The Journal of Infectious Diseases

MAJOR ARTICLE

Vitamin D Supplementation and Antibiotic Use in Older

Australian Adults: An Analysis of Data From the D-Health
Trial

Downloaded from https://academic.oup.com/jid/article/226/6/949/6628138 by National Science & Technology Library user on 26 June 2023
Hai Pham,1,2, Mary Waterhouse,1 Catherine Baxter,1 Briony Duarte Romero,1 Donald S. A. McLeod,1,3 Bruce K. Armstrong,4,5 Peter R. Ebeling,6
Dallas R. English,7,8 Gunter Hartel,1 Michael G. Kimlin,9 Rachel L. O’Connell,10 Jolieke C. van der Pols,11 Alison J. Venn,12 Penelope M. Webb,1,2
David C. Whiteman,1 and Rachel E. Neale1,2,
1
Population Health Department, QIMR Berghofer Medical Research Institute, Brisbane, Australia; 2School of Public Health, University of Queensland, Brisbane, Australia; 3Department of
Endocrinology and Diabetes, Royal Brisbane and Women’s Hospital, Brisbane, Australia; 4School of Public Health, University of Sydney, Sydney, Australia; 5School of Population and Global Health,
University of Western Australia, Perth, Australia; 6Department of Medicine, School of Clinical Sciences, Monash University, Melbourne, Australia; 7Melbourne School of Population Health,
University of Melbourne, Melbourne, Australia; 8Cancer Epidemiology Division, Cancer Council Victoria, Melbourne, Australia; 9School of Biomedical Sciences, Queensland University of
Technology, Brisbane, Australia; 10NHMRC Clinical Trials Centre, University of Sydney, Sydney, Australia; 11School of Exercise and Nutrition Sciences, Faculty of Health, Queensland University of
Technology, Brisbane, Australia; and 12Menzies Institute for Medical Research, University of Tasmania, Hobart, Australia

(See the Editorial Commentary by Peter Bergman, on pages 947–8.)

Background. Vitamin D supplementation may reduce the risk or severity of infection, but this has been investigated in few large
population-based trials. We analyzed data from the D-Health Trial, using prescription of antibiotics as a surrogate for infection.
Methods. The D-Health Trial is a randomized, double-blind, placebo-controlled trial in which 21 315 Australians aged
60–84 years were randomized to 60 000 IU of supplementary vitamin D3 or placebo monthly for 5 years. For this analysis, the
primary outcome was the number of antibiotic prescription episodes; secondary outcomes were total number of prescriptions,
repeat prescription episodes, and antibiotics for urinary tract infection. We estimated incidence rate ratios (IRRs) using negative
binomial regression, and odds ratios using logistic regression.
Results. Vitamin D supplementation slightly reduced the number of prescription episodes (IRR, 0.98; 95% confidence interval
[CI], .95–1.01), total prescriptions (IRR, 0.97; 95% CI, .93–1.00), and repeat prescription episodes (IRR, 0.96; 95% CI, .93–1.00).
There was stronger evidence of benefit in people predicted to have insufficient vitamin D at baseline (prescription episodes IRR,
0.93; 95% CI, .87–.99).
Conclusions. Vitamin D may reduce the number of antibiotic prescriptions, particularly in people with low vitamin D status.
This supports the hypothesis that vitamin D has a clinically relevant effect on the immune system.
Clinical Trials Registration. Australian New Zealand Clinical Trials Registry: ACTRN12613000743763. https://www.anzctr.
org.au/.
Keywords. antibiotic use; vitamin D supplementation; randomized controlled trial; urinary tract infection.

Infectious diseases contribute greatly to global disease burden Bacteria are one cause of infection, but it is difficult to capture
[1], with infants aged <1 year and adults aged ≥65 years expe­ the population incidence of bacterial infection. Antibiotics are
riencing the highest burden [2]. On average, there are approx­ used to treat bacterial infections [8, 9] and, although sometimes
imately 1500 hospitalizations for infection per 100 000 used prophylactically or in the context of a viral infection, they
population each year in high-income countries [3, 4]. Acute re­ may be a surrogate indicator of bacterial infection. On average,
spiratory tract infection (ARTI), urinary tract infection (UTI), each person in Australia has 1 antibiotic prescription per year;
and infections of the skin, soft tissue, and gastrointestinal tract adults aged ≥65 years have the highest rate of dispensing at ap­
are common in people aged ≥65 years [5–7]. proximately 1.3 prescriptions per year [10].
Vitamin D, best known for its role in serum calcium regula­
tion, may also regulate the immune system, upregulating innate
immunity and downregulating aspects of acquired immunity
Received 12 April 2022; editorial decision 26 May 2022; accepted 30 June 2022; published
[11, 12]. Systematic reviews and meta-analyses of observational
online 3 July 2022 studies report an inverse correlation between serum
Correspondence: R. E Neale, PhD, QIMR Berghofer Medical Research Institute, 300 Herston
25-hydroxy vitamin D concentration (25(OH)D; the most
Road, Herston 4006, Queensland, Australia ([email protected]).
The Journal of Infectious Diseases® 2022;226:949–57
commonly used marker of vitamin D status) and risk and se­
© The Author(s) 2022. Published by Oxford University Press on behalf of Infectious Diseases verity of infections [13, 14].
Society of America. All rights reserved. For permissions, please e-mail: journals.permissions
Although many studies have investigated the effect of vitamin
@oup.com
https://doi.org/10.1093/infdis/jiac279 D supplementation on infectious diseases, convincing evidence

