UPDATE Open Access
S a is ical analysis plan o he con ol o
blood p essu e and isk a enua ion- u al
Bangladesh, Pakis an, S i Lanka (COBRA-
BPS) ial: a clus e andomized ial o a
mul icomponen in e en ion e sus usual
ca e in hype ensi e pa ien s
Mihi Gandhi
1,2,3
, P yseley Nkouibe Assam
1,2*
, Elizabe h L. Tu ne
4,5
, Donald E. Mo isky
6
, Edwin Chan
7
,
Tazeen H. Ja a
5,8*
, on behal o he COBRA-BPS S udy G oup
Abs ac
Backg ound: In u al sou h Asia, hype ension emains a signi ican public heal h issue wi h sub-op imal blood
p essu e (BP) con ol a es. The goal o he ial is o e alua e he e ec i eness and cos -e ec i eness o a
mul icomponen in e en ion (MCI) compa ed o usual ca e on lowe ing BP among adul s wi h hype ension in
u al sou h-Asian communi ies. This a icle desc ibes he s a is ical analysis plan o he p ima y and seconda y
objec i es ela ed o in e en ion e ec i eness based on clinical and pa ien - epo ed endpoin s.
Me hods/Design: The s udy is a clus e andomized ial which will en oll 2550 pa icipan s aged ≥40 yea s wi h
hype ension om u al communi ies in Bangladesh, Pakis an, and S i Lanka. The uni o andomiza ion is a clus e
de ined by 250–300 households. Thi y clus e s, 10 om each coun y, a e andomized in a 1:1 a io o ei he MCI o
usual ca e, s a i ied by coun y and hei dis ance om he clinic. All pa icipan s will be assessed e e y six mon hs
o e a wo-yea pe iod a e baseline wi h measu emen s o sys olic and dias olic BP, an ihype ensi e and s a in
medica ion use, medica ion adhe ence, physical ac i i y le el, an h opome ic pa ame e s, smoking s a us, and
die a y habi s. The p ima y objec i e is o assess he e ec i eness o MCI as compa ed wi h usual ca e in e ms o
mean change in sys olic BP om baseline o inal ollow-up a wo yea s. The p ima y ou come will be modelled
using a gene alized linea mixed-model o epea ed measu es based on a pa icipan -le el analysis. The model will
include clus e andom-e ec s and will use a non-independence esidual co ela ion ma ix o accoun o epea ed
measu es on he same pa icipan . Sensi i i y analyses o he p ima y endpoin will be based on mul iple impu a ion
as well as pa e n mix u e model ipping poin analyses. Seconda y ou comes will be analyzed using he same
modeling app oach as o he p ima y ou come, wi h app op ia e dis ibu ions wi hin he exponen ial amily and
co esponding link unc ions.
(Con inued on nex page)
* Co espondence: [email p o ec ed]; azeen.ja [email p o ec ed]
1
Depa men o Bios a is ics, Singapo e Clinical Resea ch Ins i u e, #02-01,
Nanos, 31 Biopolis Way, Singapo e, Singapo e
5
Duke Global Heal h Ins i u e, Duke Uni e si y, T en Hall, 310 T en D i e,
Du ham, NC, USA
Full lis o au ho in o ma ion is a ailable a he end o he a icle
© The Au ho (s). 2018 Open Access This a icle is dis ibu ed unde he e ms o he C ea i e Commons A ibu ion 4.0
In e na ional License (h p://c ea i ecommons.o g/licenses/by/4.0/), which pe mi s un es ic ed use, dis ibu ion, and
ep oduc ion in any medium, p o ided you gi e app op ia e c edi o he o iginal au ho (s) and he sou ce, p o ide a link o
he C ea i e Commons license, and indica e i changes we e made. The C ea i e Commons Public Domain Dedica ion wai e
(h p://c ea i ecommons.o g/publicdomain/ze o/1.0/) applies o he da a made a ailable in his a icle, unless o he wise s a ed.
Gandhi e al. T ials (2018) 19:658
h ps://doi.o g/10.1186/s13063-018-3022-8
(Con inued om p e ious page)
Discussion: The a p io i s a is ical analysis plan will a oid epo ing bias and da a-d i en analysis o he p ima y and
key seconda y ou comes. The esul s o he s udy will p o ide e idence o he bene i s and isks o he MCI o BP
con ol in u al communi ies in sou h Asian coun ies wi h low- esou ced public heal h in as uc u e.
T ial egis a ion: Clinical ials.go , NCT02657746. Regis e ed on 14 Janua y 2016.
Keywo ds: Clus e andomized ial, Hype ension, S a is ical analysis plan, Blood p essu e
In oduc ion
The con ol o blood p essu e and isk a enua ion- u al
Bangladesh, Pakis an, S i Lanka (COBRA-BPS) ial is a
clus e andomized clinical ial o compa e a mul icom-
ponen in e en ion (MCI) o usual ca e. The o e all goal
is o e alua e he e ec i eness and cos -e ec i eness o
MCI in adul s aged ≥40 yea s wi h hype ension who
eside in u al communi ies o Bangladesh, Pakis an,
and S i Lanka. MCI comp ises he ollowing i e compo-
nen s: (1) home heal h educa ion (HHE) by go e nmen
communi y heal h wo ke s (CHWs); (2) blood p essu e
(BP) moni o ing and s epped-up e e al o a ained
gene al p ac i ione (GP) using a checklis ; (3) ained
public and p i a e p o ide s in managemen o hype en-
sion and using a checklis ; (4) designa ed hype ension
iage coun e and hype ension ca e coo dina o s in
go e nmen clinics; and (5) a inancing model o compen-
sa e o addi ional heal h se ices and p o ide subsides o
low-income indi iduals wi h poo ly con olled hype en-
sion. A de ailed desc ip ion o he COBRA-BPS ial
p o ocol has al eady been published [1]. I con ained a
b ie desc ip ion o he p ima y e ec i eness analysis. The
cu en a icle desc ibes a mo e de ailed s a is ical analysis
plan o he p ima y and seconda y objec i es o he
in e en ion e ec i eness based on clinical and pa ien - e-
po ed ou comes. A he ime o w i ing, no pos -baseline
ou comes o e ec i eness has been analyzed in he ial.
