1. Effect of Immediate-initiation ART on risk of severe bacterial infections in HIV positive people with CD4 count > 500/mm3
Jemma O’Connor, Fred Gordin, Andrew Phillips, Brian Angus, David Cooper, Beatriz Grinsztejn, Gustavo Lopardo, Satyajit Das, Aimee Wilkin, Elbushra Herieka, David Jilich, Hartwig Klinker, Pacharee Kantipong, Karin Klingman, Jens Lundgren for the INSIGHT START study group

2. Background
The primary publication for the START study reported a reduced risk of bacterial infectious disorder (BID) events in the immediate-initiation group1.
Bacterial diseases are common opportunistic diseases that occur frequently in PLWH2,3. It is therefore important to characterise the effect of early ART on the incidence of bacterial diseases.
Neutrophils play a key role in the immune response to bacterial infections4. We therefore hypothesised that ART may induce a beneficial effect on neutrophil count and that this might mediate the relationship between ART and reduced risk of BID.
1Lundgren D, Babiker AG, Gordin F, Emery S, Sharma S, Avihingsanon AC, et al. Initiation of Antiretroviral Therapy in Early Asymptomatic HIV Infection. N Engl J Med 2015;373(9):
2Anglaret X, Messou E, Ouassa T, Toure S, Dakoury-Dogbo N, Combe P, et al. Pattern of bacterial diseases in a cohort of HIV-1 infected adults receiving cotrimoxazole prophylaxis in Abidjan, Cote D'Ivoire. AIDS 2003;17(4):
3Bjerk SM, Baker JV, Emery S, Neuhaus J, Angus B, Gordin FM, et al. Biomarkers and Bacterial Pneumonia Risk in Patients with Treated HIV Infection: A Case-Control Study. Plos One 2013;8(2).
4Elliott A, Schoenlaub L, Freches D, Mitchell W, Zhang GQ. Neutrophils Play an Important Role in Protective Immunity against Coxiella burnetii Infection. Infection and Immunity 2015;83(8):

3. Objective
We expand on the findings presented in the START publication by reporting on all bacterial events including bacterial pneumonia events, previously unpublished.
Our aim was to investigate whether the reduction in bacterial events observed in the immediate-initiation group is explained by potential increases in neutrophil counts, and the increases in CD4 counts that were observed in this group.
Also, to determine other factors associated with the incidence of bacterial events.

4. Methods (1)
START randomised people with CD4 counts above 500 cells/mm3 to either start ART immediately, or defer ART until CD4 count dropped to 350 cells/mm3 - or until the development of an AIDS-defining event or a serious non-AIDS condition which required treatment with ART.
Severe bacterial infection (SBI) was defined as a composite of bacterial pneumonia (as reviewed by the INSIGHT Endpoint Review Committee (ERC)) or any BID (as reported in the START primary publication which satisfied at least one of the following three criteria: grade 4 severity, requiring unscheduled hospitalisations or causing death.
Follow-up was censored at the time of the SBI event, the date of the last study visit or the date of death (censoring date for the analysis).

5. Methods (2)
A longitudinal mixed model was fit using PROC MIXED in SAS to assess the difference in the mean change in laboratory markers from study entry, between the treatment groups.
A separate mixed model was fit for each of the following laboratory markers: CD4 count, neutrophil count, total lymphocyte count, haemoglobin, platelet count, and albumin level, using change since study entry as the dependent variable.
Models were adjusted for the baseline value of the laboratory marker and the visit number.

6. Methods (3)
Cox proportional hazards regression was used to model the time to SBI.
Baseline categorical covariates:
Treatment group (immediate-initiation, deferred-initiation)
Sex (male, female)
Mode of HIV acquisition (MSM, opposite sex, IDU/other/unknown)
Race (Asian, Black, Latino/Hispanic, Other, White)
Baseline continuous covariates:
Age
Time since HIV diagnosis
Time-updated categorical covariates:
BMI (kg/m2) (<18.5, , , ≥35)
smoking status
Reference groups for categorical covariates are emboldened.
All models were stratified by a six-category variable describing the geographic region of residence (Africa, Asia, Australia, Europe and Israel, North America, South America and Mexico).
Time-updated continuous covariates
Total lymphocyte count
CD4 count
Haemoglobin
HIV RNA viral load level (log10)
Platelet count
Neutrophil count
Albumin level

7. Results (1) Characteristic N (%) or Median [IQR]
All participants (n=4685)
Treatment group
Immediate-initiation ART
Deferred-initiation ART
2326 (49.6)
2359 (50.4)
Age
36 [29, 44]
Female sex
1257 (26.8)
Race
Black
White
1410 (30.1)
2086 (44.5)
Body mass index (kg/m2)
24.7 [22.3, 28.0]
Mode of HIV acquisition
MSM
Sex with person of opposite sex
2586 (55.2)
1790 (38.2)
Smoking status
1317 (28.1)
All 4,685 START participants were included in the analyses. Table 1 displays baseline and time-updated participant characteristics.

