UBC Faculty Research and Publications

Population attributable fraction of type 2 diabetes due to physical inactivity in adults: a systematic… Al Tunaiji, Hashel; Davis, Jennifer C; Mackey, Dawn C; Khan, Karim M May 18, 2014

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
52383-12889_2013_Article_6790.pdf [ 436.53kB ]
Metadata
JSON: 52383-1.0223534.json
JSON-LD: 52383-1.0223534-ld.json
RDF/XML (Pretty): 52383-1.0223534-rdf.xml
RDF/JSON: 52383-1.0223534-rdf.json
Turtle: 52383-1.0223534-turtle.txt
N-Triples: 52383-1.0223534-rdf-ntriples.txt
Original Record: 52383-1.0223534-source.json
Full Text
52383-1.0223534-fulltext.txt
Citation
52383-1.0223534.ris

Full Text

RESEARCH ARTICLE Open AccessPopulation attributable fradiabetes due to physical i3 apfrglobal mortality attributable to physical inactivity is ap-proximately 3.3 million [5]. Globally, physical inactivitycant health and economic burden on North Americanhealth care system [8]. In the US alone (2011), the age–Al Tunaiji et al. BMC Public Health 2014, 14:469http://www.biomedcentral.com/1471-2458/14/469Full list of author information is available at the end of the articleis identified as the fourth leading risk factor for mortalityamong adults [5]; it is an independent risk factor formajor chronic diseases [6,7]. Physical inactivity is alsoassociated with substantial economic burden across theadjusted incidence increased 117% from 3.5 to 8.3 per1,000 persons between 1980 and 2011 [9]. The cases ofT2DM were projected to increase from 12 million in 2000to 39 million by 2050 (i.e. a prevalence increase from 4.4%to 9.7% in 2050) [10]. The direct costs associated withT2DM was approximately $US 44.1 billion per year oralmost $US 6000 per person per year (1997 prices) [11].Furthermore, the cost of T2DM attributable to physicalinactivity (absence of leisure-time activity) ranged from* Correspondence: karim.khan@ubc.ca1Centre for Hip Health and Mobility, University of British Columbia,Vancouver Coastal Health Research Institute (VCHRI), British Columbia,Canada5Aspetar - Orthopaedic and Sports Medicine Hospital, Doha, Qatarwas to determine the best estimate of PAF for T2DM attributable to physical inactivity and absence of sportparticipation or exercise for men and women.Methods: We conducted a systematic review that included a comprehensive search of MEDLINE, EMBASE,SportDiscus, and CINAHL (1946 to April 30 2013) limited by the terms adults and English. Two reviewers screenedstudies, extracted PAF related data and assessed the quality of the selected studies. We reconstructed 95% CIs forstudies missing these data using a substitution method.Results: Of the eight studies reporting PAF in T2DM, two studies included prospective cohort studies (3 total) andsix were reviews. There were distinct variations in quality of defining and measuring physical inactivity, T2DM andadjusting for confounders. In the US, PAFs for absence of playing sport ranged from 13% (95% CI: 3, 22) in men and29% (95% CI: 17, 41) in women. In Finland, PAFs for absence of exercise ranged from 3% (95% CI: -11, 16) in men to7% (95% CI: -9, 20) in women.Conclusions: The PAF of physical inactivity due to T2DM is substantial. Physical inactivity is a modifiable risk factorfor T2DM. The contribution of physical inactivity to T2DM differs by sex; PAF also differs if physical inactivity isdefined as the absence of ‘sport’ or absence of ‘exercise’.Keywords: Population attributable fraction (PAF), Physical inactivity, Type 2 diabetes (DM-2), Systematic reviewBackgroundPhysical inactivity, a global pandemic [1], is one of themost serious public health problems of the 21st centuryin terms of consequences and cost [2-4]. Annually, theglobe, accounting for instance annual direct cost of SFr1.6 billion in Switzerland (1999 prices) to $US 24 billionin the USA (1999 prices) [4].Type 2 diabetes mellitus (T2DM) also imposes a signifi-systematic reviewHashel Al Tunaiji1,2, Jennifer C Davis1,4, Dawn C Mackey1,AbstractBackground: Physical inactivity is a global pandemic. Themellitus (T2DM) associated with physical inactivity ranges© 2014 Al Tunaiji et al.; licensee BioMed CentrCommons Attribution License (http://creativecreproduction in any medium, provided the orDedication waiver (http://creativecommons.orunless otherwise stated.ction of type 2nactivity in adults: and Karim M Khan1,5,6,7*opulation attributable fraction (PAF) of type 2 diabetesom 3% to 40%. The purpose of this systematic reviewal Ltd. This is an Open Access article distributed under the terms of the Creativeommons.org/licenses/by/2.0), which permits unrestricted use, distribution, andiginal work is properly credited. The Creative Commons Public Domaing/publicdomain/zero/1.0/) applies to the data made available in this article,Al Tunaiji et al. BMC Public Health 2014, 14:469 Page 2 of 9http://www.biomedcentral.com/1471-2458/14/469$US 1.90 billion to $US 13.20 billion per year (2007prices) [12].Physical activity benefits at least 23 different healthconditions [13,14]. Despite this, fewer than 50% of thepeople engage in sufficient physical activity to reap suchbenefits [15,16]. Prospective studies demonstrate thatphysical inactivity is an independent and modifiable riskfactor for T2DM [17]. Specifically, physical activity inter-ventions reduced the risk of developing diabetes [18-20].A method of quantifying the burden of T2DM attrib-utable to physical inactivity is population attributablefraction (PAF). PAF takes into account the degree of asso-ciation between a risk factor and the incidence of a disease(i.e., relative risk) and the public health importance of thisrisk factor at a population level. Specifically, PAF estimatesthe proportion of disease cases (i.e., T2DM cases) that areattributable to a risk factor of interest (i.e., physical in-activity) among all disease cases in a population [21].To date, PAF estimates for the excess cases of T2DMvary from 3% due to lack of exercise in Finland to 40%in Canada due to lack of moderate-vigorous physical ac-tivity [13-15]. Some of this variability is due to variation incalculating PAF based on age-, gender-, region-specificfactors. But there has been no systematic review that hasassessed the PAF of T2DM attributable to physical inactiv-ity in men and women. Also, none has used recent ad-vances in PAF as outlined by Laaksonen [13]. Examiningthe quality of these discrepant estimates and underlyingreasons for the observed variation is important as it willprovide policymakers with a guide to which of the originalstudies should carry most weight. Hence, our primary ob-jective was to quantify the PAF of T2DM attributable tophysical inactivity and absence of sport participation orexercise for men and women.MethodsData sources and search strategyIn accordance with the Preferred Reporting Items forSystematic Reviews and Meta-Analysis (PRISMA) state-ment [22] and Cochrane Collaboration guidelines [23],we [HAT, JCD, KMK] conducted a comprehensivesearch of MEDLINE, EMBASE, SportDiscus, andCINAHL. We limited our search results to adults aged19-65 years and studies published in English. We in-cluded search terms the MeSH headings: diabetes,physical activity, fitness, risk assessment. The searchstrategy detailed in Figure 1(B) includes studies pub-lished between 1946 and April 30 2013. We manuallysearched all references of articles selected for full textreview to identify additional relevant papers.Study selection and eligibility criteriaWe (HAT, JCD) included peer reviewed, published stud-ies that (i) estimated PAF or population attributable ratio(PAR) using modeling on raw data from a prospectivecohort design or (ii) published adjusted relative risk(RRadj) and prevalence of the risk factor of interest –physical inactivity [24]. Of note, review studies were in-cluded if their RRadj estimates were based on prospectivecohort data. Based on title and abstract review, we ex-cluded studies that: 1) used an exposure unrelated tophysical inactivity), 2) used an outcome that was notT2DM, 3) used an inappropriate study design for esti-mating PAF/PAR (i.e., cross-sectional, case-control orretrospective studies). Based on full text review, we ex-cluded studies that: 1) did not contain a PAF estimate,2) did not detail the independent contribution of phys-ical inactivity, 3) used an inappropriate study design forestimating PAF/PAR (i.e., cross-sectional, case-controlor retrospective studies), 4) the primary outcome wasnot T2DM, 5) were duplicates. Eight full-text articlesmet the inclusion criteria – four from our search strat-egy and four from our review of the reference lists of allarticles selected for full text review. All discrepancieswere resolved by discussion and consultation with a co-author (KMK). Figure 1(A) details the process of studyselection for this systematic review.Data extractionTwo raters (HAT, JCD) independently extracted datafrom each study and any discrepancies were discussedand reviewed by a third party (KMK). We developed alist of data extraction topics for the studies included inthis systematic review (Additional file 1: Table S1 andAdditional file 2: Table S2). These items were: author’sname, year of publication, country, journal name, studydesign, sample size, sample characteristic, length of fol-low up, operational definition for exposure (physical in-activity), operational definition for outcome (T2DM),level of adjustment for confounders, PAF estimates andcalculation method used to estimate PAF (Additional file 1:Table S1 and Additional file 2: Table S2).Exposure, outcome and outcome measures for datasynthesisOur primary exposure of interest for population at-tributable fraction (PAF) estimates was physical inactivity.Physical inactivity was defined as total physical activ-ity insufficient to meet recommended guidelines, thatis ≤ 150 minutes of moderate-intensity or ≤ 75 minutesof vigorous-intensity aerobic physical activity per week inbouts of at least 10 minutes duration accumulated acrossoccupational, transport-related, domestic or leisure-timedomains [5]. Leisure –time activity domains includes exer-cise, sport and unstructured recreation [14,25,26]. Exercise[27] is a planned, structured and repetitive physical activ-ity with the purpose of improving and/or maintainingphysical fitness. i.e. both exercise and sport are subsets ofPapers identified through MEDLINE, EMBASE, SPORTDiscuss & CINAHL databases search(n = 7652)Additional papers identified through reference lists & web of science (n = 4)Records after duplicates removed(n = 6230)Title & Abstract screening(n = 71)Records excluded with reasons(n = 22)Reasons:1. Inapproriate exposure (n=10)2. Inapproriate outcome (n=7)3. Inappropriate design (n=3)4. Not English (n=2) Full-text articles assessed for eligibility(n = 49)Full-text articles excluded with reasons(n = 41)Reasons:1. PAF not calculated (n= 33)2. PAF for joint risk factors (n=3)3. Cross sectional design (n=2)4. PAF for mortality (n=1)5. Not English (n=1)6. Duplicate (n=1)Studies included in the review(n = 8)IdentificationScreeningEligibilityIncluded1     diabetes.mp. (379144)2     diabetic$.mp. (192522)3     diabetes mellitus/ or *diabetes mellitus, type 2/ or *prediabetic state/ (146769)4     1 or 2 or 3 (429966)5     physical activit$.mp. (51021)6     physical inactivity.mp. (3541)7     fitness.mp. (48520)8     Physical Fitness/ (20584)9     Sedentary Lifestyle/ (1816)10   cardiorespiratory fitness.mp. (1697)11   Motor Activity/ (71159)12   exp sports/ or running/ or jogging/ or walking/ or weight lifting/ (103461)13   exp exercise/ or muscle stretching exercises/ or plyometric exercise/ or resistance training/ or physical fitness/ (115143)14  *Physical Exertion/ (31429)15    5 or 6 