The prevalence of Type 2 Diabetes among older people In Africa: A Systematic Review

Full text

www.thelancet.com/diabetes-endocrinology Published online November 5, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00363-0
1
Review
The prevalence of type 2 diabetes among older peopl e in
Africa: a systematic review
Mahmoud Werfalli, Mark E Engel, Alfred Musekiwa, Andre P Kengne, Naomi S Levitt
Little information is available on the prevalence of diabetes in people aged 55 years or older living on the African
continent. We did a systematic review of the prevalence of type 2 diabetes in studies reported from Jan 1, 2000, to
June 30, 2015, to provide accurate data for monitoring future trends. We did a comprehensive literature search using
an African search fi lter and extracted and synthesised data from full papers. Among 1473 identifi ed citations,
41 studies providing 49 separate data contributions involving 16 086 individuals met the inclusion criteria. The
overall prevalence of diabetes was 13·7% (95% CI 11·3–16·3) and was higher in studies based on the oral glucose
tolerance test (23·9%, 17·7–30·7, 12 contributions with 3415 participants) than fasting blood glucose criteria (10·9%,
8·9–13·0, 37 contributions with 12 671 participants; p<0·001). Prevalence was also higher in non-STEPS than in
STEPS studies (17·1%, 95% CI 13·6−20·9) vs 9·6%, 6·6–13·0, p=0·003) and in urban than in rural settings (19·7%,
15·0–24·9 vs 7·9%, 4·6–12·0, p=0·0002), but did not diff er signifi cantly across age groups, sex, sample size, year of
publication, region, or population coverage. These data highlight the need to reduce diabetes risk factors and
implement adequate management strategies. In addition, they suggest that uniform diagnostic methods should be
used across African countries and elsewhere to enable assessment of trends in diabetes prevalence and the success
of diabetes prevention strategies. A collaborative initiative is required between key international and national
diabetes and geriatric organisations to improve diabetes care for the older population in Africa and worldwide.
Introduction
The International Diabetes Federation (IDF) estimated
prevalence for type 2 diabetes in 2013, showed that the
number of people aff ected worldwide has doubled over
the past 20 years. Most (80%) live in low-income and
middle-income countries.1 In Africa, where all countries
fall into these economic categories, diabetes already
con tributes substantially to morbidity and mortality,
and the age-specifi c mortality rate is the highest in the
world.2–7 The rise in the number of individuals with
type 2 diabetes in Africa, which is similar to that which
has occurred in low-income and middle-income
countries elsewhere in the world, has been attributed to
ageing of the population and rapid change in environ-
mental factors,2 such as urbanisation, increasingly
sedentary lifestyles, and unhealthy eating patterns.
Although behaviour patterns and obesity can potentially
be modifi ed, ageing, which is one of the main drivers of
diabetes, cannot.3 In 2013, most individuals with
diabetes in Africa were younger than 60 years, and the
highest proportion (43·2%) comprised people aged
40–59 years. The small proportion of people aged
60–79 years in the region probably accounted for the
fact that only 18·8% of people with diabetes fell into
this age group.1
Africa is often referred to as the youngest continent,
which might contribute to the low prioritisation of
ageing issues in national policies.8 Yet the annual
growth rate of people older than 55 years in Africa was
estimated to be 3·1% greater than the global average
between 2007 and 2015, and is predicted to be 3·3%
greater between 2015 and 2050. Thus, there are around
64·5 million people in Africa aged 55 years or older in
2015, and there are likely to be more than 103 million
and 205 million in 2030 and 2050, respectively.7
Consequently, the diabetes prevalence in Africa is
expected to be highest in the oldest individuals by
2035.1 We did a systematic review to investigate the
prevalence of type 2 diabetes in Africa in individuals
older than 55 years, with the aim of providing accurate
data for monitoring of future trends.
Lancet Diabetes-Endocrinol 2015
Published Online
November 5, 2015
http://dx.doi.org/10.1016/
S2213-8587(15)00363-0
Chronic Disease Initiative for
Africa (CDIA) (M Werfalli MPH,
Prof N S Levitt MD), Division of
Diabetic Medicine and
Endocrinology (M Werfalli,
Prof N S Levitt), Department of
Medicine (M E Engel PhD), and
Non-Communicable Diseases
Research Unit (A P Kengne PhD),
University of Cape Town,
Cape Town, South Africa; and
Faculty of Medicine and Health
Sciences, Centre for Evidence
Based Health Care, Stellenbosch
University, Cape Town, South
Africa (A Musekiwa MSc).
Correspondence to:
Prof Naomi S Levitt,
J 47 Room 86, Old Groote Schuur
Hospital Building, Department of
Medicine, Faculty of Health
Panel: Basic concept, framework, and key goals of the WHO STEPwise approach
to surveillance11,12
The STEPwise approach to Surveillance (STEPS) is WHO’s recommended tool for
surveillance of chronic non-communicable diseases and their risk factors. It aims to
provide an entry point for low-income and middle-income countries to start chronic
disease surveillance. It is also designed to help countries build and strengthen
surveillance capacity. In all instances, STEPS targets adults aged 25–64 years and uses a
representative sample of the study population, which allows generalisation of the
results to the whole. STEPS has some fl exibility, which enables each country to expand
on the core variables and risk factors and to incorporate optional modules relevant to
local or regional interests. The STEPS instrument has three diff erent levels (or steps), all
of which have core, expanded, and optional modules of risk factor assessment that are
used dependent on what can be accomplished in a given country. Step 1 uses a
standardised questionnaire to gather demographic and behavioural information
(tobacco and alcohol use, nutrition, and physical activity) in a household setting.
Several extended options can be obtained, such as for demographics, ethnic origin,
employment status, and household income; for behaviour, binge drinking, smokeless
tobacco, and ex-smokers; and for diet, oil and fat consumption. Optional factors include
mental and oral health and objective measures of physical activity. Step 2 contains
simple physical measurements to assess anthropometry and blood pressure, which can
also be obtained in the household setting (heart rate, hip circumference, and [optional]
skin-fold thickness and physical fi tness). Step 3 consists of biochemical measurements:
“fasting blood sugar” is the core measurement; the extended tests include
measurement of total cholesterol, HDL-cholesterol, and fasting triglycerides; and
optional tests are the oral glucose tolerance test, urine examination, and measurement
of salivary cotinine to assess tobacco intake.

2
www.thelancet.com/diabetes-endocrinology Published online November 5, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00363-0
Review
See Online for appendix
Methods
Literature search
We aimed to identify prevalence studies in Africa
published from Jan 1, 2000, to June 30, 2015, with use of
the African search fi lter developed by Eisinga and
colleagues.9 The fi lter comprises African country names
and truncated terms, such as “north* Africa”, to ensure
that records indexed with regional rather than country-
specifi c terms are also retrieved. We combined database
medical subject headings (MeSH in PubMed⁄MEDLINE,
CINAHL, and Google Scholar) with a range of search
terms (appendix pp 1–3).
African country names were included in English and
languages relevant to each country for example, “Ivory
Coast” and “Côte d’Ivoire”. Where country names have
changed over time, old and new names were included,
such as “Zaire” and “Democratic Republic of Congo.10 We
searched for publications by key authors with citation
searches on the websites of WHO and the IDF, the latter
for the STEPwise approach to surveillance (STEPS)
surveys studies in Africa. The panel shows the basic
concept, framework, and key goals of STEPS surveys.11,12
We also searched the ISI Web of Knowledge. No language
restrictions were applied. An expert librarian designed
the search strategy framework and applied the
appropriate bibliographic software.
To be included in this systematic review, primary studies
had to have used cross-sectional or population-based
designs to assess the prevalence of type 2 diabetes among
older adults (described in the report as older adults, or at
least 70% of the study population aged 55 years or older)
who were resident in countries in sub-Saharan or north
Africa, irrespective of ethnic, socioeconomic, and
educational back grounds. The diagnosis of diabetes had
to have been made by a physician or defi ned based on
measured fasting plasma glucose (FPG), oral glucose
tolerance test (OGTT), or self-report, according to WHO
criteria.13 Studies had to report numeric data to enable the
calculation of prevalence. Those that used denominator
data from other studies to calculate prevalence were
excluded. Full-text articles identifi ed as meeting the
inclusion criteria on the basis of their titles and abstracts
were obtained for further assessment by two reviewers
(MW and AM), and those that did not meet the selection
criteria were excluded (fi gure 1). Disagreements were
resolved through discussions between reviewers until
consensus was reached.
Assessment of risk bias in included studies
We assessed the methodological quality of included
studies in terms of internal validity, external validity,
response rate, and generalisability of study results. We
used the ten-item rating system developed by Hoy and
colleagues14 and modifi ed by Werfalli and colleagues15
(appendix, pp 4–5) to assess sampling, the sampling
frame and size, outcome measurement, outcome
assessment, response rate, and statistical reporting.14
Each item was assigned a score of 1 (yes) or 0 (no), and
scores were summed across items to generate an overall
quality score that ranged from 0 to 10. Each study was
rated as having a low, moderate, or high risk of bias
dependent on the number of questions answered as “yes
(low risk)”: studies at low risk of bias had scores higher
than 8, moderate a score of 6–8, and high a score of 5 or
lower.15 Risk of selection and attrition biases were
assessed according to the Cochrane guidelines, in
Review Manager, version 5.2. Two reviewers (MW and
MEE) independently assessed study quality, with
disagreements being resolved by consensus.