Vitamin D and Antibiotic Use • JID 2022:226 (15 September) • 949

of benefit exists only for infection of the respiratory tract. participants were randomly assigned to 2 groups of equal
Meta-analyses of randomized controlled trials (RCTs) show size. Once the code was verified, we applied it to the dataset
that vitamin D supplementation reduces the risk of ARTI [15, with participants returned to their original study groups.
16]. It also increases the sputum culture conversion rate in pa­ Analyses performed after this have been declared exploratory.
tients with pulmonary tuberculosis [17, 18]. Five RCTs have
Baseline Characteristics
studied the effect of vitamin D supplementation on UTI and
the results are heterogeneous, but as the studies were relatively Participants were asked to complete a survey at baseline that
small (n < 600) and performed in selected populations, the re­ asked about sociodemographic characteristics, lifestyle factors,
and health. We did not measure serum 25(OH)D concentra­
sults may not be generalizable to the broader population [19–23].

Downloaded from https://academic.oup.com/jid/article/226/6/949/6628138 by National Science & Technology Library user on 26 June 2023
tion at baseline. Instead, we predicted whether deseasonalized
Four previous studies have investigated the effect of vitamin
D supplementation on antibiotic use; 3 studies were among baseline serum 25(OH)D concentration was <50 nmol/L using
a model that was developed and validated using data collected
older adults (aged ≥50 years, all n ≤ 5108) [24–26] and 1 was
from placebo participants during the trial (area under the re­
conducted in patients with antibody deficiency or frequent re­
spiratory tract infections (n = 124) [27]. These studies generat­ ceiver operating characteristic curve, 0.71) [30].
ed inconclusive results [24–27]. We used data from the large Ascertaining Antibiotic Use
population-based D-Health Trial to assess the effect of vitamin We asked participants to consent to linkage of their records
D supplementation on antibiotic prescriptions as a surrogate with Australia’s Pharmaceutical Benefits Scheme (PBS) dataset.
indicator of bacterial infection. The PBS is part of Australia’s universal health insurance
scheme, which subsidies the cost of prescription pharmaceuti­
METHODS cals for all Australian citizens and permanent residents. The
PBS dataset contains information on medicines that qualify
Trial Design, Participants, and Intervention
for a government subsidy and for which a claim has been pro­
We have previously published detailed trial methods [28].
cessed; it captures almost all antibiotics, with the exception of
Briefly, the D-Health Trial was a randomized, placebo-
some antibiotics administered in public hospitals. Services
controlled, double-blind trial with 2 parallel arms. Australians
Australia supplied the PBS data, from which we extracted all re­
aged between 60 and 79 years, randomly selected from a popu­
cords pertaining to general anti-infectives for systemic use
lation register, were invited to participate. Volunteers aged 60 to
(ie, Anatomical Therapeutic Chemical classification code J01)
84 years were also included. People were ineligible if they re­
that were dispensed over an approximately 61-month period,
ported a history of hypercalcemia, hyperparathyroidism, kidney
from 14 days after the first study tablet was due to be taken until
stones, osteomalacia or sarcoidosis, or were taking >500 inter­
1 month after the due date for the final tablet.
national units (IU) of supplementary vitamin D per day. We
Every 12 months after randomization we asked participants
randomly allocated participants in a 1:1 ratio to monthly doses
(including withdrawn participants) to complete an annual survey
of either 60 000 IU of cholecalciferol (vitamin D3) or placebo
that asked: “In the last 12 months have you been prescribed anti­
[28], using automated computer-generated permuted block
biotics?” Those who answered “yes” were asked to indicate the
randomization, stratified by age, sex, and state of residence; staff
reason(s) antibiotics were prescribed (Supplementary Material 1).
and investigators did not have access to the allocation list.
The primary outcome of D-Health was all-cause mortality Primary Outcome
[29]; secondary outcomes were incidence of cancer at all sites The primary outcome for this analysis was the total number of
(excluding keratinocyte cancer) and cancer of the colorectum antibiotic prescription episodes, under the assumption that this
specifically. Antibiotic use and infections were prespecified ter­ is a surrogate for the number of clinically relevant bacterial in­
tiary outcomes. The QIMR Berghofer Medical Research fections experienced. A prescription episode was defined as
Institute Human Research Ethics Committee approved the trial comprising one or more dispensed antibiotic prescriptions
and all participants gave consent to participate. with fewer than 3 weeks between consecutive prescriptions.
For each prescription episode, the first prescription dispensed
Blinding was assumed to be the initiation of treatment of an infection,
Participants were unblinded after all had completed the inter­ and any other prescriptions dispensed in the episode (regard­
vention period (March 2020), using processes to ensure that less of antibiotic type) were assumed to be for continued treat­
study staff, analysts, and investigators remained blind until af­ ment of the same infection (ie, repeat prescriptions).
ter the analysis of the primary outcome was complete. The cur­
rent analysis was performed blind to group allocation. We Secondary Outcomes
developed and tested code for all prespecified analyses using Due to the frequency with which UTIs occur in the elderly (we
a dataset in which group allocation had been removed, and have previously reported on ARTI [31]), we analyzed use of