or 7 or 8 or 9 or 10 or 11 or 12 or 13 or 14 (303015)16    4 and 15 (13110)17    population attributable fraction.mp. (410)18    excess risk.mp. (4051)19    attributable risk.mp. (2272)20    population attributable risk.mp. (1198)21    exp morbidity/ or incidence/ or prevalence/ (339813)22    exp risk/ or logistic models/ or risk assessment/ or risk factors/ (764962)23    17 or 18 or 19 or 20 or 21 or 22 (989461)24    16 and 23 (4748)25    limit 24 to "all adult (19 plus years)" (3329)ABFigure 1 A Flow diagram showing study selection (1A) and database search (1B) for systematic review of studies on populationattributable fraction (PAF) of type 2 diabetes (T2DM) due to physical inactivity in adults.Al Tunaiji et al. BMC Public Health 2014, 14:469 Page 3 of 9http://www.biomedcentral.com/1471-2458/14/469Al Tunaiji et al. BMC Public Health 2014, 14:469 Page 4 of 9http://www.biomedcentral.com/1471-2458/14/469leisure-time domain and are not interchangeable [27].Sport is a subset of exercise undertaken either individuallyor as a part of a team where participants adhere to a com-mon set of rules or expectation and a defined goal to win[28]. Physical inactivity was either self-reported or directlymeasured by accelerometry.Our primary outcome of interest for estimating popu-lation attributable fraction (PAF) was T2DM definedas: 1) hyperglycemia ascertained by fasting plasmaglucose ≥ 7.0 mmol/l (126 mg/dl) or 2-h plasma glu-cose ≥ 11.1 mmol/l (200 mg/dl) or 2) self report withvalidation from a registry, medical record or reim-bursement plan [29,30].Population attributable fraction (PAF) or proportion(PAR) was defined as the excess number of cases ofT2DM attributable to physical inactivity or low physicalfitness that is estimated by the following formula or oneof its variant [31]:PAFPe RRadj‐1ð ÞRRadj‐1ð Þ  100Where, Pe is population prevalence of exposure andRRadj is an adjusted relative risk.Of note, we estimated the confidence intervals (95%CI) for PAF using the substitution method when thesedata were not reported [32]. All calculations done by theauthors are labeled with an ‘a’ in Additional file 2: Table S2.This method used the upper and lower limits of RR inattributable risk (AR) formula.Due to study design, sample and analytic heterogen-eity, a meta-analysis of these data to determine PAF forT2DM was not conducted.Quality assessmentBecause our systematic review consisted of both prospect-ive cohort studies and reviews, a published quality assess-ment checklist suitable for this study was not available.Therefore, we developed a seven-item quality assessmentform. This form was created after reviewing potentiallyrelevant checklists such as the STROBE [33,34]. Fromthese examples, we created and modified questions rele-vant to assessing the quality of the PAF estimates includedin this systematic review. The questions were structuredso that they could be applied across all included studiesand study designs (Additional file 3: Table S3). All qualityassessment questions were reviewed by an expert in thefield. This quality assessment was not validated. We useddichotomized answers (+: yes, -: no) for the quality assess-ment questions to create a score out of 7. Two authors(JCD, HAT) independently evaluated each study and anydiscrepancies were discussed and reviewed by a thirdauthor (KMK). Below, we outline each of the criteriaincluded in the quality assessment.Quality assessment questionsQuestion 1: Was a clear definition provided for theexposure (physical inactivity)? Physical inactivity wasdefined as the total activity that does not meet the recom-mended guidelines of ≤ 150 minutes of moderate-intensityor ≤ 75 minutes of vigorous-intensity aerobic physicalactivity per week in bouts of at least 10 minutes durationaccumulated across occupational, transport-related, do-mestic or leisure-time activity domains [5].Question 2: Was the exposure (physical inactivity)measured objectively? Physical inactivity can either bemeasured subjectively using validated self reported ques-tionnaires or objectively using accelerometers [35].Question 3: Was a clear clinical definition providedfor the outcome (type 2 diabetes)? T2DM was definedusing hyperglycemia cutoffs as listed above.Question 4: Was the outcome ascertained by objectivemeasures or if self reported confirmed by othermeasures? The current diagnostic criteria is requires afasting plasma glucose ≥ 7.0 mmol/l (126 mg/dl) or a 2-hplasma glucose ≥ 11.1 mmol/l (200 mg/dl) [29,30].Question 5: Was the analysis based on raw data froma prospective cohort study? One of the PAF assumptionsis causality; therefore, only prospective studies weredeemed appropriate for estimating PAF [21].Question 6: Was the follow up time provided? PAF issubject to follow up time bias [21]. Specifically, a shorterfollow up time is associated with an overestimated PAFwhile a longer follow up time is associated with an under-estimated PAF.Question 7: Was population attributable fraction(PAF) or proportion (PAR) fully adjusted? PAFs aresubject to confounding bias [24]. The partial adjustmentmethod is a popular method of calculating PAF. It usespublished adjusted RR and prevalence of exposure inthis formula [24]:PAF ¼ Pe RRadj‐1ð ÞRRadj‐1 100The partially adjusted method can yield severely biasedPAF estimates [36] because the confounding variablesare not adequately adjusted. For instance, incomplete ad-justment for confounding by age and sex can lead to17% overestimation in PAF [36]. Therefore modeling allknown confounders (i.e., full adjustment modelingmethod) is a better approach [24].file 2: Table S2) [14,26]. Bull’s [26] PAF estimates for total13% (95% CI: 3, 22) in men and 29% (95% CI: 17, 41) inAl Tunaiji et al. BMC Public Health 2014, 14:469 Page 5 of 9http://www.biomedcentral.com/1471-2458/14/469ResultsOverview of studiesAfter critical review of the 49 