E3463 Full

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

PNAS PLUS

Global increase and geographic convergence in


antibiotic consumption between 2000 and 2015
Eili Y. Kleina,b,c,1, Thomas P. Van Boeckeld, Elena M. Martineza, Suraj Panta, Sumanth Gandraa, Simon A. Levine,f,g,1,
Herman Goossensh, and Ramanan Laxminarayana,f,i
a
Center for Disease Dynamics, Economics & Policy, Washington, DC 20005; bDepartment of Emergency Medicine, Johns Hopkins School of Medicine,
Baltimore, MD 21209; cDepartment of Epidemiology, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD 21205; dInstitute of Integrative
Biology, ETH Zürich, CH-8006 Zürich, Switzerland; eDepartment of Ecology and Evolutionary Biology, Princeton University, Princeton, NJ 08544; fPrinceton
Environmental Institute, Princeton University, Princeton, NJ 08544; gBeijer Institute of Ecological Economics, SE-104 05 Stockholm, Sweden; hLaboratory of
Medical Microbiology, Vaccine & Infectious Diseases Institute, University of Antwerp, 2610 Antwerp, Belgium; and iDepartment of Global Health, University
of Washington, Seattle, WA 98104

Contributed by Simon A. Levin, February 23, 2018 (sent for review October 3, 2017; reviewed by Bruce R. Levin and Dominique L. Monnet)

Tracking antibiotic consumption patterns over time and across to reduce antibiotic resistance. Given the urgency of the threat posed
countries could inform policies to optimize antibiotic prescribing by rising AMR levels (2), and in the absence of global, publicly
and minimize antibiotic resistance, such as setting and enforcing per funded, harmonized surveillance data on antibiotic use, alternative
capita consumption targets or aiding investments in alternatives to sources of data on antibiotic use must be used to track antibiotic
antibiotics. In this study, we analyzed the trends and drivers of consumption patterns across countries. Here, we use pharmaceutical
antibiotic consumption from 2000 to 2015 in 76 countries and
sales data to document global trends in antibiotic consumption.
projected total global antibiotic consumption through 2030. Be-
There have been few attempts to assess antibiotic consump-
tween 2000 and 2015, antibiotic consumption, expressed in defined
daily doses (DDD), increased 65% (21.1–34.8 billion DDDs), and the tion globally or in multiple countries (14–17), none of which has
antibiotic consumption rate increased 39% (11.3–15.7 DDDs per reported data later than 2010. The largest prior study reported
1,000 inhabitants per day). The increase was driven by low- and that antibiotic consumption increased 36% between 2000 and
middle-income countries (LMICs), where rising consumption was 2010 in 71 countries (15). However, the results from this study
correlated with gross domestic product per capita (GDPPC) growth cannot be directly compared with other studies (14, 16) because
(P = 0.004). In high-income countries (HICs), although overall con- the data were not reported as defined daily doses (DDDs), the
sumption increased modestly, DDDs per 1,000 inhabitants per day most commonly used metric to measure antibiotic consumption.
fell 4%, and there was no correlation with GDPPC. Of particular

ENVIRONMENTAL
In this study, we report on antibiotic consumption in DDDs for
concern was the rapid increase in the use of last-resort compounds, an expanded number of countries (n = 76) and over a longer

SCIENCES
both in HICs and LMICs, such as glycylcyclines, oxazolidinones, car-
time period (2000–2015). In addition, we assess differences in
bapenems, and polymyxins. Projections of global antibiotic con-
sumption in 2030, assuming no policy changes, were up to 200% consumption between high-income countries (HICs) and low-
higher than the 42 billion DDDs estimated in 2015. Although anti- and middle-income countries (LMICs), identify drivers of anti-
biotic consumption rates in most LMICs remain lower than in HICs biotic use in each income group from a set of socioeconomic and
despite higher bacterial disease burden, consumption in LMICs is
rapidly converging to rates similar to HICs. Reducing global con- Significance
sumption is critical for reducing the threat of antibiotic resistance,
but reduction efforts must balance access limitations in LMICs and Antibiotic resistance, driven by antibiotic consumption, is a
take account of local and global resistance patterns. growing global health threat. Our report on antibiotic use in
76 countries over 16 years provides an up-to-date compre-
|
antimicrobial resistance low-income countries | defined daily doses | hensive assessment of global trends in antibiotic consumption.
|
antibiotic stewardship antibiotics We find that the antibiotic consumption rate in low- and
middle-income countries (LMICs) has been converging to (and

