Factors associated with non-completion of follow-up: 33-month latent tuberculous infection treatment trial

SUMMARY

SETTING:

A post-hoc exploratory analysis of a randomized, open-label clinical trial that enrolled 8053 participants from the United States, Canada, Brazil, and Spain.

OBJECTIVE:

To assess factors associated with non­completion of study follow-up (NCF) in a 33-month latent tuberculous infection treatment trial, PREVENT TB.

DESIGN:

Participants were randomized to receive 3 months of weekly directly observed therapy vs. 9 months of daily self-administered therapy. NCF was defined as failing to be followed for at least 993 days (33 months) from enrollment. Possible factors associated with NCF were analyzed using univariate and multivariate regression via Cox proportional hazard model.

RESULTS:

Of 7061 adults selected for analysis, 841 (11.9%) did not complete study follow-up. Homelessness, young age, low education, history of incarceration, smoking, missing an early clinic visit, receiving isoniazid only, and male sex were significantly associated with NCF. Similar results were found in the North American region (United States and Canada) only. In Brazil and Spain, the only significant factor was missing an early clinic visit.

CONCLUSIONS:

Study subjects at higher risk for NCF were identified by characteristics known at enrollment or in early follow-up. Evaluation of follow-up in other trials might help determine whether the identified factors consistently correlate with retention.

AN IMPORTANT GOAL of every clinical trial is to achieve completion of follow-up, which is especially challenging in studies with long follow-up, such as those required for tuberculosis (TB). Failure to retain participants results in smaller sample size, fewer outcomes, and a potentially under-powered study. It can also introduce bias, and impair the validity and generalizability of the results.^1–5

Extensive research has focused on factors associated with non-completion of treatment. However, less research has been devoted to reasons for non­completion of follow-up (NCF). Such information is particularly important for trials of preventive treatment for TB, in which endpoints (e.g., development of TB) are rare and follow-up is lengthy.

Differences among participants who do vs. those who do not complete the protocol-defined follow-up period may include demographic, clinical, social, and behavioral characteristics.^6,7 Among 1075 participants enrolled in a TB treatment trial at US and Canadian sites, factors associated with NCF were cultural or linguistic and lack of stable housing.¹ In a 3-year TB prevention trial in Botswana among human immunodeficiency virus (HIV) infected adults, men were less likely to complete treatment and follow-up.²

The purpose of the present analysis was to identify factors associated with NCF in a 33-month clinical trial, the PREVENT TB study.

STUDY POPULATION AND METHODS

We performed a post-hoc exploratory analysis based on a plan developed after completion of the PREVENT TB trial, a Phase III open-label, randomized trial of latent tuberculous infection (LTBI) treatment. Participants infected with Mycobacterium tuberculosis who were at high risk of developing TB were enrolled from June 2001 through February 2008 at 28 sites in low (the United States, Canada, and Spain) and high (Brazil) TB burden countries.^8,9 Participants aged <18 years were excluded from this analysis because completion of follow-up might reflect characteristics of the parents rather than the child.

Participants were assigned to either 12 directly observed doses of once-weekly isoniazid (INH) (900 mg) and rifapentine (RPT) (900 mg) (3HP-DOT) or 9 months of daily self-administered INH (300 mg) (9HSAT), following unrestricted randomization, stratified by enrolling site and by participant HIV status. Within household clusters of close contacts, randomization was performed for the first participant enrolled; subsequent participants from the same household received the same regimen. Participants were evaluated monthly during treatment. Those receiving the 3HP-DOT regimen had scheduled visits for evaluation at weeks 4, 8, and 12, while participants receiving the 9H-SAT regimen had nine monthly visits during treatment. All participants were evaluated every 3 months after completion or discontinuation of treatment until month 21, then every 6 months until month 33 of the trial. The assessment at each visit included weight, signs and symptoms of TB, any anti-tuberculosis drug received, and new diagnosis or hospitalization.¹⁰

Adherence to the assigned regimen was assessed at each visit: by observed treatment records for 3HPDOT, and by pill count and self-report for 9H-SAT. This trial compared the effectiveness of the combined regimen with that of 9H-SAT after participants had been followed for 33 months from enrollment. Enrollment started at 22 North American sites in 2001; two sites were added in 2002 and one each in 2003 and 2008. Non-North American sites (in Brazil and Spain) were added in 2003 and 2004.

All participants provided written informed consent. Institutional review boards at the Centers for Disease Control and Prevention (CDC), Atlanta, GA, USA, and at the participating clinical sites, approved the study protocol.

Factors possibly associated with NCF were organized as demographic (age, sex, race, ethnicity, country of origin, education, unemployment, homelessness, history of incarceration, and need for an interpreter), clinical (body mass index, HIV status, cirrhosis [by self-report], methadone treatment, and potential for pregnancy [defined below]), social (alcohol consumption [defined below], injection drug use ever, and current smoking at the time of enrollment), behavioral (missing an early clinic visit [defined below]), regimen assigned, and enrolling site. Potential for pregnancy was categorized as confirmed pregnancy, participants for whom pregnancy was not possible (women aged >51 years or men), or women of reproductive age (between 18 and 51 years of age). Alcohol use was defined as an affirmative response to a question asking whether the participant ever drank alcoholic beverages. Alcohol abuse was defined as a score of 72 on the CAGE questionnaire.¹¹ Participants were considered to have missed an early visit on the 3HP-DOT regimen if they missed at least one of the first three observed doses, followed by receiving an observed dose at any time during the treatment period or, on the 9H-SAT regimen, if they missed at least one of the first three monthly visits, followed by a monthly visit at any time during the treatment period.

Non-completion of a 3HP-DOT or 9H-SAT regimen was defined as not taking at least 11 of 12 doses within 10–16 weeks for 3HP-DOT or at least 240 of 270 doses within 35–52 weeks for 9H-SAT. Participants were encouraged to continue study follow-up after completion or discontinuation of treatment. Non-completion of study follow-up was defined as failing to be followed for at least 993 days (33 months [equal to 9-month regimen plus 2 years of follow-up]) from enrollment, if TB or death did not occur before that time. Follow-up evaluations were conducted by telephone, e-mail, regular mail, or in person. Participants who were followed by e-mail or regular mail received a questionnaire with standard study follow-up questions. Follow-up continued for participants who moved or were incarcerated. An in-person visit was scheduled if the participant reported symptoms of TB at any time during the study, or reported symptoms of an adverse event within 60 days after the last study drug dose, and for the final study visit. Participants who completed at least 993 days of follow-up, developed TB, or died, were considered completers of follow-up. The analysis was conducted using SAS for Windows, Version 9.3 (Statistical Analysis System, Cary, NC, USA).

Proportions of NCF were calculated for all potential factors outlined. Age was stratified above and below the median age of the study population. We evaluated clinical and demographic characteristics of the combined cohorts, the North American (United States and Canada) and non-North American (Brazil and Spain) cohorts, and each cohort separately. We calculated hazard ratios (HRs), 95% confidence intervals (CIs), and P values for potential factors associated with NCF. The proportional hazard assumption was evaluated for each factor using a Kaplan-Meier graph, and by performing the proportionality test of binomial variables. We evaluated collinearity of all factors with P ⩽ 0.2 in the univariate analyses using the proc reg procedure in SAS (condition index around 10). Interaction terms were selected based on expert opinion (Appendix).^* A final Cox proportional hazard model was defined in which backward elimination was performed at P < 0.05.

Duration of retention was divided into three equal periods: 1–331, 332–662, and 663–993 days. Proportions of clinical and demographic characteristics in these three groups, as well as time of follow-up in person-months, were calculated separately.

We used Pearson’s correlation coefficient and a linear regression model to evaluate the relationship between the proportions of NCF and participants enrolled by site. Univariate and multivariate analyses of the North American cohort were also performed among the first participants enrolled of a cluster and among participants who completed at least 80% of the study follow-up (at least 793 days from enrollment). The latter was intended to evaluate whether early vs. late loss to follow-up differed between study regimens. Most sites provided compensation for study participants for the time and effort required to participate; however, the methods and amounts were too variable for a meaningful analysis of their influence on completion of study follow-up.

RESULTS

Of 8053 participants enrolled in the PREVENT TB trial, 992 were excluded from the analysis: 706 were aged <18 years, 259 were determined ineligible after enrollment, and 27 were unable to complete follow-up due to closure of the enrolling site. The remaining 7061 participants made up the cohort selected for analysis; 841 (11.9%) did not complete study follow-up. NCF was higher among participants who did not complete treatment (419/1601, 26.2%) than among those who completed the assigned regimen (422/5460, 7.7%) (P < 0.001) (Figure and Table 1). No clear violation of the proportional hazard assumption was found. The univariate analysis showed that receiving 9H-SAT (9H-SAT, 448/3418 [13.1%] vs. 3HP-DOT, 393/3643 [10.8%]), young age, male sex, Hispanic ethnicity, low education, history of incarceration, homelessness, alcohol abuse, injection drug use, current smoking at the time of enrollment, enrolling site, and missing an early clinic visit were associated with NCF. In a multivariate logistic regression analysis, after controlling for several factors, we found that the hazard of NCF after completion or discontinuation of LTBI treatment was 1.9 times higher among participants who experienced homelessness; 1.4 times higher among participants aged <37 years, those with <9^th grade education, those with a history of incarceration, those who reported tobacco smoking at the time of enrollment, and those who missed an early clinic visit; and 1.2 times higher among participants receiving 9H-SAT than among those receiving 3HP-DOT, and among males. Enrolling site was statistically significant, indicating variability across the sites even after controlling for other factors (P < 0.001) (Table 2; interaction terms are evaluated in Appendix Table A.1).