full text manuscripts, eightstudies met our inclusion criteria (Figure 1A, Additional file 1:Table S1 and Additional file 2: Table S2). There weredistinct variations in quality across studies with respectto defining and measuring physical inactivity, definingand measuring T2DMand adjusting for confounders inthe final model for calculating PAF and follow up time(Additional file 1: Table S1 and Additional file 2: Table S2).Of the eight studies, three focused on the exposure of‘total physical inactivity’ [26,37,38], three on leisure-time activity and two on subsets of leisure-time activity -specifically ‘exercise’ [39] and ‘sport’ [40] (Additional file 2:Table S2). Of the eight studies, two described three differ-ent prospective cohorts and six were reviews of publisheddata. The two prospective cohort studies (included threeprospective cohorts) [14] estimated PAF using full ad-justment modeling. The six reviews estimated PAF usingpublished data of adjusted relative risk (RRadj) from previ-ously published cohort studies and estimated the preva-lence of physical inactivity (Pe) from cross-sectional data.Physical inactivity was self- reported in all studies exceptone that used data on prevalence of physical inactivitymeasured by accelerometry [37].Prospective cohort studies (2 studies, 3 prospectivecohorts)The three prospective cohorts scored the highest onquality assessment, Additional file 3: Table S3. The PAFfor physical inactivity ranged from 3% (95% CI: -11, 16)to 29% (95% CI: 17, 41). In Finland, the PAF from twoprospective cohort studies for exercise, a subset ofleisure-time domain, ranged from 3% (95% CI: -11, 16)to 7% (95% CI: -9, 20) [13]. The cumulative incidenceranged from 2.6 to 3.9 per 100 people, the adjusted rela-tive risk (RRadj) ranged from 1.28 (95% CI: 0.99, 1.48) to1.35 (95% CI: 0.97, 1.6) and the prevalence of physicalinactivity (Pe) ranged from 24.1% and 36.5%. In theUSA, the PAF for sport, subset of leisure-time domain,to range from 13% to 29%: 13% (95% CI: 3, 22) in menand 29% (95% CI: 17, 41) in women [40]. The cumula-tive incidence was 7.6 per 100 person, the adjusted rela-tive risk (RRadj) was 1.21 (95% CI: 1.1, 1.35) for menand 1.43 (95% CI: 1.21, 1.68) for women and the preva-lence of physical activity (Pe) was 55.2% for men and66.3% for women.Country-specific reviews on published data (4 studies)The PAF estimates from these four studies ranged from20.1% (17.8, 30.1) [37] to 39% (95% CI: 35.9, 41.7) [38]for total physical inactivity and 19.9% (95% CI: 11, 27.1)[41] to 21.1% (16.5, 25.2) [42] for leisure-time activity.The 95% confidence intervals were constructed for allwomen. In Finland, Finland, the PAF of T2DM due tophysical inactivity for the occasional exerciser (≤30 min/day,subset of leisure-time activity domain) ranged from 3%(95% CI: -11, 16) to 7% (95% CI: -9, 20). The PAF esti-mates for T2DM attributable to physical inactivity variedwidely. Specifically, further variation is notable across studydesign, countries and sex. Such divergence may be ex-plained by the distinct inconsistency in quality acrossstudies. Below we elaborate on how two categories relat-physical inactivity ranged from 5.2% (95% CI: 2.2, 8.2) inCanada to 13% (95% CI: 4.8, 16.6) in Finland for totalphysical inactivity while Lee [14] estimated PAFs forleisure-time to range from 7% (95% CI: 0.8, 14.4) inCanada to 10.7% (95% CI: 5.4, 16.8) in South Africa. Inone review [26] the 95% CI intervals were not reportedtherefore we reconstructed them using the substitutionmethod [32]. The adjusted relative risk (RRadj) 1.24(1.1, 1.39) and the prevalence of physical inactivity rangedfrom 23% to 61%.DiscussionA review of the variation that exists in PAF across theexistent literatureThe PAF estimates for T2DM that is attributable to phys-ical inactivity varied widely from 3%-39% across studies(Janssen & Laksoonen). As determined from the perform-ance on our quality assessment, the best quality data inthis systematic review suggest that the PAF of T2DM dueto physical inactivity in the USA for a non sport partici-pant (never engaged in strenuous sports) ranged fromPAF estimates using the substitution method [32]. Theadjusted relative risk (RRadj) ranged from 1.24 (95% CI:1.1, 1.39) to 1.74 (95% CI: 1.65, 1.83) and the prevalenceof physical inactivity ranged from 19.8% to 82% for menand 26.8% to 86.3% for women. The ranges of PAF,RRadj and Pe estimates from these country-specificstudies were narrower than estimates generated fromthe three prospective cohort studies.Global review on published data (2 studies)In general, the global review studies [14,26] reported lowerPAFs than the country-specific reviews and the prospect-ive cohort studies except for Finland. The review studieshad different definitions for physical inactivity [14,26]. Bull[26] defined physical inactivity as total physical inactivitywhile Lee [14] referred to leisure-time activity alone. Fur-ther, these two reviews used different formulas containingdifferent denominators to calculate PAF from previouslypublished data (Additional file 1: Table S1 and Additionaling to study methodology and statistical analysis con-tribute to the observed variation in PAF estimates.inactivity is assessed. For example, using an objectiveAl Tunaiji et al. BMC Public Health 2014, 14:469 Page 6 of 9http://www.biomedcentral.com/1471-2458/14/469Analysis of the potential explanations for thedemonstrated variation in PAFTwo main factors explain the wide variation we observein the PAF estimates for T2DM attributable to physicalinactivity: heterogeneous study methodology (i.e., studydesign, exposure and outcome measurement) and statisticalmethodology.MethodologyChoice of study designThe choice of study design is a key factor that may explainsubstantial variation PAF estimate. More recently, meth-odological advances demonstrate that