A ntibiotic resistance—the ability of microbes to evolve and


withstand the effects of antibiotics—is a significant cause of
morbidity and mortality globally (1–3). Antibiotic consumption is
in some countries surpassing) levels typically observed in high-
income countries. However, inequities in drug access persist, as
many LMICs continue to be burdened with high rates of in-
a primary driver of antibiotic resistance (4). The association be- fectious disease-related mortality and low rates of antibiotic
tween antibiotic consumption and resistance is well documented consumption. Our findings emphasize the need for global
across spatial and temporal scales at individual hospitals (5), surveillance of antibiotic consumption to support policies to
nursing homes (6), primary care facilities (7), and communities reduce antibiotic consumption and resistance while providing
(8), as well as across countries (9). Many countries have adopted access to these lifesaving drugs.
national action plans on antimicrobial resistance (AMR) that aim
to reduce per capita antibiotic consumption. The Global Action Author contributions: E.Y.K. and R.L. designed research; E.Y.K. performed research;
E.Y.K., E.M.M., and S.P. analyzed data; and E.Y.K., T.P.V.B., E.M.M., S.G., S.A.L., H.G.,
Plan on Antimicrobial Resistance endorsed by the member states and R.L. wrote the paper.
of the World Health Organization (WHO) and affirmed at the Reviewers: B.R.L., Emory University; and D.L.M., European Centre for Disease Prevention
high-level meeting on antimicrobial resistance during the 71st and Control.
General Assembly of the United Nations (10), recommends that The authors declare no conflict of interest.
all countries collect and report antibiotic consumption data (11). This open access article is distributed under Creative Commons Attribution License 4.0 (CC
Surveillance data on country-level antibiotic use are needed to BY).
(i) monitor national and global trends over time; (ii) compare an- 1
To whom correspondence may be addressed. Email: [email protected] or slevin@princeton.
tibiotic use among countries; (iii) provide a baseline for the evalu- edu.
ation of future efforts to reduce antibiotic use; (iv) enable This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
epidemiological analysis of the association between antibiotic use 1073/pnas.1717295115/-/DCSupplemental.
and resistance over time (12, 13); and (v) support policies that aim Published online March 26, 2018.

www.pnas.org/cgi/doi/10.1073/pnas.1717295115 PNAS | vol. 115 | no. 15 | E3463–E3470


Fig. 1. Global antibiotic consumption by country: 2000–2015. (A) Change in the national antibiotic consumption rate between 2000 and 2015 in DDDs per
1,000 inhabitants per day. For Vietnam, Bangladesh, The Netherlands, and Croatia, change was calculated from 2005, and for Algeria from 2002 as data
before those years for those countries were not available. (B) Antibiotic consumption rate by country for 2015 in DDDs per 1,000 inhabitants per day. Data
source: IQVIA MIDAS, 2000–2015, IQVIA Inc. All rights reserved (https://www.iqvia.com/solutions/commercialization/geographies/midas).

medical indicators, and project future growth in global antibiotic by interpolation, using the ratio of antibiotic consumption in the hospital and
consumption. retail sectors for the years for which data had been reported.
Data on antibiotic sales in standard units (SUs) and kilograms were pur-
Methods chased under license from IQVIA. An SU is an IQVIA designation that rep-
resents a single-dose unit such as a pill, capsule, or equal amount of liquid.
We estimated global antibiotic consumption using the IQVIA MIDAS data- Sales expressed in kilograms were converted into DDDs using the Anatomical
base. IQVIA uses national sample surveys of antibiotic sales to develop es- Therapeutic Chemical Classification System (ATC/DDD, 2016) developed by
timates of the total volume of sales of each antibiotic molecule (or combination the WHO Collaborating Centre for Drug Statistics Methodology. For mole-
of molecules). For each country, antibiotic consumption was reported by month cules not included in the ATC/DDD index, DDD values were estimated from
or quarter and broken down between the retail and hospital sectors. We other sources or as the average of DDD unit values by class (SI Appendix, SI
obtained data for 76 countries from 2000 through 2015. Central America (Costa Methods). Data for SUs were available for all years, whereas kilogram data
Rica, El Salvador, Guatemala, Honduras, Nicaragua, and Panama) and French were available only for the period 2005–2015. The ratio of SUs to kilograms
for 2005–2015 was used to estimate kilograms and DDDs for 2000–2004. A
West Africa (Benin, Burkina Faso, Cameroon, Chad, Côte d’Ivoire, Republic of
country’s annual antibiotic consumption rate in DDDs per 1,000 inhabitants
Congo, Guinea, Mali, Niger, Senegal, and Togo) were included as two indi-
per day was calculated using population estimates from the World Bank
vidual groups of countries with aggregated sales for these regions. Sixty-six DataBank. Consumption rates were subsequently compared between groups
countries had data available for every year between 2000 and 2015, while data of countries based on their World Bank income classification in 2007.
on the remaining countries covered partial time periods (SI Appendix, Table Fixed-effects panel regression analysis was used to quantify the association
S1). In countries where both hospital and retail data were reported for some between economic and health indicators and the antibiotic consumption
but not all years (2000–2015), consumption in the missing sector was estimated rate. The explanatory variables included per capita gross domestic product

E3464 | www.pnas.org/cgi/doi/10.1073/pnas.1717295115 Klein et al.


