiology of an endocrine network of mineral metabolism . Annu Rev Physiol . 2013 ; 75 : 503 - 33 .
4. Lim K , Lu TS , Molostvov G , et al. Vascular Klotho deficiency potentiates the development of human artery calcification and mediates resistance to fibroblast growth factor 23 . Circulation . 2012 ; 125 ( 18 ): 2243 - 55 .
5. Martin A , David V , Quarles LD . Regulation and function of the FGF23/klotho endocrine pathways . Physiol Rev . 2012 ; 92 ( 1 ): 131 - 55 .
6. Block GA , Hulbert-Shearon TE , Levin NW , Port FK . Association of serum phosphorus and calcium x phosphate product with mortality risk in chronic hemodialysis patients: a national study . Am J Kidney Dis . 1998 ; 31 ( 4 ): 607 - 17 .
7. Dhingra R , Sullivan LM , Fox CS , et al. Relations of serum phosphorus and calcium levels to the incidence of cardiovascular disease in the community . Arch Intern Med . 2007 ; 167 ( 9 ): 879 - 85 .
8. Tonelli M , Sacks F , Pfeffer M , Gao Z , Curhan G , Cholesterol And Recurrent Events Trial Investigators. Relation between serum phosphate level and cardiovascular event rate in people with coronary disease . Circulation . 2005 ; 112 ( 17 ): 2627 - 33 .
9. Onufrak SJ , Bellasi A , Cardarelli F , et al. Investigation of gender heterogeneity in the associations of serum phosphorus with incident coronary artery disease and all-cause mortality . Am J Epidemiol . 2009 ; 169 ( 1 ): 67 - 77 .
10. Chang AR , Grams ME . Serum phosphorus and mortality in the Third National Health and Nutrition Examination Survey (NHANES III): effect modification by fasting . Am J Kidney Dis . 2014 ; 64 ( 4 ): 567 - 73 .
11. Chang AR , Lazo M , Appel LJ , Gutie?rrez OM, Grams ME . High dietary phosphorus intake is associated with all-cause mortality: results from NHANES III . Am J Clin Nutr . 2014 ; 99 ( 2 ): 320 - 7 .
12. Ikram MA , Brusselle GGO , Murad SD , et al. The Rotterdam Study: 2018 update on objectives, design and main results . Eur J Epidemiol . 2017 ; 32 ( 9 ): 807 - 50 .
13. Bolland MJ , Grey AB , Ames RW , et al. The effects of seasonal variation of 25-hydroxyvitamin D and fat mass on a diagnosis of vitamin D sufficiency . Am J Clin Nutr . 2007 ; 86 ( 4 ): 959 - 64 .
14. Tomson J , Emberson J , Hill M , et al. Vitamin D and risk of death from vascular and non-vascular causes in the Whitehall Study and meta-analyses of 12000 deaths . Eur Heart J . 2013 ; 34 ( 18 ): 1365 - 74 .
15. Terzikhan N , Verhamme KM , Hofman A , Stricker BH , Brusselle GG , Lahousse L . Prevalence and incidence of COPD in smokers and non-smokers: the Rotterdam Study . Eur J Epidemiol . 2016 ; 31 ( 8 ): 785 - 92 .
16. Stevens PE , Levin A , Kidney Disease: Improving Global Outcomes Chronic Kidney Disease Guideline Development Work Group Members . Evaluation and management of chronic kidney disease: synopsis of the kidney disease: improving global outcomes 2012 clinical practice guideline . Ann Intern Med . 2013 ; 158 ( 11 ): 825 - 30 .
17. Levey AS , Stevens LA , Schmid CH , CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration) , et al. A new equation to estimate glomerular filtration rate . Ann Intern Med . 2009 ; 150 ( 9 ): 604 - 12 .
18. Campos-Obando N , Castano-Betancourt MC , Oei L , et al. Bone mineral density and chronic lung disease mortality: the Rotterdam Study . J Clin Endocrinol Metab . 2014 ; 99 ( 5 ): 1834 - 42 .
19. Meng J , Ohlsson C , Laughlin GA , Osteoporotic Fractures in Men (MrOs) Study Group , et al. Associations of estradiol and testosterone with serum phosphorus in older men: the Osteoporotic Fractures in Men study . Kidney Int . 2010 ; 78 ( 4 ): 415 - 22 .