The a p io i s a is ical analysis plan will a oid epo ing bias
and da a-d i en analysis. This a icle does no include an
analysis plan o o he objec i es such as e alua ing cos -e -
ec i eness o he in e en ion and pa ien s’expe ience du -
ing he cou se o in e en ion h ough quali a i e da a.
These opics will be co e ed in sepa a e pape s.
Randomiza ion
The ial is conduc ed in Bangladesh, Tangail Dis ic
(popula ion 3.2 million), and Munshiganj Dis ic (popu-
la ion 1.4 million); Pakis an, Tha a Dis ic (popula ion
1.5 million); and S i Lanka, Pu alam Dis ic (popula ion
1.6 million). The uni o andomiza ion was a clus e
de ined by 250–300 households as de ined by local admin-
is a ion acco ding o CHW ca chmen a ea (each se ed
by 1–2 CHWs). These clus e s we e g ouped in geog aph-
ically con iguous adminis a i e uni s (AUs) as de ined by
he local go e nmen s (12 sub-dis ic s in Tangail, six
subdis ic s in Munshignaj, 30 union councils in Tha a,
12 medical o ice s o he heal h di ision in Pu alam
Dis ic ) such ha each uni is se ed by one go e nmen
clinic. Fi s , in he selec ed dis ic o each coun y, 10
adminis a i e uni s we e delibe a ely sampled, and he
espec i e go e nmen clinic was de e mined. Wi hin each
AU in each coun y, eligible clus e s we e iden i ied (one
clus e is de ined as a illage o Bangladesh, 2–5 neigh-
bo ing illages o Pakis an, and wo G ama Niladha i
[GN] di isions o S i Lanka). Each coun y measu ed he
dis ance o clus e s om he espec i e go e nmen clinic
by a GPS de ice. In each AU, clus e s we e s a i ied in o
wo s a a acco ding o hei dis ance o espec i e
go e nmen clinic: FAR and NEAR (a dis ance o ≤2km
was de ined as NEAR and > 2 km as FAR). In each a m
(usual ca e o MCI), h ee o he i e AUs we e andomly
sampled o be NEAR AUs so ha he emaining wo we e
FAR. Then, one NEAR clus e om each NEAR AU and
one FAR clus e om each FAR AU we e andomly
selec ed o pa icipan ec ui men . A minimum dis ance
o 10 km be ween andomized clus e s we e ensu ed.
Figu e 1shows a schema ic diag am o clus e selec ion
c i e ia. In summa y, andomiza ion was s a i ied by
coun y as well as by he dis ance om he go e nmen
clinic, and AUs (equi alen ly, he sampled clus e s)
we e andomized in a 1:1 a io o ei he MCI o usual
ca e wi hin he six s a a de ined by he combina ion o
coun y and dis ance (NEAR e sus FAR) using a com-
pu e -gene a ed andomiza ion p og am a Singapo e
Clinical Resea ch Ins i u e, Singapo e.
S udy assessmen s
S udy assessmen s a e summa ized in Table 1.Themajo i y
o he assessmen will be pe o med a six-mon hly home
isi s up o wo yea s om he baseline. The accep able
ole ance in he six-mon hly isi s is ± 2 mon hs. All
assessmen s will be pe o med by independen assesso s
(masked o andomiza ion s a us). Mo e de ails o s udy
assessmen o ms, ques ionnai es, and checklis s a e a ail-
able in he published p o ocol [1].
Sample size
The planned sample size is 2550 pa icipan s wi h hype -
ension, co esponding o a a ge sample size o 85
hype ensi e pa icipan s pe clus e o each o he 10
Gandhi e al. T ials (2018) 19:658 Page 2 o 11
clus e s pe coun y in each o he h ee coun ies. Based
on indings om a p e ious easibili y s udy in u al
Bangladesh, Pakis an, and S i Lanka [2], a conse a i e
in aclass co ela ion coe icien (ICC) o 0.02 was
conside ed. Assuming 80% pa icipan e en ion a e
pe clus e a wo yea s a e baseline (68 hype ensi e
pa icipan s pe clus e ) and a wo-sided ype I e o
a e o 5%, he ial will p o ide > 99% powe o he
o e all es o de ec a di e ence be ween a ms in SBP
educ ion as small as 4 (SD 11) mm Hg [3–5]. The
s udy will use 5 mmHg as he clinically meaning ul
di e ence be ween he wo a ms o educ ion in SBP.