PNAS PLUS
(GDP; purchasing power parity); imports of goods and services as a per- The antibiotic consumption rate of broad-spectrum penicillins,
centage of GDP (as a measure of trade); measles vaccination coverage in the most commonly consumed class of antibiotics (39% of total
children between the ages of 12 and 23 mo [lack of pneumococcal conjugate DDDs in 2015), increased 36% between 2000 and 2015 globally.
vaccination (PCV) coverage information limited our ability to include PCV
The greatest increase was in LMICs (56%), although the antibiotic
coverage as a variable]; physician density per 1,000 population; and per-
consumption rate in HICs increased 15% (Fig. 3A). While the
centage of the population living in urban areas (as a measure of health
system access). All explanatory variables were pooled by country/year. Re-
antibiotic consumption rate of the next three most consumed
gression models analyzed data for HICs and LMICs separately. Health and eco- classes—cephalosporins (20% of total DDDs), quinolones (12% of
nomic factors were obtained from the World Development Indicators in the total DDDs), and macrolides (12% of total DDDs)—all increased
World Bank DataBank (18). Serial correlation was assessed using the Wooldridge overall, the antibiotic consumption rate decreased in HICs. In
test (19). Errors were clustered by country to account for high serial correlation. LMICs, the antibiotic consumption rate increased 399, 125, and
STATA version 14.1 was used for all statistical analyses. 119% for cephalosporins, quinolones, and macrolides, respectively,
Global antibiotic consumption was calculated by extrapolating use for while the antibiotic consumption rate of these three drugs in HICs
countries not included in the IQVIA database. Extrapolations were based on decreased by 18, 1, and 25%, respectively (Fig. 3 B–D).
the average per capita antibiotic use for countries from the same income Consumption of newer and last-resort antibiotic classes in-
group. We then projected global antibiotic use until 2030, assuming constant creased across all country income groups between 2000 and 2015.
per capita use rates for all countries at current levels, with total use increasing The United States was the largest consumer of glycylcyclines
only through population growth. In addition to this baseline projection, we
(tigecycline) and oxazolidinones (primarily linezolid as tedizolid
modeled two other scenarios: (i) no policy changes, where countries’ anti-
biotic consumption rates for 2016 through 2030 were assumed to continue
was not introduced until 2014) through the late 2000s. However,
to change based on their compounded growth rate between 2010 and the antibiotic consumption rate of these drugs in the United States
2015 and (ii) a target policy in which countries were assumed to converge to began declining in 2009 (Fig. 4 A and B), and in 2015, Taiwan,
the 2015 global median antibiotic consumption rate by 2020 (SI Appendix, SI Italy, Turkey, and Austria all had higher consumption rates for
Methods). Population projections were retrieved from the World Bank glycylcyclines than the United States, while India surpassed the
DataBank (18) except for Taiwan, for which data were obtained from the United States antibiotic consumption rate for oxazolidinones in
Taiwan National Development Council. 2012 to become the highest consumer. The antibiotic consumption
rate of carbapenems increased greatly in LMICs between 2000 and
Results 2015 but remained far below consumption rates in HICs (Fig. 4C).
Global antibiotic consumption increased by 65% between 2000 and Similarly, the antibiotic consumption rate of polymyxins (largely
2015, from 21.1 to 34.8 billion DDDs, while the antibiotic con- colistin) also increased in LMICs, particularly in LMICs-UM
sumption rate increased 39% from 11.3 to 15.7 DDDs per countries, but remained much lower than in HICs (Fig. 4D).
1,000 inhabitants per day over the study period. The mean anti- The highest polymyxin consumption rates were in Spain, the
biotic consumption rate across countries increased 28% from United Kingdom, and Ireland, all of which had rates greater than

ENVIRONMENTAL
16.4 (SD 9.9) DDDs per 1,000 inhabitants per day to 20.9 (SD 9.8), 0.05 DDDs per 1,000 inhabitants per day in 2015.

SCIENCES
and the median antibiotic consumption rate increased 25% from We found a significant positive association between GDP per
15.5 to 19.5 DDDs per 1,000 inhabitants per day. capita and changes in the antibiotic consumption rate in LMICs
The increase in global consumption was primarily driven by (P = 0.004), although no statistically significant association was
increased consumption in LMICs. In 2000, HICs, led by France, found between these factors in HICs (P = 0.52). Other indica-
New Zealand, Spain, Hong Kong, and the United States, had the tors, including the measles vaccination rate (which is a proxy for
highest antibiotic consumption rates. In 2015, four of the six public health intervention capability), imports as a percentage of
countries with the highest consumption rates were LMICs (Turkey, GDP, and physician density, were not correlated with changes in
Tunisia, Algeria, and Romania; Fig. 1 and SI Appendix, Fig. S1). In per capita antibiotic use across countries, irrespective of income
HICs, although the total amount of antibiotics consumed increased group (Table 1).
6% between 2000 and 2015, from 9.7 to 10.3 billion DDDs, the Between 2000 and 2015, the estimated total global antibiotic
antibiotic consumption rate decreased by a modest 4%, from 26.8 to consumption rate (including countries not reported in the
25.7 DDDs per 1,000 inhabitants per day (Fig. 2A). In LMICs, an- IQVIA database) decreased slightly in HICs from 27.0 to 25.7
tibiotic consumption increased 114%, from 11.4 to 24.5 billion DDDs, DDDs per 1,000 inhabitants per day in HICs and increased
and the antibiotic consumption rate increased 77%, from 7.6 to from 8.6 to 13.9 DDDs per 1,000 inhabitants per day in LMICs
13.5 DDDs per 1,000 inhabitants per day. Low- and lower-middle- (SI Appendix, Fig. S2). Total global antibiotic consumption in
income countries (LMICs-LM) accounted for a greater share of this
increase than upper-middle-income countries (LMICs-UM): total
Table 1. Fixed-effects regression analysis of factors associated
antibiotic consumption in LMICs-LM increased 117%, from 8.1 to with global antibiotic consumption (DDD per capita): 2000–2015
17.5 billion DDDs, while, in LMICs-UM, antibiotic consumption in-
creased 110%, from 3.3 to 6.9 billion DDDs (Fig. 2B). The antibiotic Coefficient (SD)*
consumption rate in both LMICs-UM and LMICs-LM increased
Low- and
78%, from 12.0 to 21.3 DDDs per 1,000 inhabitants per day, and from
middle-income High-income
6.7 to 11.9 DDDs per 1,000 inhabitants per day, respectively. Factor countries countries
In 2015, the leading HIC consumers of antibiotics were the
United States, France, and Italy, while the leading LMIC con- Log(GDP per capita) 3.14 (1.00)† 0.56 (0.70)
sumers were India, China, and Pakistan. Whereas antibiotic Percentage of children (12–23 mo) 0.04 (0.05) 0.07 (0.06)
consumption in the three leading HICs marginally increased, the vaccinated for measles
highest-consuming LMICs saw large increases. Between 2000 Log(Imports as percentage of GDP) −1.01 (1.01) −0.20 (1.16)
and 2015, antibiotic consumption increased from 3.2 to 6.5 billion Physician density per 1,000 population 1.39 (0.73) 0.49 (0.34)
DDDs (103%) in India, from 2.3 to 4.2 billion DDDs (79%) in Observations 302 305
China, and from 0.8 to 1.3 billion DDDs (65%) in Pakistan. The Countries 39 32
antibiotic consumption rate increased from 8.2 to 13.6 DDDs per Data source: IQVIA MIDAS, 2000–2015, IQVIA Inc. All rights reserved
1,000 inhabitants per day (63%) in India, from 5.1 to 8.4 DDDs per (https://www.iqvia.com/solutions/commercialization/geographies/midas).
1,000 inhabitants per day (65%) in China, and from 16.2 to *Both regressions are clustered by country to adjust for high serial correlation.