Data extraction
Two reviewers (MW and AM) independently selected
studies and extracted relevant information. Disagree-
ments were resolved by consensus or consultation with a
third reviewer (NSL). Study characteristics documented
included country name, year of publication, national
population, region (rural or urban), age range, sex, study
design, criteria for sample selection, sample size,
ascertainment of diabetes status, and diagnostic criteria.
Data synthesis and analysis
Three reviewers (MW, APK, and NSL) did the statistical
analysis and data synthesis. Unadjusted prevalence
estimates and SEs were recalculated for type 2 diabetes
in people aged 55 years or older (number of cases/sample
size) based on the information on crude numerators and
81 records identified through
searches of other sources
43 Google free search
27 WHO database
11 contact with authors
640 records screened after duplicates removed
165 full-text articles assessed for eligibility
1473 records identified
through database
searches
475 records excluded
41 articles included in synthesis
124 full-text articles excluded
28 no primary research
8 no study design
30 no prevalence data
16 no population
12 not on type 2 diabetes
23 not published within review
period (2000–13)
2 not retrieved
5 studies with poor quality
rating
Figure 1: Selection of articles for inclusion in the systematic review
Science, University of Cape Town,
Observatory, Cape Town, 7935,
South Africa
naomi.levitt@uct.ac.za

www.thelancet.com/diabetes-endocrinology Published online November 5, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00363-0
3
Review
denominators provided in the individual studies. To keep
the eff ect of studies with extremely small or extremely
large prevalence estimates on the overall estimate to a
minimum, we stabilised the variance of the study-specifi c
prevalence with the Freeman-Tukey single arcsine
transformation16 before pooling the data with the
random-eff ects meta-analysis model (appendix 9–12).17
Heterogeneity between studies was assessed with
Cochran’s Q statistic and the I² statistic,18 which estimates
the percentage of total variation across studies due to
true between-study diff erences rather than chance.
I² values greater than 60–70% generally indicate the
presence of substantial heterogeneity. We explored
sources of heterogeneity by comparing diabetes
prevalence between subgroups defi ned by several study-
level characteristics. We assessed the presence of
publication bias using the Egger test of bias.19 We did all
analyses with the meta package in R (version 3.0.3).
Findings
Search results
The searches identifi ed 1554 citations. After screening of
titles and abstracts and removal of duplicates, 640 studies
were selected for further scrutiny, of which 165 were
selected for full-text review. Of these, 41 met the inclusion
criteria and were included in this systematic review
Sampling Age
range
(years)
Prevalence (95%CI) Criteria*
Size Strategy Response
rate (%)
Urban Rural Urban and rural
Male Female Total Male Female Total Male Female Total
Algeria20 1457 Random
sampling
90·0% 55–64,
≥60
·· ·· ·· ·· ·· ·· ·· ·· 11·2%
(7·5–15·8),
12·0%
WHO
1998
Algeria21 7656 Random
sampling
NR ≥60 17·4% 5·7% 10·3% 10·2% 7·2% 8·3% 14·2%
(11·9–16·8)
6·4%
(5·2–7·9)
10·5%
(8·6–12·2)
WHO
1985
Angola22 709 Random
sampling
71·0% 60–69 ·· ·· 14·6%
(8·9–22·1)
·· ·· ·· ·· ·· ·· WHO
1998
Burkina Faso23 467 Random
sampling
100% ≥60 ·· ·· 31·8%
(13·9–54·9)
·· ·· ·· ·· ·· ·· WHO
1999
Cameroon24 1279 Multistage
sampling
>95·0% ≥55 15·2%
(9·0–23·6)
10·8%
(5·3 18·9)
13·1%
(8·8–18·6)
·· ·· ·· ·· ·· ·· WHO
1998
Cameroon25 1702 1702 100% ≥55 ·· ·· ·· ·· ·· ·· ·· ·· 24·7%
(15·3–36·1)
WHO
1999
Canary
Islands26
1193 Random
sampling
86·3% ≥55 ·· ·· ·· ·· ·· ·· 39·6%
(29·7–50·1)
26·3%
(17·9–36·1)
32·8%
(26·3–
39·9)
WHO
1999
Democratic
Republic of
Congo27
711 Multistage
sampling
98·6% ≥60 ·· ·· ·· ·· ·· ·· ·· ·· 8·3%
(6·7–10·2)
WHO
1999
Ethiopia28 5500 Random
sampling
81·3% ≥55 ·· ·· ·· ·· ·· ·· ·· ·· 6·7%
(4·4–9·7)
WHO
1999
Guinea29 2000 Multistage
sampling
77·0% ≥55,
55–64
·· ·· ·· ·· ·· ·· ·· ·· 9·7%
(7·3–12·6,
7·8%
(5·0–11·6)
WHO
1999
Kenya30 2061 Cluster
sampling
99·0% ≥55 7·6%
(2·5–16·8)
16·7%
(7·0–31·4)
11·1%
(5·9–18·6)
·· ·· ·· ·· ·· ·· WHO
1999
Libya31 1002 Multistage
sampling
86·6% ≥60 ·· ·· 34·9% ·· ·· 28·3% 30·2%
(18·3–44·3)
17·8%
(8·0–32·1)
24·5%
(16·4–34·2)
WHO
1998
Mayotte32 1268 Random
sampling
70·0% ≥60 25·9%
(17·0–36·5)
·· ·· ·· ·· ·· ·· ·· ·· WHO
1999
Mozambique33 2343 Random
sampling
NR 55–64 ·· ·· 8·7%
(5·1–13·8)
·· ·· 5·7%
(2·3 11·5)
·· ·· ·· WHO
1999
Nigeria34 2780 Random
sampling
97·3% 55–64 - - - - - - 14·3%
(10·9–18·2)
34·8%
(28·3–41·8)
21·5%
(18·2–25·0)
WHO
1999
Réunion35 3600 Random
sampling
80·6% ≥60 ·· ·· ·· ·· ·· ·· 34·6%
(28·2–41·4)
40·0%
(34·3–45·9)
37·7%
(33·4–42·1)
WHO
1999
Seychelles36 1255 Random
sampling
80·2% 55–64 22·0%
(15·7–29·5)
26·5%
(20·2 −33·6)
24·5%
(19·9–29·5)
·· ·· ·· ·· ·· ·· WHO
1999
(Table 1 continues on next page)

4
www.thelancet.com/diabetes-endocrinology Published online November 5, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00363-0
Review
(fi gure 1).20–60 The reasons for exclusion of 124 studies are
detailed in the appendix (pp 13–21). Of the 41 studies
included, two21,35 provided prevalence estimates separately
for urban and rural participants, another four
28,43,47,60
provided estimates separately for two non-overlapping age
groups, and one51 provided prevalence estimates separately
for two non-overlapping age groups for urban participants
and two non-overlapping age groups for rural participants.
These datasets were counted separately, leading to 49 data
contributions being assessed in the main analyses.
Generally, the overall risk of bias was low in studies (n=31),
but was moderate in 11 and high in fi ve (appendix pp 6–8).
25 of the studies included in this systematic review
were published in peer-reviewed journals (for the
purpose of this Review we referred to them as non-
STEPS studies),20–43 16 were STEPS studies,44–60 and one
was a thesis (tables 1, 2).23 For STEPS studies published
in peer-reviewed journals, we used the latest published
version that included the complete dataset. Of the
54 countries of the African continent, 30 (57%),
accounting for 74·3% of the total population
(7535 million of 1136 billion),60 were represented in this
systematic review: three studies from Algeria, two from
Benin, two from Cameroon, three from Democratic
Republic of the Congo, two from Ethiopia, two from
Libya, one from Mozambique, three from South Africa,
two from Tunisia, and one from each of Angola,
Botswana, Burkina Faso, Canary Islands, Egypt, Gabon,
Guinea, Kenya, Mayotte, Malawi, Mauritania, Mauritius,
Niger, Nigeria, Réunion, Seychelles, Sudan, Tanzania,
Togo, Uganda, and Zimbabwe. The analytical sample
size ranged from 467 to 10 000. All STEPS studies used
multistage cluster sampling techniques, and the
response rates were 54·6–100%. The non-STEPS studies
used random sampling or multistage cluster sampling
techniques and had response rates of 70–99%.
12 studies used OGTT as the method of diabetes
diagnosis, and 37 studies used FPG. 48 studies used WHO
1998/1999 diagnostic criteria for type 2 diabetes and one
study used WHO 1985 criteria. 30 (73%) studies were done
in urban and rural areas, 17 (41%) were done only in urban
areas, and eight (19%) only in rural areas. The defi nition of
older people was 55–64 years in 21 (43%) studies, 55 years
or older in six (12%) studies, 60 years or older in ten (20%)
studies, 65 years or older in four (8%) studies. Some
studies used more than one defi nition; for these studies
and when the age bands were not mutually exclusive, the
age band that included the greatest age range was used in
the analysis (ie, ≥55 years when provided along with
55–64 years, or ≥60 years when provided along with
≥65 years old). Thus, our main analysis included 41 studies
rather than the 49 data contributions.
