Table 2 Number of infected individuals for different hypotheses and comparison with clinical trial data

From: Model-based predictions of protective HIV pre-exposure prophylaxis adherence levels in cisgender women

Scenario

 

PrEP efficacy (bottom-up)

91.1 (29.3, 99.9)

91.0 (28.9, 99.9)

93.4 (34.8, 100)

93.5 (37.0, 100)

35.8 (2.6, 60.4)

58.0 (12.4, 84.3)

48.2 (3.3, 71.5)

65.0 (15.7, 87.6)

Study

Infclin_sim

3 (0, 17)

HPTN 084

Infsim

4 (1, 9)

4 (1, 9)

3 (0, 7)

3 (0, 7)

27 (15, 42)

18 (9, 29)

22 (12, 35)

15 (7, 25)

P value

0.5

0.4996

0.5007

0.4911

0.0204

0.0903

0.0483

0.1399

Infclin_sim

3 (0, 7)

TDF2

Infsim

1 (0, 5)

1 (0, 6)

1 (0, 4)

1 (0, 4)

11 (2, 27)

7 (1, 18)

9 (1, 22)

6 (1, 16)

P value

0.3986

0.3975

0.3487

0.3460

0.1412

0.2586

0.1936

0.3246

Infclin_sim

2 (0, 7)

Partners-PrEP

Infsim

3 (0, 8)

3 (0, 8)

2 (0, 6)

2 (0, 6)

20 (7, 41)

14 (4, 28)

16 (5, 34)

11 (3, 24)

P value

0.4412

0.4437

0.5106

0.5116

0.0204

0.0606

0.0357

0.0972

Infclin_sim

8 (0, 26)

VOICE

Infsim

2 (0, 5)

2 (0, 6)

1 (0, 4)

1 (0, 4)

14 (7, 23)

9 (4, 17)

12 (5, 20)

8 (3, 15)

P value

0.3102

0.3123

0.2952

0.2881

0.2958

0.4448

0.3766

0.5039

PrEP efficacy (bottom-up)

93.1 (38.8, 99.9)

93.0 (38.4, 99.9)

95.1 (45.4, 100)

95.2 (47.4, 100)

37.6 (3.9, 60.9)

60.0 (16.1, 84.6)

50.4 (5.1, 71.8)

67.1 (20.5, 87.8)

FEM-PrEP

Infclin_sim

8 (0, 17)

Infsim

1 (0, 4)

1 (0, 4)

1 (0, 3)

1 (0, 3)

9 (3, 17)

6 (1, 12)

7 (2, 14)

5 (1, 10)

P value

0.2171

0.2193

0.2029

0.2039

0.4529

0.4435

0.5122

0.3976

  1. The first row depicts the distinct hypotheses in ‘traffic-light’ notation and the second row the ‘bottom-up’ estimated mean PrEP efficacy (95% CI) in individuals with detectable plasma TFV for that hypothesis (compare Fig. 4). The columns show the mean number of infected individuals (95% CI), from ‘top-down’ clinical trial simulation, Infclin_sim (depicted in bold; compare Fig. 1), and from ‘bottom-up’ simulation, Infsim, with deduced PrEP efficacies for the distinct hypotheses. The P value tests for differences in the number of infected individuals deduced from bottom-up modeling versus clinical data. The P value was empirically calculated by computing the proportion of 106 simulation pairs, for which the null hypothesis was true (that is, H0: P = no. of simulations where infected individuals from hypothesis X was equal to or less than clinical estimate/total no. of simulations; H1: no. of infected individuals from hypothesis X more than the corresponding clinical estimate). Crosses visually indicate whether the statistical test provided trends (single unfilled red cross) or statistically different predictions at P < 0.05 (filled single- or double-red cross).
Back to article page