19.6 DDDs per 1,000 inhabitants per day (21%) in Pakistan. P < 0.01.

Klein et al. PNAS | vol. 115 | no. 15 | E3465


Fig. 2. Global antibiotic consumption by country income classification: 2000–2015. (A) Graph showing how the antibiotic consumption rate in DDDs per
1,000 inhabitants per day has rapidly increased for LMICs, while remaining nearly constant for HICs. However, as shown in B, the larger population sizes in
many LMICs result in greater total antibiotic consumption (DDDs) in LMICs even though their consumption rate (and thus per capita use) is lower. In B, each
bar reflects total consumption in the specified year for that country or group of countries. Data source: IQVIA MIDAS, 2000–2015, IQVIA Inc. All rights reserved
(https://www.iqvia.com/solutions/commercialization/geographies/midas).

2015 was estimated to be 42.3 billion DDDs (15.8 DDDs per we assumed no policy changes and constant antibiotic consumption
1,000 inhabitants per day)—10.7 billion DDDs in HICs and rates set at current levels of use, global antibiotic use is projected to
31.6 billion DDDs in LMICs. In our baseline condition, where increase 15% between 2015 and 2030. If all countries continue to

Fig. 3. Antibiotic consumption rate for HICs, LMICs-UM, and LMICs-LM of the four most-consumed therapeutic classes of antibiotics in DDDs per 1,000 inhabitants per day.
(A) Broad-spectrum penicillins, which correspond to the Anatomical Therapeutic Chemical (ATC) classification of penicillins with extended spectrum (J01CA) excluding car-
benicillins. (B) Cephalosporins, which correspond to the ATC classification codes J01DB, J01DC, J01DD, and J01DE for the four generations of cephalosporins. (C) Macrolides,
which correspond to the ATC classification for macrolides, lincosamides, and streptogramins (J01F). (D) Quinolones, which correspond to the ATC classification for quinolone
antibacterials (J01M). Data source: IQVIA MIDAS, 2000–2015, IQVIA Inc. All rights reserved (https://www.iqvia.com/solutions/commercialization/geographies/midas).

E3466 | www.pnas.org/cgi/doi/10.1073/pnas.1717295115 Klein et al.


PNAS PLUS
ENVIRONMENTAL
SCIENCES
Fig. 4. Antibiotic consumption rate for HICs, LMICs-UM, and LMICs-LM of new and last-resort antibiotics in DDDs per 1,000 inhabitants per day. (A) Gly-
cylcyclines, which correspond to the ATC classification for tigecycline (J01AA12). (B) Oxazolidinones, which correspond to the ATC classifications for linezolid
(J01XX08) and tedizolid (J01XX11). (C) Carbapenems, which correspond to the ATC classification for carbapenems (J01DH). (D) Polymyxins, which correspond
to ATC classification for polymyxins (J01XB). Data source: IQVIA MIDAS, 2000–2015, IQVIA Inc. All rights reserved (https://www.iqvia.com/solutions/
commercialization/geographies/midas).