Pe aining o he e ogenei y in he in e en ion e ec
among he h ee coun ies, we assume ha < 3 mmHg
di e ence in SBP is no clinically meaning ul. The e o e,
he in e en ion e ec will be conside ed he e ogeneous
among coun ies i he di e ence in SBP educ ion
be ween any wo coun ies is ≥3 mmHg (SD 11) ( o
example, a educ ion in SBP o 3 mmHg in one coun y
and 9 mmHg in he o he wo coun ies; o educ ion in
SBP is 3 mmHg in one coun y, 6 mmHg he in second,
and 9 mmHg in he hi d). The planned sample size p o-
ides > 80% powe o de ec he e ogenei y (as de ined
abo e) in in e en ion e ec s based on he ollowing
assump ions: ICC o 0.02 and ype I e o a e o 0.16%
Fig. 1 Schema ic diag am o clus e selec ion c i e ia in he
andomiza ion Mul icomponen in e en ion clus e ; Usual ca e
clus e ; Go e nmen clinic nea es o he andomized clus e ;
Non- andomized clus e ; Go e nmen clinic ou side 10 km adius
o andomized clus e . Rec angles ep esen adminis a i e uni s.
Do ed line su ounding a clus e ep esen s a 10-km adius o he
clus e . Do ed line su ounding a go e nmen clinic ep esen s a
2-km adium o he clinic. No usual ca e clus e s a e wi hin a 10-km
adius o MCI clus e s. Any usual ca e clus e should no be nea e o
a MCI go e nmen clinic han i s own; simila ly, any MCI clus e
should no be nea e o a usual ca e go e nmen clinic han i s own
Table 1 S udy assessmen s schedule
Assessmen s S udy isi s
Sc eening Baseline 6-mon hly (un il
2 yea s om baseline)
In o med consen ✓
Demog aphics
cha ac e is ics
✓
Sys olic and dias olic blood
p essu es
a
✓✓✓
Cu en blood p essu e
medica ions
✓
Socioeconomic
cha ac e is ics
✓
Medical his o y ✓
Concomi an medica ions ✓✓
Family medical his o y ✓
Tobacco smoking s a us ✓✓
In e na ional physical
ac i i y ques ionnai e
✓✓
Die a y ques ionnai e ✓✓
EQ-5D-5L ques ionnai e ✓✓
b
MMAS-8 o
an ihype ensi e
medica ion adhe ence
✓✓
MMAS-8 o s a ins
adhe ence
✓
Adiposi y measu es (BMI,
wais ci cum e ence)
✓✓
Labo a o y es s ✓✓
b
Se um c ea inine
Fas ing blood glucose
To al choles e ol
High densi y lipop o ein
choles e ol
Low densi y lipop o ein
choles e ol
T iglyce ides
U ine spo albumin
U ine spo sodium
U ine spo c ea inine
Ad e se/se ious ad e se
e en s
✓✓
Dea h ✓
a
Only measu emen s aken by independen assesso s who will be masked o
andomiza ion will be used o he analysis
b
To be assessed only a inal ollow-up isi
Gandhi e al. T ials (2018) 19:658 Page 3 o 11
(based on a Bon e oni adjus men o h ee pai wise
compa isons).
In addi ion o his, he ial has > 80% powe o de ec a
di e ence o 4 mmHg (SD 11) in SBP educ ion be ween
he MCI and usual ca e a ms o each coun y sepa a ely,
o an ICC o 0.02, a wo-sided ype I e o a e o 5%, and
10 clus e s o size 85 pa icipan s pe coun y wi h 80%
pa icipan e en ion a e pe clus e a wo yea s pos
baseline (68 pa icipan s pe clus e ) (Fig. 2). Fu he mo e,
he high powe also ensu es ha he main e ec is
adequa ely powe ed e en a e adjus ing o d opou s.
The e o e, he s udy is no o e -powe ed o he e ogenei y
o d opou s (missing da a). Powe and Sample Size (PASS)
e sion 14 so wa e was used o he powe calcula ions.
Based on ou p e ious wo k in u ban Pakis an, and
expec ing a somewha lowe a i ion a e in u al a eas,
he a i ion a e is likely o be < 15% a he end o
wo yea s in he o e all s udy; he e o e, ou assump ions
abou ollow-up a es a e conse a i e [4].
S udy objec i es
P ima y e ec i eness objec i e
To assess he e ec i eness o MCI compa ed o usual
ca e in e ms o mean change in sys olic blood p essu e
(SBP) om baseline o inal ollow-up a wo yea s pos
baseline.
Seconda y e ec i eness objec i es
To compa e mean change om baseline, o p opo ion,
a he inal ollow-up a wo yea s pos baseline be ween
MCI and usual ca e a m o he ollowing ou comes:
BP con olled o a ge (SBP < 140 mmHg and
dias olic blood p essu e [DBP] < 90 mmHg)
Response (SBP < 140 mmHg and dias olic BP
< 90 mmHg, o ≥5 mmHg educ ion in SBP a wo
yea s om he baseline)
Poo ly con olled BP (SBP ≥160 mmHg o DBP
≥100 mmHg)
DBP
An i-hype ension and s a in medica ions use
An i-hype ensi e medica ion and s a ins adhe ence
sco e assessed by he eigh -i em Mo isky Medica ion
Adhe ence Scale (MMAS-8)
Physical ac i i y sco e assessed by In e na ional
Physical Ac i i y Ques ionnai e (IPAQ)
Ca dio ascula e en s isk sco e assessed by
INTERHEART “Choles e ol”modi iable isk sco e
Heal h- ela ed quali y o li e assessed by 5-Le el
Eu oQol-5 Dimension (EQ-5D-5L) ques ionnai e
Body mass index (BMI)
Wais ci cum e ence
Cu en smoking s a us
F equency o ege ables and ui s in ake pe week
Fig. 2 S a is ical powe o indi idual coun y and o e all s udy a he planned sample size
Gandhi e al. T ials (2018) 19:658 Page 4 o 11
Sal in ake assessed by u ine spo sodium- o-
c ea inine a io and 24-h u ine sodium
Inciden diabe es
Choles e ol le el assessed by o al choles e ol, high
densi y lipop o ein choles e ol, low densi y
lipop o ein choles e ol, and iglyce ides
Es ima ed glome ula il a ion a e (eGFR)
U ine albumin exc e ion
Table 2lis s de ini ions o he abo e ou comes a a
pa icula isi .