Figure.

Flow chart of participants evaluated for factors associated with non-completion of study follow-up in a 33-month Clinical Trial of Latent Tuberculosis Infection Treatment, the PREVENT TB Trial. This figure shows the number of participants who were enrolled in the trial, the groups that were excluded from the analysis, and the remaining cohort that was evaluated for factors associated with non-completion of study follow-up, by region. Inclusion criteria were as follows: 1) close contact with a culture-confirmed PTB patient (within 2 years before enrollment) and TST-positive; 2) recent conversion to TST positivity; 3) HIV infection plus TST-positive or recent close contact with a culture-confirmed PTB patient regardless of TST results; or 4) TST-positive with fibrotic changes on chest radiograph consistent with previously untreated TB. Exclusion criteria were as follows: 1) current confirmed TB; 2) suspected TB; 3) TB resistant to INH or RMP in the source case; 4) history of treatment for >14consecutive days with a rifamycin or >30 consecutive days with INH during the previous 2 years; 5) documented history of completing adequate treatment for active TB or latent M. tuberculosis infection in an HIV-negative person; 6) history of sensitivity/intolerance to INH or rifamycins, 7) serum AST >5 times the upper limit of normal if AST was determined, 8) pregnant or lactating females (not excluded in this cohort for analysis), 9) persons currently receiving or planning to receive HIV-1 therapy within 90 days of enrollment, or 10) weight <10.0 kg. PREVENT TB Trial reasons for ineligibility: source case INH/RMP-resistant (50%), source case culture-negative (31%), other (19%). This analysis includes participants from all enrolling sites: the United States, Canada, Brazil and Spain. * This analysis considers only adults aged ⩾ 18 years. ^†Microbiological and non-microbiological ineligibles. ^‡Participants who did not have the opportunity to complete at least 993 days of study follow-up due to closure of enrolling site; clustering below and above 900 days: 59.3% (16/27) were followed in the trial between 687 and 889 days and 40.7% (11/27) between 915 and 991 days. PTB = pulmonary tuberculosis; TST = tuberculosis skin test; HIV = human immunodeficiency virus; INH = isoniazid; RMP = rifompicin; AST = aspartate immunotransferase.

Table 1.

Completion of follow-up status by completion of treatment in the North American and non-North American regions combined^*

	Did not complete treatment (n = 1601)		Completed treatment (n = 5460)
	Did not complete follow-up (n = 419)	Completed follow-up (n = 1182)	Did not complete follow-up (n = 422)	Completed follow-up (n = 5038)
Regimen	n (%)	n (%)	n (%)	n (%)
3HP-DOT (n = 3643)†	155 (37.0)	485 (41.0)	238 (56.4)	2765 (54.9)
9H-SAT (n = 3418)†	264 (63.0)	697 (59.0)	184 (43.6)	2273 (45.1)
Total (n = 7061)	419 (26.2)‡	1182 (73.8)‡	422 (7.7)§	5038 (92.3)§

^*Proportion of non-completion of follow-up between those who did not complete treatment (419/1601, 26.2%)vs. those who completed treatment (422/5460, 7.7%) was statistically significant (P < 0.001).

^†3HP-DOT = 3 months of directly observed once-weekly RPT (maximum dose, 900 mg) plus INH (maximum dose, 900 mg); 9H-SAT = 9 months of daily self­administered INH (maximum dose, 300 mg). Participants had the opportunity to continue study follow-up after completion or discontinuation of treatment.

^‡The denominator for the percentage is all participants who did not complete treatment (n = 1601).

^§The denominator for the percentage is all participants who completed treatment (n =5460).

H, INH = isoniazid; P, RPT = rifapentine; DOT = directly observed therapy; SAT = self-administered therapy.

Table 2.

Univariate and multivariate analyses of factors associated with non-completion of study follow-up (n = 841) compared to those who completed study follow-up (n = 6220) in the North American and non-North American regions combined: Cox proportional hazard model (n= 7061)

	Non-completion of follow-up (%)	Multivariate
		HR (95%CI)	aHR (95%CI)
Regimen*
3HP-DOT (n = 3643)	10.8	Reference	Reference
9H-SAT (n = 3418)	13.1	1.2 (1.07–1.41)	1.2 (1.04–1.37)
Age, years (median = 37)
<37 (n = 3413)	13.5	1.3 (1.15–1.50)	1.4 (1.21–1.62)
⩾37 (n = 3648)	10.4	Reference	Reference
Sex
Female (n = 3194)	10.4	Reference	Reference
Male (n = 3867)	13.2	1.3 (1.14–1.50)	1.2 (1.01–1.36)
Race
White (n = 3530)	13.2	Reference	Reference
Black (n = 1642)	12.8	0.98 (0.83–1.15)	0.87 (0.72–1.05)
Asian (n = 888)	7.4	0.5 (0.42–0.71)	0.6 (0.45–0.80)
Other (n = 1001)†	9.8	0.7 (0.58–0.90)	0.9 (0.57–1.31)
Ethnicity
Non-Hispanic (n = 3668)	10.9	Reference
Hispanic (n = 2650)	14.1	1.3 (1.14–1.51)
Brazil (n = 248)	10.5
Spain (n = 495)	8.9
HIV status
Negative (n = 3509)	12.3	Reference
Positive (n = 206)	12.1	1.0 (0.67–1.49)
Unknown (n = 3346)	11.5	0.9 (0.81–1.06)
Country of origin
Non-US (n = 4542)	12.1	Reference
US (n = 2519)	11.5	1.0 (0.83–1.10)
Educational level
⩽8th grade (n = 1444)	13.0	1.4(1.12–1.78)	1.4 (1.05–1.73)
9th grade, some college (n = 4384)	12.3	1.3 (1.08–1.61)	1.1 (0.92–1.40)
College or higher (n = 1233)	9.4	Reference	Reference
Incarceration‡
No (n = 6665)	11.5	Reference	Reference
Yes (n = 396)	18.7	1.7 (1.36–2.19)	1.4 (1.03–1.82)
Unemployed
No (n = 6252)	11.7	Reference	Reference
Yes (n = 809)	13.4	1.2 (0.95–1.43)	0.8 (0.60–0.99)
Homeless§
No (n = 6553)	11.3	Reference	Reference
Yes (n = 508)	19.5	1.8 (1.49–2.26)	1.9 (1.42–2.47)
Alcohol consumption¶
No (n = 3306)	11.1	Reference
Use (n = 3229)	12.3	1.1 (0.97–1.29)
Abuse (n = 526)	14.8	1.4(1.10–1.80)
Injection drug use
No (n = 6780)	11.7	Reference
Yes (n = 281)	16.4	1.4(1.07–1.95)
Chronic liver disease (cirrhosis)
No (n = 6787)	11.9	Reference
Yes (n = 274)	11.3	1.0 (0.67–1.37)
Current smoker
No (n = 4941)	10.6	Reference	Reference
Yes (n = 2120)	14.9	1.5 (1.27–1.67)	1.4 (1.16–1.59)
Methadone treatment
No (n = 6920)	11.9	Reference
Yes (n = 141)	11.4	1.0 (0.58–1.57)
Potential for pregnancy
Not possible (n = 4447)	12.5	Reference
Female (n = 2506)	11.1	0.9 (0.75–1.0)
Pregnancy (n = 108)	8.3	0.6 (0.33–1.23)
Enrolling site#
Missing clinic visit**
No (n = 6633)	11.5	Reference	Reference
Yes (n = 428)	18.9	1.7 (1.37–2.17)	1.4 (1.07–1.73)

^*3HP-DOT = 3 months of directly observed once-weekly RPT (maximum dose, 900 mg) plus INH (maximum dose, 900 mg); 9H-SAT = 9 months of daily self-administered INH (maximum dose, 300 mg).

^†Includes North American Indian and other participants in the United States and Canada.

^‡History of living in a correctional institution for ⩾1 month prior to enrollment.

^§History of homelessness or living in a shelter or single room occupancy for ⩾6 months prior to enrollment.

^¶Use: affirmative response to a question asking whether the participant ever drank alcoholic beverages. Abuse: score of ⩾2 with the CAGE (Cut down, Annoyed, Guilty, Eye-opener) questionnaire.

^#Enrolling site was analyzed after a random selection of the reference value; overall P < 0.001.

^**Missed at least one of the first three DOT visits for the 3HP-DOT regimen and at least one of the three monthly clinic visits for the 9H-SAT regimen, followed by a DOT or monthly visit after the missing DOT/visit respectively (includes those who did not receive any study dose).

Note: Body mass index was explored, but due to a lack of statistical significance and lack of a meaningful association with the outcome, it was not included in the table/analyses.

HR = hazard ratio; CI = confidence interval; aHR = adjusted HR; H, INH = isoniazid; P, RPT = rifapentine; DOT = directly observed therapy; SAT = self-administered treatment; HIV = human immunodeficiencyvirus.

From the total cohort, 6318 participants were enrolled at North American sites (United States and Canada) and 743 at non-North American sites (Brazil and Spain). Differences in clinical and demographic characteristics between regions are shown in Table 3.

Table 3.