increase their antibiotic consumption rates at their compounded an- sistance levels has become increasingly visible. Despite the
nual growth rates, we estimate that total consumption would increase emergence and spread of nearly untreatable infections, the global
202% to 128 billion DDDs, while the antibiotic consumption rate response to this public health crisis remains slow and inadequate.
would increase 161% to 41.1 DDDs per 1,000 inhabitants per day. Reducing antibiotic consumption rates in HICs and slowing the
Finally, if all countries converge on the global median antibiotic growth rate of consumption in LMICs is urgently needed to
consumption rate in 2015 of 17.9 DDDs per 1,000 inhabitants contain the problem of resistance, particularly given the long
per day by 2020, we estimate global antibiotic consumption timescales and resources necessary for development of new anti-
would increase 32% to 55.6 billion DDDs (Fig. 5). biotics. However, there is a need to balance access to essential
medications, particularly in LMICs where the burden of infectious
Discussion diseases likely still outweighs the burden of resistant infections and
Using a global database of antibiotic sales, we found that anti- where in many countries there is a significant unmet need for
biotic consumption rates increased dramatically in LMICs be- antibiotics. Stewardship can improve judicious use without
tween 2000 and 2015, and in some LMICs have reached levels diminishing access to effective medications. Efforts to reduce
previously reported only in HICs. Overall consumption has also unnecessary or inappropriate use based on awareness campaigns
greatly increased, and the total amount of antibiotics consumed have resulted in lower antibiotic consumption rates in some high-
in LMICs, which was similar to HICs in 2000, was nearly consuming HICs (20). However, maintaining those efforts in the
2.5 times that in HICs in 2015. Rising incomes are a major driver long run has proven challenging (21, 22), and the methods used in
of increased antibiotic consumption in LMICs. Thus, although HICs may not be appropriate or feasible in LMICs. Research is
rates of antibiotic consumption in most LMICs remain below the urgently needed to understand the most effective methods for
general rate in HICs, barring major policy changes, they are implementing stewardship programs in LMICs from the local to
expected to increase over time and converge, and possibly sur- the national level in a manner that does not restrict antibiotics
pass, antibiotic consumption rates in HICs, in part due to the from those most burdened by treatable diseases.
higher burden of infectious diseases in LMICs. Numerous studies have examined the drivers of consumption
Tracking rates of antibiotic use is vitally important because of within and between HICs; however, the large variations in anti-
the well-quantified relationship between antibiotic use and re- biotic consumption among countries are poorly explained. In our
sistance. However, although data on the burden of resistant bac- study, increases in the antibiotic consumption rate in LMICs were
terial infections is limited, both in HICs and especially in LMICs positively correlated with per capita GDP growth rates, but no
(2), the magnitude of the challenge posed by rising antibiotic re- similar relationship could be identified for HICs. Increases in

Klein et al. PNAS | vol. 115 | no. 15 | E3467


We observed prominent differences in the rate of change in
classes of antibiotics consumed in HICs and LMICs, particularly
cephalosporins, consumption of which increased rapidly in LMICs
while declining in HICs. These changes are concerning because
the consumption of cephalsoporins, particularly third-generation
cephalsoporins, is associated with emergence of extended spectrum
beta-lactamase–producing bacteria (30). Consumption of other
broad-spectrum agents like fluoroquinolones, macrolides, and
second-line agents like oxazolidinones also increased in LMICs and
decreased in HICs. In contrast, consumption of broad-spectrum
penicillins, carbapenems, and polymyxins increased in both HICs
and LMICs, although the rate of increase was faster in LMICs-UM
than LMICs-LM. Changes in consumption patterns in LMICs
likely reflect increasing access due to economic growth as noted
above, as well as patent expiration, which reduces barriers to access
as well as the cost of medications. However, the changing com-
position of consumption may also reflect alterations in patterns of
resistance. For example, India, which had the greatest increase in
antibiotic consumption in LMICs (65% between 2000 and 2015),
had sustained economic growth (∼10% annual increase in GDP)
through the 2000s. A large portion of this increase in consumption
was due to an increase in the use of cephalosporins, which was
likely due not only to economic growth, but also to changing pre-
scribing practices for respiratory tract infections, skin and soft tis-
sue infections, gonococcal infections, and enteric fever, where
cephalosporins have replaced penicillins and quinolones for in-
Fig. 5. Projected total global antibiotic consumption (billions of DDDs): 2000–
fection management due to rising resistance (31). Similarly, with
2030. Estimated global antibiotic consumption in all countries in billions of
DDDs for three scenarios: (i) all countries continue to consume at current per
increasing quinolone resistance in Salmonella Typhi (32), cepha-
capita rates; (ii) consumption of all countries continues to change at current losporins have become the treatment of choice for this infection in
compound annual growth rates; and (iii) all countries converge to the global both outpatient (cefixime, cefpodoxime) and inpatient (ceftriax-
median antibiotic consumption rate. Estimates were produced using antibiotic one) settings (33). LMICs, with higher disease burdens, may be
use data for 2000–2015 from the IQVIA MIDAS database and World Bank more sensitive to rising rates of resistance, which may drive local
DataBank population estimates and projections for 2000–2030. Data source: changes in antibiotic consumption patterns. However, a lack of
IQVIA MIDAS, 2000–2015, IQVIA Inc. All rights reserved (https://www.iqvia. surveillance for resistant infections makes it difficult to ascribe
com/solutions/commercialization/geographies/midas). consumption changes to infectious disease burdens or resistance
changes to overall consumption levels. Increased surveillance for
resistance and infectious disease burden can improve the scientific
economic growth provide access to goods and services, including
basis for reducing antibiotic consumption.
antibiotics, which provides the most likely explanation for the
Globally, the consumption of last-resort antibiotics—carbape-
positive relationship found between increasing wealth and in-
nems and colistin—has been increasing. This rise is consistent with
creasing antibiotic consumption in LMICs. However, GDP growth a well-documented increase in the number of infections resistant to
in LMICs is also linked with increasing urbanization, which can carbapenems and colistin (34, 35). Given the recent finding that
facilitate the transmission of infectious diseases (23) and may plasmid-mediated colistin resistance is spreading globally (35), this
contribute to the link between GDP growth and antibiotic con- drug should be used prudently in humans and animals. Although
sumption. Increasing incidence of bacterial infections, such as consumption of the newer second-line drugs glycylcyclines and
enteric fever (24), has been associated with rapid urbanization and oxazolidinones declined in the United States after the Food and
increased consumption of antibiotics. Furthermore, rising in- Drug Administration warned of a higher risk of death compared
cidence of nonbacterial infections associated with urbanization, with other drugs used to treat similar infections (36, 37), globally
such as dengue, chikungunya (25, 26), and viral diarrheal illnesses the consumption of these drugs has increased. With rapidly
(27), is a significant driver of inappropriate consumption of anti- growing incomes in LMICs, continued high disease burden, and
biotics in LMICs. Additionally, declining air quality associated increasing rates of resistance, an accelerated uptake of these and
with urbanization and continued use of solid fuels for cooking also other new drugs can be expected compared with prior uptake
drives antibiotic consumption in LMICs by increasing the in- rates for older drugs. This may lead to shorter time frames of
cidence of acute respiratory-tract infections (28). effectiveness for new drugs, and as newer drugs are often less
The lack of a relationship between economic growth and anti- well-tolerated, an increase in adverse drug-related events.
biotic consumption in HICs suggests that access is not the leading While rising incomes provide increased access to medications,
factor driving differences between countries. Rather differences in in resource-limited settings, financial barriers continue to restrict
consumption rates are driven by social and cultural norms regarding access to antibiotics particularly for the most vulnerable. This is
attitudes toward prescribing and use of antibiotics [see, e.g., further compounded by rising resistance to affordable first-line
Blommaert et al. (29)]. Therefore, embedding judicious use as a treatments, which already prevents the most vulnerable pop-
normative value in LMICs could prevent the future inappropriate ulations from accessing effective treatments in some countries
use of antibiotics that currently plagues HICs. In addition, un- (1). Innovative pricing mechanisms should be developed to im-
derstanding the factors that result in lower antibiotic consumption prove access to lifesaving drugs for vulnerable populations (38)
rates in some HICs may help identify future policy solutions to without compromising their future efficacy (31). Reducing the
promote similar trends in LMICs. However, additional studies in burden of disease is an alternative mechanism for reducing an-
LMICs are needed to identify the regulatory and policy factors tibiotic consumption, particularly in countries where antibiotics
that result in lower consumption across these settings and whether are used as surrogates for other infection control measures (39).
they differ from higher-income settings. Investments in sanitation and improved hygiene measures were a