950 • JID 2022:226 (15 September) • Pham et al

antibiotics to treat a UTI in the year prior to an annual survey who were missing data for any of the confounders included
(pooled across 5 surveys). For this analysis we combined data in the multivariable models were excluded (n = 1407; 7%).
from annual surveys and the PBS (Supplementary Material 2).
In sensitivity analyses, we analyzed the total number of anti­ Subgroup Analyses
biotics dispensed to assess the effect of excluding records that We assessed whether the effect of vitamin D supplementation
were assumed to be repeat prescriptions. We did not account on our primary outcome varied across prespecified subgroups
for prophylactic use of antibiotics because the percentage of par­ of: (1) age (<70; ≥70 years); (2) sex; (3) body mass index (<25;
ticipants who reported prophylactic use was very similar be­ ≥25 kg/m2); (4) self-rated overall health (very good or excel­
tween the vitamin D and placebo groups (Supplementary lent; good, fair, or poor); and (5) predicted baseline serum

Downloaded from https://academic.oup.com/jid/article/226/6/949/6628138 by National Science & Technology Library user on 26 June 2023
Table 1). 25(OH)D concentration (<50; ≥50 nmol/L). We tested wheth­
er these characteristics modified the effect of vitamin D supple­
Monitoring Compliance and Adverse Events mentation on antibiotic use by fitting interaction terms.
To monitor compliance, we measured serum 25(OH)D concen­
tration in blood collected during the intervention from a ran­ Exploratory Analyses

domly selected sample of participants; selection was stratified We hypothesized that the observed effect of vitamin D on the
by study group, age, sex, state, and month of annual survey com­ total number of prescriptions might be due, in part, to differ­
pletion. Each annual survey asked participants to report the ences in the number of repeat prescriptions between the 2 study
number of study tablets they had taken in the previous year arms. We therefore estimated the effect of vitamin D supple­
and about their use of off-study supplements containing vita­ mentation on the number of repeat antibiotic prescription ep­
min D. We asked participants to report any adverse events via isodes (ie, antibiotic prescribing episodes where antibiotics
telephone or email, and also captured diagnoses of hypercalce­ were dispensed more than once). In a sensitivity analysis, we
mia, kidney stones, and hyperparathyroidism in annual surveys. excluded any repeat episodes where all prescriptions were dis­
pensed on the same day.
Statistical Methods We also assessed whether the effect of vitamin D supplemen­
We followed a predefined analysis plan and used SAS version tation on total antibiotic prescriptions and repeat antibiotic pre­
9.4 (SAS Institute) and R version 3.6.1 (R Foundation for scription episodes varied across subgroups, as defined above.
Statistical Computing). Analyses of the effect of vitamin D sup­ Additionally, we conducted an exploratory analysis in which
plementation on antibiotic use followed the intention-to-treat we examined the effect of vitamin D supplementation on the
principle and were adjusted for randomization stratification number of prescriptions of individual classes of antibiotics.
factors of age, sex, and state of residence. We present effect es­
timates and their 95% confidence intervals (CIs), and use a sig­ RESULTS
nificance level of P < .05 with no adjustment for multiple Approximately 8.5% of participants from each study group did
testing [32, 33]. not give permission to access their PBS data (Figure 1). There
All D-Health participants (n = 21 315), excluding the 5 peo­ were differences between participants who did and did not con­
ple who requested that their data be deleted, were eligible for sent to linkage (Supplementary Table 2), but the baseline char­
this analysis. Participants who did not consent to linkage acteristics of included participants were well balanced
with PBS data were excluded. For analyses of UTI, we further (Table 1). For the analysis of UTI, a further 89 people were ex­
excluded participants with missing data for all 5 annual sur­ cluded (vitamin D, n = 38; placebo, n = 51). Compliance and
veys. We compared baseline characteristics between those in­ adverse events have been published [29]. Briefly, more than
cluded and excluded in the primary analysis using χ2 tests. 80% of participants reported taking ≥80% of their study cap­
We analyzed count variables (ie, number of prescription epi­ sules, and there was little difference between the 2 groups
sodes, number of prescriptions) using negative binomial re­ with respect to adverse events. Female sex, older age
gression, estimating incidence rate ratios (IRRs) and rate (≥70 years), markers of poorer health, and having predicted se­
differences per person year at risk (RDs). Occurrence of UTIs rum 25(OH)D concentration <50 nmol/L were associated with
was analyzed according to whether a UTI requiring antibiotic more antibiotic prescription episodes (Supplementary Table 3).
treatment had occurred each year, using logistic regression to
calculate odds ratios (ORs); generalized estimating equations Effect of Vitamin D Supplementation on Antibiotic Use
with an exchangeable correlation matrix accounted for cluster­ We observed a slightly lower rate of antibiotic prescription ep­
ing across surveys. isodes in the intervention than in the placebo group (0.94 vs
We assessed the association between baseline factors and the 0.96 episodes per person year), although the CIs were compat­
number of antibiotic prescription episodes using univariable ible with no effect (IRR, 0.98 [95%, CI, .95 to 1.01]; RD, −0.02
and multivariable negative binomial models. Participants [95%, CI, −.05 to .01]; Table 2). The total number of antibiotics