20. Collet D. Modelling survival data in medical research . 3rd ed. London: Chapman & Hall/CRC; 2014 .
21. Putter H , Fiocco M , Geskus RB . Tutorial in biostatistics: competing risks and multi-state models . Stat Med . 2007 ; 26 ( 11 ): 2389 - 430 .
22. Austin PC , Fine JP . Accounting for competing risks in randomized controlled trials: a review and recommendations for improvement . Stat Med . 2017 ; 36 ( 8 ): 1203 - 9 .
23. Lee DM , Vanderschueren D , Boonen S , European Male Ageing Study Group, et al. Association of 25- hydroxyvitamin D , 1 , 25 - dihydroxyvitamin D and parathyroid hormone with mortality among middle-aged and older European men . Age Ageing . 2014 ; 43 ( 4 ): 528 - 35 .
24. Nilsson IL , Yin L , Lundgren E , Rastad J , Ekbom A. Clinical presentation of primary hyperparathyroidism in Europe-nationwide cohort analysis on mortality from nonmalignant causes . J Bone Miner Res . 2002 ; 17 ( Suppl 2 ): N68 - 74 .
25. Isakova T , Wahl P , Vargas GS , et al. Fibroblast growth factor 23 is elevated before parathyroid hormone and phosphate in chronic kidney disease . Kidney Int . 2011 ; 79 ( 12 ): 1370 - 8 .
26. Dhayat NA , Ackermann D , Pruijm M , et al. Fibroblast growth factor 23 and markers of mineral metabolism in individuals with preserved renal function . Kidney Int . 2016 ; 90 ( 3 ): 648 - 57 .
27. Yang B , Lu C , Wu Q , Zhan J , Zhao H , Cao Y. Parathyroid hormone, cardiovascular and all-cause mortality: a meta-analysis . Clin Chim Acta . 2016 ; 455 : 154 - 60 .
28. Urakawa I , Yamazaki Y , Shimada T , et al. Klotho converts canonical FGF receptor into a specific receptor for FGF23 . Nature . 2006 ; 444 ( 7120 ): 770 - 4 .
29. Yang H , Luo H , Tang X , et al. Prognostic value of FGF23 among patients with end-stage renal disease: a systematic review and meta-analysis . Biomark Med . 2016 ; 10 ( 5 ): 547 - 56 .
30. Munoz Mendoza J , Isakova T , Cai X , CRIC Study Investigators , et al. Inflammation and elevated levels of fibroblast growth factor 23 are independent risk factors for death in chronic kidney disease . Kidney Int . 2017 ; 91 ( 3 ): 711 - 9 .
31. Qin Z , Liu X , Song M , et al. Fibroblast growth factor 23 as a predictor of cardiovascular and all-cause mortality in prospective studies . Atherosclerosis . 2017 ; 261 : 1 - 11 .
32. Scialla JJ , Lau WL , Reilly MP , Chronic Renal Insufficiency Cohort Study Investigators , et al. Fibroblast growth factor 23 is not associated with and does not induce arterial calcification . Kidney Int . 2013 ; 83 : 1159 - 68 .
33. Yamada S , Giacelli CM . Vascular calcification in CKD-MBD: roles for phosphate, FGF23 and Klotho . Bone. 2017 ; 100 : 87 - 93 .
34. Marc?ais C , Maucort-Boulch D , Drai J , ARNOGENE project , et al. Circulating klotho associates with cardiovascular morbidity and mortality during hemodialysis . J Clin Endocrinol Metab . 2017 ; 102 ( 9 ): 3154 - 61 .
35. Wu M , Rementer C , Giachelli CM . Vascular calcification: an update on mechanisms and challenges in treatment . Calcif Tissue Int . 2013 ; 93 ( 4 ): 365 - 73 .
36. Roma?n-Garc? ?a P, Carrillo-Lo?pez N, Ferna?ndez-Mart??n JL , NavesD??az M, Ruiz-Torres MP , Cannata-And??a JB. High phosphorus diet induces vascular calcification, a related decrease in bone mass and changes in the aortic gene expression . Bone . 2010 ; 46 ( 1 ): 121 - 8 .