E3468 | www.pnas.org/cgi/doi/10.1073/pnas.1717295115 Klein et al.


PNAS PLUS
major driving force in reducing the burden of infectious diseases ber of units per package and the amount of active substance per
in HICs in the 20th century. Large infrastructure projects in unit has increased over time in Europe and perhaps also in other
LMICs that improve the delivery of clean, safe water could re- continents (45). Despite these limitations, reporting of antibiotic
duce the burden of disease without increasing the use of anti- consumption rates on a global scale is critical for carrying out
biotics. For example, a recent study in India demonstrated that effective policies to reduce consumption, such as setting and
improved water quality could significantly reduce the burden of enforcing targets for antibiotic consumption based on current
childhood diarrheal diseases, which in turn would reduce anti- consumption levels (46) and inducing change by identifying the
biotic consumption (40). Improved quality and quantity of water heaviest consumers of antibiotics (13). As part of the global action
would likely improve hand hygiene compliance, particularly in plan on AMR adopted by WHO, UN member states were urged
healthcare settings. In HICs, numerous studies have found that
to establish national action plans to combat antibiotic resistance,
improved hand hygiene helps reduce the consumption of anti-
including reporting antibiotic consumption (11). However, addi-
biotics in the clinical setting (41). Access to vaccines and point-
tional study is needed to link antibiotic consumption rates of
of-care diagnostics could similarly reduce unnecessary antibiotic
consumption in resource-poor settings (42), both directly through specific drugs and resistance rates of target pathogens to better
vaccination against bacterial illnesses, such as the PCV (43, 44), inform policies to reduce antibiotic consumption.
or indirectly by reducing viral illnesses that often are treated With antibiotic consumption increasing worldwide, the challenge
unnecessarily with antibiotics. Policies to promote these alter- posed by antibiotic resistance is likely to get worse. As with cli-
natives to reduce consumption should be a major part of efforts mate change, there may be an unknown tipping point (47), and
to reduce antibiotic resistance. To aid policy development in this this could herald a future without effective antibiotics. Even in
arena, future research efforts should quantify the economic the absence of tipping points, the decline of antibiotic effec-
benefit of investments by developed countries in these types of tiveness represents a major threat to human health. Radical re-
interventions relative to investments in the discovery of new thinking of policies to reduce consumption is necessary, including
antibiotic compounds. major investments in improved hygiene, sanitation, vaccination,
To the best of our knowledge, IQVIA currently provides the and access to diagnostic tools both to prevent unnecessary anti-
only source of harmonized data on global antibiotic consumption. biotic use and to decrease the burden of infectious disease. While
Without an alternative surveillance system to estimate global more study is needed to understand the risks of radical reduc-
consumption, it is not possible to assess potential systematic bias in tions in consumption, immediate strategies are necessary to re-
the database used for this study; however, our estimates are duce mortality among the millions of people who die from
strongly correlated with data from the European Surveillance of resistant infections annually.
Antimicrobial Consumption Network (ESAC-Net) (SI Appendix,
Fig. S3), as well as others that have reported global antibiotic ACKNOWLEDGMENTS. R.L., E.Y.K., and S.G. were supported by grants from