Vitamin D and Antibiotic Use • JID 2022:226 (15 September) • 951

 Downloaded from https://academic.oup.com/jid/article/226/6/949/6628138 by National Science & Technology Library user on 26 June 2023
Figure 1. Participant flow (CONSORT flow diagram). Abbreviations: EOI, expression of interest; ITT, intention to treat; PBS, Pharmaceutical Benefits Scheme.

dispensed in the vitamin D arm was slightly lower than in the antibiotic prescriptions and the number of repeat antibiotic ep­
placebo arm (1.46 vs 1.51 prescriptions per person year; IRR, isodes (Supplementary Figures 1 and 2, and Supplementary
0.97 [95% CI, .93 to 1.00]; RD, −0.05 [95% CI, −.10 to .00]), Tables 6 and 7).
and they also had fewer repeat prescription episodes (0.32 vs
0.34 repeat episodes per person year; IRR, 0.96 [95% CI, .93
DISCUSSION
to 1.00]; RD, −0.01 [95% CI, −.03 to .00]; Table 2). Vitamin
D supplementation did not reduce the odds of antibiotic use These analyses indicate that supplementation with 60 000 IU of
for UTI (pooled OR, 1.01; 95% CI, .94 to 1.09; Table 2). In ex­ vitamin D per month may lead to a small reduction in antibi­
ploratory analyses, vitamin D supplementation reduced the otic use among older Australians, particularly in those aged
number of antibiotic prescriptions in all classes with the excep­ <70 years, who reported poorer overall health, or who had
tion of penicillins and other antibacterials (Supplementary low predicted baseline vitamin D status.
Table 4). The rate of antibiotic prescribing in the D-Health cohort was
The effect of vitamin D supplementation was modified by consistent with expectations. Almost half of Australians aged
baseline age, overall health, and predicted serum 25(OH)D ≥45 years had at least 1 antibiotic prescribed in 2015, and the
concentration (P interactions .055, .04, and .052, respectively; number of antibiotic prescriptions dispensed increased with
Figure 2). Vitamin D supplementation reduced the number age [10]. In line with this, 47% of D-Health participants had
of antibiotic prescription episodes in people aged <70 years at least 1 antibiotic prescription in 2015. The rate of antibiotic
(IRR, 0.95 [95% CI, .91 to .99]; RD −0.04 [95% CI, −.08 to dispensing was similar to that in the large population-based 45
−.01]), with poorer overall health (IRR, 0.95 [95% CI, .91 to and Up study (1.51 vs 1.64 prescriptions per person per year for
1.00]; RD, −0.05 [95% CI, −.10 to .00]), and with predicted se­ the D-Health and 45 and Up studies, respectively) [10].
rum 25(OH)D concentration <50 nmol/L (IRR, 0.93 [95% CI, Consistent with other studies [10, 34], antibiotic prescribing
.87 to .99]; RD, −0.07 [95% CI, −.13 to −.01]; Figure 2 and was more common amongst older people, women, and those
Supplementary Table 5). There was negligible effect in alterna­ with poorer self-reported overall health.
tive categories of these variables. Similar patterns were ob­ The reduction in the total number of prescriptions and in the
served for the exploratory analyses of the total number of number of repeat prescription episodes in the vitamin D

952 • JID 2022:226 (15 September) • Pham et al

Table 1. Baseline Characteristics for Participants Included in the 95% CI, .93–1.04) [26]. One other trial (n = 124) in people
Analysis According to Randomization Group with antibody deficiency and frequent respiratory tract infec­
tion also found that vitamin D supplementation reduced the
Placebo, No. (%) Vitamin D, No. (%)
Characteristic (n = 9735) (n = 9762) number of days on antibiotics (mean difference, 17 days per
Sex
year; P = .02) [27]. The findings of these studies and the
Men 5327 (54.7) 5336 (54.7) D-Health Trial may suggest a benefit of vitamin D supplemen­
Women 4408 (45.3) 4426 (45.3) tation on the length or severity of infection.
Age, y Ongoing symptoms of infection resulting in longer antibiot­
60–64 2399 (24.6) 2387 (24.5)
ic treatment could be triggered by hyperinflammatory re­