Clinical and demographic characteristics of participants stratified by region of enrollment (n = 7061)

	Region of enrollment
	North America (n = 6318)	Non-North America (n = 743)
	n (%)	n (%)
Regimen
3HP-DOT (n = 3643)*	3254 (51.5)	389 (52.4)
9H-SAT (n = 3418)*	3064 (48.5)	354 (47.6)
Age, years (median = 37)
<37 (n = 3413)	3048 (48.2)	365 (49.1)
⩾37 (n = 3648)	3270 (51.8)	378 (50.9)
Sex
Female (n = 3194)	2838 (44.9)	356 (47.9)
Male (n = 3867)	3480 (55.1)	387 (52.1)
Race
African (n = 1642)	1642 (26)	—
Asian (n = 888)	888 (14.1)	—
Other (n = 1001)	258 (4.1)	743 (100)
White (n = 3530)	3530 (55.9)	—
Ethnicity
Hispanic (n = 2650)	2650 (41.9)	—
Non-Hispanic (n = 3668)	3668 (58.1)	—
Spain (n = 248)	—	248 (33.4)
Brazil (n = 495)	—	495 (66.6)
Country of birth
US (n = 2519)	2519 (39.9)	—
Non-US (n = 4542)	3799 (60.1)	743 (100)
HIV status
Negative (n = 3509)	3142 (49.7)	367 (49.4)
Positive (n = 206)	148 (2.3	58 (7.8)
Unknown (n = 3346)	3028 (47.9)	318 (42.8)
Body mass index, kg/m2
<18.5 (n = 131)	108 (1.7)	23 (3.1)
18.5–24.9 (n = 2146)	1813 (28.7)	333 (44.8)
25–29.9 (n = 2503)	2258 (35.7)	245 (33)
⩾30 (n = 2281)	2139 (33.9	142 (19.1)
Educational level
⩽8th grade (n = 1444)	1145 (18.1)	299 (40.2)
9th grade, some college (n = 4384)	3997 (63.3)	387 (52.1)
College or higher (n = 1233)	1176 (18.6)	57 (7.7)
History of incarceration
No (n = 6665)	5925 (93.8)	740 (99.6)
Yes (n = 396)	393 (6.2)	3 (0.4)
Unemployment
No (n = 6252)	5534 (87.6)	718 (96.6)
Yes (n = 809)	784 (12.4)	25 (3.4)
Homelessness
No (n = 6553)	5815 (92.0)	738 (99.3)
Yes (n = 508)	503 (8.0)	5 (0.7)
Alcohol consumption
Use (n = 3229)	2925 (46.3)	304 (40.9)
Abuse (n = 526)	484 (7.7)	42 (5.7)
No (n = 3306)	2909 (46.0)	397 (53.4)
Injection drug use ever
No (n = 6780)	6052 (95.8)	728 (98.0)
Yes (n = 281)	266 (4.2)	15 (2.0)
Chronic liver disease
No (n = 6787)	6068 (96.0)	719 (96.8)
Yes (n = 274)	250 (4.0)	24 (3.2)
Current smoker
No (n = 4941)	4394 (69.5)	547 (73.6)
Yes (n = 2120)	1924 (30.5)	196 (26.4)
Potential for pregnancy
Female (n = 2506)	2243 (35.5)	263 (35.4)
Pregnancy (n = 108)	105 (1.7)	3 (0.4)
Not possible (n = 4447)	3970 (62.8)	477 (64.2)
Methadone use
No (n = 6920)	6182 (97.8)	738 (99.3)
Yes (n = 141)	136 (2.2)	5 (0.7)
Concomitant medications ⩾4
No (n = 6069)	5367 (84.9)	702 (94.5)
Yes (n = 992)	951 (15.1)	41 (5.5)
Contact TB
No (n = 2198)	2114(33.5)	84(11.3)
Yes (n = 4863)	4204 (66.5)	659 (88.7)
TST conversion
No (n = 4749)	4074 (64.5)	675 (90.8)
Yes (n = 2312)	2244 (35.5)	68 (9.2)
Fibrosis
No (n = 6873)	6133 (97.1)	740 (99.6)
Yes (n = 188)	185 (2.9)	3 (0.4)
HIV/LTBI treatment
No (n = 6865)	6178 (97.8)	687 (92.5)
Yes (n = 196)	140 (2.2)	56 (7.5)

^*3HP-DOT = 3 months of directly observed once-weekly RPT (maximum dose, 900 mg) plus INH (maximum dose, 900 mg); 9H-SAT = 9 months of daily self­administered INH (maximum dose, 300 mg).

H, INH = isoniazid; P, RPT=rifapentine; DOT = directly observed therapy; SAT = self-administered treatment; HIV = human immunodeficiency virus; LTBI = latent tuberculous infection.

Among the 6318 participants enrolled at North American sites, 771 did not complete study follow-up (771/6318, 12.2%) (Appendix Table A.2). The univariate and multivariate analyses found similar results compared to the total trial cohort (Table 4; interaction terms are evaluated in Appendix Table A.3). Among 743 participants enrolled in non-North American sites, 70 did not complete study follow-up (70/743, 9.4%; Appendix Table A.4). The only factor significantly associated with NCF in the univariate and multivariate analyses was missing an early clinic visit (adjusted HR 6.1) (Table 5).

Table 4.

Univariate and multivariate analysis of factors associated with non-completion of study follow-up (n = 771) compared to those who completed study follow-up (n = 5547) in the North American region: Cox proportional hazard model (n = 6318)

	Non-completion of follow-up %	Univariate* HR (95%CI)	Multivariate aHR (95%CI)
Regimen†
3HP-DOT (n = 3254)	11.2	Reference	Reference
9H-SAT (n = 3064)	13.3	1.2 (1.05–1.39)	1.2 (1.05–1.39)
Age, years (median = 37)
<37 (n = 3048)	13.9	1.3 (1.14–1.52)	1.4 (1.20–1.61)
⩾37 (n = 3270)	10.6	Reference	Reference
Sex
Female (n = 2838)	10.7	Reference
Male (n = 3480)	13.5	1.3 (1.12–1.49)
Race
White (n = 3530)	13.2	Reference	Reference
African American (n = 1642)	12.8	1.0 (0.83–1.16)	0.9 (0.78–1.14)
Asian (n = 888)	7.4	0.5 (0.42–0.71)	0.5 (0.39–0.70)
Other (n = 258)‡	10.9	0.8 (0.55–1.19)	0.8 (0.55–1.26)
Ethnicity
Non-Hispanic (n = 3668)	10.9	Reference
Hispanic (n = 2650)	14.1	1.3 (1.13–1.50)
HIV status
Negative (n = 3142)	12.6	Reference
Positive (n = 148)	12.2	1.0 (0.61–1.57)
Unknown (n = 3028)	11.8	0.9 (0.81–1.08)
Country of origin
Non-US (n = 3799)	12.7	Reference	Reference
US (n = 2519)	11.5	0.9 (0.79–1.05)	0.6 (0.50–0.73)
Educational level
⩽8th grade (n = 1145)	14.4	1.6 (1.27–2.05)
9th grade, some college (n = 3997)	12.5	1.4(1.12–1.70)
College or higher (n = 1176)	9.2	Reference
Incarceration§
No (n = 5925)	11.8	Reference	Reference
Yes (n = 393)	18.8	1.7 (1.34–2.16)	1.4 (1.08–1.91)
Unemployed
No (n = 5534)	12.0	Reference
Yes (n = 784)	13.5	1.1 (0.93–1.41)
Homeless¶
No (n = 5815)	11.6	Reference	Reference
Yes (n = 503)	19.5	1.8 (1.44–2.21)	2.0 (1.50–2.58)
Alcohol consumption#
No (n = 2909)	11.2	Reference
Use (n = 2925)	12.6	1.1 (0.98–1.32)
Abuse (n = 484)	15.7	1.5 (1.15–1.90)
Injection drug use
No (n = 6052)	12.0	Reference
Yes (n = 266)	16.2	1.4(1.02–1.89)
Chronic liver disease (cirrhosis)
No (n = 6068)	12.2	Reference
Yes (n = 250)	12.0	1.0 (0.69–1.44)
Current smoker
No (n = 4394)	10.9	Reference	Reference
Yes (n = 1924)	15.3	1.5 (1.27–1.70)	1.5 (1.28–1.77)
Methadone treatment
No (n = 6182	12.2	Reference
Yes (n = 136)	11.8	1.0 (0.59–1.59)
Potential for pregnancy
Not possible (n = 3970)	12.7	Reference
Female (n = 2243)	11.5	0.9 (0.77–1.03)
Pregnancy (n = 105)	8.6	0.6 (0.33–1.25)
Enrolling site**
Missing an early clinic visit††‡‡
No (n = 5906)	11.8	Reference
Yes (n = 412)	18.0	1.6 (1.24–2.0)

^*In univariate analysis, among 2498 non-US/Canada-born participants, 34 used the services of an interpreter, while 2464 did not; respectively 11.9% and 14.7% of those who did not complete study follow-up did not need and did need an interpreter; this difference was non-significant (HR 1.3, 95%CI 0.53–3.10, P = 0.58).

^†3HP-DOT = 3 months of directly observed once-weekly RPT (maximum dose, 900 mg) plus INH (maximum dose, 900 mg); 9H-SAT = 9 months of daily self-administered INH (maximum dose, 300 mg).

^‡Includes North American Indian and other participants in the United States and Canada.

^§History of living in a correctional institution for ⩾1 month prior to enrollment.

^¶History of homelessness or living in a shelter or single room occupancy for ⩾6 months prior to enrollment.

^#Use: affirmative response to a question asking whether the participant ever drank alcoholic beverages. Abuse: score of ⩾2 the CAGE (Cut down, Annoyed, Guilty, Eye-opener) questionnaire.