8. Results (2) Characteristic N (%) or Median [IQR]
All participants (n=4685)
CD4 count (cells/mm3)
696 [571, 839]
HIV RNA (log10 copies/mL)
3.7 [3.0, 4.3]
Neutrophil count (cells/mm3 x102)
29.6 [23.0, 37.7]
Total lymphocyte count (cells/mm3 x102)
21.6 [18.4, 25.6]
Haemoglobin (g/dL x10)
1.5 [1.3, 1.5]
Platelet count (cells/mm3x105)
2.4 [2.0, 2.8]
Albumin level (g/dL)
4.4 [4.2, 4.6]
All 4,685 START participants were included in the analyses. Table 1 displays baseline and time-updated participant characteristics.

9. Results (3)
Ninety-four people experienced at least one SBI (immediate-initiation group n=28, deferred-initiation group n=66) during 13,890 person-years at risk (incidence of 0.68 per 100 person-years of follow-up (PYFU), 95% CI 0.54–0.81).
Of the 94 people who developed an SBI, 50 experienced a BID (immediate-initiation group n=14, deferred-initiation group n=36) and 48 developed bacterial pneumonia (immediate-initiation group n=14, deferred-initiation group n=34).
The two components of SBI do not total 94 because four people developed both a BID and bacterial pneumonia.

10. Results (4)
Baseline median (interquartile range) values of biomarkers investigated for their association with SBI were as follows CD4 count, 651 cells/mm3 (584, 765); neutrophil count, cells/mm3 (2148.8, ); total lymphocyte count, cells/mm3 (1830.0, ); haemoglobin, 14.5 g/dL (13.4, 15.3); platelet count, cells/mm3 x 103 (195.0, 273.0); and albumin level, 4.4 g/dL (4.1, 4.6).
During follow-up, the change in biomarker values since study entry was significantly higher in the immediate-initiation group, than in the deferred-initiation group, for all biomarkers investigated. The average estimated change from baseline (95% confidence interval (CI)) was as follows: CD4 count, cells/mm3 (185.1, 203.3), p<.0001; neutrophil count, cells/mm3 (258.6, 384.2), p<.0001; total lymphocyte count, 55.3 cells/mm3 (23.3, 87.3), p=0.0007; haemoglobin, 0.19 g/dL (0.14, 0.24), p<.0001; platelet count, 17.8 cells/mm3 x 103 (15.3, 20.2), p<.0001; and albumin level, 0.09 g/dL (0.07, 0.10), p<.0001.

11. Results (5)
Univariable and multivariable hazard ratios (HRs) and 95% CIs for a selection of factors that were included in the Cox model are presented in Figure 2. In univariable analysis immediate-initiation ART was associated with a reduced risk of SBI (HR 0.41, 95% CI 0.27–0.65, p<.0001). However, in multivariable analysis (after adjustment for all covariates listed in the methods) the HR was no longer significantly below one (HR 0.75, 95% CI 0.42–1.32, p=0.30). Adjustment for time-updated CD4 count in the multivariable model was responsible for the attenuation of the HR for treatment group.
In multivariable analysis a BMI above 35 kg/m2 (HR 3.38, 95% CI 1.74–6.57, p<.0001) was associated with an increased risk of SBI when compared to a reference group of a BMI between 18.5 and 24.9 kg/m2. Higher time-updated CD4 count (HR per 100 cells/mm3 0.83, 95% CI 0.74–0.92, p=0.0004) was associated with a reduced risk of SBI. Time-updated neutrophil count was not associated with risk of SBI (HR per 10,000 cells/mm3 0.94, 95% CI 0.20–4.32, p=0.96). Other factors that were included in the multivariable model but were not associated with SBI include age (p=0.17), sex (p=0.24), race (p=0.14), smoking status (p=0.34), mode of HIV infection (p=0.48), time since HIV diagnosis (p=0.82), hepatitis B (p=0.93) and hepatitis C status (p=0.43), HIV RNA viral load (p=0.46), total lymphocyte count (p=0.09), hemoglobin (p=0.66), platelet count (p=0.55), and albumin level (p=0.43).

12. Conclusions (1)
Adjustment for time-updated CD4 count in the multivariable model was responsible for the attenuation of the treatment group HR which suggests that the effect of immediate ART is at least partially mediated by the increase in CD4 count in the immediate ART group.
In contrast, we found no evidence that the effect was mediated by the increases seen in neutrophil count with early ART, as we initially hypothesised.
Our findings are consistent with previous studies which demonstrated the efficacy of early ART in reducing the risk of bacterial diseases5,6.
5 Danel C, Moh R, Gabillard D, Badje A, Le Carrou J, Ouassa T, et al. A Trial of Early Antiretrovirals and Isoniazid Preventive Therapy in Africa. N Engl J Med 2015;373(9):
6 Severe P, Juste MAJ, Ambroise A, Eliacin L, Marchand C, Apollon S, et al. Early versus Standard Antiretroviral Therapy for HIV-Infected Adults in Haiti. N Engl J Med 2010;363(3):

13. Conclusions (2)
Our findings provide further evidence of a clinical benefit of early initiation of ART. As demonstrated in Figure 1, biomarkers in untreated HIV infection begin to decline even in people with high CD4 counts.
Controlling viral replication and maintaining good immunological function at early stages of HIV infection is important for prevention of a range of bacterial infections.
In summary, our study demonstrates the protective effect of immediate ART in reducing the risk of a broad spectrum of serious bacterial infections among HIV positive people.
The effect of immediate ART appears to be mediated in part by ART-induced increases in the CD4 count, but not by ART-induced increases in neutrophil count.

14. Acknowledgements