Explo a o y e ec i eness objec i es
To compa e mean change in SBP a he inal ollow-up
a wo yea s pos baseline be ween MCI and usual ca e
a ms o he ollowing sub-g oups de ined acco ding o
baseline cha ac e is ics:
Table 2 Ou come de ini ions o e ec i eness objec i es
Ou come Measu emen and de ini ion
Sys olic and dias olic blood p essu e
(SBP/DBP)
Measu ed using a calib a ed au oma ed de ice (Om on HEM-7300 Blood P essu e Moni o ) wi h he
indi idual in he si ing posi ion. Th ee eading a e aken a leas 3 min apa . Mean o las wo eadings
will be used as he inal measu emen .
Blood p essu e con olled o a ge SBP (mean o las 2 o 3 eadings) is < 140 mmHg and DBP (mean o las 2 o 3 eadings) is < 90 mmHg.
Response SBP (mean o las 2 o 3 eadings) is < 140 mmHg and DBP (mean o las 2 o 3 eadings) is < 90 mmHg, o
change in mean o las wo eadings om he mean o las wo eadings om he baseline is ≥5 mmHg.
Poo ly con olled blood p essu e SBP (mean o las 2 o 3 eadings) is ≥160 mmHg o DBP (mean o las 2 o 3 eadings) is ≥100 mmHg.
An i-hype ensi e and s a in
medica ion usage
In o ma ion on any ongoing an i-hype ensi e and s a in medica ions will be classi ied in o one o he
ollowing medica ion classes: Angio ensin II Recep o Blocke o Angio ensin-Con e ing Enzyme Inhibi o ,
Be a Blocke , Calcium Channel Blocke , diu e ics, and s a ins.
Eigh -i em Mo isky Medica ion
Adhe ence Scale (MMAS-8) sco es
Sel - epo ed medical adhe ence is measu ed by he MMAS-8, sepa a ely o an i-hype ensi e medica ion
and s a ins [13–16]. MMAS-8 sco es a e calcula ed by summing all coded answe s.
In e na ional Physical Ac i i y
Ques ionnai e (IPAQ) sco es
To al physical ac i i y sco e (MET-min/week) and ac i i y classi ica ion (Inac i e, Minimally Ac i e, and Highly
Ac i e) a e de i ed acco ding o he IPAQ sco ing guideline [17].
Ca dio ascula e en s isk sco e The INTERHEART “Choles e ol”modi iable isk sco e p o ides a comp ehensi e nume ic assessmen o isk
ac o s o ca dio ascula e en s [18]. The sco e is he sum o poin s o ques ions co esponding o
ca ego ies o hese isk ac o s.
Fi e-le el Eu oQol-5 Dimension
(EQ-5D-5L) ques ionnai e index
The EQ-5D-5L is adminis e ed o assess a pa icipan ’s heal h s a us on he day o assessmen . In addi ion, i
has a isual analogue scale (VAS) measu ing heal h on a scale o 0 (The wo s heal h you can imagine) o
100 (The bes heal h you can imagine) [19].
The EQ-5D-5L index summa izing heal h s a us o he pa icipan s is calcula ed using he EQ-5D-5L alue
se o England [20]. Cu en ly, he e is no alue se a ailable o Bangladesh, Pakis an, o S i Lanka. Howe e ,
i any mo e sui able alue se becomes a ailable be o e he inal da a analysis, i will be used. The EQ-5D-5L
VAS is also conside ed as an addi ional heal h- ela ed quali y-o -li e measu e.
Body mass index Calcula ed as weigh (kg) di ided by heigh
2
(m). Heigh is measu ed using s anda dized Po able
S adiome e (Model SECA 213) in cm wi h g adua ion o 1 mm. Weigh is measu ed using s anda dized
OMRON Digi al Weigh Scale (Model HN-286).
Wais ci cum e ence Measu ed as pe he WHO STEPS p o ocol [21]. The measu emen o wais ci cum e ence is made a he
app oxima e midpoin be ween he lowe ma gin o he las palpable ib and he op o he iliac c es .
Cu en smoking s a us Indi iduals smoking obacco on a daily basis, including ciga e e, pipes, ciga s, che oo s, ciga illos, and wa e
pipe smoking sessions, a e conside ed cu en smoke s.
F ui s and ege ables in ake A die a y ques ionnai e is adminis e ed o collec in o ma ion on die a y habi s ela ed o ui s and
ege ables in ake. A leas one in ake pe week will be conside ed an indica o o each ype o die a y
in ake.
Sal in ake Measu ed in e ms o u ine spo sodium- o-c ea inine a io and 24-h u ine sodium es ima ion by Kawaskai
o mula [22]. (Chemis y analyze [u ine spo sodium]: Beckman Synch on Cx-7 by Ion Elec ode; Regen
[u ine spo sodium]: aluminum silica e; Chemical analyze [u ine spo c ea inine]: Synch on Cx-7/Del a;
Regen [u ine spo c ea inine]: THC2)
Inciden diabe es The use o hypoglycemic agen s o as ing blood glucose ≥126 mg/dL a any ime du ing he wo-yea
ollow-up pe iod o all pa icipan s wi hou p e alen diabe es a en ollmen . (Chemis y analyze : Beckman
Synch on Cx-7/Del a; Reagen : GLUCm)
Choles e ol le el Measu ed in e ms o o al choles e ol, high densi y lipop o ein choles e ol, low densi y lipop o ein
choles e ol, and iglyce ides. (Chemis y analyze : Roche Hi achi 912; Reagen : Roche eagen s)
Es ima ed glome ula il a ion a e Es ima ed using CKD-EPI equa ion [23].