prospective cohortstudies are preferable for PAF estimation because the cal-culations rely on censored time to event data [43,44,28].Historically, there is a large body of literature estimatingPAF from case-control and cross sectional data [24]. Forexample, only two of the eight studies included in this sys-tematic review reported three prospective cohort studiesthat were designed to estimate PAF as a primary outcomemeasure. As such, we observed wide variation in PAF esti-mates due to fundamental differences in study design. Sec-ond, PAF is based on multiple assumptions. One of theseassumptions is that PAF assumes that risk factors precedeand be causally related to the outcome. This assumptionrequires a longitudinal study design–a prospective cohortstudy. Ignoring such assumptions can lead to inaccurateestimations and hence incorrect interpretation of PAFestimates. Lastly, length of followup is another criticalfactor in accurately valuing PAF. In this systematic review,the follow up period ranged from 5 to 20 years overall andfrom 7 to 12 years in the three prospective cohort studies.Importantly, short follow up times tend to overestimatePAF and longer followup times generally underestimatePAF [21].Measurement of exposure (domain-specific PAF)Another reason that could explain the observed degreeof variation in PAF is the use of different definitions forthe physical inactivity. Physical inactivity occurs whentotal activity fails to meet the recommended guidelinesof ≥ 150 minutes of moderate-intensity or ≥ 75 minutesof vigorous-intensity aerobic physical activity per weekin bouts of at least 10 minutes duration accumulated acrossoccupational, transport-related, domestic or leisure-time ac-tivity domains [5]. Leisure–time activity consists of exercise,sport [14,25,26]. Specifically, exercise and sports are uniquesubsets of the leisure-time activity domain; they are notinterchangeable [27]. Therefore, acknowledging distinc-tion between is essential in our interpretation of results[27]. Two studies reporting three prospective cohortsscored high in our quality assessment. Despite this, thePAF estimates varied widely from 13% (3, 22) to 29%(17, 41) for occasional exerciser (≤30 min/day) [13] andmeasure such as accelerometry is more likely to capturetotal physical activity compared than a subjective meas-ure (i.e., self report). Self reporting of physical inactivityis prone to measurement error (i.e., often underestima-tion of physical inactivity) and consequently biased PAF(i.e., often overestimation) estimates. In a systematic re-view, Prince [35] reported low-to-moderate correlationsbetween self-report and direct measures of physical in-activity that ranged from -0.71 to 0.96. A clear trend forthe mean differences was not present. However, self-report measures were 44% (range: -78% to 500%) higherthan those measured directly by accelerometers. This sug-gests there is a trend of self-report measures over report-ing physical activity leading to an under-estimation ofboth physical inactivity and subsequent PAF estimates.Measurement of outcomeA third reason that could explain PAF estimate variationis the use of different definitions for T2DM [30,45].Current diagnostic criteria are fasting plasma ≥ 7.0 mmol/l(126 mg/dl) or 2-h plasma glucose≥ 11.1 mmol/l (200 mg/dl)[29,30]. Among the studies we reviewed, there were somedifferences in methods of diagnosis of T2DM. None of thestudies included in this review was based solely on plasmaglucose.Self- reported T2DM is also subject to measurementbias. For instance, the accuracy of self-reported T2DM isgood (kappa = 0.78) and of moderate sensitivity (73%)[46,47]. However, T2DM can remain asymptomatic forat least 4 to 7 years before a clinical diagnosis is made.[48]. As a result, T2DM may be undiagnosed in up to50% of cases [49,50]. This underestimation of the inci-dence of T2DM leads to an underestimate of RR and3% (-11, 16) to 7% (-9, 20) for non sport participants.This could partially be explained by the use differentsubsets definition of leisure-time domain. In the fourcountry-specific reviews, only two studies [41,42] usedsimilar definitions for the physical inactivity of theleisure-time activity domain. In the two global reviewstudies, the PAFs ranged from 5.2% (2.2, 8.2) to 10.9%(4.8, 16.6). These studies [14,26] also have differentdefinitions for physical inactivity. For example, Bull[26] estimated PAF for total physical inactivity whileLee [14] estimated PAF based primarily on the leisure-time domain.Another factor that could explain variation in PAF isthat physical activity was self reported in all studies ex-cept one [37]. A higher PAF of 39% (35.9, 41.7) wasbased on Canadian data [15]. One explanation for thehigher PAF observed may be due in part to how physicalPAF. Therefore, objective measurement of T2DM is de-sirable for accurate PAF estimates.Al Tunaiji et al. BMC Public Health 2014, 14:469 Page 7 of 9http://www.biomedcentral.com/1471-2458/14/469Statistical analysisThere are two published modeling techniques PAF: thefull adjustment method and the partial adjustment method.Below we discuss the pros and cons of these methods inthe context of estimating the PAF of T2DM attributableto physical inactivity.Full adjustment method (modeling techniques)In the two prospective studies PAF different modelingtechniques were used. Laaksonen used a piecewise constanthazard model while Steinbrecher used Cox proportionalhazard model [21,24]. To reduce bias in PAF estimates andaccount for death, Laaksonen [21] suggests using piecewiseover Cox model when the outcome of interest is disease.Partial adjustment method (crude formula)In the four country-specific review studies, the PAF wascalculated from published data of adjusted relative risk(RRadj) using previously published cohort studies andthe prevalence of physical inactivity (prevalence of expos-ure, Pe) was estimated from previously published crosssectional surveys. In the presence of confounding, a popu-lar method of calculating PAF is to use published adjustedRR and estimated prevalence in the crude formula 1 [31]:PAF ¼ Pe RRadj‐1ð ÞPe RRadj‐1ð Þ½  þ 1 100This method is called partial adjustment. Partial ad-justment is a common method when data on all knownconfounders are not available or not measured. However,formula 1 should only be used in the absence of con-founding, because it assumes no confounding of theexposure-outcome association [25]. Four of the countryspecific review studies in this review used formula 1. Inthe presence of confounding another variant formula isrecommended, formula 2 [31]:PAF ¼ Pe RRadj‐1ð ÞRRadj 100Only one global review study [14] used formula 2.Severe confounding bias may occur with partial adjust-ment method, especially formula 1, because the fractionof the outcome that is attributable to the confoundingvariables is not adequately adjusted [36]. For example,one study demonstrated that partial adjustment forconfounding by age and sex yielded a 17% overestimationin PAF [36]. Hence, the full adjustment method thatadjusts for all known confounders is a better choicefor estimating PAF.Adjustment for confoundersIn this review, over-adjustment or under-adjustment(most likely) of known confounders varied explainingsome of the variation in PAF estimates [51]. For instance,adjusting for intermediate variables as confounders canlead to over-estimated or null-biased PAF [52]. There-fore, adjustment should be limited to known evidencebased confounders.Subgroup analysis (sex specific PAF)PAF integrates and is directly related to relative risk(RR) and the prevalence of physical inactivity (Pe) in thepopulation [53]. Thus, for a given RR, different preva-lence estimates for physical inactivity yield different PAFestimates and vice versa in a non linear fashion [53]. Inthis review, one high quality prospective study reportedwidely variable sex specific PAFs for non sport partici-pants [40]: 29% (95% CI: 17, 41) for women and 13%(95% CI: 3, 22) for men. In women, both the RRadj 1.43(95% CI: 1.21, 168) and Pe 66.3% were higher than men:RRadj 1.21 (1.1, 1.35) and Pe 55.2%, respectively. Thiscould explain sex difference observed in PAF estimates.For example, Flegal [36] showed that a small differenceof 3% in age subgroup between the source populationand the target population lead to a 42% overestimationin PAF. In addition PAF is sensitive to minor changes inRR. A difference of 0.20 in RR almost doubled the PAFestimate. This highlights the important of accuratelyquantifying the RR and Pe prior to estimating PAF.Limitations and strengthsThis systematic review did not include a meta-analysisbecause pooling was not appropriate due to the hetero-geneity of studies at conceptual, operational, design andstatistical levels. Study heterogeneity was due in part tothe inclusion criteria for this systematic review. Specific-ally, we included studies that estimated PAF or PARusing modeling on raw data from a prospective cohortdesign or (ii) that used published adjusted relative risk.Further, data from each study on physical inactivity werecollected from different populations using different sam-pling and estimation methods. These differences con-tribute to the wide variation in PAF T2DM attributableto physical inactivity. This is the first systematic reviewthat has ascertained the PAF T2DM attributable to phys-ical inactivity. We believe the results of this systematicreview provide an essential platform for understandingmethodological and statistical reasons that underpincurrent and widely varying PAF estimates. Further, thisstudy provides an initial step toward developing criteria toreport and evaluate PAFs in the future.ConclusionsThe best quality data from this systematic review indi-cate the PAF of T2DM attributable to physical inactivityshould be considered and interpreted by domain and/orsubset of physical inactivity. In the USA, PAFs for sportAl Tunaiji et al. BMC Public Health 2014, 14:469 Page 8 of 9http://www.biomedcentral.com/1471-2458/14/469ranged from 13% (95% CI: 3, 22) to 29% (95% CI: 17, 41):13% (95% CI: 3, 22) in men and 29% (95% CI: 17, 41) inwomen. In Finland, the PAFs for exercise ranged from 3%(95% CI: -11, 16) to 7% (95% CI: -9, 20). The best studydesign for estimating PAF is the prospective cohort. Toobtain the most accurate estimate of PAF the followingneed to be implemented: objective measurement for ex-posure (physical inactivity), objective measurement of out-come (T2DM), full adjustment method that adjusted forall known confounder and a piecewise model.PAF is a valuable statistic in ascertaining burden of adisease due to a specific risk factor from a public healthperspective only when it is accurately calculated using anappropriate study design (i.e., a prospective cohort study).Future studies estimating PAF could reduce the wide vari-ability we currently observe in PAF data by using validand reliable methods to measures physical inactivity andby using consistent ‘best practice’ methodology forreporting PAF [21,54]. Such improvements in study de-sign methodology and consistent cutting edge method-ology will facilitate appropriate and well-informed publichealth decision making choices.Additional filesAdditional file 1: Table S1. Characteristics of studies and outcomemeasure.Additional file 2: Table S2. Summary estimate of prevalence ofexposure (Pe), adjusted relative risk (RRadj), population attributablefraction (PAF) and calculation methods of PAF for physical inactivitydomains.Additional file 3: Table S3 Quality assessment* of the eight studies.Competing interestsThe authors declare that they have no competing interests.Authors’ contributionsHAT and JCD searched for relevant literature and wrote the manuscript.Together with KMK, they conceived the study idea. KMK and DCM helpedwith drafting and revisions. KMK has given the final approval of the versionto be published. All authors read and approved the final manuscript.AcknowledgementsThis work was supported by the CIHR Emerging Teams grant (KK) - Mobilityin Aging (Institute of Aging). JCD is funded by CIHR and MSFHR PostdoctoralFellowships.Author details1Centre for Hip Health and Mobility, University of British Columbia,Vancouver Coastal Health Research Institute (VCHRI), British Columbia,Canada. 