Downloaded from https://academic.oup.com/jid/article/226/6/949/6628138 by National Science & Technology Library user on 26 June 2023
65–69 2677 (27.5) 2690 (27.6)
70–74 2657 (27.3) 2657 (27.2)
sponses [35]. In vitro studies have shown that the active form
≥75 2002 (20.6) 2028 (20.8) of vitamin D (calcitriol) modulates the anti-inflammatory pro­
Body mass index, kg/m² cess by downregulating inflammatory cytokines and upregulat­
<25 2869 (29.6) 2985 (30.7) ing anti-inflammatory cytokines [36]. Thus, one possible
≥25 6820 (70.4) 6738 (69.3) explanation for the reduced number of repeat antibiotic pre­
Missing 46 39
scription episodes in the D-Health Trial, and for fewer days
Smoking history
Never 5253 (54.4) 5315 (54.9)
on antibiotics in other trials, is the potential anti-inflammatory
Ex-smoker 3959 (41.0) 3998 (41.3) properties of vitamin D.
Current 438 (4.5) 377 (3.9) We did not observe an effect of vitamin D supplementation
Missing 85 72 on self-reported UTI requiring antibiotic treatment. Five other
Self-rated overall health
RCTs have considered the effect of vitamin D supplementation
Excellent or very good 5323 (55.5) 5396 (56.2)
Good, fair, or poor 4271 (44.5) 4205 (43.8)
on UTI, most in selected population subgroups [19–23].
Missing 141 161 Weekly doses of vitamin D supplements (20 000 IU) reduced
Self-rated quality of life the risk of self-reported UTI in prediabetic adults (n = 511)
Excellent or very good 6459 (67.8) 6444 (67.5) [19], and daily doses of 600 IU vitamin D reduced the risk of
Good, fair, or poor 3070 (32.2) 3100 (32.5) clinically confirmed UTI in men with benign prostatic hyper­
Missing 206 218
plasia (n = 389) [22]. The DO-HEALTH trial (a multicenter
Predicted serum 25(OH)D concentration, nmol/L
<50 2393 (24.6) 2325 (23.8)
RCT among 2157 people aged ≥70 years) found a lower rate
≥50 7342 (75.4) 7437 (76.2) of clinically confirmed UTI in people supplemented with
2000 IU vitamin D per day than the placebo group (n = 542;
IRR, 0.64; 95% CI, .42–.96) [23]. In contrast, supplementation
compared with the placebo group was slightly more marked of postmenopausal women (n = 297) with 20 000 IU twice
than the reduction in the number of prescription episodes. weekly did not alter the number of self-reported UTIs in the
While this may be a chance finding, it might also indicate previous 12 months [21], and a small study in infants (n =
that vitamin D affects the length or severity, but not the inci­ 65) found no difference in the frequency of UTI between the
dence, of infection. A previous analysis of D-Health Trial vitamin D and placebo groups [20]. Overall, there is limited ev­
data found that vitamin D supplementation slightly reduced idence of a specific effect of vitamin D supplementation on
the duration and severity of ARTI [31]. Further, among a sub­ UTI.
group of participants who completed a respiratory symptom In our subgroup analyses, vitamin D supplementation only
diary and had an ARTI (n = 814), a smaller percentage of peo­ reduced the number of antibiotic prescription episodes in peo­
ple in the vitamin D group than in the placebo group received ple aged <70 years, those with predicted baseline serum
antibiotic treatment for ARTI, although this was not statisti­ 25(OH)D concentration <50 nmol/L, and in those who report­
cally significant (18% vs 23%; OR, 0.79; 95% CI, .54–1.15) [31]. ed poorer overall health. In contrast to our results, 2 other trials
Previous RCTs have generated inconsistent results regarding found an effect of vitamin D supplementation on antibiotic
the effect of vitamin D supplementation on antibiotic prescrip­ prescriptions in people aged ≥70 years, but not in younger peo­
tions. Three studies found that vitamin D supplementation ple [25, 26]. The ViDA Trial did not observe any interaction
substantially reduced antibiotic use, albeit with relatively with baseline vitamin D status [26].
wide CIs (OR, 0.37 [95% CI, .15–.87]; OR, 0.84 [95% CI, .64– Results from the ViDA Trial suggested that vitamin D sup­
1.09]; and OR, 0.72 [95% CI, .48–1.07]) [24, 25, 27]. In the plements reduced only the number of tetracycline prescriptions
population-based ViDA trial (n = 5108), 100 000 IU/month (IRR, 0.81; 95% CI, .70–.93), with no effect on other antibiotic
of vitamin D over 3.3 years had a more marked effect on the classes [26]. According to current clinical guidelines, most
number of days on antibiotics (IRR, 0.90; 95% CI, .82–.98) common antibiotics (such as penicillin and cephalosporin)
than on the number of antibiotic prescriptions (IRR, 0.98; have indications for infections with different types of bacteria

Vitamin D and Antibiotic Use • JID 2022:226 (15 September) • 953

Table 2. Effects of Vitamin D Supplementation on Antibiotic Prescriptions Dispensed

Placebo (n = 9735, PYAR = 47 306) Vitamin D (n = 9762, PYAR = 47 491)

Outcomes
No. of People Incidence Rate Incidence Rate Rate Differencea
With ≥1 No. of per 1 PYARa No. of People With No. of per 1 PYARa IRRa per 1 PYAR
Episode Episodes (95% CI) ≥1 Episodes Episodes (95% CI) (95% CI) (95% CI)

Antibiotic prescription 8292 45 367 0.96 (.94–.98) 8288 44 754 0.94 (.92–.96) .98 (.95–1.01) −0.02 (−.05 to .01)
episodesb
Total antibiotic 8292 71 098 1.51 (1.47–1.55) 8288 69 467 1.46 (1.42–1.49) .97 (.93–1.00) −0.05 (−.10 to .00)
prescriptions