^**Enrolling site was analyzed after a random selection of the reference value; overall P < 0.001.

^††Missing at least one of the first three DOT visits for the 3HP-DOT regimen and at least one of the three monthly clinic visits for the 9H-SAT regimen, followed by a DOT or monthly visit after the missing DOT/visit respectively (includes those who did not receive any study dose).

^‡‡The relation between missing an early clinic visit and history of incarceration or homelessness was non-significant. Note: Body mass index was explored, but due to a lack of statistical significance and lack of a meaningful association with the outcome, it was not included in the table/analyses.

HR=hazard ratio; CI =confidence interval; aHR=adjusted HR; H, INH =isoniazid; P, RPT=rifapentine; DOT=directly observed therapy; SAT =self-administered treatment; HIV = human immunodeficiency virus.

Table 5.

Univariate and multivariate analysis of factors associated with non-completion of study follow-up (n = 70) compared to those who completed study follow-up (n = 643) in the Non-North American region (n = 743): Cox proportional hazard model

Characteristic	Non-completion of follow-up %	Univariate* HR (95%CI)	Multivariate aHR (95%CI)
Regimen†
3HP-DOT (n = 389)	7.7	Reference
9INH-SAT (n = 354)	11.3	1.5 (0.93–2.40)
Age, years (median = 37)
<37 (n = 365)	10.4	1.3 (0.78–2.0)
⩾37 (n = 378)	8.5	Reference
Sex
Female (n = 356)	7.9	Reference
Male (n = 387)	10.9	1.4 (0.87–2.27)
HIV status
Negative (n = 367)	9.8	Reference
Positive (n = 58)	12.1	1.2 (0.56–2.81)
Unknown (n = 318)	8.5	0.9 (0.53–1.43)
Educational level
⩽ 8th grade (n = 299)	7.7	0.5 (0.24–1.17)
9th grade, some college (n = 387)	10.1	0.7 (0.32–1.46)
College or higher (n = 57)	14.0	Reference
Incarceration‡
No (n = 740)	9.5	Reference
Yes (n = 3)	0.0	<0.001 (<0.001->999)
Unemployed
No (n = 718)	9.5	Reference
Yes (n = 25)	8.0	0.9 (0.21–3.52)
Homeless§
No (n = 738)	9.4	Reference
Yes (n = 5)	20.0	2.3 (0.32–16.81)
Alcohol consumption¶
No (n = 397)	10.1	Reference
Use (n = 304)	9.2	0.9 (0.57–1.48)
Abuse (n = 42)	4.8	0.5 (0.12–1.96)
Injection drug use
No (n = 728)	9.2	Reference
Yes (n = 15)	20.0	2.3 (0.71–7.20)
Chronic liver disease (cirrhosis)
No (n = 719)	9.6	Reference
Yes (n = 24)	4.2	0.4 (0.06–2.96)
Current smoker
No (n = 547)	8.8	Reference
Yes (n = 196)	11.2	1.3 (0.77–2.12)
Methadone treatment
No (n = 738)	9.5	Reference
Yes (n = 5)	0.0	<0.001 (<0.001->999.9)
Potential for pregnancy
Not possible (n = 477)	10.7	Reference
Female (n = 263)	7.2	0.7 (0.39–1.11
Pregnancy (n = 3)	0.0	<0.001 (<0.001->999.9)
Enrolling site
Spain (n = 248)	3.5	1.18 (0.72–1.91)
Brazil (n = 495)	8.9	Reference
Missing clinic visit¶
No (n = 727)	8.7	Reference	Reference
Yes (n = 16)	43.8	6.1 (2.80–13.34)	6.1 (2.80–13.34)

^*In univariate analysis, only 5 participants required the services of an interpreter among participants enrolled at non-US/Canada sites (n = 743). The association between need for an interpreter and non-completion of follow-up was not significant (P = 0.98).

^†3HP-DOT = 3 months of directly observed once-weekly RPT (maximum dose, 900 mg) plus INH (maximum dose, 900 mg); 9H-SAT = 9 months of daily self­administered INH (maximum dose, 300 mg).

^‡History of living in a correctional institution for ⩾1 month prior to enrollment.

^§History of homelessness or living in a shelter or single room occupancy for ⩾6 months prior to enrollment.

^¶Use: affirmative response to a question asking whether the participant ever drank alcoholic beverages. Abuse: score of ⩾2 on the CAGE (Cut down, Annoyed, Guilty, Eye-opener) questionnaire.

^#Missing at least one of the first three DOT visits for the 3HP-DOT regimen and at least one of the three monthly clinic visits for the 9H-SAT regimen, followed by a DOT or monthly visit after the missing DOT/visit, respectively (includes those who did not receive any study dose). No statistically significant interactions resulted between missing an early clinic visit and factors with P ⩽ 0.2 in univariate analysis.

Note: Body mass index was explored, but due to a lack of statistical significance and lack of a meaningful association with the outcome, it was not included in the table/analyses.

HR = hazard ratio; CI = confidence interval; aHR = adjusted HR; H, INH = isoniazid; P, RPT = rifapentine; DOT = directly observed therapy; SAT =self-administered treatment; HIV = human immunodeficiency virus.

In most cases, it was not possible to determine the reason for NCF (Table 6). No major differences in clinical and demographic characteristics were found among participants who discontinued follow-up within the first 331 days, 332–662 days, or 663–993 days. However, the percentage of participants who discontinued follow-up during all three intervals was higher among participants allocated to the 9HSAT regimen (Appendix Table A.5).

Table 6.

Reasons for non-completion of study follow-up by region and by regimen

	North-America (n = 771)		Non North-America (n = 70)
	3HP-DOT* (n = 363)	9H-SAT* (n = 408)	3HP-DOT* (n = 30)	9H-SAT* (n = 40)
Reason	n (%)	n (%)	n (%)	n (%)
Withdrew consent (n = 139)	58 (16.0)	72 (17.6)	6 (20.0)	3(7.5)
Developed active TB (n = 5)†	3 (0.83)	2 (0.49)	—	—
Refusal of further follow-up (n = 7)	4 (1.1)	3 (0.74)	—	—
Lost to follow-up (n = 410)‡	187 (51.5)	194 (47.5)	11 (36.7)	18 (45.0)
Other (n = 147)	56 (15.4)	76 (18.6)	7 (23.3)	8 (20.0)
Missing (n = 133)	55 (15.2)	61 (15.0)	6 (20.0)	11 (27.5)
Total (n = 841)	363 (47.1)	408 (52.8)	30 (42.9)	40 (57.1)

^*3HP-DOT = 3 months of directly observed once-weekly RPT (maximum dose, 900 mg) plus INH (maximum dose, 900 mg); 9H-SAT = 9 months of daily self-administered INH (maximum dose, 300 mg).

^†Classified as 3HP-DOT probable clinical TB in adult (n=2), spine disease (discitis) not characteristic of TB spondylitis (n = 1)and 9H-SAT probable clinical TB in adult (n = 2).

^‡3 participants in the 3HP group and 4 in the 9H group died after month 33, and were lost at the last follow-up evaluation. They were therefore added to the lost to follow-up category.

H, INH = isoniazid; P, RPT = rifapentine; DOT = directlyobserved therapy; SAT = self-administered treatment.

No association was found between NCF proportion and the number of participants enrolled by site (Pearson’s correlation coefficient = −0.15; P = 0.44, R² = 0.02) (Appendix Table A.6). The proportion of NCF by enrolling site, stratified by regimen, did not show any pattern among high enrolling sites (Appendix Table A.7).

Among 5228 participants who were the first enrolled in a cluster in the North American region, 1247 did not complete treatment and 3981 did complete treatment; 330 (26.5%) participants not completing treatment and 320 (8.0%) participants completing treatment did not complete study follow-up (P < 0.001) (Appendix Table A.8). Results of univariate and multivariate analyses were similar to those shown in Tables 2 and 4 (Appendix Table A.9).

In the sensitivity analysis evaluating failure to complete at least 80% of follow-up, among participants enrolled in the North American sites, 1481 did not complete treatment and 4837 completed treatment; 321 (21.7%) treatment non-completers and 185 (3.8%) treatment completers did not complete study follow-up (P < 0.001) (Appendix Table A.10). Results of univariate and multivariate analysies were similar to those shown on Tables 2 and 4 (Appendix Table A.11).

In this study, collinearity was not found among factors that resulted in P ⩽ 0.2 in the univariate analysis.

DISCUSSION

Our study found that in a large (n = 7061) clinical trial of LTBI treatment requiring lengthy follow-up, the proportion that did not complete at least 993 days follow-up after enrollment was 11.9% (North American sites 12.2%, non-North American sites 9.4%). This proportion is similar to the 10.2% NCF rate found in a clinical trial in the United States and Canada with 2 years of post-TB treatment follow-up.1 Retention is especially challenging for an LTBI treatment trial, considering the asymptomatic nature of the condition being treated.¹²

The analyses conducted in two separate regions showed different results. Differences in sociodemo­graphic characteristics as well as the small sample size of the non-North American cohort, with the small number of observations for some factors, might explain these differences. The HRs of some factors (e.g., young age, male sex, homelessness, injection drug use, and smoking) were >1, but were not statistically significant.

