Skip to main content
. 2011 Feb 15;52(4):493–506. doi: 10.1093/cid/ciq167

Table 2.

Reported Associations with Pharmacy-Based Adherence Measures (PAMs) in Low- and Middle-Income Countries

Study (year) Design Type of care Region ART naive ART regimen (%)a PAM category PAM definition in study PAM monthsb Sample size, no. of persons Key findingsc
Nachega et al [13] (2006) Retrospective cohort Private Sub-Saharan Africa (multiple countries) Yes NNRTI (82), PI MPR Months ART claims submitted (entire regimen)/months from start to death, withdrawal or censor Variable;
median, 22
6288 1. PAM <80% predicted death and death + LTFU (P < .01)
2. compared with PAM adherence of 100%, decreasing PAM strata increasingly predicted death (P < .01), except for PAM adherence of 80%–99%
MPR Months ART claims submitted (entire regimen)/Months in the interval 12 (0–12) 3267 PAM <80% in first 12 months predicted death (P < .01)
Weidle et al [26] (2006) Clinical trial Home based Uganda Yes NNRTI (100) PCf (Days 3TC delivered - days 3TC returned)/days in the interval 3 (3–6)
3 (9–12)
913
894
1. PAM <95% predicted VFd at 6 or 12 months (P < .01)
2. self-report predicted VF at 12 (P < .05) but not 6 months
PC (3TC Pills delivered – 3TC pills returned)/3C pills delivered 3 (3–6)
3 (9–12)
913
894
PAM <95% predicted VFd at 6 or 12 months (P < .05)
Nachega et al [4] (2007) Retrospective cohort Private Sub-Saharan Africa (multiple countries) Yes NNRTI (100) MPR Months ART claims submitted (all ARVs)/months from start to death/leaving/censor Variable
median, 26
2821 PAM strata >50% increasingly predicted sustained VL suppression (P < .01), shorter time to VL suppression (P < .05), and increased time to viral rebound e (P < .05)
Bisson et al [17] (2008) Retrospective cohort Private Sub-Saharan Africa (multiple countries) Yes NNRTI (100) MPR Months ART claims submitted (all ARVs)/months from start to study endpoint 6 (0–6)
12 (0–12)
958
872
1. PAM <90% predicted VFc at 6 and 12 months (P < .01)
2.it was better than changes in the CD4 cell count at predicting VFd at 6 and 12 months (P < .01)
Variable
median, 20
1101 1. PAM <90% predicted viral rebounde (P < .05)
2.not different than changes in the CD4 cell count from maximum on-treatment value in predicting viral rebounde
3 (0–3)
3 (6–9)
958
872
PAM was no better than changes in the CD4 cell count over first 6 or 12 months in predicting VFc at 6 or 12 months
Bisson et al [18] (2008) Case-control Public Botswana No NNRTI (100) PC Sum of (days ART prescribed – remnant days ART) between last and 3 prior fills/days between last and 3 prior fills 3 (varied) 302 1. Decreasing PAM rates (90%-95%, 80%-90%, and <80%
2. (P < .05) and PAM <95% (P < .01) in 3 months prior to recruitment predicted VF,d compared with PAM >95%
Goldman et al [23] (2008) Retrospective cohort (all clinical or IF) Public Zambia Yes NNRTI (100) MPR 100% - [(days late to pharmacy visits – 3)/days on ART] g Variable
Median, 24
913 1. Lower PAM (<80%, 80%-94%, and >95%) more likely to predict VFd at time of the VL test (P < .05)
2. Self-reported adherence did not predict VFd
San Lio et al [25] (2008) Prospective cohort NGO, free Mozambique No NNRTI (100) PC (Days pills prescribed – days pills returned)/days between appointments 12 (varied) 394 PAM <95% predicted VFh after 12 months of follow-up (P < .05)
Toure et al [20] (2008) Retrospective cohort Public, private and NGO Cote d'Ivoire Yes NNRTI (96), PI, 3NRTI MPR Days ART given to patient/days since ART start to last visit, or censor if last visit was after censor date Variable
median, 8
10211 1. PAM <80% predicted increases in the CD4 cell count of <50 cells after 6 months (P < .01)
2. PAM <80% predicted LTFU (P < .01) but not death over a period of 16 months
Chi et al [19] (2009) Retrospective cohort Public Zambia Yes NNRTI (100) MPR 100% - [(days late to pharmacy visits – 3)g/days on ART] 12 (0–12) 27115 1. PAM <80% predicted lower CD4 cell counts after 18–36 months (P < .01)
2. decreasing PAM adherence rates (>95%, 80%-94%, and <80%) predicted LTFU after 12-36 months of ARTh (P < .01) PAM <80% predicted death (P < .01) at 12–36 months but higher strata (80%-94% and >95%) did not
Danel et al [22] (2009) Clinical trial (one or both of VF or IF) Free Cote d'Ivoire Yes NNRTI (87), PI MPR Days ART delivered/days in the interval 6 (0–6) 208 PAM of >90% did not predict CD4 cell counts of >350 cells/μL plus VL suppression at 36 months
30 (6–36) 208 PAM >90% predicted either or both of the following: a CD4 cell count >350 cells/μL and VL suppression at 36 months (P < .01)
Rougemont et al [24] (2009) Prospective cohort Private Cameroon Yes NNRTI (99), PI PPU “Nonadherent was defined as being >2 weeks late to pick-up medication or as “abandoned ART” on phone tracing 6 (0–6) 194 ”Nonadherent” status predicted VFd (P < .01); no different than CD4 cell count change over 6 months at predicting VFd; day 30 Self-reported adherence did not predict 6-month VFd
Ross-Degnan et al [21] (2010) Retrospective cohort Public, private, and NGO Sub-Saharan Africa (multiple countries) Yes NNRTI, PI (NR) MPR Days with ART/days since ART start Variable
median, 6
409 1. PAM <80% (but not 80%-90% or 90%-100%) predicted lower CD4 cell counts (at 4-9 months), compared with PAM 100% (P < .05)
2. PAMs were not directly compared with self-reported adherence
MPR ”Nonadherent” was defined as >30 days without receipt of ART Variable
Median, 6
409 “Nonadherent” status predicted lower CD4 cell counts at 4-9 months (P < .05)

NOTE. ART, antiretroviral therapy; ARV, antiretroviral; IF, immunological failure; LTFU, lost to follow-up; MPR, medication possession ratio; NGO, nongovernmental organization; NNRTI, nonnucleoside reverse-transcriptase inhibitor; NR, not reported; PC, pill count; PI, protease inhibitor; 3NRTI, triple nucleoside reverse-transcriptase inhibitor; 3TC, lamivudine; VF, virological failure; PPU, pill pick-up; VL, viral load.

a

Data are ART regimens for that study. Number in parentheses represents percentage of subjects receiving the predominant regimen.

b

Duration of adherence assessment, with the months over which assessed in parentheses. If there was a variable duration of adherence assessment, than the median, mean, or range is listed.

c

Number after PAM is the percentage adherence.

d

Single viral load above threshold.

e

Single viral load above threshold after previous VL suppression.

f

Because remnant pills were counted to determine adherence, this measure comes under the PC category despite being referred to as medication possession ratio in the study.

g

Subjects not late to pharmacy visit until after 3 days, to account for routine provision of 3 days extra ART.

h

Statistical significance for association was not reported, so we determined statistical significance using raw data with the χ2 test.