ENVIRONMENTAL
consumption (14, 16). While the DDD consumption data that we the Bill & Melinda Gates Foundation to the Global Antibiotic Resistance

SCIENCES
report permit our estimates to be directly compared with other Partnership (OPP1112355) at the Center for Disease Dynamics, Economics
sources of antibiotic consumption, including those from ESAC- & Policy. R.L. and S.A.L. were supported by the Grand Challenges in Health
Program at Princeton University. T.P.V.B. was supported by the ETH Zürich
Net, DDDs are not a perfect outcome measure of antibiotic
postdoctoral fellowship program and by ETH Zürich’s Center for Adaptation
prescribing, particularly for penicillins (14, 45). Indeed, for these to a Changing Environment. E.M.M. and S.P. were supported by Intergovern-
drugs, the DDD is much lower than the actual prescribed dose, mental Personnel Agreements from the US Centers for Disease Control and
therefore overestimating antibiotic consumption. Also, the num- Prevention (16IPA1609425, 16IPA1609424).

1. Laxminarayan R, et al. (2016) Access to effective antimicrobials: A worldwide chal- 12. World Health Organization (2013) Integrated surveillance of antimicrobial resistance.
lenge. Lancet 387:168–175. Available at apps.who.int/iris/bitstream/10665/91778/1/9789241506311_eng.pdf. Ac-
2. Laxminarayan R, et al. (2013) Antibiotic resistance: The need for global solutions. cessed September 13, 2016.
Lancet Infect Dis 13:1057–1098. 13. Collineau L, et al. (2017) Guidance on the selection of appropriate indicators for
3. Lim C, et al. (2016) Epidemiology and burden of multidrug-resistant bacterial in- quantification of antimicrobial usage in humans and animals. Zoonoses Public Health
fection in a developing country. Elife 5:e18082. 64:165–184.
4. Malhotra-Kumar S, Lammens C, Coenen S, Van Herck K, Goossens H (2007) Effect of 14. Versporten A, et al.; WHO/Europe-ESAC Project Group (2014) Antibiotic use in Eastern
azithromycin and clarithromycin therapy on pharyngeal carriage of macrolide- Europe: A cross-national database study in coordination with the WHO Regional
resistant streptococci in healthy volunteers: A randomised, double-blind, placebo-
Office for Europe. Lancet Infect Dis 14:381–387.
controlled study. Lancet 369:482–490. 15. Van Boeckel TP, et al. (2014) Global antibiotic consumption 2000 to 2010: An analysis
5. Fridkin SK, et al.; Intensive Care Antimicrobial Resistance Epidemiology (ICARE)
of national pharmaceutical sales data. Lancet Infect Dis 14:742–750.
Project and the National Nosocomial Infections Surveillance (NNIS) System Hospitals
16. Högberg LD, Muller A, Zorzet A, Monnet DL, Cars O (2014) Antibiotic use worldwide.
(2001) The effect of vancomycin and third-generation cephalosporins on prevalence
Lancet Infect Dis 14:1179–1180.
of vancomycin-resistant enterococci in 126 U.S. adult intensive care units. Ann Intern 17. Col NF, O’Connor RW (1987) Estimating worldwide current antibiotic usage: Report of
Med 135:175–183.
Task Force 1. Rev Infect Dis 9(Suppl 3):S232–S243.
6. Daneman N, et al. (2015) Variability in antibiotic use across nursing homes and the risk
18. World Bank (2017) World DataBank: World Development Indicators. Available at
of antibiotic-related adverse outcomes for individual residents. JAMA Intern Med 175:
databank.worldbank.org/data/reports.aspx?source=world-development-indicators.
1331–1339.
Accessed May 31, 2017.
7. Costelloe C, Metcalfe C, Lovering A, Mant D, Hay AD (2010) Effect of antibiotic pre-
19. Drukker DM, et al. (2003) Testing for serial correlation in linear panel-data models.
scribing in primary care on antimicrobial resistance in individual patients: Systematic
Stata J 3:168–177.
review and meta-analysis. BMJ 340:c2096.
20. Sabuncu E, et al. (2009) Significant reduction of antibiotic use in the community after
8. Steinke D, Davey P (2001) Association between antibiotic resistance and community
a nationwide campaign in France, 2002-2007. PLoS Med 6:e1000084.
prescribing: A critical review of bias and confounding in published studies. Clin Infect
21. Institut de veille sanitaire (InVS) et Agence nationale de sécurité du médicament et
Dis 33(Suppl 3):S193–S205.
des produits de santé (ANSM) (2015) Consommation d’antibiotiques et résistance aux
9. Goossens H, Ferech M, Vander Stichele R, Elseviers M; ESAC Project Group (2005)
Outpatient antibiotic use in Europe and association with resistance: A cross-national antibiotiques en France : Nécessité d’une mobilisation déterminée et durable. Bilan
database study. Lancet 365:579–587. des Données de Surveillance. [Antibiotic consumption and antibiotic resistance in
10. United Nations (2016) Draft political declaration of the high-level meeting of the France: The need for a determined and sustainable mobilization. Review of Surveil-
General Assembly on antimicrobial resistance. Available at https://www.un.org/pga/ lance Data] (Institut de veille sanitaire, Saint-Maurice, France), pp 1-16. French.
71/wp-content/uploads/sites/40/2016/09/DGACM_GAEAD_ESCAB-AMR-Draft-Politi- 22. Shekarchian S, Schwartz BS, Teherani A, Irby D, Chin-Hong PV (2016) Is it time for a
cal-Declaration-1616108E.pdf. Accessed June 27, 2017. coordinated and longitudinal approach to antibiotic stewardship education? Clin
11. World Health Organization (2015) Global action plan on antimicrobial resistance. Infect Dis 63:848–849.
Available at www.who.int/antimicrobial-resistance/publications/global-action-plan/en/. 23. Alirol E, Getaz L, Stoll B, Chappuis F, Loutan L (2011) Urbanisation and infectious
Accessed May 31, 2017. diseases in a globalised world. Lancet Infect Dis 11:131–141.