U ine albumin exc e ion Measu ed in e ms o u ine albumin- o-c ea inine a io de ined as a a io o spo u ine albumin di ided by
spo u ine c ea inine exp essed as mg/g. (Chemis y analyze : Beckman Synch on Cx-7/Del a; Regen :
Py ogallol ed plus sodium molybda e)
Gandhi e al. T ials (2018) 19:658 Page 5 o 11

Pa icipa ing coun y (Bangladesh, Pakis an, S i
Lanka)
Clus e dis ance om he p ima y ca e clinic (Fa , Nea )
Gende (Male, Female)
Cu en ly on an i-hype ensi e medica ion (Yes, No)
Poo ly con olled BP (SBP ≥160 mmHg o DBP ≥
100 mmHg)
Socioeconomic le el (Poo , Non-poo )
Addi ional explo a o y analysis may be pe o med o
e alua e he in e en ion e ec o o he ou comes o
he abo emen ioned sub-g oups.
Sa e y objec i es
To compa e he MCI and usual ca e a ms a he inal
ollow-up a wo yea s pos andomiza ion o he ollowing
endpoin s:
P opo ion o pa icipan s who expe ienced any
se ious ad e se e en (SAE)
P opo ion o pa icipan s who expe ienced any SAE
o special in e es (dea h [all cause], hospi al
admission due o co ona y hea disease, hea
ailu e, o s oke)
On e en o an AE, he si e in es iga o will decide i s
ca ego y and sys em o gan class and also e alua e
whe he an AE is an SAE i i leads o o is classi ied in o
one o mo e o ollowing ca ego ies: dea h; li e- h ea ening;
disabili y o pe manen damage; hospi aliza ion (excludes
eme gency oom isi s); p olonga ion o hospi al s ay (≥
24 h); equi ed in e en ion o p e en pe manen impai -
men o damage; o he SAEs. Table 3lis s he p ede ined
ca ego ies and sys em o gan class o AEs.
SAEs epo ed wi h an onse da e be o e he baseline
isi o a e he inal assessmen isi a wo yea s will
no be included in he analysis.
Po en ial co a ia es
The ollowing baseline a iables may be conside ed as
po en ial co a ia es in he suppo i e analyses:
Age (in yea s)
Gende (male, emale)
Educa ion le el (no o mal educa ion, o mal
educa ion)
Ma i al s a us (single [ne e ma ied, di o ced,
sepa a ed, widowed], no single [ma ied])
Socioeconomic le el (poo , middle, high)
BMI (obese/o e weigh [≥23.5 kg/m
2
BMI],
non-obese/no -o e weigh [< 23.5 kg/m
2
BMI]) [6]
Wais ci cum e ence
Diabe es (yes, no)
Ch onic diseases (yes [hea disease, ch onic kidney
disease, s oke], no)
Cu en ly using an an i-hype ensi e medica ion
(yes, no)
Cu en smoking s a us (yes, no)
Physical ac i i y sco e (inac i e/minimally ac i e,
highly ac i e)
Sal in ake
○U ine spo sodium- o-c ea inine a io
○24-h u ine sodium
Choles e ol le el
○To al choles e ol
○High densi y lipop o ein choles e ol
○Low densi y lipop o ein choles e ol
○ iglyce ides
Kidney unc ion
○Es ima ed glome ula il a ion a e (eGRF)
○U ine spo albumin- o-c ea inine a io
Po en ial mode a o s
The ollowing a iables eco ded o e ime may be
conside ed as po en ial mode a o s in he suppo i e
analyses:
BMI
Wais ci cum e ence
Adhe ence o an i-hype ensi e medica ion
Cu en smoking s a us
Physical ac i i y le el
Popula ions
In en - o- ea (ITT) popula ion
The ITT popula ion consis ed o all en olled pa icipan s
wi h a baseline isi (i.e. assessed o he p ima y and
seconda y ou comes). Pa icipan s om he MCI clus e s
will be included in he MCI a m e en i hey did no
ecei e he MCI. Simila ly, pa icipan s om he usual
ca e clus e s will be included in he usual ca e a m e en i
hey a e exposed o he MCI.
Table 3 P ede ined ca ego ies and sys em o gan classes o
ad e se e en s
Ca ego ies Sys em o gan classes
Angioedema and anaphylac ic eac ion
Pe iphe al edema
Hypo ension
Co ona y hea disease
Hea ailu e
S oke o ansien ischemic a ack
Headache, dizziness, o ligh headedness
Flushing
Cough a e ini ia ing an ihype ensi e
Abdominal pain
Muscle pain
Falls and auma
O he
Sys emic eac ions
Ca dio ascula sys em
Ne ous sys ems
Skin and appendages
Respi a o y sys em
Gas oin es inal and
hepa obilia y sys em
O he
Gandhi e al. T ials (2018) 19:658 Page 6 o 11
T ea ed popula ion
The MCI a m includes all en olled pa icipan s who
ha e a ended a leas one in e iew on HHE o is-
i ed ained GPs (i.e. based on eceip o physician’s
managemen checklis ) as a pa o he MCI in e en ion.