The inclusion of persons experiencing homelessness in TB clinical trials is important because they are disproportionally affected by TB. They represent a challenge for recruitment, initiation and completion of treatment, and for retention in research settings, due to unstable housing, food insecurity, and high prevalence of chronic mental health conditions.^1,13,14 Data from the PREVENT TB study, North American region, showed that the proportion of non-completion of LTBI treatment not associated with adverse events was 57% higher in participants with history of homelessness before enrollment than in participants with no such history (P < 0.001).¹⁵ The same participants analyzed in this study showed a high rate of failure of retention to the end of the trial (19.5%). Given the complexity of working with people who have unstable housing, appropriate management is needed to maximize adherence in research settings.^16,17

Self-reported smoking at the time of enrollment is also associated with both non-completion of LTBI treatment¹⁵ and NCF. Other risk factors are also associated with smoking: among 2120 participants who reported smoking at enrollment, 14% had a history of incarceration and 18% had experienced homelessness. Although a smoking cessation program might not succeed during the relatively short treatment phase, it might be feasible during a lengthy follow-up trial.¹⁸

The analyses performed indicate that missing an early clinic visit is a factor significantly associated with NCF after completion or discontinuation of LTBI treatment. This result is consistent with a separate analysis that indicated that participants who missed an early clinic visit for reasons other than an adverse event, but who returned at least once later, were more likely to discontinue the LTBI treatment regimen.¹⁵ Focused interventions to increase both completion of treatment and retention to the end of the trial might target participants who miss an early visit.

Incentives in clinical trials may be useful for recruitment and retention of participants.^4,19–21 An effort was made to analyze the influence of compensation provided by enrolling sites to participants of this trial; however, a meaningful analysis of influence on completion of follow-up was not possible due to the high variability among methods and amounts of compensation.

In conclusion, clinical trials require retention through both treatment and follow-up phase to be successful. Evaluation of follow-up in other trials might help determine whether the identified factors in this study consistently correlate with NCF.

APPENDIX

Table A.1.

Univariate and multivariate analyses of factors associated with non-completion of study follow-up (n = 841) compared to those who completed study follow-up (n = 6220) in the North American and non-North American regions combined (n = 7061): Cox proportional hazard model

	Non-completion of follow-up %	Univariate HR (95%CI)	Multivariate aHR (95%CI)
Regimen*
3HP-DOT (n = 3643)	10.8	Reference	Reference
9H-SAT (n = 3418)	13.1	1.2 (1.07–1.41)
Age, years (median = 37)
<37 (n = 3413)	13.5	1.3 (1.15–1.50)	1.4 (1.21–1.61)
⩾37 (n = 3648)	10.4	Reference	Reference
Sex
Female (n = 3194)	10.4	Reference
Male (n = 3867)	13.2	1.3 (1.14–1.50)
Race
White (n = 3530)	13.2	Reference	Reference
Black (n = 1642)	12.8	0.98 (0.83–1.15)	0.9 (0.72–1.03)
Asian (n = 888)	7.4	0.5 (0.42–0.71)	0.6 (0.44–0.77)
Other (n = 1001)†	9.8	0.7 (0.58–0.90)	0.8 (0.55–1.27)
Ethnicity
Non-Hispanic (n = 3668)	10.9	Reference
Hispanic (n = 2650)	14.1	1.3 (1.14–1.51)
Brazil (n = 248)	10.5
Spain (n = 495)	8.9
HIV status
Negative (n = 3509)	12.3	Reference
Positive (n = 206)	12.1	1.0 (0.67–1.49)
Unknown (n = 3346)	11.5	0.9 (0.81–1.06)
Country of origin
Non-US (n = 4542)	12.1	Reference
US (n = 2519)	11.5	1.0 (0.83–1.10)
Educational level
⩽8th grade (n = 1444)	13.0	1.4(1.12–1.78)
9th grade, some college (n = 4384)	12.3	1.3 (1.08–1.61)
College or higher (n = 1233)	9.4	Reference
Incarceration‡
No (n = 6665)	11.5	Reference
Yes (n = 396)	18.7	1.7 (1.36–2.19)
Unemployed
No (n = 6252)	11.7	Reference
Yes (n = 809)	13.4	1.2 (0.95–1.43)
Homeless§
No (n = 6553)	11.3	Reference	Reference
Yes (n = 508)	19.5	1.8 (1.49–2.26)	1.7 (1.32–2.23)
Alcohol consumption¶
No (n = 3306)	11.1	Reference
Use (n = 3229)	12.3	1.1 (0.97–1.29)
Abuse (n = 526)	14.8	1.4(1.10–1.80)
Injection drug use
No (n = 6780)	11.7	Reference
Yes (n = 281)	16.4	1.4(1.07–1.95)
Chronic liver disease (cirrhosis)
No (n = 6787)	11.9	Reference
Yes (n = 274)	11.3	1.0 (0.67–1.37)
Current smoker
No (n = 4941)	10.6	Reference	Reference
Yes (n = 2120)	14.9	1.5 (1.27–1.67)	1.4 (1.17–1.62)
Methadone treatment
No (n = 6920)	11.9	Reference
Yes (n = 141)	11.4	1.0 (0.58–1.57)
Potential for pregnancy
Not possible (n = 4447)	12.5	Reference
Female (n = 2506)	11.1	0.9 (0.75–1.0)
Pregnancy (n = 108)	8.3	0.6 (0.33–1.23)
Enrolling site#
Missing clinic visit**
No (n = 6633)	11.5	Reference	Reference
Yes (n = 428)	18.9	1.7 (1.37–2.17)	1.4 (1.07–1.72)
Interaction terms
IDU × regimen
3HP-DOT: IDU vs. no IDU			0.6 (0.38–0.81)
9H-SAT: IDU vs. no IDU			1.7 (0.95–2.96)
Sex × alcohol
Female: abuse vs. no alcohol			0.3 (0.12–0.89)
Male: abuse vs. no alcohol			1.2 (0.91–1.68)

^*3HP-DOT = 3 months of directly observed once-weekly RPT (maximum dose, 900 mg) plus INH (maximum dose, 900 mg); 9H-SAT = 9 months of daily self-administered INH (maximum dose, 300 mg).

^†Includes North American Indian and other participants in the United States and Canada.

^‡History of living in a correctional institution for ⩾1 month prior to enrollment.

^§History of homelessness or living in a shelter or single room occupancy for ⩾6 months prior to enrollment.

^¶Use: affirmative response to a question asking whether the participant ever drank alcoholic beverages. Abuse: score of ⩾2 with the CAGE (Cut down, Annoyed, Guilty, Eye-opener) questionnaire.

^#Enrolling site was analyzed after a random selection of the reference value; overall P < 0.001.

^**Missing at least one of the first three DOT visits for the 3HP-DOTregimen and at least one of the three monthly clinic visits for the 9H-SAT regimen, followed by a DOT or a monthly visit after the missing DOT/visit, respectively (includes those who did not receive any study dose).

Note: Body mass index was explored, but due to a lack of statistical significance and lack of a meaningful association with the outcome, it was not included in the table/analyses.

HR = hazard ratio; CI = confidence interval; aHR = adjusted HR; H, INH = isoniazid; P, RPT = rifapentine; DOT = directly observed therapy; SAT =self-administered treatment; HIV = human immunodeficiency virus.

Table A.2.

Completion of follow-up status by completion of treatment in the North American region (n = 6318)

	Did not complete treatment (n = 1481)		Completed treatment (n = 4837)
	Did not complete follow-up* (n = 385)	Completed follow-up (n = 1096)	Did not complete follow-up* (n = 386)	Completed follow-up (n = 4451)
Regimen	n (%)	n (%)	n (%)	n (%)
3HP-DOT (n = 3254)†	141 (36.6)	450 (41.1)	222 (57.5)	2441 (54.8)
9H-SAT (n = 3064)†	244 (63.4)	646 (58.9)	164 (42.5)	2010 (45.2)
Total (n = 6318)	385 (26.0)‡	1096 (74.0)‡	386 (8.0)§	4451 (92.0)§

^*Participants had the opportunity to continue study follow-up after completion or discontinuation of treatment; non-completion of follow-up = (385 + 386)/6318 = 12.2%. The difference in the proportion of non-completion of follow-up between those who did not complete treatment (385/1481, 26.0%) vs. those who completed treatment (386/4837, 7.7%) was statistically significant (P < 0.001).

^†3HP-DOT = 3 months of directly observed once-weekly RPT (maximum dose, 900 mg) plus INH (maximum dose, 900 mg); 9H-SAT = 9 months of daily self-administered INH (maximum dose, 300 mg).

^‡The denominator for the percentage is all participants who did not complete treatment (n = 1481).

^§The denominator for the percentage is all participants who completed treatment (n = 4837).

H, INH = isoniazid; P, RPT = rifapentine; DOT = directly observed therapy; SAT = self-administered treatment.

Table A.3.