Klein et al. PNAS | vol. 115 | no. 15 | E3469


24. Steele AD, Hay Burgess DC, Diaz Z, Carey ME, Zaidi AKM (2016) Challenges and op- 36. Food and Drug Administration (2013) FDA Drug Safety Communication: FDA warns of
portunities for typhoid fever control: A call for coordinated action. Clin Infect Dis increased risk of death with IV antibacterial Tygacil (tigecycline) and approves new
62(Suppl 1):S4–S8. boxed warning. Available at https://www.fda.gov/drugs/drugsafety/ucm369580.htm.
25. Gubler DJ (2011) Dengue, urbanization and globalization: The unholy trinity of the Accessed September 8, 2017.
21(st) century. Trop Med Health 39(Suppl):3–11. 37. Senior K (2007) FDA issue linezolid warning. Lancet Infect Dis 7:310.
26. Weaver SC (2013) Urbanization and geographic expansion of zoonotic arboviral 38. Laxminarayan R (2014) Antibiotic effectiveness: Balancing conservation against in-
diseases: Mechanisms and potential strategies for prevention. Trends Microbiol 21: novation. Science 345:1299–1301.
360–363. 39. Laxminarayan RP, Malani AP, Howard DP, Smith DLP (2007) Extending the Cure: Policy
27. Neiderud C-J (2015) How urbanization affects the epidemiology of emerging in- Responses to the Growing Threat of Antibiotic Resistance (Resources for the Future,
fectious diseases. Infect Ecol Epidemiol 5:27060. Washington, DC).
28. Brugha R, Grigg J (2014) Urban air pollution and respiratory infections. Paediatr 40. Nandi A, Megiddo I, Ashok A, Verma A, Laxminarayan R (2017) Reduced burden of
Respir Rev 15:194–199. childhood diarrheal diseases through increased access to water and sanitation in In-
29. Blommaert A, et al. (2014) Determinants of between-country differences in ambu- dia: A modeling analysis. Soc Sci Med 180:181–192.
latory antibiotic use and antibiotic resistance in Europe: A longitudinal observational 41. Lawes T, et al. (2015) Effects of national antibiotic stewardship and infection control
study. J Antimicrob Chemother 69:535–547. strategies on hospital-associated and community-associated meticillin-resistant
30. Paterson DL, Bonomo RA (2005) Extended-spectrum β-lactamases: A clinical update. Staphylococcus aureus infections across a region of Scotland: A non-linear time-
Clin Microbiol Rev 18:657–686. series study. Lancet Infect Dis 15:1438–1449.
31. Kotwani A, Holloway K (2014) Antibiotic prescribing practice for acute, uncompli- 42. Do NTT, et al. (2016) Point-of-care C-reactive protein testing to reduce inappropriate
cated respiratory tract infections in primary care settings in New Delhi, India. Trop use of antibiotics for non-severe acute respiratory infections in Vietnamese primary
Med Int Health 19:761–768. health care: A randomised controlled trial. Lancet Glob Health 4:e633–e641.
32. Gandra S, et al. (2016) Trends in antibiotic resistance among major bacterial patho- 43. Lee GC, et al. (2014) Outpatient antibiotic prescribing in the United States: 2000 to
gens isolated from blood cultures tested at a large private laboratory network in 2010. BMC Med 12:96.
India, 2008-2014. Int J Infect Dis 50:75–82. 44. Poehling KA, et al. (2004) Population-based impact of pneumococcal conjugate vac-
33. Capoor MR, Nair D (2010) Quinolone and cephalosporin resistance in enteric fever. cine in young children. Pediatrics 114:755–761.
J Glob Infect Dis 2:258–262. 45. Bruyndonckx R, et al. (2014) Measuring trends of outpatient antibiotic use in Europe:
34. Johnson AP, Woodford N (2013) Global spread of antibiotic resistance: The example Jointly modelling longitudinal data in defined daily doses and packages. J Antimicrob
of New Delhi metallo-β-lactamase (NDM)-mediated carbapenem resistance. J Med Chemother 69:1981–1986.
Microbiol 62:499–513. 46. Laxminarayan R, et al. (2016) UN high-level meeting on antimicrobials: What do we
35. Liu Y-Y, et al. (2016) Emergence of plasmid-mediated colistin resistance mechanism need? Lancet 388:218–220.
MCR-1 in animals and human beings in China: A microbiological and molecular bi- 47. Lenton TM, et al. (2008) Tipping elements in the Earth’s climate system. Proc Natl
ological study. Lancet Infect Dis 16:161–168. Acad Sci USA 105:1786–1793.

E3470 | www.pnas.org/cgi/doi/10.1073/pnas.1717295115 Klein et al.

You might also like