The usual ca e a m includes all en olled pa icipan s.
Pa icipan s om an MCI clus e who a e conside ed o
ha e no been “ ea ed”(i.e. ha e no a ended any in e -
iew on HHE and ha e no isi ed a ained GP as a pa
o he MCI in e en ion) will be analyzed wi h he usual
ca e a m.
Pe -p o ocol popula ion
The pe -p o ocol popula ion consis s o all en olled
pa icipan s who do no ha e any signi ican p o ocol
de ia ions (desc ibed below).
Signi ican p o ocol iola ion/de ia ions including hose
which could ha e an impac on he p ima y e ec i eness
measu es, hose which p esen a sa e y isk o he pa ici-
pan s, and/o hose ha a e o e hical conce n will be
iden i ied du ing a blinded da a e iew be o e da abase
lock.
Gene al signi ican p o ocol iola ion/de ia ion c i e ia
a e lis ed below:
Eligibili y de ia ions (included in he s udy despi e
mee ing ollowing c i e ia)
1) W ong hype ensi e diagnosis: nei he has
pe sis en ly uncon olled BP (SBP ≥140 mmHg
o DBP ≥90 mmHg) no on an i-hype ensi e
medica ions.
2) Unde age: aged < 40 yea s.
On-s udy de ia ions (con inued he s udy despi e
mee ing ollowing c i e ia)
1) P egnancy
2) Any majo medical sys emic illness which
p ecludes con inua ion
3) E o in in e en ion assignmen : pa icipan s
ea ed wi h in e en ion a m di e en om he
assigned clus e ’s in e en ion a m.
4) Wi hd aw consen o los o ollow-up
The abo e lis o signi ican p o ocol de ia ion c i e ia
may be ex ended as app op ia e.
All e ec i eness analyses will be pe o med using he
ITT popula ion. Howe e , conside ing he na u e o he
s udy (communi y-based clus e andomized ial wi h da a
collec ion in u al pa s o h ee de eloping coun ies),
he e is a chance ha a small p opo ion o pa icipan s
mayno p o ideany ollow-upda aandhence heywillno
con ibu e o he e alua ion o in e en ion e ec i eness. I
his p opo ion o pa icipan s is sizable, pa icipan dispos-
i ion, and demog aphic and baseline pa icipan cha ac e is-
ics will be analyzed using he ITT popula ion, as well as
excluding pa icipan s who ha e no con ibu ed in he
e ec i eness analyses. The ITT and pe -p o ocol popu-
la ions may be used o addi ional suppo i e analysis
o e ec i eness endpoin s. S udy in e en ion exposu e
and sa e y analyses will be pe o med using he ea ed
popula ion. The ITT popula ion may also be used o
suppo i e sa e y analyses.
S a is ical analyses
Gene al me hods and da a handling ules
All obse ed da a will be included in he ITT popula ion,
wi h he excep ion o da a collec ed ou side he accep able
assessmen window (± 2 mon hs) o each ime poin and
pa icipan s wi h no ollow-up da a. The equency dis i-
bu ion o he e ec i eness ou comes will be e iewed, e.g.
using boxplo s and his og ams. Con inuous a iables wi h
excessi e skewness and/o ku osis will be analyzed using
app op ia e me hods o asymme ic da a o conside ed
o ans o ma ion. All p alues will be wo-sided. A p
alue < 0.05 o he p ima y analysis will be conside ed
s a is ically signi ican , in line wi h he p especi ied le el
used in he sample size calcula ion. All con idence in e -
als (CI) will be a he 95% le el. All s a is ical analyses
will be ca ied ou using SAS so wa e (SAS Ins i u e, NC,
USA). A e he s a is ical plan has been w i en and
signed o , and a e he da abase o he inal analysis has
been locked, he indi idual clus e ’s assigned in e en ion
will be made known o he s udy eam.
T ial p o ile
The numbe o pa icipan s en olled in o he s udy a
sc eening, easons o sc eening ailu e, numbe o pa ici-
pan s en olled, numbe o pa icipan s who comple ed he
baseline and ollow-up isi s, mean and SD o clus e size
a each isi will be summa ized by in e en ion a m using
he CONSORT low cha [7]. The dis ibu ion o baseline
cha ac e is ics will be summa ized by: (1) in e en ion
a m; and (2) coun y and in e en ion a m, wi h desc ip-
i e s a is ics o he ITT popula ion.
In e en ion exposu e
MCI exposu e is e alua ed using he HHE session deli e y
a e, physician e e al a e, and physician’s e alua ion
a e. Table 4de ines hese ideli y measu es. The ideli y
measu es a e es ima ed along wi h co esponding 95% CI
o coun y o he MCI a m based on he ea ed
popula ion. Exposu e and adhe ence o an i-hype ensi e
medica ions is summa ized as seconda y ou comes.
P ima y e ec i eness analysis
This p ima y analysis will be pe o med on he ITT
popula ion. Change in SBP a wo yea s om baseline is
he p ima y ou come. The ou six-mon hly change-
Gandhi e al. T ials (2018) 19:658 Page 7 o 11
om-baseline measu emen s (six mon hs, 12 mon hs, 18
mon hs, and 24 mon hs) om all pa icipan s will be
modelled simul aneously using a likelihood-based gene al-
ized linea mixed-model o epea ed measu es (MMRM)
based on a pa icipan -le el analysis, inco po a ing a clus e
andom-e ec , using Gaussian dis ibu ion and iden i y link
unc ion [8,9]. App op ia e dis ibu ions wi hin he
exponen ial amily and co esponding link unc ions
will be employed o he p ima y ou come in case o
non-no mali y. An uns uc u ed ma ix will be used o
model he esidual a iance-co a iance s uc u e wi hin
pa icipan . I his model ails o con e ge, he e ogeneous
oepli z, he e ogeneous au o eg essi e o o de one, au o-
eg essi e o o de one, and compound symme y s uc-
u es will be conside ed in he speci ied o de o model
he co ela ion be ween ime poin s om he same pa ici-
pan . MMRM accoun s o missing da a and is alid unde
he missing a andom (MAR) assump ion.