In the assessment of methodological quality, fi ve
contributions were deemed to be of poor methodological
quality and were excluded from the analysis. Of the
remaining 46 contributions included, 11 were deemed
Sampling Age
range
(years)
Prevalence (95%CI) Criteria*
Size Strategy Response
rate (%)
Urban Rural Urban and rural
Male Female Total Male Female Total Male Female Total
(Continued from previous page)
South Africa37 642 Random
sampling
87·6% ≥60 51·9%
(33·1–69·8)
41·3%
(31·9–51·1)
43·6%
(35·2–52·2)
·· ·· ·· ·· ·· ·· WHO
1999
South Africa38 1099 Random
sampling
86% 55–64,
65–74
17·2%
(8·6–29·4),
25·2%
(10·7–44·9)
26·1%
(17·5–36·3),
34·7%
(21·7–49·6)
22·7%
(16·2–30·2),
38·2%
(21·1–42·7)
·· ·· ·· ·· ·· ·· WHO
1999
South Africa39 1300 Cluster
sampling
78·9% ≥55 ·· ·· ·· 7·4%
(2·8–15·4)
6·4%
(3·9–9·7)
6·6%
(4·3–9·6)
·· ·· ·· WHO
1998
Tunisia40 3729 multistage
stratifi ed
85·1% ≥60 19·8% 27·4% 24·3% 10·6% 12·4% 11·5% 15·8%
(12·2–19·9)
21·9%
(18·2–25·9)
19·2%
(16·6–22·0)
WHO
1999
Tunisia41 598 multistage
cluster
sampling
96·3% ≥65 32·0% 28·0% 29·3% 15·2% 16·3% 15·9% 25·0%
(10·7–44·9)
26·5%
(22·2–31·2)
27·4%
(23·9–31·2)
WHO
1999
Tanzania42 640 multistage
random
sampling
100% 55–64,
≥60
25·0%
(12·1–42·2),
11·4%
(3·2–26·7)
30·6%
(19·6–43·7),
11·1%
(3·1–26·1
28·6%
(19·9–38·6),
11·3%
(5·0–21·0)
·· ·· ·· ·· ·· ·· WHO
1999
Uganda43 1497 Random
sampling
90·4% 55–64 ·· ·· ·· ·· ·· ·· ·· ·· 9·2%
(5·1–15·0)
WHO
1999
*1985 criteria, fasting plasma glucose ≥7·8 mmol/L, 2 h post-challenge glucose concentration ≥11·1 mmol/L, or both; 1998 criteria, fasting plasma glucose ≥7·0 mmol/L, 2 h plasma glucose ≥11·1 mmol/L, or
both; 1999 criteria fasting plasma glucose ≥7·0 mmol/L, 2 h post-challenge glucose concentration ≥11·1 mmol/L, or both.
Table 1: Prevalence of type 2 diabetes in older populations assessed in studies not using the WHO STEPwise approach to surveillance

www.thelancet.com/diabetes-endocrinology Published online November 5, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00363-0
5
Review
to be of moderate quality and 30 of high methodological
quality (appendix p 22).
Prevalence by age, sex, region, and type of residency
The overall prevalence of type 2 diabetes across the
49 contributions (n=16 086 participants) was 13·7%
(95% CI 11·3–16·3; appendix p 9, p 23) and did not diff er
signifi cantly by age group (p=0·187) or sex (p=0·611;
appendix p 9, pp 24–25). When assessed by sex, the
prevalence of diabetes was 14·3% (95% CI 11·9–17·0)
overall (n=12 373), 13·6% (10·7–16·9) for men (n=5345),
and 15·0% (11·2–19·2) in women (n=7028, appendix p 9).
Prevalence of type 2 diabetes did not diff er signifi cantly
(p=0·181) between sub-Saharan and north Africa (13·8%,
95% CI 13·2–14·3, 39 studies vs 16·6%, 11·7–22·3,
ten studies; appendix p 10). By contrast, the overall
prevalence of diabetes was signifi cantly higher in urban
than rural populations (19·7%, 15·0–24·9, 17 studies vs
7·9%, 4·6–12·0, eight studies; p=0·0002; appendix
pp 9–10).
Prevalence by method of data collection (STEPS vs
non-STEPS)
The prevalence of type 2 diabetes was signifi cantly lower
in STEPS studies (p=0·003) than in non-STEPS studies
(9·6%, 95% CI 6·6–13·0, 20 contributions vs 17·1%,
13·6–20·9, 29 contributions; fi gures 2, 3; appendix p 10).
There was a signifi cant diff erence across age groups in
the non-STEPS studies (p<0·0001), but not across the
STEPS surveys (p=0·550, appendix p 10). No diff erence
in prevalence was seen between studies larger and
smaller than the median sample size (p=0·592), although
there was a signifi cant age diff erential among studies
above the median sample size (p=0·008, appendix p 11).
Heterogeneity remained signifi cant (p<0·0001) within
the STEPS and non-STEPS age groups (fi gures 2, 3).
Sampling Age range
(years)
Prevalence in urban and rural areas (95%CI) Criteria*
Size Strategy Rate of response (%) Male Female Total
Algeria44 4097 Multistage
cluster sampling
3820 (93·2%) 55–64 6·4% (3·7–10·2) 5·1% (2·9–8·0) 5·7% (3·9 −7·8) WHO 1999
Benin45 2568 Multistage
cluster sampling
2568 (100%) 55–64 11·4% (6·8–17·6) 8·3% (5·6–11·7) 15·9% (11·8- 2 0·7) WHO 1999
Benin46 4597 Multistage
cluster sampling
4597 (100%) 55–64 ·· ·· 14·4% (10·5–19·2)
Botswana47 4003 Multistage
cluster sampling
2820 (70·4%) 55–64 4·4% (1·6–9·4) 8·3% (5·6–11·7) 12·3% (8·7–16·7) WHO 1999
Democratic
Republic of Congo48
1943 Multistage
cluster sampling
1123 (57·8%) 55–64, ≥65 5·2% (1·4–12·8) 2·1 6·2% (2·1 −14·0),
3·7% (0·5–12·7)
5·7% (1·4–12·8),
3·0 (0·6–8·4)
WHO 1999
Democratic
Republic of Congo49
9970 Multistage
sampling
90·3% 55−64,
65–98
37·7%, 36·1% 44·5%, 30·6% 41·4%, 33·4% WHO 1999
Egypt50 10 000 Multistage
cluster sampling
9730 (97·3%) 55–64 17·2% (13·9–20·9) 28·7% (25·1–41·9) 22·6% (19·9–25·5) WHO 1999
Ethiopia51 2200 Multistage
cluster sampling
2141 (97·3%) 55–64 Urban 5·1% (0·6–17·3),
rural 1·0% (0·0–5·3)
Urban 30·6% (19·6–43·7),
rural 12·2% (7·3–18·9)
Urban 10·7% (6·6–16·2),
rural 1·1% (0·1–3·8)
WHO 1999
Gabon52 2800 Multistage
cluster sampling
2708 (96·7%) 55–64 ·· ·· 22·2% WHO 1999
Libya53 3625 Multistage
cluster sampling
2646 (73·0%) 55–64 25·6% (18·5–33·2) 33·4% (25·1–41·9) 29·0% (23·7–34·7) WHO 1999
Malawi54 5206 Multistage
cluster sampling
2842 (54·6%) 55–64 7·5% (3·6–13·3) 6·1% (3·8–9·6) 6·6% (4·5–9·4) WHO 1999
Mauritania55 2600 Multistage
cluster sampling
2500 (96·1%) 55–64 3·7% (1·4–7·8) 3·2% (1·0–7·3) 3·4% (1·7–6·1) WHO 1999
Mauritius56 4500 Multistage
cluster sampling
4200 (93·3%) 60–69, ≥70 36·2%, 36·6% 38·7%, 42·9% 37·6%, 39·7% WHO 1999
Niger57 3060 Multistage
cluster sampling
2778 (90·8%) 55–64 4·4% (2·2–7·8) 0·8% (0·0–4·4) 3·2% (1·7–5·5) WHO 1999
Sudan58 1573 Multistage
cluster sampling
921 (58·5%) 55–64 19·9% (13·7–27·3) 16·2% (10·9–22·6) 17·9% (13·8–22·6) WHO 1999
Togo59 4800 Multistage
cluster sampling
4370 (91·0%) 55–64 11·1% (6·7–17·0) 3·8% (1·4–8·0) 7·5% (4·8–10·9) WHO 1999
Zimbabwe60 3000 Multistage
cluster sampling
3081 (100%) 55–64 15·7% (8·1–26·4) 14·0% (9·5–19·6) 14·4% (10·5–19·2) WHO 1999
*1999 criteria FPG ≥7·0 mmol/L, 2 h post-challenge glucose concentration ≥11·1 mmol/L, or both.
Table 2: Prevalence of type 2 diabetes in older populations assessed in studies using the WHO STEPwise approach to surveillance

6
www.thelancet.com/diabetes-endocrinology Published online November 5, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00363-0
Review
Prevalence by diagnostic method (FPG vs OGTT)
When diagnosed with FPG the prevalence of type 2
diabetes was 10·9% (95% CI 8·9–13·0; 37 contributions)
and when diagnosed with OGTT was 23·9% (17·7–30·7;
12 contributions, p value for diff erence <0·0001;
fi gures 4, 5, appendix p 12). When studies that used the
same diagnostic method were compared, prevalence did
not diff er between age groups (p=0·410 for FPG and
p=0·117 for OGTT). When the prevalence of diabetes was
compared across the diff erent regions (eastern, western,
central, northern, and southern) after stratifi cation by
diagnostic method, prevalence did not diff er (p=0·139 for
OGTT and p=0·070 for FPG), in line with the main
analysis.