Univariate and multivariate analysis of factors associated with non-completion of study follow-up (n = 771) compared to those who completed study follow-up (n = 5547) in the North American region: Cox proportional hazard model (n = 6318)

	Non-completion follow-up %	Univariate* HR (95%CI)	Multivariate with interaction terms aHR (95%CI)
Regimen†
3HP-DOT (n = 3254)	11.2	Reference	Reference
9H-SAT (n = 3064)	13.3	1.2 (1.05–1.39)	1.2 (1.01–1.36)
Age, years (median = 37)
<37 (n = 3048)	13.9	1.3 (1.14–1.52)
⩾37 (n = 3270)	10.6	Reference
Sex
Female (n = 2838)	10.7	Reference
Male (n = 3480)	13.5	1.3 (1.12–1.49)
Race
White (n = 3530)	13.2	Reference	Reference
African American (n = 1642)	12.8	1.0 (0.83–1.16)	1.0 (0.83–1.29)
Asian (n = 888)	7.4	0.5 (0.42–0.71)	0.6 (0.39–0.77)
Other (n = 258)‡	10.9	0.8 (0.55–1.19)	0.8 (0.55–1.27)
Ethnicity
Non-Hispanic (n = 3668)	10.9	Reference
Hispanic (n = 2650)	14.1	1.3 (1.13–1.50)
HIV status
Negative (n = 3142)	12.6	Reference
Positive (n = 148)	12.2	1.0 (0.61–1.57)
Unknown (n = 3028)	11.8	0.9 (0.81–1.08)
Country of origin
Non-US (n = 3799)	12.7	Reference
US (n = 2519)	11.5	0.9 (0.79–1.05)
Educational level
⩾8th grade (n = 1145)	14.4	1.6 (1.27–2.05)
9th grade some college (n = 3997)	12.5	1.4 (1.12–1.70)
College or higher (n = 1176)	9.2	Reference
Incarceration§
No (n = 5925)	11.8	Reference
Yes (n = 393)	18.8	1.7 (1.34–2.16)
Unemployed
No (n = 5534)	12.0	Reference
Yes (n = 784)	13.5	1.1 (0.93–1.41)
Homeless¶
No (n = 5815)	11.6	Reference	Reference
Yes (n = 503)	19.5	1.8 (1.44–2.21)	2.1 (1.59–2.75)
Alcohol consumption#
No (n = 2909)	11.2	Reference
Use (n = 2925)	12.6	1.1 (0.98–1.32)
Abuse (n = 484)	15.7	1.5 (1.15–1.90)
Injection drug use
No (n = 6052)	12.0	Reference
Yes (n = 266)	16.2	1.4 (1.02–1.89)
Chronic liver disease (cirrhosis)
No (n = 6068)	12.2	Reference
Yes (n = 250)	12.0	1.0 (0.69–1.44)
Current smoker
No (n = 4394)	10.9	Reference	Reference
Yes (n = 1924)	15.3	1.5 (1.27–1.70)	1.5 (1.30–1.78)
Methadone treatment
No (n = 6182)	12.2	Reference
Yes (n = 136)	11.8	1.0 (0.59–1.59)
Potential for pregnancy
Not possible (n = 3970)	12.7	Reference
Female (n = 2243)	11.5	0.9 (0.77–1.03)
Pregnancy (n = 105)	8.6	0.6 (0.33–1.25)
Enrolling site**
Missing an early clinic visit††
No (n = 5906)	11.8	Reference	Reference
Yes (n = 412)	18.0	1.6 (1.24–2.0)	1.3 (1.004–1.65)
Interaction terms
Country of birth × age
US: age <37 years vs. ⩾37 years			1.7 (1.35–2.20)
Non-US: age <37 years vs. ⩾37 years			1.2 (1.0–1.45)
Incarceration × ethnicity
Incarceration: Hispanic vs. non-Hispanic			2.1 (1.19–3.56)
No incarceration: Hispanic vs. non-Hispanic			1.1 (0.84–1.38)

^*In univariate analysis, among 2498 non-US/Canada-born participants, 34 used the services of an interpreter, while 2464 did not; respectively 11.9% and 14.7% of those who did not complete study follow-up did not need and did need an interpreter; this difference was non-significant (HR 1.3, 95%CI 0.53–3.10, P = 0.58).

^†3HP-DOT = 3 months of directly observed once-weekly RPT (maximum dose, 900 mg) plus INH (maximum dose, 900 mg); 9H-SAT = 9 months of daily self-administered INH (maximum dose, 300 mg).

^‡Includes North American Indian and other participants in the United States and Canada.

^§History of living in a correctional institution for ⩾1 month prior to enrollment.

^¶History of homelessness or living in a shelter or single room occupancy for ⩾6 months prior to enrollment.

^#Use: affirmative response to a question asking whether the participant ever drank alcoholic beverages. Abuse: score of ⩾2 with the CAGE (Cut down, Annoyed, Guilty, Eye-opener) questionnaire.

^**Enrolling site was analyzed after a random selection of the reference value; overall P < 0.001.

^††Missing at least one of the first three DOT visits for the 3HP-D0T regimen and at least one of the three monthly clinic visits for the 9H-SAT regimen, followed by a DOT or a monthly visit after the missing DOT/visit, respectively (includes those who did not receive any study dose). No significant interaction resulted between missing an early clinic visit and history of incarceration or homelessness.

Note: Body mass index was explored, but due to a lack of statistical significance and lack of a meaningful association with the outcome, it was not included in the table/analyses.

HR=hazard ratio; CI =confidence interval; aHR=adjusted HR; H, INH =isoniazid; P, RPT=rifapentine; DOT=directly observed therapy; SAT =self-administered treatment; HIV = human immunodeficiency virus.

Table A.4.

Completion of follow-up status by completion of treatment in the non-North American region (n = 743)^*

	Did not complete treatment (n = 120)		Completed treatment (n = 623)
	Did not complete follow-up (n = 34)	Completed follow-up (n = 86)	Did not complete follow-up (n = 36)	Completed follow-up (n = 587)
Regimen	n (%)	n (%)	n (%)	n (%)
3HP-DOT (n = 389)†	14 (41.2)	35 (40.7)	16 (44.4)	324 (55.2)
9H-SAT (n = 354)†	20 (58.8)	51 (59.3)	20 (55.6)	263 (44.8)
Total (n = 743)	34 (28.3)‡	86 (71.7)‡	36 (5.8)§	587 (94.2)§

^*Participants had the opportunity to continue study follow-up after completion or discontinuation of treatment. Non-completion of follow-up = (34 + 36)/743 = 9.4%. The proportion of non-completion of follow-up between those who did not complete treatment (34/120, 28.3%) vs. those who completed treatment (36/623, 5.8%) was statistically significant (P < 0.001).

^†3HP-DOT = 3 months of directly observed once-weekly RPT (maximum dose, 900 mg) plus INH (maximum dose, 900 mg); 9H-SAT = 9 months of daily self-administered INH (maximum dose, 300 mg).

^‡The denominator for the percentage is all participants who did not complete treatment (n = 120).

^§The denominator for the percentage is all participants who completed treatment (n = 623)

H, INH = isoniazid; P, RPT =rifapentine; DOT = directly observed therapy; SAT = self-administered treatment.

Table A.5.

Clinical and demographic characteristics of participants who did not complete study follow-up, by number of days of follow-up in the trial

	NCF within the first 331 days (n = 352)	NCF between 332 and 662 days (n = 167)	NCF between 663 and 993 days (n = 322)
	n (%)	n (%)	n (%)
Person-month	1214	2892	9538
Regimen*
3HP-DOT	168 (47.7)	78 (46.7)	145 (45.7)
9H-SAT	184 (52.3)	89 (53.3)	175 (54.4)
Age, years, median [IQR]	35 [26–45]	35 [26–45]	36 [26–45]
Male sex	228 (64.8)	112 (67.1)	170 (52.8)
Race
White	186 (52.8)	96 (57.5)	185 (57.5)
African American	95 (27.0)	42 (25.2)	73 (22.7)
Asian	33 (9.4)	14 (8.4)	19 (5.9)
Other†	38 (10.8)	15 (9.0)	45 (14.0)
Ethnicity
Hispanic	146 (41.5)	80 (47.9)	147 (45.7)
Non-Hispanic	180 (51.1)	78 (46.7)	140 (43.5)
Spain	9 (2.6)	4 (2.4)	13 (4.0)
Brazil	17 (4.8)	5 (3.0)	22 (6.8)
HIV status
Positive	11 (3.1)	4 (2.34)	10 (3.1)
Negative	167 (47.4)	87 (52.1)	179 (55.6)
Unknown	174 (49.4)	76 (45.5)	133 (41.3)
Country of birth
US	114 (32.4)	55 (32.9)	121 (37.6)
Non-US	238 (67.6)	112 (67.1)	201 (62.4)
Body mass index, median [IQR]	27 [24–30]	27 [24–31]	27 [24–31]
Education
⩽8th grade	81 (23.0)	37 (22.2)	70 (21.7)
9th grade, some college	216 (61.4)	110 (65.9)	211 (65.5)
College or higher	55 (15.6)	20 (12.0)	41 (12.7)
History of incarceration	33 (9.4)	17 (10.2)	24 (7.5)
Unemployed	43 (12.2)	24 (14.4)	41 (12.7)
Homeless	42 (11.9)	28 (16.8)	29 (9.0)
Alcohol consumption
No	147 (41.8)	60 (35.9)	160 (49.7)
Use	169 (48.0)	83 (49.7)	144 (44.7)
Abuse	36 (10.2)	24 (14.4)	18 (5.6)
Injection drug use	22 (6.3)	5 (3.0)	19 (5.9)
Current smoker	137 (38.9)	76 (45.5)	103 (32.0)

^*3HP-DOT = 3 months of directly observed once-weekly RPT (maximum dose, 900 mg) plus INH (maximum dose, 900 mg); 9H-SAT = 9 months of daily self­administered INH (maximum dose, 300 mg).

^†Includes North American Indian and other participants in the United States and Canada.

NCF=non-completion of follow-up; H, INH=isoniazid; P, RPT=rifapentine; DOT=directly observed therapy; SAT=self-administered treatment; IQR=interquartile range: HIV = human immunodeficiency virus.

Table A.6.