2Zayed Military Hospital, Abu Dhabi, United Arab Emirates.3Department of Biomedical Physiology and Kinesiology, Simon FraserUniversity, British Columbia, Canada. 4Centre for Clinical Epidemiology andEvaluation, School of Population and Public Health, University of BritishColumbia, British Columbia, Canada. 5Aspetar - Orthopaedic and SportsMedicine Hospital, Doha, Qatar. 6Department of Family Practice, Faculty ofMedicine, University of British Columbia, British Columbia, Canada. 7Centrefor Hip Health and Mobility, Robert H.N. Ho Research Centre, 769-2635 LaurelStreet, Vancouver, BC V6H 2K2, Canada.Received: 16 October 2013 Accepted: 15 April 2014Published: 18 May 2014References1. Kohl HW 3rd, Craig CL, Lambert EV, Inoue S, Alkandari JR, Leetongin G,Kahlmeier S: The pandemic of physical inactivity: global action for publichealth. Lancet 2012, 380(9838):294–305.2. Blair SN: Physical inactivity: the biggest public health problem of the 21stcentury. Br J Sports Med 2009, 43(1):1–2.3. Colditz GA: Economic costs of obesity and inactivity. Med Sci Sports Exerc1999, 31(11 Suppl):S663–S667.4. Pratt M, Norris J, Lobelo F, Roux L, Wang G: The cost of physical inactivity:moving into the 21st century. Br J Sports Med 2014, 48(3):171–173.5. World Health Organization (WHO): Global recommendations on physicalactivity for health.http://www.who.int/dietphysicalactivity/factsheet_recommendations/en/ (Accessed Jan 13 2014).6. Beunen G: Physical activity, fitness, and health - International Proceedingsand Consensus Statement - Bouchard, C., Shephard, RJ., Stephens, T. Am JHum Biol Offic J Hum Biol Counc 1994, 6(5):675–676.7. Blair SN, Cheng Y, Holder JS: Is physical activity or physical fitness moreimportant in defining health benefits? Med Sci Sports Exerc 2001,33(6 Suppl):S379–S399. discussion S419-320.8. Wild S, Roglic G, Green A, Sicree R, King H: Global prevalence of diabetes:estimates for the year 2000 and projections for 2030. Diabetes Care 2004,27(5):1047–1053.9. Center for disease control and prevention (CDC): Diabetes data and trends.2013. Accessed June 18, 2013.10. Honeycutt AA, Boyle JP, Broglio KR, Thompson TJ, Hoerger TJ, Geiss LS,Narayan KM: A dynamic Markov model for forecasting diabetesprevalence in the United States through 2050. Health Care Manag Sci2003, 6(3):155–164.11. Ettaro L, Songer TJ, Zhang P, Engelgau MM: Cost-of-illness studies indiabetes mellitus. Pharmacoeconomics 2004, 22(3):149–164.12. Shah P: The conomic cost of type 2 diabetes attributable to physicalinactivity in the United States and Texas in 2007. In Texas Medical CenterDissertations (via ProQuest). ; 2010.13. Laaksonen MA, Knekt P, Rissanen H, Harkanen T, Virtala E, Marniemi J,Aromaa A, Heliovaara M, Reunanen A: The relative importance ofmodifiable potential risk factors of type 2 diabetes: a meta-analysis oftwo cohorts. Eur J Epidemiol 2010, 25(2):115–124.14. Lee IM, Shiroma EJ, Lobelo F, Puska P, Blair SN, Katzmarzyk PT: Effect ofphysical inactivity on major non-communicable diseases worldwide: ananalysis of burden of disease and life expectancy. Lancet 2012,380(9838):219–229.15. U.S. Department of Health & Human Services: Physical Activity GuidelinesAdvisory Committee report. In 2008. cited 15 Jan 2012).http://www.health.gov/paguidelines/ (accessed 12 Jan Sept 2014.16. World Health Organization: Political declaration of the High-level Meeting ofthe General Assembly on the Prevention and Control of Non-communicableDiseases. 2011. cited 15 Jan 2012) http://www.who.int/nmh/events/un_ncd_summit2011/en/ (accessed 14 Jan 2014.17. Gill JM, Cooper AR: Physical activity and prevention of type 2 diabetesmellitus. Sports Med 2008, 38(10):807–824.18. Pan XR, Li GW, Hu YH, Wang JX, Yang WY, An ZX, Hu ZX, Lin J, Xiao JZ,Cao HB, Liu PA, Jiang XG, Jiang YY, Wang JP, Zheng H, Zhang H, Bennett PH,Howard BV: Effects of diet and exercise in preventing NIDDM in peoplewith impaired glucose tolerance. The Da Qing IGT and Diabetes Study.Diabetes Care 1997, 20(4):537–544.19. Tuomilehto J, Lindstrom J, Eriksson JG, Valle TT, Hamalainen H, Ilanne-Parikka P,Keinanen-Kiukaanniemi S, Laakso M, Louheranta A, Rastas M, Salminen V,Uusitupa M, Finnish Diabetes Prevention Study Group: Prevention of type2 diabetes mellitus by changes in lifestyle among subjects with impairedglucose tolerance. N Engl J Med 2001, 344(18):1343–1350.20. Knowler WC, Barrett-Connor E, Fowler SE, Hamman RF, Lachin JM, Walker EA,Nathan DM: Reduction in the incidence of type 2 diabetes with lifestyleintervention or metformin. N Engl J Med 2002, 346(6):393–403.21. Laaksonen MA, Harkanen T, Knekt P, Virtala E, Oja H: Estimation ofpopulation attributable fraction (PAF) for disease occurrence in a cohortstudy design. Stat Med 2010, 29(7–8):860–874.22. Liberati A, Altman DG, Tetzlaff J, Mulrow C, Gotzsche PC, Ioannidis JP,Clarke M, Devereaux PJ, Kleijnen J, Moher D: The PRISMA statementfor reporting systematic reviews and meta-analyses of studies thatevaluate health care interventions: explanation and elaboration.PLoS Med 2009, 6(7):e1000100.population attributable fraction for mortality in a cohort study using apiecewise constant hazards model. Am J Epidemiol 2010, 171(7):837–847.doi:10.1186/1471-2458-14-469Cite this article as: Al Tunaiji et al.: Population attributable fraction oftype 2 diabetes due