Investigation of the sources of heterogeneity
Substantial heterogeneity was seen across the con-
tributing studies overall, within subgroups for sex, age,
residence, and median sample size, across diagnostic
methods, and across data collection methods (all
p<0·0001 for heterogeneity). We found no evidence of
publication bias (Egger score) overall and in most age
groups (appendix pp 9–12).
Discussion
Our systematic review showed that type 2 diabetes is not
rare in people aged 55 years and older across Africa. The
estimated prevalence of diabetes was two times higher in
studies that used OGTT than in those that used FPG to
diagnose diabetes, and was nearly twice as high in urban
than in rural settings. The two main methods used to
survey diabetes prevalence were also associated with
diff erent results: non-STEPs studies showed a
prevalence 1·6 times higher than STEPS studies.
Prevalence did not, however, diff er by sex, sample
size, year of publication, region, or population coverage.
When stratifi ed by age group, prevalence of diabetes
varied within some major subgroups, but not all.
The prevalence and number of people aff ected by
diabetes varies worldwide. The latest IDF Atlas61 reports
that the highest number of people with diabetes between
ages 20 and 79 years is in the western Pacifi c Region,
where 138 million people are aff ected, although
prevalence is 8·6%, which is close to the worldwide
estimate. The North American and Caribbean region has
the highest prevalence at 11%, with an estimated
37 million people aff ected, followed by the Middle East
and north African region, where prevalence is 9·2% and
35 million people are aff ected.61 In Europe the prevalence
and number aff ected are estimated to be 8·5% and
56 million people, in South and Central America the
values are 8·1% and 25 million people, and in Africa
4·9% and 22 million people, which are the lowest
values.61 According to the Atlas, the three African
countries with the highest prevalence are Gabon (10·7%),
Réunion (15·4%), and Seychelles (12·1%).62 The 2014
Atlas did not report the data stratifi ed by age group or
diagnostic method.61
Age group and country
Age 55–64 years
Algeria44
Benin45
Botswana47
Democratic Republic of Congo48
Egypt50
Libya53
Malawi54
Mauritania55
Niger57
Sudan58
Togo59
Zimbabwe60
Benin46
Ethiopia51
Ethiopia51
Random effects model
Heterogeneity: I2=95·6%, tau2=0·016, p<0·0001
Age ≥65 years
Democratic Republic of Congo48
Zimbabwe60
Benin46
Ethiopia51
Ethiopia51
Random effects model
Heterogeneity: I2=90·5%, tau2=0·0114, p<0·0001
Overall random effects model
Heterogeneity: I2=94·8%, tau2=0·0145, p<0·0001
Sample size
570
277
277
157
884
276
437
320
374
313
321
270
384
178
187
5225
102
301
415
165
143
1126
6351
Prevalent diabetes
32
44
34
9
200
80
29
11
12
56
24
39
49
19
2
640
3
39
66
16
3
127
767
Prevalence (95% CI)
5·6 (3·9–7·8)
15·9 (11·8–20·7)
12·3 (8·7–16·7)
5·7 (2·7–10·6)
22·6 (19·9–25·5)
29·0 (23·7–34·7)
6·6 (4·5–9·4)
3·4 (1·7–6·1)
3·2 (1·7–5·5)
17·9 (13·8–22·6)
7·5 (4·8–10·9)
14·4 (10·5–19·2)
12·8 (9·6–16·5)
10·7 (6·6–16·2)
1·1 (0·1–3·8)
10·1 (6·5–14·4)
2·9 (0·6–8·4)
13·0 (9·4–17·3)
15·9 (12·5–19·8)
9·7 (5·6–15·3)
2·1 (0·4–6·0)
8·0 (3·5–14·2)
9·6 (6·6–13·0)
Weight (%)
Prevalence (95% CI)
020304010
5·2%
5·0%
5·0%
4·8%
5·2%
5·0%
5·1%
5·1%
5·1%
5·1%
5·1%
5·0%
5·1%
4·9%
4·9%
75·7%
4·6%
5·1%
5·1%
4·8%
4·8%
24·3%
100%
Figure 2: Meta-analysis results for prevalence of type 2 diabetes in studies using the WHO STEPwise approach to surveillance

www.thelancet.com/diabetes-endocrinology Published online November 5, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00363-0
7
Review
The overall estimated prevalence in this review of
African studies fell within that for people aged
60–79 years in 13 European countries in the DECODE
study63 and those aged 60–69 years in the Chinese and
Japanese cohorts in the DECODA study.64 The prevalence
was, however, lower than that in people older than
65 years from the USA NHANES survey in 2006,65 and in
people aged 60–69 years in the Indian cohorts of the
DECODA study (31⋅0% in the former, and 41·4% and
34·6% in men and women in the latter).
The two times higher overall estimated prevalence of
diabetes we found in African studies when diagnosis was
made with OGTT than with FPG criteria is in line with
fi ndings in previous reports. Many studies have reported
that FPG and 2 h glucose concentrations from OGTT are
not consistent in detecting diabetes in the same people.
For example, in the DECODE study,63 of the 1517 people
with newly diagnosed diabetes, 40% met only the FPG
criterion, 31% met only the 2 h plasma glucose criterion,
and 28% met both. Therefore, use of only FPG would not
have identifi ed around 30% of people with diabetes.
Those investigators, therefore, concluded that diabetes
would be underestimated in the older population in
Europe if diagnoses were based on FPG. In the Rancho
Bernardo study in the USA,66 new diagnoses of diabetes
were made more frequently with OGTT than with FPG;
70% of women and 48% of men aged 50–89 years were
diagnosed by 2 h plasma glucose alone.66 The researchers
noted that diabetes defi ned by OGTT alone is common
in older adults, since the results with OGTT increase
with increasing age, which more than doubles the risk of
fatal cardiovascular disease and of heart disease in older
women. The use of FPG alone for screening or diagnosis
of diabetes might not identify older adults in these risk
Age group and country
Age 55–64 years
Algeria20
Guinea29
Mozambique33
Mozambique33
Nigeria34
Seychelles36
South Africa38
Uganda43
Tanzania42
Random effects model
Heterogeneity: I2=91·4%, tau2=0·0116, p<0·0001
Age ≥55 years
Cameroon24
Canary Islands26
Ethiopia28
Guinea29
Kenya30
South Africa39
Cameroon25
Random effects model
Heterogeneity: I2=93·1%, tau2=0·0136, p<0·0001
Age ≥60 years
Algeria21
Algeria21
Angola22
Democratic Republic of Congo49
Libya31
Mayotte32
Réunion35
South Africa34
Tunisia40
Burkina Faso23
Tanzania42
Random effects model
Heterogeneity: I2=97%, tau2=0·0195, p<0·0001
Age ≥65 years
South Africa38
Tunisia41
Random effects model
Heterogeneity: I2=0%, tau2=0, p=0·4967
Overall random effects model
Heterogeneity: I2=95·4%, tau2=0·016, p<0·0001
Sample size
242
282
183
122
582
331
150
152
98
2142
198
195
375
525
108
380
73
1854
1158
1030
123
1000
98
85
499
140
838
22
71
5064
77
598
675
9735
Prevalent diabetes
27
22
16
7
125
81
34
14
28
354
26
64
25
51
12
25
18
221
119
85
18
83
24
22
188
61
161
7
8
776
24
164
188
1539
Prevalence (95% CI)
11·2 (7·5–15·8)
7·8 (5·0–11·6)
8·7 (5·1–13·8)
5·7 (2·3–11·5)
21·5 (18·2–25·0)
24·5 (19·9–29·5)
22.7(16·2–30·2)
9·2 (5·1–15·0)
28·6 (19·9–38·6)
14·6(9·8–20·3)
13·1 (8·8–18·6)
32·8 (26·3–39·9)
6·7 (4·4–9·7)
9·7 (7·3–12·6)
11·1 (5·9–18·6)
6·6 (4·3–9·6)
24·7 (15·3–36·1)
13·6 (8·1–20·4)
10·3 (8·6–12·2)
8·3 (6·6–10·1)
14·6 (8·9–22·1)
8·3 (6·7–10·2)
24·5 (16·4–34·2)
25·9 (17·0–36·5)
37·7 (33·4–42·1)
43·6 (35·2–52·2)
19·2 (16·6–22·0)
31·8 (13·9–54·9)
11·3 (5·0–21·0)
19·8 (13·4–27·1)
31·2 (21·1–42·7)
27·4 (23·9–31·2)
27·8 (24·5–31.3)
17·1 (13·6–20·9)
Weight (%)
Prevalence (95% CI)
020304010
3·6%
3·6%
3·5%
3·4%
3·7%
3·6%
3·4%
3·4%
3·3%
31·5%
3·5%
3·5%
3·6%
3·7%
3·3%
3·6%
3·1%
24·4%
3·7%
3·7%
3·4%
3·7%
3·3%
3·2%
3·7%
3·4%
3·7%
2·2%
3·1%
37·2%
3·2%
3·7%
6·9%
100%
Figure 3: Meta-analysis results for prevalence of type 2 diabetes in studies not using the WHO STEPwise approach to surveillance

8
www.thelancet.com/diabetes-endocrinology Published online November 5, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00363-0
Review
groups and, therefore, they should be reassessed if a
diagnosis was made with this method. These diff erences
have important implications for Africa, where STEPS
surveys were based on protocols that recommended the
use of FPG as a core component of biochemical
measurement, as such studies are becoming a growing
source of information about diabetes prevalence data.