Proportion of non-completion of study follow-up by site, sorted by number of participants enrolled^*

	Non-completion of follow-up	Participants enrolled (n = 7062)
Site	n (% enrollment)	n (%)
G	145 (12.6)	1,153 (16.3)
R	69 (13.4)	515 (7.3)
Y	50 (10.0)	502 (7.1)
P	44 (8.9)	495 (7.0)
AA	37 (10.3)	359 (5.1)
B	69 (20.8)	331 (4.7)
K	48 (15.0)	319 (4.5)
I	14 (4.6)	306 (4.3)
F	25 (8.5)	294 (4.2)
E	18 (6.3)	288 (4.1)
L	18 (6.5)	276 (3.9)
Q	26 (10.5)	248 (3.5)
M	26 (11.8)	221 (3.1)
C	17 (8.2)	208 (2.9)
O	35 (17.8)	197 (2.8)
T	10 (5.3)	188 (2.7)
J	17 (9.8)	173 (2.4)
V	23 (13.5)	170 (2.4)
Z	13 (8.3)	157 (2.2)
N	18 (15.1)	119 (1.7)
D	37 (32.2)	115 (1.6)
W	25 (22.3)	112 (1.6)
U	29 (30.5)	95 (1.3)
H	3 (3.5)	85 (1.2)
X	11 (19.3)	57 (0.8)
S	2 (4.4)	45 (0.6)
A	12 (36.4)	33 (0.5)
KK	0	1 (0.0)

^*No association between proportions of non-completion of study follow-up and number of participants enrolled by site: Pearson’s correlation coefficient = −0.15, P = 0.44, R² = 0.02.

Table A.7.

Proportion of non-completion of study follow-up by site and by regimen

	3HP-DOT*		9H-SAT*
	Non-completion of follow-up	Enrollments (n = 3644)	Non-completion of follow-up	Enrollments (n = 3418)
Enrolling site	n (% of enrollments)	n (%)	n (% of enrollments)	n (%)
G	73 (12.5)	584 (16.0)	72 (12.7)	569 (16.6)
R	36 (13.6)	265 (7.3)	33 (13.2)	250 (7.3)
B	34 (20.6)	165 (4.5)	35 (21.1)	166 (4.9)
Y	26 (10.4)	250 (6.9)	24 (9.5)	252 (7.4)
P	19 (7.8)	245 (6.7)	25 (10.0)	250 (7.3)
D	17 (33.3)	51 (1.4)	20 (31.3)	64 (1.9)
O	16 (18.2)	88 (2.4)	19 (17.4)	109 (3.2)
M	14 (10.1)	138 (3.8)	12 (14.6)	82 (2.4)
U	14 (31.1)	45 (1.2)	15 (30.0)	50 (1.5)
E	13 (7.4)	176 (4.8)	5 (4.5)	112 (3.3)
AA	13(7.5)	174 (4.8)	24 (13.0)	185 (5.4)
J	12 (13.2)	91 (2.5)	5(6.1)	82 (2.4)
F	11 (6.6)	166 (4.6)	14 (10.9)	128 (3.7)
N	11 (16.7)	66 (1.8)	7(13.2)	53(1.6)
Q	11 (7.6)	144 (4.0)	15 (14.4)	104 (3.0)
W	11 (22.4)	49 (1.3)	14 (22.2)	63 (1.8)
K	10 (6.0)	168 (4.6)	38 (25.2)	151 (4.4)
V	9 (9.4)	96 (2.6)	14 (18.9)	74 (2.2)
C	8 (8.1)	99 (2.7)	9 (8.3)	109 (3.2)
L	8(5.6)	142 (3.9)	10 (7.5)	134 (3.9)
I	7 (4.5)	155 (4.3)	7 (4.6)	151 (4.4)
Z	7 (9.6)	73 (2.0)	6(7.1)	84 (2.5)
A	6 (30.0)	20 (0.5)	6 (46.2)	13 (0.4)
X	3 (10.7)	28 (0.8)	8 (27.6)	29 (0.8)
T	2 (2.2)	93 (2.6)	8 (8.4)	95 (2.8)
H	1 (2.1)	48 (1.3)	2 (5.4)	37 (1.1)
S	1 (4.3)	23 (0.6)	1 (4.5)	22 (0.6)
KK	0	1 (0.0)	0	0

^*3HP-DOT = 3 months of directly observed once-weekly RPT (maximum dose, 900 mg) plus INH (maximum dose, 900 mg); 9H-SAT = 9 months of daily self-administered INH (maximum dose, 300 mg).

H, INH = isoniazid; P, RPT = rifapentine; DOT = directly observed therapy; SAT = self-administered treatment.

Table A.8.

Completion of follow-up status by completion of treatment among first participants enrolled in a cluster in the North American region (n = 5228)^*

	Did not complete treatment (n = 1247)		Completed treatment (n = 3981)
	Did not complete follow-up (n = 330)	Completed follow-up (n = 917)	Did not complete follow-up (n = 320)	Completed follow-up (n = 3661)
Regimen†	n (%)	n (%)	n (%)	n (%)
3HP-DOT (n = 2590)	117 (35.5)	366 (39.9)	178 (55.6)	1929 (52.7)
9H-SAT (n = 2638)	213 (64.6)	551 (60.1)	142 (44.4)	1732 (47.3)
Total (n = 5228)	330 (26.5)‡	917 (73.5)‡	320 (8.0)§	3661 (92.0)§

^*Participants had the opportunity to continue study follow-up after completion or discontinuation of treatment. The proportion of non-completion of follow-up between those who did not complete treatment (26.5%) vs. those who completed treatment (8.0%) was statistically significant (P < 0.001).

^†3HP-DOT = 3 months of directly observed once-weekly RPT (maximum dose, 900 mg) plus INH (maximum dose, 900 mg); 9H-SAT = 9 months of daily self-administered INH (maximum dose, 300 mg).

^‡The denominator is all participants who did not complete treatment (n = 1247).

^§The denominator is all participants who completed treatment (n = 3981).

H, INH = isoniazid; P, RPT = rifapentine; DOT = directly observed therapy; SAT = self-administered treatment.

Table A.9.

Univariate and multivariate analyses of factors associated with non-completion of study follow-up (n = 650) compared to those who completed study follow-up (n = 4578) among the first participants enrolled in a cluster in the North American region: Cox proportional hazard model (n = 5228)

	Non-completion of follow-up %	Univariate HR (95%CI)	Multivariate aHR (95%CI)
Regimen*
3HP-DOT (n = 2590)	11.4	Reference	Reference
9H-SAT (n = 2638)	13.5	1.2 (1.02–1.39)
Age, years (median = 37)
<37 (n = 2447)	14.8	1.5 (1.24–1.69)	1.6 (1.33–1.85)
⩾37 (n = 2781)	10.4	Reference	Reference
Sex
Female (n = 2407)	11.1	Reference
Male (n = 2821)	13.6	1.3 (1.08–1.48)
Race
White (n = 2826)	13.7	Reference	Reference
Black (n = 1487)	12.7	0.9 (0.79–1.12)	1.0 (0.80–1.21)
Asian (n = 722)	7.5	0.5 (0.40–0.71)	0.5 (0.37–0.70)
Other (n = 193)†	10.4	0.8 (0.48–1.18)	0.8 (0.48–1.27)
Ethnicity
Non-Hispanic (n = 3247)	10.9	Reference
Hispanic (n = 1981)	15.0	1.4 (1.20–1.63)
HIV status
Negative (n = 2729)	12.8	Reference
Positive (n = 146)	12.3	1.0 (0.61–1.58)
Unknown (n = 2353)	12.0	0.9 (0.81–1.11)
Country of origin
Non-US (n = 2966)	13.3	Reference	Reference
US (n = 2262)	11.3	0.8 (0.72–0.99)	0.6 (0.50–0.77)
Educational level
⩽8th grade (n = 799)	16.4	1.9 (1.46–2.46)	1.5 (1.11–1.96)
9th grade, some college (n = 3320)	12.6	1.4 (1.15–1.78)	1.1 (0.91–1.44)
College or higher (n = 1109)	9.0	Reference	Reference
Incarceration‡
No (n = 4890)	12.0	Reference	Reference
Yes (n = 338)	18.9	1.7 (1.29–2.16)	1.5 (1.09–2.0)
Unemployed
No (n = 4568)	12.2	Reference
Yes (n = 660)	13.8	1.1 (0.92–1.43)
Homeless§
No (n = 4813)	11.9	Reference	Reference
Yes (n = 415)	19.0	1.7 (1.33–2.14)	1.8 (1.37–2.48)
Alcohol consumption¶
No (n = 2302)	11.8	Reference
Use (n = 2539)	12.5	1.1 (0.91–1.26)
Abuse (n = 387)	16.3	1.5 (1.11–1.93)
Injection drug use
No (n = 4984)	12.3	Reference
Yes (n = 244)	14.8	1.2 (0.87–1.71)
Chronic liver disease (cirrhosis)
No (n = 5008)	12.5	Reference
Yes (n = 220)	11.8	1.0 (0.65–1.42)
Current smoker
No (n = 3611)	11.1	Reference	Reference
Yes (n = 1617)	15.3	1.4 (1.23–1.68)	1.5 (1.27–1.80)
Methadone treatment
No (n = 5095)	12.4	Reference
Yes (n = 133)	12.0	1.0 (0.59–1.59)
Potential for pregnancy
Not possible (n = 3225)	12.9	Reference
Female (n = 1923)	11.8	0.9 (0.76–1.05)
Pregnancy (n = 80)	7.5	0.6 (0.25–1.24)
Enrolling site#
Missing clinic visit**
No (n = 4880)	12.0	Reference	Reference
Yes (n = 349)	19.5	1.7 (1.32–2.19)	1.5 (1.14–1.91)

^*3HP-DOT = 3 months of directly observed once-weekly RPT (maximum dose, 900 mg) plus INH (maximum dose, 900 mg); 9H-SAT = 9 months of daily self-administered INH (maximum dose, 300 mg).