Downloaded from https://academic.oup.com/jid/article/226/6/949/6628138 by National Science & Technology Library user on 26 June 2023
Repeat episodes, 5762 16 083 0.34 (.33–.35) 5707 15 614 0.32 (.32–.33) .96 (.93–1.00) −0.01 (−.03 to .00)
exploratory analysisc
Repeat episodes, 5218 12 900 0.27 (.26–.28) 5148 12 435 0.26 (.25–.27) .96 (.92–1.00) −0.01 (−.02 to .00)
sensitivity analysisd
Placebo, No. Episodes/No. Vitamin D, No. Episodes/No. ORe (95% CI)
Surveys (%) (n = 9684) Surveys (%) (n = 9724)
Antibiotics for UTIf 2775/47 028 (5.9) 2826/47 238 (6.0) 1.01 (.94–1.09)
Abbreviations: CI, confidence interval; IRR, incidence rate ratio; OR, odds ratio; PYAR, person year at risk; UTI, urinary tract infection.
a
Incidence rates, rate differences, and incidence rate ratios from negative binomial regression models that included the randomization stratification variables of sex, age, and state of residence
at baseline.
b
Prescriptions dispensed within 21 days of another prescription comprised a single episode.
c
A repeat antibiotic prescription episode was defined as a prescription episode with more than 1 record.
d
Repeat episodes based on antibiotics prescribed on the same day were excluded.
e
Outcomes pooled across 5 surveys; OR estimated using logistic regression models (using generalized estimating equations with an exchangeable correlation matrix) with adjustment for sex,
age, and state of residence at baseline.
f
People with missing data for all 5 annual surveys were excluded (n = 89).

occurring in a range of organs. Furthermore, prescriptions of status, as the positive predictive value for 25(OH)D concentra­
antibiotics are subject to practitioner preference, so there is tion <50 nmol/L was modest (23%).
no clear link between antibiotic class and type of infection. D-Health participants were somewhat healthier than the old­
This, combined with the mechanism of action of vitamin D er Australian population. The mean 25(OH)D concentration in
on the immune system, suggests it is unlikely that vitamin D the placebo group was a little higher than reported in the 2011/
would have a specific effect on use of any particular antibiotic 2012 Australian Health survey (77 vs 69 nmol/L) [37], but this
class. We found evidence of an effect across most classes, albeit may be due to differences in the geographic distribution of par­
with the exception of penicillin and other antibiotics, confirm­ ticipants or the timing of blood sampling. Importantly, these
ing this hypothesis. results are unlikely to be generalizable to populations with a
Two recent meta-analyses of RCTs suggested that daily dos­ much greater prevalence of vitamin D deficiency.
ing was more effective than intermittent dosing for the preven­ Applying these findings to the Australian population, we es­
tion of ARTI. However, the P value for interaction was not timate that routine vitamin D supplementation of older
significant in the 2021 meta-analysis and both meta-analyses Australian adults may reduce the total number of antibiotic
showed that the effect of vitamin D on risk of ARTI was signifi­ prescriptions from 1.51 million to 1.46 million per million peo­
cant in children aged 1 to 16 years old but not in adults [17, 18]. ple per year (ie, by approximately 250 000 prescriptions per an­
Thus, the effect on ARTI of daily versus bolus doses of vitamin num). In older people with 25(OH)D concentration <50 nmol/
D remains inconclusive. L, there would be 138 000 fewer antibiotic prescriptions (8% re­
This analysis has a number of limitations. Firstly, we used duction) per million people per year. While these relatively
antibiotics as a surrogate for infection; a limitation is that we small differences may have limited clinical significance at an in­
did not obtain clinical confirmation of infectious episodes. dividual level, at a population level fewer antibiotic prescrip­
However antibiotic use was ascertained via linkage with tions may lower the chance of developing antibiotic
Australia’s PBS data, minimizing any risk of selection bias resistance. The possible stronger effect in people who were pre­
due to incomplete follow-up. Secondly, we did not measure dicted to have lower vitamin D status at baseline suggests that
baseline 25(OH)D concentration, because the trial was not avoiding vitamin D deficiency may reduce the need for antibi­
powered for subgroup analyses of our primary outcome (mor­ otics to treat infection.
tality) so we could not justify the expense. Misclassification of In conclusion, the results of the D-Health Trial support the
participants’ baseline vitamin D status may have attenuated notion that vitamin D plays a role in infectious diseases.
the effect estimate in people predicted to have low vitamin D While the effect in this largely vitamin D-replete population

954 • JID 2022:226 (15 September) • Pham et al

 Downloaded from https://academic.oup.com/jid/article/226/6/949/6628138 by National Science & Technology Library user on 26 June 2023
Figure 2. Effect of vitamin D supplementation on the number of antibiotic prescription episodes for all participants and within participant subgroups, defined according to
baseline characteristics. Incidence rate ratios were estimated using negative binomial models, adjusting for baseline age, sex, and state of residence at randomization.
Abbreviations: CI, confidence interval; IR, incidence rate.