Our data and those of previous studies suggest that if use
of FPG alone were followed in practice, half of the
diabetes burden in the people older than 55 years would
be missed.
Comparison of mortality in people with a diagnosis of
diabetes based on FPG or OGTT has consistently shown
worse clinical outcomes with the latter. For example, in
the DECODE study,63 the hazard ratios for death in
people with diabetes diagnosed by FPG were 1·6 (95% CI
1·4–1·8) for all-cause mortality, 1·6 (1·3–1·9) for
cardiovascular mortality, and 1·6 (1·4–1·9) for non-
cardiovascular mortality. For OGTT, the equivalent
hazard ratios were 2·0 (1·7–2·3), 1·9 (1·5–2·4), and
2·1 (1·7–2·5), respectively.63 On the basis of
fi ndings from three population-based longitudinal
Age group and country
Age 55–64 years
Algeria44
Benin45
Botswana47
Democratic Republic of Congo48
Egypt50
Guinea29
Libya53
Malawi54
Mauritania55
Mozambique33
Mozambique33
Niger57
Sudan58
Togo59
Uganda43
Zimbabwe60
Benin46
Tanzania42
Ethiopia51
Ethiopia51
Random effects model
Heterogeneity: I2=94·6%, tau2=0·0146, p<0·0001
Age ≥55 years
Cameroon24
Ethiopia28
Guinea29
Kenya30
Cameroon25
Random effects model
Heterogeneity: I2=79%, tau2=0·0041, p=0·0008
Age ≥60 years
Algeria21
Algeria21
Democratic Republic of Congo49
Mayotte32
Tunisia40
Burkina Faso23
Tanzania42
Random effects model
Heterogeneity: I2=92·7%, tau2=0·0059, p<0·0001
Age ≥65 years
Democratic Republic of Congo48
Zimbabwe60
Benin46
Ethiopia51
Ethiopia51
Random effects model
Heterogeneity: I2=90·5%, tau2=0·0114, p<0·0001
Overall random effects model
Heterogeneity: I2=92·7%, tau2=0·0095, p<0·0001
Sample size
570
277
277
157
884
282
276
437
320
183
122
374
313
321
152
270
384
98
178
187
6062
198
375
525
108
73
1279
1158
1030
1000
85
838
22
71
4204
102
301
415
165
143
1126
12671
Prevalent diabetes
32
44
34
9
200
22
80
29
11
16
7
12
56
24
14
39
49
28
19
2
727
26
25
51
12
18
132
119
85
83
22
161
7
8
485
3
39
66
16
3
127
1471
Prevalence (95% CI)
5·6 (3·9–7·8)
15·9 (11·8–20·7)
12·3 (8·7–16·7)
5·7 (2·7–10·6)
22·6 (19·9–25·5)
7·8 (5·0–11·6)
29·0 (23·7–34·7)
6·6 (4·5–9·4)
3·4 (1·7– 6·1)
8·7 (5·1–13·8)
5·7 (2·3–11·5)
3·2 (1·7– 5·5)
17·9 (13·8–22·6)
7·5 (4·8–10·9)
9·2 (5·1–15·0)
14·4 (10·5–19·2)
12·8 (9·6–16·5)
28·6 (19·9–38·6)
10·7 (6·6–16·2)
1·1 (0·1– 3·8)
10·3 (7·2–13·9)
13·1 (8·8–18·6)
6·7 (4·4– 9·7)
9·7 (7·3–12·6)
11·1 (5·9–18·6)
24·7 (15·3–36·1)
11·7 (7·8–16·2)
10·3 (8·6–12·2)
8·3 (6·6–10·1)
8·3 (6·7–10·2)
25·9 (17·0–36·5)
19·2 (16·6–22·0)
31·8 (13·9–54·9)
11·3 (5·0–21·0)
13·8 (9·7–18·5)
2·9 (0·6–8·4)
13·0 (9·4–17·3)
15·9 (12·5–19·8)
9·7 (5·6–15·3)
2·1 (0·4–6·0)
8·0
(3·5–14·2)
10·9 (8·9–13·0)
Weight (%)
Prevalence (95% CI)
020304010
2·9%
2·8%
2·8%
2·6%
3·0%
2·8%
2·8%
2·9%
2·8%
2·7%
2·5%
2·9%
2·8%
2·8%
2·6%
2·8%
2·9%
2·4%
2·7%
2·7%
55·4%
2·7%
2·9%
2·9%
2·5%
2·3%
13·2%
3·0%
3·0%
3·0%
2·3%
3·0%
1·4%
2·2%
18·0%
2·4%
2·8%
2·9%
2·6%
2·6%
13·4%
100%
Figure 4: Meta-analysis results for prevalence of type 2 diabetes in studies in which diabetes diagnosis was based on fasting plasma glucose

www.thelancet.com/diabetes-endocrinology Published online November 5, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00363-0
9
Review
studies in Mauritius, Fiji, and Nauru, Shaw and
colleagues67 reported a 2·7 times increase in risk for all-
cause mortality in men and 2·0 times in women with
newly diagnosed diabetes based on OGTT, compared
with people with normal glucose tolerance. By contrast,
people with diagnoses based on FPG had no increased
risk. Similar fi ndings have been reported in the Hoorn
study.68 Thus, OGTT should be used rather than FPG to
diagnose new cases of diabetes.
The higher prevalence of diabetes in urban than in
rural settings is in agreement with fi ndings from studies
done in other countries.69–72 Urbanisation in Africa, as
elsewhere, is associated with notable changes in lifestyle,
including decreased physical activity. In rural African
communities, full-time labour-intensive sub sistence
farming is a common means of livelihood and transport
is active—ie, walking or cycling—particularly in
individuals with low socioeconomic status. In urban
areas, although there is still little leisure time, active
transport and the levels of occupation-related and total
physical activity are substantially reduced compared with
the rural setting. Additionally, living in rural and
traditional environments is associated with consuming
diets incorporating more fruit and vegetables than in
urban settings.62,73–76 By contrast, diets in urban areas are
associated with increased intake of animal fat and refi ned
carbohydrates, including fast foods and sugar-sweetened
foods and drinks, since, notwithstanding the widespread
poverty in urban areas, access to these types of food is
much easier than in rural areas.77,78
Food insecurity, a risk factor for obesity, is an important
issue for the urban poor in Africa. Although studies have
revealed a strong positive relation between obesity and
high socioeconomic status, the obesity burden might be
shifting to sections of the poor urban population, where
people might have little knowledge and fi nancial
resources to adopt healthier lifestyles.78–84 The implications
of the rural–urban gradient in diabetes prevalence are
profound for Africa, where countries are currently facing
one of the most rapid rates of increased urbanisation
worldwide. At present, the average annual rate of
urbanisation is 3·2%, and it is predicted that 47·7% of
the African population will be living in urbanised areas
within 15 years, which equates to an estimated 744 million
people living in cities by 2030. Therefore, a striking
increase in the prevalence and number of adults with
type 2 diabetes and associated risk factors is likely.85
Our fi ndings have important policy implications for
Africa. Although the attention of policy makers is fi nally
extending beyond HIV/AIDs, tuberculosis, and malaria to
non-communicable diseases, the enormity of this latter
epidemic, including diabetes, is not fully appreciated.