to physical inactivity in adults: a systematic review.BMC Public Health 2014 14:469.Al Tunaiji et al. BMC Public Health 2014, 14:469 Page 9 of 9http://www.biomedcentral.com/1471-2458/14/46923. Higgins JP, Green S: Cochrane handbook for systematic reviews ofinterventions. England: Wiley Online Library; 2008.24. Benichou J: A review of adjusted estimators of attributable risk. StatMethods Med Res 2001, 10(3):195–216.25. Caspersen CJ: Physical activity epidemiology: concepts, methods, andapplications to exercise science. Exerc Sport Sci Rev 1989, 17:423–473.26. Bull FC: Comparative quantification of health risks. Geneva: World HealthOrganization; 2004.27. Caspersen CJ, Powell KE, Christenson GM: Physical activity, exercise, andphysical fitness: definitions and distinctions for health-related research.Public Health Rep 1985, 100(2):126–131.28. Khan KM, Thompson AM, Blair SN, Sallis JF, Powell KE, Bull FC, Bauman AE:Sport and exercise as contributors to the health of nations. Lancet 2012,380(9836):59–64.29. World Health Organization (WHO): Definition and diagnosis of diabetesmellitus and intermediate hyperglycemia. ; 2006.30. Diagnosis and classification of diabetes mellitus. Diabetes Care 2012,35(Suppl 1):S64–S71.31. Rockhill B, Newman B, Weinberg C: Use and misuse of populationattributable fractions. Am J Public Health 1998, 88(1):15–19.32. Daly LE: Confidence limits made easy: interval estimation using asubstitution method. Am J Epidemiol 1998, 147(8):783–790.33. von Elm E, Altman DG, Egger M, Pocock SJ, Gotzsche PC, Vandenbroucke JP:The Strengthening the Reporting of Observational Studies in Epidemiology(STROBE) statement: guidelines for reporting observational studies. Lancet2007, 370(9596):1453–1457.34. Drummond M, Manca A, Sculpher M: Increasing the generalizability ofeconomic evaluations: recommendations for the design, analysis, andreporting of studies. Int J Technol Assess Health Care 2005, 21(2):165–171.35. Prince SA, Adamo KB, Hamel ME, Hardt J, Connor Gorber S, Tremblay M:A comparison of direct versus self-report measures for assessingphysical activity in adults: a systematic review. Int J Behav Nutr PhysActiv 2008, 5:56.36. Flegal KM, Graubard BI, Williamson DF: Methods of calculating deathsattributable to obesity. Am J Epidemiol 2004, 160(4):331–338.37. Janssen I: Health care costs of physical inactivity in Canadian adults. ApplPhysiol Nutr Metab Physiologie Appl Nutr Metab 2012, 37(4):803–806.38. Joubert J, Norman R, Bradshaw D, Goedecke JH, Steyn NP, Puoane T:Estimating the burden of disease attributable to excess body weight inSouth Africa in 2000. South Afr Med J Suid-Afrikaanse Tydskrif VirGeneeskunde 2007, 97(8 Pt 2):683–690.39. Laaksonen DE, Lindstrom J, Lakka TA, Eriksson JG, Niskanen L, Wikstrom K,Aunola S, Keinanen-Kiukaanniemi S, Laakso M, Valle TT, Ilanne-Parikka P,Louheranta A, Hamalainen H, Rastas M, Salminen S, Cepaitis Z, Hakumaki M,Kaikkonen H, Harkonen P, Sundvall J, Tuomilehto J, Uusitupa M: Physicalactivity in the prevention of type 2 diabetes: the Finnish diabetesprevention study. Diabetes 2005, 54(1):158–165.40. Steinbrecher A, Morimoto Y, Heak S, Ollberding NJ, Geller KS, Grandinetti A,Kolonel LN, Maskarinec G: The preventable proportion of type 2 diabetesby ethnicity: the multiethnic cohort. Ann Epidemiol 2011, 21(7):526–535.41. Katzmarzyk PT, Gledhill N, Shephard RJ: The economic burden of physicalinactivity in Canada. CMAJ: Can Med Assoc J J Assoc Med Canadienne 2000,163(11):1435–1440.42. Katzmarzyk PT, Janssen I: The economic costs associated with physicalinactivity and obesity in Canada: an update. Can J Appl Physiol RevueCanadienne Physiol Appl 2004, 29(1):90–115.43. Chen YQ, Hu C, Wang Y: Attributable risk function in the proportionalhazards model for censored time-to-event. Biostatistics 2006, 7(4):515–529.44. Samuelsen SO, Eide GE: Attributable fractions with survival data. Stat Med2008, 27(9):1447–1467.45. Weyer C, Bogardus C, Mott DM, Pratley RE: The natural history of insulinsecretory dysfunction and insulin resistance in the pathogenesis of type2 diabetes mellitus. J Clin Invest 1999, 104(6):787–794.46. Huerta JM, Tormo MJ, Egea-Caparros JM, Ortola-Devesa JB, Navarro C:Accuracy of self-reported diabetes, hypertension and hyperlipidemiain the adult Spanish population. DINO Stud Find Rev Esp Cardiol 2009,62(2):143–152.47. Martin LM, Leff M, Calonge N, Garrett C, Nelson DE: Validation of self-reportedchronic conditions and health services in a managed care population.Am J Prev Med 2000, 18(3):215–218.Submit your next manuscript to BioMed Centraland take full advantage of: • Convenient online submission• Thorough peer review• No space constraints or color figure charges• Immediate publication on acceptance• Inclusion in PubMed, CAS, Scopus and Google Scholar• Research which is freely available for redistribution48. Harris MI, Klein R, Welborn TA, Knuiman MW: Onset of NIDDM occurs atleast 4-7 yr before clinical diagnosis. Diabetes Care 1992, 15(7):815–819.49. Harris MI: Undiagnosed NIDDM: clinical and public health issues. DiabetesCare 1993, 16(4):642–652.50. Harris MI, Hadden WC, Knowler WC, Bennett PH: Prevalence of diabetesand impaired glucose tolerance and plasma glucose levels in U.S.population aged 20-74 yr. Diabetes 1987, 36(4):523–534.51. Porta MD: A dictionary of epidemiology. ; 2008.52. Schisterman EF, Cole SR, Platt RW: Overadjustment bias and unnecessaryadjustment in epidemiologic studies. Epidemiology 2009, 20(4):488–495.53. Powell KE, Blair SN: The public health burdens of sedentary living habits:theoretical but realistic estimates. Med Sci Sports Exerc 1994, 26(7):851–856.54. Laaksonen MA, Knekt P, Harkanen T, Virtala E, Oja H: Estimation of theSubmit your manuscript at www.biomedcentral.com/submit

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.52383.1-0223534/manifest

Comment

Related Items