was small, these findings support policies to minimize the prev­ Acknowledgments. We acknowledge the D-Health Trial staff
alence of vitamin D deficiency in older people. and members of the data and safety monitoring board (Patricia
Valery, Ie-Wen Sim, and Kerrie Sanders); Services Australia for
Supplementary Data supplying Pharmaceutical Benefits Schedule data; the D-Health
Supplementary materials are available at The Journal of Trial participants who committed to this research; and the staff
Infectious Diseases online. Consisting of data provided by the of the data linkage units of the State and Territory health de­
authors to benefit the reader, the posted materials are not copy­ partments (Western Australia, Victoria, South Australia,
edited and are the sole responsibility of the authors, so ques­ Northern Territory, New South Wales, Queensland, and
tions or comments should be addressed to the corresponding Tasmania) for the linkage of the date of death data used in
author. this study. Furthermore, we thank the data custodians from
The State Registries of Births, Deaths, and Marriages.
Notes Financial support. This work was supported by National
Author contributions. D. W., P. W., G. H., D. E., M. K., R. O., Health and Medical Research Council (NHMRC) (grant num­
J. V., A. V., C. B., B. D. R., P. E., D. M., B. A., and R. N. were bers GNT1046681 and GNT1120682). P. R. E., R. E. N.,
involved in designing the trial. C. B., B. D. R., M. W., D. M., P. M. W., and D. C. W. are/were supported by fellowships
and R. N. were involved in the recruitment, data collection, from the NHMRC (grant numbers GNT1197958,
and curation. H. P., M. W., and R. N. were involved in the in­ GNT1060183, GNT1173346, and GNT1155413). D. S. A.
vestigations, formal analysis, and validation of the study. H. P. M. is supported by a Metro North Clinician Research
wrote the first draft of the report with input from M. W. and Fellowship (Metro North Hospital andHealth Service) and
R. N. All authors were involved in writing-review and editing Queensland Advancing Clinical Research Fellowship
the final draft. M. W. and R. N. provided supervision. (Queensland Government). H. P. is supported by a University

Vitamin D and Antibiotic Use • JID 2022:226 (15 September) • 955

of Queensland PhD Scholarship. The vitamin D assays was per­ 10. Chen Y, Kirk MD, Stuart R, et al. Socio-demographic and
formed at the University of Western Australia, supported by in­ health service factors associated with antibiotic dispensing
frastructure funding from the Western Australian State in older Australian adults. PLoS One 2019; 14:e0221480.
Government in partnership with the Australian Federal 11. Hewison M. Vitamin D and immune function: an over­
Government, through Bioplatforms Australia and the view. Proc Nutr Soc 2012; 71:50–61.
National Collaborative Research Infrastructure Strategy. 12. Golpour A, Bereswill S, Heimesaat MM. Antimicrobial and
Potential conflicts of interest. R. E. N. has received funding immune-modulatory effects of vitamin D provide promis­
from Viatris for an unrelated study of pancreatic ing antibiotics-independent approaches to tackle bacterial
cancer. P. M. W. has received funding from Astra Zeneca for infections—lessons learnt from a literature survey. Eur J

Downloaded from https://academic.oup.com/jid/article/226/6/949/6628138 by National Science & Technology Library user on 26 June 2023
an unrelated study of ovarian cancer. P. R. E. reports grants Microbiol Immunol (Bp) 2019; 9:80–7.
from Amgen, grants from Sanofi, and grants from Alexion. 13. Autier P, Boniol M, Pizot C, Mullie P. Vitamin D status and
All other authors declare no potential conflicts of interests. ill health: a systematic review. Lancet Diabetes Endocrinol
All authors have submitted the ICMJE Form for Disclosure 2014; 2:76–89.
of Potential Conflicts of Interest. Conflicts that the editors con­ 14. Deng QF, Chu H, Wen Z, Cao YS. Vitamin D and urinary
sider relevant to the content of the manuscript have been tract infection: a systematic review and meta-analysis. Ann
disclosed. Clin Lab Sci 2019; 49:134–42.
15. Jolliffe DA, Camargo CA Jr, Sluyter JD, et al. Vitamin D
supplementation to prevent acute respiratory infections:
References a systematic review and meta-analysis of aggregate data
1. Vos T, Lim SS, Abbafati C, et al. Global burden of 369 dis­ from randomised controlled trials. Lancet Diabetes
eases and injuries in 204 countries and territories, 1990– Endocrinol 2021; 9:276–92.
2019: a systematic analysis for the Global Burden of 16. Martineau AR, Jolliffe DA, Hooper RL, et al. Vitamin D
Disease Study 2019. Lancet 2020; 396:1204–22. supplementation to prevent acute respiratory tract infec­
2. Cassini A, Colzani E, Pini A, et al. Impact of infectious dis­ tions: systematic review and meta-analysis of individual
eases on population health using incidence-based participant data. BMJ 2017; 356:i6583.
disability-adjusted life years (DALYs): results from the 17. Karbalaei M, Ghazvini K, Keikha M. Clinical efficacy of vi­
burden of communicable diseases in Europe study. tamin D supplementation on pulmonary TB patients: the
European Union and European Economic Area countries evidence of clinical trials. J Clin Tuberc Other Mycobact
2009 to 2013. Euro Surveill 2018; 23:17. Dis 2020; 20:100174.
3. Australian Government, Australian Institute of Health and 18. Zhang J, Chen C, Yang J. Effectiveness of vitamin D supple­
Wellbeing. Australia’s health 2020, 2020. https://www. mentation on the outcome of pulmonary tuberculosis
aihw.gov.au/reports/australias-health/infectious-and-comm treatment in adults: a meta-analysis of randomized con­
unicable-diseases. Accessed 3 January 2022. trolled trials. Chin Med J Engl 2019; 132:2950–9.
4. Kennedy JL, Haberling DL, Huang CC, et al. Infectious dis­ 19. Jorde R, Sollid ST, Svartberg J, Joakimsen RM, Grimnes G,
ease hospitalizations: United States 2001 to 2014. Chest Hutchinson MY. Prevention of urinary tract infections
2019; 156:255–68. with vitamin D supplementation 20,000 IU per week for
5. Chami K, Gavazzi G, Carrat F, et al. Burden of infections five years. Results from an RCT including 511 subjects.
among 44,869 elderly in nursing homes: a cross-sectional Infect Dis (Lond) 2016; 48:823–8.
cluster nationwide survey. J Hosp Infect 2011; 79:254–9. 20. Merrikhi A, Ziaei E, Shahsanai A, Kelishadi R,
6. Choy CS, Chen H, Yau CS, Hsu EK, Chik NY, Wong AT. Maghami-Mehr A. Is vitamin D supplementation effective
Prevalence of infections among residents of residential care in prevention of recurrent urinary tract infections in the
homes for the elderly in Hong Kong. Hong Kong Med J pediatrics? A randomized triple-masked controlled trial.
2016; 22:347–55. Adv Biomed Res 2018; 7:150.
7. Hepper HJ, Sieber C, Walger P, Bahrmann P, Singler K. 21. Oberg J, Verelst M, Jorde R, Cashman K, Grimnes G. High
Infections in the elderly. Crit Care Clin 2013; 29:757–74. dose vitamin D may improve lower urinary tract symptoms
8. Australian Commission on Safety and Quality in Health in postmenopausal women. J Steroid Biochem Mol Biol
Care (ACSQHC). AURA 2016—first Australian report on 2017; 173:28–32.
antimicrobial use and resistance in human health. 22. Safwat AS, Hasanain A, Shahat A, et al. Cholecalciferol for
Sydney: ACSQHC, 2016. the prophylaxis against recurrent urinary tract infection
9. Silverman M, Povitz M, Sontrop JM, et al. Antibiotic pre­ among patients with benign prostatic hyperplasia: a ran­
scribing for nonbacterial acute upper respiratory infections domized, comparative study. World J Urol 2019; 37:
in elderly persons. Ann Intern Med 2017; 166:765–74. 1347–52.