Increased survival in people taking anti retroviral therapy
and increased longevity in those who are HIV negative,
together with urbanisation, will place an ever-increasing
burden on the already stretched health-care services in
Age group and country
Age 55–64 years
Algeria20
Nigeria34
Seychelles36
South Africa38
Random effects model
Heterogeneity: I2=849%, tau2=0·0047, p=0·0002
Age ≥55 years
Canary Islands26
South Africa39
Random effects model
Heterogeneity: I2=984%, tau2=0·0605, p<0·0001
Age ≥60 years
Angola22
Libya31
Réunion35
South Africa37
Random effects model
Heterogeneity: I2=922%, tau2=0·0171, p<0·0001
Age ≥65 years
South Africa38
Tunisia41
Random effects model
Heterogeneity: I2=0%, tau2=0, p=0·4967
Overall random effects model
Heterogeneity: I2=949%, tau2=0·017, p<0·0001
Sample size
242
582
331
150
1305
195
380
575
123
98
499
140
860
77
598
675
3415
Prevalent diabetes
27
125
81
34
267
64
25
89
18
24
188
61
291
24
164
188
835
Prevalence (95% CI)
11·2 (7·5–15·8)
21·5 (18·2–25·0)
24·5 (19·9–29·5)
22·7(16·2–30·2)
19·7 (14·2–25·8)
32·8 (26·3–39·9)
6·6 (4·3– 9·6)
17·7 (0·8–49·2)
14·6 (8·9–22·1)
24·5 (16·4–34·2)
37·7 (33·4–42·1)
43·6 (35·2–52·2)
29·6 (18·2–42·5)
31·2 (21·1–42·7)
27·4 (23·9–31·2)
27·8 (24·5–31·3)
23·9 (17·7–30·7)
Weight (%)
Prevalence (95% CI)
020304010
8·5%
8·8%
8·6%
8·2%
34·0%
8·4%
8·6%
17·0%
8·0%
7·8%
8·7%
8·1%
32·7%
7·5%
8·8%
16·3%
100%
Figure 5: Meta-analysis results for prevalence of type 2 diabetes in studies in which diabetes diagnosis was based on using the oral glucose tolerance test

10
www.thelancet.com/diabetes-endocrinology Published online November 5, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00363-0
Review
Africa, unless aggressive preventive strategies are put in
place. If monitoring of secular trends in the incidence
and prevalence of diabetes associated with the
epidemiological transition across all African countries is
to be successful, use of standardised methods will be
important. African countries that have done STEPS
surveys must be encouraged to move towards having a
regular cycle of risk factor surveillance refl ected in
national reporting of non-communicable diseases, and
especially plans of action for diabetes. The questionnaire
items and measures used in STEPS and the indicators
reported from STEPS surveys need to be periodically
reviewed to adapt to the latest scientifi c standards and
policy needs in diff erent countries.86,87
The issue of which test to use for diabetes prevalence
studies (OGTT, FPG, or HbA1c) in Africa is complex,
because decisions have to take into account cost,
convenience, and reliability. Measurement of trends
over time will be made diffi cult if diagnostic criteria
change. There is little doubt that HbA1c would be the
simplest indicator for diabetes because it avoids the
need for fasting and a measurement of glucose
concentrations at 2 h for an OGTT. HbA1c also has less
within-individual variation and better predicts micro-
vascular and macrovascular complications than FPG
and OGTT. The cutoff s for diabetes with HbA1c, however,
have not been established for African populations,
which is clearly important to address. The use of HbA1c
in epidemiological studies will only be viable once
reliable, cost-eff ective, point-of-care equipment has
been developed. Further more, if HbA1c were to become
the standard test for diabetes prevalence studies, it
would only be possible to assess trends once new
baseline estimates with this method had been
established. Irrespective of the diagnostic criteria or test
used, future studies must assess the prevelance of type 2
diabetes according to uniform case defi nition and
diagnostic methods and provide standardised prevalence
values for older people across Africa to enable
comparisons across the continent.
African countries also need encouragement to move
from subnational implementation of STEPS surveys to
capturing national prevalence data. Increased collabo-
ration between African governments and WHO is also
needed to make the data collected from STEPS widely
available so that they can be used to strengthen strategies
for prevention and management of non-communicable
diseases, including type 2 diabetes.87 Indeed, there is a
clear obligation to invest in surveillance, diabetes
prevention, and creation of aff ordable, innovative models
of health-care access and systems to halt the growing
burden of diabetes in Africa. In the meantime, primary
prevention should focus on targets suggested by
evidence, such as improving access to health care, health
education, and countering risk factors for vascular
disease, including, hypertension and obesity in middle
age, smoking, and physical inactivity.
Our systematic review has several strengths. We used a
comprehensive review protocol15 and made extensive
eff ort to identify all the available evidence without
language restrictions. We reported this systematic review
according to PRISMA guidelines. We applied an African
search fi lter, searched multiple electronic databases, and
used a rigorous approach to select studies for inclusion.
Another important strength is that we controlled for the
eff ect of multiple-publication bias in the analysis of the
results by avoiding inclusion of duplicate publications
that might have skewed the interpretation of the
prevalence estimates. Similarly, we appraised the
methodological quality of individual studies with
standard quality assessment tools for prevalence studies.
Use of an arcsine transformation to stabilise the variance
of primary studies before pooling limited the eff ects of
studies with small and large prevalence values on the
pooled estimates overall and across subgroups.
Our fi ndings should be interpreted in the context of
some limitations. First, data on the prevalence of diabetes
in people aged 55 years and older in Africa over the past
15 years are scarce and, therefore, describing trends of
diabetes prevalence over time was not possible. Second,
our ability to assess the quality of studies that we identifi ed
was limited by the methodological information provided,
some of which was incomplete. Diff ering study sample
sizes might partly explain variability in diabetes estimates.
Third, we found notable heterogeneity in prevalence
measures of type 2 diabetes, which was incompletely
explained by subgroup analyses. Despite substantial
diff erences, we could pool data to provide useful
estimates. Fourth, most primary studies lacked data on
key covariates that could have been used in metaregression
analyses to further explore and control for the sources of
variations in prevalence between studies.
Our cutoff of age 55 years for the older population
could be questioned. In reality, the defi nition of older is
somewhat arbitrary. For example the United Nations
uses a cutoff of 60 years, whereas for research purposes,
the cutoff for older populations is frequently 60 or
65 years.88 Traditional African defi nitions of a so-called
older or elderly person correlate with the chronological
ages of 50–65 years, dependent on the setting, region,
and country.89 Age 50 years was used as the lower cutoff
for the Minimum Data Set Project on Ageing in sub-
Saharan Africa90 because this threshold was thought to be
more realistic than age 60 years for African populations.
Furthermore, collection of data from the age of 50 years
might show emerging trends that would be of assistance
to policy makers and for health-care planning.90
Many parts of Africa have only recently entered
demographic transition, which accounts for the evidence
that older people represent just over 5% of the African
population in 2014.91,92 Yet, the ageing population in
Africa is growing at a much faster pace than in any other
continent, and has been accompanied by an increase in
the median population age and changes in the

www.thelancet.com/diabetes-endocrinology Published online November 5, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00363-0
11
Review
dependency ratio. Thus, the proportion of the population
composed of children has decreased, and that of people
aged 60 years and older has increased. By 2050, the older
population in Africa is projected to more than triple to
reach 205 million. Furthermore,the widespread access to
antiretroviral therapy over the past 4–5 years in sub-
Saharan Africa has already had a substantial eff ect on life
expectancy and will contribute to the growing number of
older people in Africa; thus, the number of people at risk
of type 2 diabetes and cardiovascular disease will also
increase. Indeed, the projected worldwide median life
expectancy in people living with HIV is 75 years, which is
only 7 years less than that for the general population.93
We cannot give an accurate assessment of the
contribution that diabetes in the older population makes
to the total adult burden of diabetes in Africa, but, on the
basis of this review and population data, we estimate that
9 million Africans aged 55 years and older are living with
diabetes in 2015. Accordingly, a concerted eff ort from
multiple sectors will be crucial to ensuring improved
health-care delivery for people with diabetes to reduce
the unacceptably high levels of morbidity and mortality
in the African region.
In conclusion, the substantial burden of diabetes in the
older population is an important health issue for African
countries. Uniform diagnostic methods are needed to
assess trends in diabetes prevalence within and between
countries. Such uniformity will better enable diabetes
prevention strategies to be put in place and collaborative
initiatives to be developed between key international and
national diabetes and geriatric organisations that can do
further research and, ultimately, improve diabetes care
for older people in Africa and worldwide.
Contributors
All authors conceived the study and were responsible for designing the
protocol. MW, MEE, APK, and NSL designed the study. MW did the
literature search and, together with AM, selected the studies, extracted
the relevant information. All authors synthesised the data. MW wrote the
fi rst draft of the paper. MEE, APK, and NSL provided critical guidance on
the analysis and overall direction of the study. All authors critically
revised successive drafts of the paper and approved the fi nal version.
Declaration of interests
We declare no competing interests.
Acknowledgments
We thank Tamzyn Suliaman, University of Cape Town, South Africa, for
technical support and assistance in the planning of the search strategy
and management of references. We also thank the Evidence-Based
Medicine Research Support Unit, Faculty of Health Sciences, University
of Cape Town, South Africa, for support.
References
1 Guariguata L, Whiting DR, Hambleton I, Beagley J, Linnenkamp U,
Shaw JE. Global estimates of prevalence of diabetes in adults for 2013
and projections to 2035. Diabetes Res Clin Pract 2014; 103: 137–49.
2 Alli F, Maharaj P. The health situation of older people in Africa.
In Maharaj P, ed. Aging and Health in Africa. New York, NY:
Springer, 2013: 53–89.
3 Steyn K, Damasceno A. Lifestyle and related risk factors for chronic
diseases. In Jamison DT, ed. Disease and mortality in sub-Saharan
Africa. New York, NY: World Bank, 2006: 247–66.
4 Levitt NS. Diabetes in Africa: epidemiology, management and
healthcare challenges. Heart 2008; 94: 1376–82.
5 Kowal P, Kahn K, Ng, N, et al. Ageing and adult health status in
eight lower-income countries: the INDEPTH WHO-SAGE
collaboration. Glob Health Action 2010; published online Sept 27.
DOI:10.3402/gha.v3i0.5302.
6 Motala AA. Diabetes trends in Africa. Diabetes Metab Res Rev 2002;
18 (suppl 3): S14–20.
7 Sobngwi E, Mauvais-Jarvis F, Vexiau P, Mbanya JC, Gautier JF.
Diabetes in Africans. Part 1: epidemiology and clinical specifi cities.
Diabetes Metab 2001; 27: 628–34.
8 Dalal S, Beunza JJ, Volmink J, et al. Non-communicable diseases in
sub-Saharan Africa: what we know now. Int J Epidemiol 2011;
40: 885–901.