All p ima y analysis models will include ixed e ec s
o baseline SBP, coun y, indica o o dis ance om
clinic ( a o nea ), age, gende , in e en ion a m, isi
numbe , and he in e en ion a m-by- isi numbe in e -
ac ion. The p ima y ou come o in e es a wo yea s
om baseline will be es ima ed wi h co esponding 95%
CIs using he app op ia e con as a he inal isi . We
will employ es ic ed/ esidual maximum likelihood wi h
he be ween-wi hin app oxima ion o deg ee o eedom
es ima ion [10].
Suppo i e e ec i eness analyses
A MMRM model will be pe o med o SBP simila o
he p ima y analysis men ioned abo e including in e ac ions
o coun y, in e en ion g oup, and ime o de e mine
whe he he e ec s di e by coun y. I he in e ac ion e ec
is ound o be clinically meaning ul, a MMRM model simila
o he p ima y analysis model will be pe o med sepa a ely
o each coun y. The coun y-speci ic analyses will use
Bon e oni co ec ed p alues and CIs o e alua ing
in e en ion e ec i eness. Fu he analysis will be pe -
o med simila o he p ima y analysis wi h each po en ial
con ounde a a ime as ixed e ec (sepa a e model o
each con ounde ) as well as all (o selec ed) po en ial
con ounde s a he same ime as ixed e ec s (single
model including mul iple con ounde s). The inal lis o
con ounde s will be decided conside ing s a is ical signi i-
cance (p< 0.1), e ec size, and clinical impo ance. A
MMRM model will be pe o med o SBP simila o he
p ima y analysis men ioned abo e, including each mode -
a o a a ime as ixed e ec as well as all mode a o s a
he same ime as ixed e ec s using he ITT popula ion.
Fu he suppo i e analysis may be pe o med using he
ITT and pe -p o ocol popula ions.
Seconda y e ec i eness analyses
Seconda y e ec i eness ou comes will be analyzed using
a simila s a egy applied o he p ima y endpoin . Tha
is, using a MMRM wi h an uns uc u ed a iance-co a i-
ance ma ix. App op ia e dis ibu ions wi hin he expo-
nen ial amily and co esponding link unc ions will be
employed o each ou come. Analysis o inciden diabe es
( o hose wi hou p e alen diabe es a baseline), sal in-
ake, and choles e ol le el will be based on only wo
ime-poin s (baseline and wo yea s) using simila MMRM
models. All seconda y analyses will be pe o med on he
ITT popula ion. Suppo i e analyses may be pe o med
using he ea ed and pe -p o ocol popula ions. Addi ional
suppo i e analyses may be pe o med o e alua ed in e -
en ion e ec a e adjus ing o po en ial con ounde s.
Explo a o y e ec i eness analyses
A MMRM model will be pe o med o SBP simila o
he p ima y analysis men ioned abo e including a ixed
e ec e m o cu en ly an i-hype ensi e medica ion
(yes/no) a baseline and i s in e ac ions wi h in e en ion
g oup and ime o de e mine whe he he in e en ion
e ec di e s by an i-hype ensi e medica ion use s a us
a baseline. Simila analysis will be pe o med o clus e
dis ance om he p ima y ca e clinic, gende , poo ly
con olled BP s a us, and socioeconomic s a us a base-
line. I he in e ac ion e ec is ound o be clinically
meaning ul, a MMRM model simila o he p ima y
analysis model will be pe o med sepa a ely o each
sub-g oup. These analyses will be pe o med on he ITT
popula ion.
Table 4 In e en ion ideli y measu es
Fideli y measu e De ini ion
Home heal h educa ion (HHE)
session deli e y a e
Calcula ed as he o al numbe o h ee-mon hly HHE sessions deli e ed a he household le el using he
Communi y Heal h Wo ke s Moni o ing and Home Heal h Educa ion Checklis di ided by he o al numbe o
planned HHE sessions un il he s udy discon inua ion/comple ion, mul iplied by 100.
Physician e e al a e Calcula ed as he o al numbe o imes hey a e e e ed o ained physicians by CHWs using he Gene al
P ac i ione Re e al Checklis di ided by he o al numbe o imes pa icipan s iden i ied wi h ha ing poo ly
con olled BP (SBP ≥160 mmHg o DBP ≥100 mmHg) du ing s udy isi s un il he s udy discon inua ion/comple ion,
mul iplied by 100.
Physician’s e alua ion a e Calcula ed as he o al numbe o imes hey a e e alua ed by ained physicians using he Gene al P ac i ione
Managemen Checklis di ided by he o al numbe o imes pa icipan s iden i ied wi h ha ing poo ly con olled BP
du ing s udy isi s un il he s udy discon inua ion/comple ion, mul iplied by 100.
Gandhi e al. T ials (2018) 19:658 Page 8 o 11
Handling o missing da a in he e ec i eness analyses
The p ima y analysis is planned wi h no impu a ion o
missing da a. The p ima y analysis, based on a likelihood-
based MMRM, is alid unde he MAR assump ion [11].