37. Kuro-o M , Matsumura Y , Aizawa H , et al. Mutation of the mouse klotho gene leads to a syndrome resembling ageing . Nature . 1997 ; 390 ( 6655 ): 45 - 51 .
38. Nakatani T , Sarraj B , Ohnishi M , et al. In vivo genetic evidence for klotho-dependent, fibroblast growth factor 23 (Fgf23)-mediated regulation of systemic phosphate homeostasis . FASEB J . 2009 ; 23 ( 2 ): 433 - 41 .
39. Suga T , Kurabayashi M , Sando Y , et al. Disruption of the klotho gene causes pulmonary emphysema in mice. Defect in maintenance of pulmonary integrity during postnatal life . Am J Respir Cell Mol Biol . 2000 ; 22 ( 1 ): 26 - 33 .
40. Ohnishi M , Razzaque MS . Dietary and genetic evidence for phosphate toxicity accelerating mammalian aging . FASEB J . 2010 ; 24 ( 9 ): 3562 - 71 .
41. Morishita K , Shirai A , Kubota M , et al. The progression of aging in klotho mutant mice can be modified by dietary phosphorus and zinc . J Nutr . 2001 ; 131 ( 12 ): 3182 - 8 .
42. European Food Safety Authority. Assessment of one published review on health risks associated with phosphate additives in food . EFSA Journal . 2013 ; 11 : 3444 . https://doi.org/10.2903/j.efsa. 2013 . 3444 .
43. Varraso R , Chiuve SE , Fung TT , et al. Alternate Healthy Eating Index 2010 and risk of chronic obstructive pulmonary disease among US women and men: prospective study . BMJ . 2015 ; 350 : h286 .
44. Jiang R , Paik DC , Hankinson JL , Barr RG . Cured meat consumption, lung function and chronic obstructive pulmonary disease among United States adults . Am J Respir Crit Care Med . 2007 ; 175 ( 8 ): 798 - 804 .
45. Varraso R , Jiang R , Barr RG , Willet WC , Camargo CA Jr. Prospective study of cured meats consumption and risk of chronic obstructive pulmonary disease in men . Am J Epidemiol . 2007 ; 166 ( 12 ): 1438 - 45 .
46. Okubo H , Shaheen SO , Ntani G , Hertfordshire Cohort Study Group, et al. Processed meat consumption and lung function: modification by antioxidants and smoking . Eur Respir J . 2014 ; 43 ( 4 ): 972 - 82 .
47. de Battle J , Mendez M , Romieu I , PAC-COPD Study Group , et al. Cured meat consumption increases risk of readmission in COPD patients . Eur Respir J . 2012 ; 40 ( 3 ): 555 - 60 .
48. Ravikumar P , Ye J , Zhang J , et al. a-Klotho protects against oxidative damage in pulmonary epithelia . Am J Physiol Lung Cell Mol Physiol . 2014 ; 307 ( 7 ): L566 - 75 .
49. Ravikumar P , Li L , Ye J , et al. aKlotho deficiency in acute kidney injury contributes to lung damage . J Appl Physiol . 2016 ; 120 ( 7 ): 723 - 32 .
50. Rzewuska-Lech E , Jayachandran M , Fitzpatrick LA , Miller VM . Differential effects of 17beta-estradiol and raloxifene on VSMC phenotype and expression of osteoblast-associated proteins . Am J Physiol Endocrinol Metab . 2005 ; 289 ( 1 ): E105 - 12 .
51. Jeon GH , Kim SH , Yun SC , Chae HD , Kim CH , Kang BM . Association between serum estradiol levels and coronary artery calcification in postmenopausal women . Menopause . 2010 ; 17 ( 5 ): 902 - 7 .
52. Collins P , Rosano GM , Sarrel PM , et al. 17 beta-Estradiol attenuates acetylcholine-induced coronary arterial constriction in women but not men with coronary heart disease . Circulation . 1995 ; 92 ( 1 ): 24 - 30 .
53. Onufrak SJ , Bellasi A , Shaw LJ , et al. Phosphorus levels are associated with subclinical atherosclerosis in the general population . Atherosclerosis . 2008 ; 199 ( 2 ): 424 - 31 .