9 Eisinga A, Siegfried N, Clarke M. The sensitivity and precision of
search terms in Phases I, II and III of the Cochrane Highly
Sensitive Search Strategy for identifying reports of randomized
trials in medline in a specifi c area of health care—HIV/AIDS
prevention and treatment interventions. Health Info Libr J 2007;
24: 103–09.
10 Pienaar E, Grobler L, Busgeeth K, Eisinga A, Siegfried N.
Developing a geographic search fi lter to identify randomised
controlled trials in Africa: fi nding the optimal balance between
sensitivity and precision. Health Info Libr J 2011; 28: 210–15.
11 WHO. Chronic diseases and health promotion: WHO STEPwise
approach to surveillance (STEPS). http://www.who.int/chp/steps/
manual/en/ (accessed May 31, 2013).
12 WHO. Chronic diseases and health promotion:STEPwise approach
to chronic disease risk factor surveillance. http://www.who.int/chp/
steps/en/ (accessed May 31, 2013).
13 WHO. Defi nition and diagnosis of diabetes mellitus and
intermediate hyperglycaemia: report of a WHO/IDF consultation.
Geneva: World Health Organization, 2006: 1–50.
14 Hoy D, Brooks P, Woolf A, et al. Assessing risk of bias in prevalence
studies: modifi cation of an existing tool and evidence of interrater
agreement. J Clin Epidemiol 2012; 65: 934–39.
15 Werfalli M, Musekiwa A, Engel ME, Ross I, Kengne AP, Levitt NS.
The prevalence of type 2 diabetes mellitus among older people in
Africa: a systematic review study protocol. BMJ Open 2014;
4: e004747.
16 Barendregt JJ, Doi SA, Lee YY, Norman RE, Vos T. Meta-analysis of
prevalence. J Epidemiol Community Health 2013; 67: 974–78.
17 DerSimonian R, Laird N. Meta-analysis in clinical trials.
Control Clin Trials 1986; 7: 177–88.
18 Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring
inconsistency in meta-analyses. BMJ 2003; 327: 557–60.
19 Egger M, Smith GD. Meta-Analysis. Potentials and promise.
BMJ 1997; 315: 1371–74.
20 Malek R, Belateche F, Laouamri S, et al. Prevalence of type 2
diabetes mellitus and glucose intolerance in the Setif area (Algeria).
Diabetes Metab 2001; 27: 164–71 (in French).
21 Zaoui S, Biémont C, Meguenni K. Epidemiology of diabetes in
urban and rural regions of Tlemcen (Western Algeria). Sante 2007;
17: 15–21 (in French).
22 Evaristo-Neto AD, Foss-Freitas MC, Foss MC. Prevalence of diabetes
mellitus and impaired glucose tolerance in Luanda—Angola.
J Diabetes Metab 2012; 3: 209.
23 Sagna Y, Tieno H, Guira O, et al. Prevalence and associated risk
factors of diabetes and impaired fasting glucose in urban
population; a study from Burkina Faso. J Diabetol 2014; 2: 4.
24 Nchanchou MV. Type 2 diabetes and its association with lifestyle
factors 2008. PhD thesis, University of Oslo, 2008.
25 Katte JC, Dzudie A, Sobngwi E, et al. Coincidence of diabetes
mellitus and hypertension in a semi-urban Cameroonian
population: a cross-sectional study. BMC Public Health 2014; 14: 696.
26 Boronat M, Varillas VF, Saavedra P, et al. Diabetes mellitus and
impaired glucose regulation in the Canary Islands (Spain):
prevalence and associated factors in the adult population of Telde,
Gran Canaria. Diabet Med 2006; 23: 148–55.
27 Katchunga P, Masumbuko B, Belma M, et al. Age and living in an
urban environment are major determinants of diabetes among
South Kivu Congolese adults. Diabetes Metab 2012; 38: 324–31.
28 Muluneh AT, Haileamlak A, Tessema F, et al. Population based survey
of chronic non-communicable diseases at Gilgel Gibe fi eld research
center, southwest Ethiopia. Ethiop J Health Sci 2012; 22: 7–18.

12
www.thelancet.com/diabetes-endocrinology Published online November 5, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00363-0
Review
29 Baldé NM, Diallo I, Baldé MD, et al. Diabetes and impaired fasting
glucose in rural and urban populations in Futa Jallon (Guinea):
prevalence and associated risk factors. Diabetes Metab 2007;
33: 114–20.
30 Ayah R, Joshi MD, Wanjiru R, et al. A population-based survey of
prevalence of diabetes and correlates in an urban slum community
in Nairobi, Kenya. BMC Public Health 2013; 13: 371.
31 Kadiki OA, Roaeid RB. Prevalence of diabetes mellitus and
impaired glucose tolerance in Benghazi Libya. Diabetes Metab 2001;
27: 647–54.
32 Solet JL, Baroux N, Pochet M, et al. Prevalence of type 2 diabetes
and other cardiovascular risk factors in Mayotte in 2008:
the MAYDIA study. Diabetes Metab 2011; 37: 201–07.
33 Silva-Matos C, Gomes A, Azevedo A, Damasceno A, Prista A,
Lunet N. Diabetes in Mozambique: prevalence, management and
healthcare challenges. Diabetes Metab 2011; 37: 237–44.
34 Ekpenyong C. NE U, Akpan E, Samson T. Double burden,
non-communicable diseases and risk factors evaluation in
sub-Saharan Africa: the Nigerian Experience. Eur J Sustain Dev
2012; 1: 249–70.
35 Favier F, Jaussent I, Moullec NL, et al, for the the REDIA Study Group.
Prevalence of type 2 diabetes and central adiposity in La Reunion
Island, the REDIA Study. Diabetes Res Clin Pract 2005; 67: 234–42.
36 Faeh D, William J, Tappy L, Ravussin E, Bovet P. Prevalence,
awareness and control of diabetes in the Seychelles and relationship
with excess body weight. BMC Public Health 2007; 7: 163.
37 Erasmus RT, Soita DJ, Hassan MS, et al. High prevalence of
diabetes mellitus and metabolic syndrome in a South African
coloured population: baseline data of a study in Bellville, Cape
Town. S Afr Med J 2012; 102: 841–44.
38 Peer N, Steyn K, Lombard C, Lambert EV, Vythilingum B, Levitt NS.
Rising diabetes prevalence among urban-dwelling black South
Africans. PLoS One 2012; 7: e43336.
39 Motala AA, Esterhuizen T, Gouws E, Pirie FJ, Omar MA. Diabetes and
other disorders of glycemia in a rural South African community:
prevalence and associated risk factors. Diabetes Care 2008; 31: 1783–88.
40 Bouguerra R, Alberti H, Salem LB, et al. The global diabetes
pandemic: the Tunisian experience. Eur J Clin Nutr 2007; 61: 160–65.
41 Hammami S, Mehri S, Hajem S, Koubaa N, Souid H, Hammami M.
Prevalence of diabetes mellitus among non institutionalized elderly
in Monastir City. BMC Endocr Disord 2012; 12: 15.
42 Ruhembe CC, Mosha TC, Nyaruhucha CM. Prevalence and
awareness of type 2 diabetes mellitus among adult population in
Mwanza city, Tanzania. Tanzan J Health Res 2014; 16: 4.
43 Mayega RW, Guwatudde D, Makumbi F, et al. Diabetes and
pre-diabetes among persons aged 35 to 60 years in eastern Uganda:
prevalence and associated factors. PLoS One 2013; 8: e72554.
44 Ministry of Health, Population, and Hospital Reform. Algeria 2003
STEPS survey report. Geneva: World Health Organization http://
www.who.int/chp/steps/STEPS_Report_Algeria.pdf?ua=1&ua=1
(accessed Sept 28, 2015).
45 Ministry of Health Benin, WHO. Geneva. Ministry of Health Benin.
Benin STEPS noncommunicable disease risk factors survey 2008.
http://ghdx.healthdata.org/record/benin-steps-noncommunicable-
disease-risk-factors-survey-2008 (accessed Sept 28, 2015).
46 Djrolo F, Adoukonou T, Houehanou C, Houinato JD, Houinato D.
Diabetes in Borgou Department in Benin: prevalence and
associated factors. J Diabetes Mellitus 2015; 5: 90–96.
47 Ministry of Health Botswana, WHO. Botswana STEPS Chronic
disease risk factor surveillance report 2007. http://www.who.int/chp/
steps/2007_STEPS_Report_Botswana.pdf (accessed Sept 28, 2015).
48 Ministry of Health, WHO. Enquête sur les facteurs de risque des
maladies non transmissibles à Kinshasa, Capitale de le RD Congo,
selon l’approche STEPS de l’OMS: rapport d’analyse. http://www.who.
int/chp/steps/STEPS_DRC_Final.pdf?ua=1 (accessed Sept 28, 2015).
49 On’Kin JBKL, Longo-Mbenza B, Okwe N, Kabangu N, Mpandamadi S,
Wemankoy O. Prevalence and risk factors of diabetes mellitus in
Kinshasa Hinterland. Int J Diabetes Metab 2008; 16: 97–106.
50 Ministry of Health and Population, Egypt Preventive Sector,
Central Epidemiology and Disease Surveillance (ESU),
Non-Communicable Disease Surveillance Unit (NCDSU), WHO.
Non-communicable diseases and their risk factors 2005–2006.
http://www.who.int/chp/steps/STEPS_Report_Egypt_2005-06.pdf
(accessed Sept 28, 2015).