The MAR assump ion means ha missingness is inde-
penden o he unobse ed ou come alues a e accoun -
ing o he app op ia e obse ed da a and co a ia es in he
model.
In o de o e alua e he obus ness o he indings o
he MAR assump ion, sensi i i y analyses will be pe -
o med unde a ying assump ions o da a conside ed
likely o be missing unde MAR, as well as missing no
a andom (MNAR). MNAR means ha missingness de-
pends on he unobse ed alues and canno be p edic ed
solely based on he pa icipan ’s obse ed da a. Se e al
ypes o s a is ical models ha e been p oposed o analyze
clinical s udy da a unde such assump ions. We will use
wo possible app oaches o e alua e he impac o
missingness.
The i s app oach ha will be implemen ed o his
s udy is he use o mul iple impu a ions and MMRM o
he p ima y ou come. Using MI da a o change in SBP
pos baseline (six mon hs, 12 mon hs, 18 mon hs, and
24 mon hs om baseline measu emen s) om all pa ic-
ipan s will be modelled simul aneously using MMRM
model same as he p ima y analysis.
I is expec ed ha he majo i y o missing da a will be
caused by pa icipan s discon inuing om he s udy
p ema u ely. The esul ing missing da a will ha e a
mono one pa e n, meaning ha once a pa icipan has
missing da a o a isi , da a will be missing o all subse-
quen isi s. I is also expec ed ha a small amoun o
non-mono one missing da a (when pa icipan s skip
in e media e isi s bu e u n o e alua ions a subse-
quen isi s) will be p esen . The in e mi en missing
da a will be impu ed using he Mon e Ca lo Ma ko
Chain (MCMC) me hod o mul iple impu a ion be o e
he impu a ion o he mono one missing da a [8].
The second app oach is he use o pa e n mix u e
models (PMMs) and mul iple impu a ions, which will be
implemen ed i he amoun o missingness on he p ima y
ou come is > 25% o he di e ence in pe cen age o missing
da a be ween he in e en ion a ms is ≥15% [8]. PMMs
ha e he ad an ages o allowing anspa en and clinically
in e p e able o mula ions o heassump ions ega ding
unobse ed da a [11,12]. PMMs wi h del a (δ)adjus men s
will be used and impu a ions will be based on an MNAR
clinical assump ion ha pa icipan s om he MCI a m
who discon inue a a gi en ime poin would ha e, on a e -
age, hei unobse ed e icacy sco e wo se by some
amoun , δ, compa ed wi h he obse ed e icacy sco e o
pa icipan s who con inue o he nex assessed ime poin .
Fo pu poses o he sensi i i y analyses, pa icipan s who
discon inue om he usual ca e a m a e ea ed as i hey
would ha e exhibi ed he same e olu ion o he disease and
same bene i om in e en ion wi h usual ca e as pa ici-
pan s ha s ayed on he s udy. Del a alues will be based
on he es ima ed in e en ion e ec aken om he p ima y
analysis in he ITT popula ion whe e alues will a y om
0 o he es ima ed in e en ion di e ence in inc emen s o
0.5 mmHg, so ha one can assess a which poin he s udy
conclusions change om a o able o un a o able, ha is,
so ha onecan inda ippingpoin .Thesewillbebasedon
10 impu a ions δ- alue. The magni ude o he ipping poin
will hen be in e p e ed clinically and he obus ness o he
s udy conclusions o missingness e alua ed.
Sa e y analysis
All on-s udy SAEs epo ed on SAE epo ing o ms will
be summa ized by in e en ion a m, using he ea ed
popula ion. The pe cen age and equency o pa icipan s
who e e epo ed each ype o SAE and SAE o special
in e es along wi h sys em o gan class will be abula ed. A
simila sa e y analysis may be pe o med on he ITT
popula ion. All dea hs, oge he wi h easons, will be
summa ized using coun s by in e en ion a m based on
he ITT popula ion.
In e im analyses
Planned in e im sa e y analyses a e o occu a e e y
six mon hs om he s a o he s udy. The in e im ana-
lyses will summa ize pa icipan baseline cha ac e is ics
and on-s udy sa e y da a by andomized g oup, as well as
pooled o e all andomized g oups, o he ea ed popula-
ion. The by andomized g oup analysis is e iewed by an
independen da a sa e y and moni o ing boa d. The s udy
eam, excep s a is icians in ol ed in pe o ming he
in e im analyses, is blinded o hese esul s. The pooled
analysis (combined o andomized g oups) is p esen ed o
he s udy eam.
Discussion
The COBRA-BPS ial is a p agma ic, mul i-coun y,
clus e andomized, con olled ial o a MCI wi h po en-
ial o implemen i on a wide scale na ionally in he
pa icipa ing coun ies and beyond. The s udy aims o
e alua e he bene i s o he MCI and moni o po en ial
sa e y conce ns o BP con ol in he u al communi ies
in sou h-Asian coun ies wi h low- esou ced public heal h
in as uc u es.
The clus e andomized s udy design is a p agma ic
s udy o mimic how he p oposed in e en ion can be
ollou p ima ily using he exis ing in as uc u e in
coun ies wi h di e en ypes o heal hca e sys ems and
a ailabili y o esou ces. The e o e, i helps o e alua e
no jus he e ec i eness o he in e en ion in he
eal-li e se ing bu also i s easibili y in implemen a ion
wi h an es ima e o equi ed esou ces, ime and cos s.
Gandhi e al. T ials (2018) 19:658 Page 9 o 11