956 • JID 2022:226 (15 September) • Pham et al

23. Bischoff-Ferrari HA, Vellas B, Rizzoli R, et al. Effect of vi­ 30. Waterhouse M, Baxter C, Duarte Romero B, et al.
tamin D supplementation, omega-3 fatty acid supplemen­ Predicting deseasonalised serum 25 hydroxy vitamin D
tation, or a strength-training exercise program on clinical concentrations in the D-health trial: an analysis using
outcomes in older adults: the DO-HEALTH randomized boosted regression trees. Contemp Clin Trials 2021; 104:
clinical trial. JAMA 2020; 324:1855–68.
106347.
24. Avenell A, Cook JA, Maclennan GS, Macpherson GC.
31. Pham H, Waterhouse M, Baxter C, et al. The effect of vita­
Vitamin D supplementation to prevent infections: a sub-
min D supplementation on acute respiratory tract infection
study of a randomised placebo-controlled trial in older
people (RECORD trial, ISRCTN 51647438). Age Ageing in older Australian adults: an analysis of data from the

Downloaded from https://academic.oup.com/jid/article/226/6/949/6628138 by National Science & Technology Library user on 26 June 2023
2007; 36:574–7. D-Health Trial. Lancet Diabetes Endocrinol 2021; 9:69–81.
25. Tran B, Armstrong BK, Ebeling PR, et al. Effect of vitamin 32. Li G, Taljaard M, Van den Heuvel ER, et al. An introduc­
D supplementation on antibiotic use: a randomized con­ tion to multiplicity issues in clinical trials: the what, why,
trolled trial. Am J Clin Nutr 2014; 99:156–61. when and how. Int J Epidemiol 2017; 46:746–55.
26. Wu Z, Camargo CA, Sluyter J, Waayer D, Toop L, Scragg R. 33. Rothman KJ. No adjustments are needed for multiple com­
Effect of monthly vitamin D supplementation on antibiotic parisons. Epidemiology 1990; 1:43–6.
prescribing in older adults: a post hoc analysis of a random­ 34. Mertz D, Tjam EY, Poss J, et al. Prevalence and predictors
ized controlled trial. Am J Clin Nutr 2021; 114:314–21. of antibiotic use in community-based elderly in Ontario,
27. Bergman P, Norlin AC, Hansen S, et al. Vitamin D3 sup­
Canada. Infect Control Hosp Epidemiol 2011; 32:710–3.
plementation in patients with frequent respiratory tract in­
35. Fajgenbaum DC, June CH. Cytokine storm. N Engl J Med
fections: a randomised and double-blind intervention
2020; 383:2255–73.
study. BMJ Open 2012; 2:e001663.
36. Prietl B, Treiber G, Pieber TR, Amrein K. Vitamin D and
28. Neale RE, Armstrong BK, Baxter C, et al. The D-health tri­
al: a randomized trial of vitamin D for prevention of mor­ immune function. Nutrients 2013; 5:2502–21.
tality and cancer. Contemp Clin Trials 2016; 48:83–90. 37. Australian Bureau of Statistics. Australian Health Survey:
29. Neale RE, Baxter C, Romero BD, et al. The D-health trial: a biomedical results for nutrients, 2011–12. http://www.abs.
randomised controlled trial of the effect of vitamin D on gov.au/ausstats/[email protected]/Lookup/4364.0.55.006main
mortality. Lancet Diabetes Endocrinol 2022; 10:120–8. +features12011-12. Accessed 23 June 2020.

Vitamin D and Antibiotic Use • JID 2022:226 (15 September) • 957