51 Abebe SM, Berhane Y, Worku A, Assefa A. Diabetes mellitus in
North West Ethiopia: a community based study.
BMC Public Health 2014; 14: 97.
52 Ministry of Public Health and Public Hygiene, WHO. Enquête sur
les facteurs de risque des maladies non transmissibles à Libreville
et Owendo http://www.who.int/chp/steps/2009_STEPS_Report_
Gabon.pdf?ua=1 (accessed Sept, 28, 2015).
53 Ministry of Health Libya, WHO. Libya STEPS Noncommunicable
Disease Risk Factors Survey 2009 (in Arabic). http://www.who.int/
chp/steps/2009_STEPS_Libya.pdf?ua=1 (accessed Sept 28, 2015).
54 Ministry of Health Malawi, WHO. Malawi STEPS survey for
chronic non-communicable diseases and their risk factors: fi nal
report. http://www.who.int/chp/steps/Malawi_2009_STEPS_
Report.pdf?ua=1 (accessed Sept 28, 2015).
55 Ministry of Health Mauritania, WHO. Enquête sur les facteurs de
risque des maladies non transmissibles selon l’approche STEPwise
de l’OMS: étude de l’hypertension arterielle, du diabete et des autre
facteurs de risque à Nouakchott Mauritanie, janvier–juin 2006.
http://www.who.int/chp/steps/STEPS_Report_Mauritania.
pdf?ua=1 (accessed Sept 28, 2015).
56 Ministry of Health Mauritius, WHO. Mauritius non-communicable
diseases survey 2004. http://www.who.int/chp/steps/STEPS_
Report_Mauritius.pdf?ua=1 (accessed Sept 28, 2015).
57 Republique du Niger Ministere de la Sante Publique, WHO.
Mesure des facteurs de risqué de maladies non transmissibles au
Niger (Approche Step “wise” de l’OMS). http://www.who.int/chp/
steps/2007_STEPS_Report_Niger.pdf ?ua=1 (accessed
Sept 28, 2015).
58 Ministry of Health Sudan, WHO. Sudan STEPS noncommunicable
disease risk factors survey (Khartoum state) 2005. http://www.who.
int/chp/steps/STEPS_Report_Sudan2005.pdf?ua=1 (accessed
Sept 28, 2015).
59 Agoudavi K, Ministry of Health Togo. Rapport fi nal de l’enquête
STEPS Togo 2010. http://www.who.int/chp/steps/2010STEPS_
Report_Togo_FR.pdf?ua=1 (accessed Sept 28, 2015).
60 Ministry of Health & Child Welfare, University of Zimbabwe,
WHO. National Survey: Zimbabwe non-communicable disease risk
factors – (ZiNCoDs) preliminary report 2005 using the WHO
STEPwise approach to surveillance of non-communicable diseases
(STEPS). http://www.who.int/chp/steps/STEPS_Zimbabwe_Data.
pdf?ua=1 (accessed Sept 28, 2015).
61 International Diabetes Federation. IDF Diabetes Atlas, Sixth
Edition: key fi ndings 2014. http://www.idf.org/diabetesatlas/
update-2014 (accessed Sept 28, 2015).
62 Bos M, Agyemang C. Prevalence and complications of diabetes
mellitus in Northern Africa, a systematic review.
BMC Public Health 2013; 13: 387.
63 DECODE Study Group. Age- and sex-specifi c prevalences of
diabetes and impaired glucose regulation in 13 European cohorts.
Diabetes Care 2003; 26: 61–69.
64 Qiao Q, Hu G, Tuomilehto J, et al. Age- and sex-specifi c prevalence
of diabetes and impaired glucose regulation in 11 Asian cohorts.
Diabetes Care 2003; 26: 1770–80.
65 Cowie CC, Rust KF, Ford ES, et al. Full accounting of diabetes and
pre-diabetes in the U.S. population in 1988–1994 and 2005–2006.
Diabetes Care 2009; 32: 287–94.
66 Barrett-Connor E, Ferrara A. Isolated postchallenge
hyperglycemia and the risk of fatal cardiovascular disease in older
women and men. The Rancho Bernardo Study. Diabetes Care 1998;
21: 1236–39.
67 Shaw JE, Hodge AM, de Courten M, Chitson P, Zimmet PZ.
Isolated post-challenge hyperglycaemia confi rmed as a risk factor
for mortality. Diabetologia 1999; 42: 1050–54.
68 de Vegt F, Dekker JM, Ruhé HG, et al. Hyperglycaemia is
associated with all-cause and cardiovascular mortality in the
Hoorn population: the Hoorn Study. Diabetologia 1999;
42: 926–31.
69 Ginter E, Simko V. Type 2 diabetes mellitus, pandemic in
21st century. Adv Exp Med Biol 2013; 771: 42–50.
70 Whiting DR, Guariguata L, Weil C, Shaw J. IDF diabetes atlas:
global estimates of the prevalence of diabetes for 2011 and 2030.
Diabetes Res Clin Pract 2011; 94: 311–21.
71 Mohan V, Sandeep S, Deepa R, Shah B, Varghese C. Epidemiology of
type 2 diabetes: Indian scenario. Indian J Med Res 2007; 125: 217–30.

www.thelancet.com/diabetes-endocrinology Published online November 5, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00363-0
13
Review
72 Li H, Oldenburg B, Chamberlain C, et al. Diabetes prevalence and
determinants in adults in China mainland from 2000 to 2010:
a systematic review. Diabetes Res Clin Pract 2012; 98: 226–35.
73 Peer N, Kengne A, Motala AA, Mbanya JC. Diabetes in the Africa
region: an update. Diabetes Res Clin Pract 2014 103: 197–205.
74 Mbanya JCN, Motala AA, Sobngwi E, Assah FK, Enoru ST. Diabetes
in sub-Saharan Africa. Lancet 2010; 375: 2254–66.
75 Badran M, Laher I. Type II diabetes mellitus in Arabic-speaking
countries. Int J Endocrinol 2012; 2012: 902873.
76 Ng M, Fleming T, Robinson M, et al. Global, regional, and national
prevalence of overweight and obesity in children and adults during
1980–2013: a systematic analysis for the Global Burden of Disease
Study 2013. Lancet 2014; 384: 766–81.
77 Ziraba AK, Fotso JC, Ochako R. Overweight and obesity in urban
Africa: a problem of the rich or the poor? BMC Public Health
2009; 9: 465.
78 Popkin BM. The nutrition transition in the developing world.
Dev Policy Rev 2003; 21: 581–97.
79 Kennedy G, Nantel G, Shetty P, for the Food and Agriculture
Organization of the United Nations. Globalization of food systems
in developing countries: impact on food security and nutrition
FAO Food Nutr Pap 2004; 83: 1–300.
80 Dixon J, Omwega AM, Friel S, Burns C, Donati K, Carlisle R.
The health equity dimensions of urban food systems.
J Urban Health 2007; 84 (suppl): i118–29.
81 York DA, Rössner S, Caterson I, et al, for the American Heart
Association. Prevention conference VII: obesity, a worldwide
epidemic related to heart disease and stroke: group I: worldwide
demographics of obesity. Circulation 2004; 110: e463–70.
82 Kruger HS, Puoane T, Senekal M, van der Merwe MT. Obesity in
South Africa: challenges for government and health professionals.
Public Health Nutr 2005; 8: 491–500.
83 Shaw JE, Sicree RA, Zimmet PZ. Global estimates of the prevalence
of diabetes for 2010 and 2030. Diabetes Res Clin Pract 2010; 87: 4–14.
84 Hilawe EH, Yatsuya H, Kawaguchi L, Aoyama A. Diff erences by sex
in the prevalence of diabetes mellitus, impaired fasting glycaemia
and impaired glucose tolerance in sub-Saharan Africa: a systematic
review and meta-analysis. Bull World Health Organ 2013;
91: 671–82D.
85 Obeng-Odoom F. The state of African cities 2014. J Asian Afr Stud
2014; published online Sept 7. DOI:10.1177/0021909614547604.
86 Alwan A, Maclean DR. A review of non-communicable disease in
low- and middle-income countries. In Health 2009; 1: 3–9.
87 WHO. 2008–2013 action plan for the global strategy for the
prevention and control of noncommunicable diseases. Geneva,
Switzerland: World Health Organization, 2008.
88 Beard JR, Bloom DE. Towards a comprehensive public health
response to population ageing. Lancet 2015; 385: 658–61.
89 WHO. Defi nition of an older or elderly person: Proposed working
defi
nition of an older person in Africa for the MDS Project.
http://www.who.int/healthinfo/survey/ageingdefnolder/en/
(accessed May 7, 2015).
90 Kowal P, Peachey K. Indicators for the minimum data set project on
ageing: a critical review in sub-Saharan Africa. Geneva: World
Health Organization, 2001.
91 Population Reference Bureau. 2014 world population data sheet.
www.prb.org/pdf14/2014-world-population-data-sheet_eng (accessed
Aug 28, 2015).
92 Gerland P, Raftery AE, Sevčíková H, et al. World population
stabilization unlikely this century. Science 2014; 346: 234–37.
93 Gedela K, Vibhuti M, Pozniak A, Ward B, Boffi to M.
Pharmacological management of cardiovascular conditions and
diabetes in older adults with HIV infection. HIV Med 2014;
15: 257–68.