^†Includes North American Indian and other participants in the United States and Canada.

^‡History of living in a correctional institution for ⩾1 month prior to enrollment.

^§History of homelessness or living in a shelter or single room occupancy for ⩾6 months prior to enrollment.

^¶Use: affirmative response to a question asking whether the participant ever drank alcoholic beverages. Abuse: score of ⩾2 on the CAGE (Cut down, Annoyed, Guilty, Eye-opener) questionnaire.

^#Enrolling site was anaylsed after a random selection of the refereence value; overall P < 0.001.

^**Missing at least one of the first three DOT visits for the 3HP-DOT regimen and at least one of the three monthly clinic visits for the 9H-SATregimen, followed by a DOT or monthly visit after the missing DOT/visit, respectively (includes those who did not receive any study dose). No statistically significant interactions resulted between missing an early clinic visit and factors with P ⩽ 0.2 in univariate analysis.

Note: Body mass index was explored, but due to a lack of statistical significance and lack of a meaningful association with the outcome, it was not included in the table/analyses.

HR = hazard ratio; CI = confidence interval; aHR = adjusted HR; H, INH = isoniazid; P, RPT = rifapentine; DOT = directly observed therapy; SAT = self-administered treatment; HIV = human immunodeficiency virus.

Table A.10.

Partial completion (80%) of study follow-up by completion of treatment in the North American region (n = 6318)^*

	Did not complete treatment (n = 1481)		Completed treatment (n = 4837)
	Did not complete follow-up	Completed follow-up	Did not complete follow-up	Completed follow-up
Regimen†	(n = 321)	(n = 1160)	(n = 185)	(n = 4652)
3HP-DOT (n = 3254)	122 (38.0)	469 (40.4)	122 (65.9)	2541 (54.6)
9H-SAT (n = 3064)	199 (62.0)	691 (59.6)	63 (34.1)	2111 (45.4)
Total (n = 6318)	321 (21.7)‡	1160 (78.3)‡	185 (3.8)§	4652 (96.2)§

^*Participants had the opportunity to continue with the study follow-up after completion or discontinuation of treatment. The proportion of non-completion of 80% of follow-up between those who did not complete treatment (21.7%) vs. those who completed (3.8%) was statistically significant (P < 0.001).

^†3HP-DOT = 3 months of directly observed once-weekly RPT (maximum dose, 900 mg) plus INH (maximum dose, 900 mg); 9H-SAT = 9 months of daily self-administered INH (maximum dose, 300 mg).

^‡The denominator is all participants who did not complete treatment (n = 1481).

^§The denominator is all participants who completed treatment (n = 4837).

H, INH = isoniazid; P, RPT = rifapentine; DOT = directly observed therapy; SAT = self-administered treatment.

Table A.11.

Univariate and multivariate analyses of factors associated with non-completion of study follow-up (n = 506) compared to those who completed 80% of study follow-up (n = 5812) in the North American region: Cox proportional hazard model (n = 6318)

	Non-completion of follow-up %	Univariate HR (95%CI)	Multivariate aHR (95%CI)
Regimen*
3HP-DOT (n = 3254)	7.5	Reference
9H-SAT (n = 3064)	8.6	1.1 (0.96–1.37)
Age, years (median = 37)
<37 (n = 3048)	9.4	1.4(1.18–1.68)	1.5 (1.28–1.86)
⩾37 (n = 3270)	6.7	Reference	Reference
Sex
Female (n = 2838)	6.1	Reference
Male (n = 3480)	9.6	1.6 (1.33–1.92)
Race
White (n = 3530)	8.4	Reference	Reference
Black (n = 1642)	8.7	1.1 (0.86–1.28)	1.0 (0.79–1.26)
Asian (n = 888)	5.5	0.7 (0.48–0.88)	0.6 (0.42–0.82)
Other (n = 258)†	7.4	0.9 (0.55–1.39)	0.8 (0.51–1.39)
Ethnicity
Non-Hispanic (n = 3668)	7.4	Reference
Hispanic (n = 2650)	8.9	1.2 (1.01–1.43)
HIV status
Negative (n = 3142)	7.9	Reference
Positive (n = 148)	7.4	1.0 (0.52–1.74)
Unknown (n = 3028)	8.2	1.0 (0.87–1.24)
Country of origin
Non-US (n = 3799)	8.7	Reference	Reference
US (n = 2519)	7.0	0.8 (0.67–0.96)	0.6 (0.44–0.71)
Educational level
⩽8th grade (n = 1145)	9.6	1.6 (1.19–2.17)
9th grade, some college (n = 3997)	8.1	1.4 (1.05–1.75)
College or higher (n = 1176)	6.0	Reference
Incarceration‡
No (n = 5925)	7.6	Reference
Yes (n = 393)	13.7	1.9 (1.41–2.48)	1.4 (1.04–2.02)
Unemployed
No (n = 5534)	7.9	Reference
Yes (n = 784)	8.8	1.1 (0.87–1.45)
Homeless§
No (n = 5815)	7.5	Reference	Reference
Yes (n = 503)	14.3	2.0 (1.54–2.54)	2.1 (1.55–2.95)
Alcohol consumption¶
No (n = 2909)	6.7	Reference
Use (n = 2925)	8.6	1.3 (1.07–1.56)
Abuse (n = 484)	12.2	1.9 (1.40–2.51)
Injection drug use
No (n = 6052)	7.9	Reference
Yes (n = 266)	9.8	1.3 (0.85–1.86)
Chronic liver disease (cirrhosis)
No (n = 6068)	8.1	Reference
Yes (n = 250)	6.0	0.7 (0.45–1.25)
Current smoker
No (n = 4394)	6.8	Reference	Reference
Yes (n = 1924)	10.8	1.6 (1.37–1.95)	1.6 (1.31–1.96)
Methadone treatment
No (n = 6182)	8.0	Reference
Yes (n = 136)	7.4	0.9 (0.49–1.73)
Pregnancy
Not possible (n = 3970)	9.0	Reference	Reference
Female (n = 2243)	6.6	0.7 (0.59–0.87)	0.8 (0.65–0.97)
Pregnancy (n = 105)	2.9	0.3 (0.10–0.97)	0.3 (0.1–0.98
Enrolling site#
Missing clinic visit**
No (n = 5906)	7.7	Reference	Reference
Yes (n = 412)	12.6	1.7 (1.25–2.22)	1.4 (1.03–1.85)

^*3HP-DOT = 3 months of directly observed once-weekly RPT (maximum dose, 900 mg) plus INH (maximum dose, 900 mg); 9H-SAT = 9 months of daily self-administered INH (maximum dose, 300 mg).

^†Includes North American Indian and other participants in the United States and Canada.

^‡History of living in a correctional institution for ⩾1 month prior to enrollment.

^§History of homelessness or living in a shelter or single room occupancy for ⩾6 months prior to enrollment.

^¶Use: affirmative response to a question asking whether the participant ever drank alcoholic beverages. Abuse: score of ⩾2 on the CAGE (Cut down, Annoyed, Guilty, Eye-opener) questionnaire.

^#Enrolling site was analyzed after a random selection of the reference value; overall P < 0.001.

^**Missing at least one of the first three DOT visits for the 3HP-DOT regimen and at least one of the three monthly clinic visits for the 9H-SAT regimen, followed by a DOT or a monthly visit after the missing DOT/visit, respectively (includes those who did not receive any study dose).

Note: Body mass index was explored, but due to a lack of statistical significance and lack of a meaningful association with the outcome, it was not included in the table/analyses.

HR = hazard ratio; CI = confidence interval; aHR = adjusted HR; H, INH = isoniazid; P, RPT = rifapentine; DOT = directly observed therapy; SAT = self-administered treatment; HIV = human immunodeficiency virus.

Table A.12.

Interaction terms evaluated in the multivariate analyses

Regimen × age

Regimen × sex

Regimen × race

Regimen × ethnicity

Regimen × country of birth (US vs. non-US)

Regimen × education

Regimen × incarceration

Regimen × unemployed

Regimen × homelessness

Regimen × alcohol use/abuse

Regimen × injection drug use

Regimen × smoking

Regimen × reproduction

Regimen × missing a clinic visit

Sex × age

Sex × race

Sex × incarceration

Sex × unemployed

Sex × homelessness

Sex × alcohol use/abuse

Sex × injection drug use

Sex × missing a clinic visit

Race × incarceration

Race × unemployment

Race × homelessness

Race × missing a clinic visit

Ethnicity × age

Ethnicity × sex

Ethnicity × country of birth (US vs. non-US)

Ethnicity × education

Ethnicity × incarceration

Ethnicity × unemployed

Ethnicity × homeless

Ethnicity × injection drug use

Ethnicity × smoking

Ethnicity × potential for reproduction

Ethnicity × missing a clinic visit

Country of birth × age

Country of birth × sex

Country of birth × incarceration

Country of birth × unemployment

Country of birth × homeless

Unemployed × alcohol use/abuse

Unemployed × missing a clinic visit

Homeless × missing a clinic visit

Injection drug use × missing a clinic visit

Injection drug use × homelessness

Smoking × age

Missing a clinic visit × age