Abstract
Objectives
Tuberculosis (TB) is common in people living with HIV, leading to worse clinical outcomes including increased mortality. We investigated risk factors for developing TB following HIV diagnosis.
Design
Adults aged at least 15 years first presenting to health services for HIV care in England, Wales or Northern Ireland from 2000 to 2014 were identified from national HIV surveillance data and linked to TB surveillance data.
Methods
We calculated incidence rates for TB occurring more than 91 days after HIV diagnosis and investigated risk factors using multivariable Poisson regression.
Results
A total of 95 003 adults diagnosed with HIV were followed for 635 591 person-years; overall incidence of TB was 344 per 100 000 person-years (95% confidence interval 330–359). TB incidence was high for people who acquired HIV through injecting drugs [PWID; men 876 (696–1104), women 605 (365–945)] and black Africans born in high TB incidence countries [644 (612–677)]. The adjusted incidence rate ratio for TB amongst PWID was 4.79 (3.35–6.85) for men and 6.18 (3.49–10.93) for women, compared with MSM. The adjusted incidence rate ratio for TB in black Africans from high-TB countries was 4.27 (3.42–5.33), compared with white UK-born individuals. Lower time-updated CD4+ cell count was associated with increased rates of TB.
Conclusion
PWID had the greatest risk of TB; incidence rates were comparable with those in black Africans from high TB incidence countries. Most TB cases in PWID were UK-born, and likely acquired TB through transmission within the United Kingdom. Earlier HIV diagnosis and quicker initiation of antiretroviral therapy should reduce TB incidence in these populations.
Keywords: cohort studies, coinfection, HIV, observational study, risk factors, tuberculosis
Introduction
Tuberculosis (TB) and HIVare leading causes of morbidity and mortality. Globally, in 2014, there were 1.2 million new cases of TB in people living with HIV (PLHIV), accounting for one in eight TB diagnoses [1]. TB was responsible for one in three HIV-related deaths in 2014.
In England, Wales and Northern Ireland, 25% of AIDS defining illnesses from 2001 to 2010 were TB [2]. The rate of TB disease in PLHIV in the United Kingdom was estimated as 328/100 000 person-years between 1996 and 2005 [excluding patients diagnosed with TB and HIV simultaneously (within 91 days)] [3], and 669/100 000 person-years across all groups 2007–2011 [4]. Estimated TB incidence in the general population is much lower; 10/100 000 population in 2015 [5].
Previous studies in the United Kingdom have found higher rates of TB in PLHIV who acquired HIVabroad, or had black African or Indian/Pakistani/Bangladeshi ethnicity, than in white and UK-born populations [3,6]. TB incidence decreased with increasing CD4+ cell count at HIV diagnosis and was lower for individuals on antiretroviral therapy (ART). However [6], was limited in its implications for UKTB–HIV control as it was restricted to heterosexuals and did not adjust for time on ART, which is known to be linked to TB incidence [4]. It also included patients diagnosed simultaneously with TB and HIV, many of whom are only diagnosed with HIV as a result of their TB diagnosis [6]. Furthermore, the UK-CHIC study [3] did not provide estimates of TB incidence in people who inject drugs (PWID).
TB incidence in HIV-positive PWID in the 1980s and 1990s was very high [7]; however, the link between TB and HIV-positive PWID in the ARTera is less clear. Five cohort studies found TB rates were elevated by a factor of 1.7–4.4 when compared with MSM or people who do not inject drugs [8–12], whereas one cohort [13] and one cross-sectional study [14] found no significantly increased risk. In the United Kingdom, PWID are typically diagnosed with HIV late [15] and have high rates of death [16,17], despite good levels of ART coverage (90%), similar to other risk groups [18]. No recent studies in the United Kingdom have investigated the risk of TB for PWID. This study aimed to investigate risk factors for developing TB following HIV diagnosis, including HIV acquisition by IDU, to address the paucity of evidence in resource-rich countries in the ART era.
Methods
Study population
Adults (aged 15 years or older) notified to Public Health England (PHE)’s HIV and AIDS Reporting System (HARS), first presenting with HIV to health services in England, Wales and Northern Ireland between 2000 and 2014, were included. HARS comprises four linked data sources: reports of all new HIV/AIDS diagnoses and deaths, national laboratory data for CD4+ cell count, annual reporting of demographic and clinical information of PLHIV from all national clinics and death reports from the Office of National Statistics [17,19].
Outcome: tuberculosis disease diagnosed from 2000 to 2014
TB cases included both culture-confirmed and presumptive (clinical and radiological signs, including a response to specific therapy) diagnoses. UK HIV and TB surveillance are undertaken separately, necessitating data linkage to analyse coinfection. TB cases across England, Wales and Northern Ireland are reported to the PHE’s Enhanced Tuberculosis Surveillance (ETS) system. To identify PLHIV with TB disease, HARS and ETS data were linked using a probabilistic matching algorithm (adapted from [20]), with supplementary deterministic matching to accept/reject borderline matches [21].
Incident TB was defined as TB disease notified to ETS or reported to HARS as a new AIDS-defining illness, which was diagnosed more than 91 days after HIV diagnosis. TB cases diagnosed within 91 days of HIV diagnosis were considered simultaneous diagnoses, to differentiate patients who were not aware of their HIV infection prior to their TB diagnosis. TB cases diagnosed more than 91 days before HIV were considered existing disease. A 91-day threshold for defining simultaneous diagnoses was a pragmatic choice to account for delays in diagnosis and reporting, and to exclude ART-induced unmasking immune reconstitution inflammatory syndrome.
Exposure variables
We included demographic [age at HIV diagnosis, sex, ethnicity, country of birth, TB incidence in country of birth, route of HIV infection, year of HIV diagnosis, index of multiple deprivation (IMD) decile] and clinical (viral load at first presentation, and time-updated CD4+ cell count and ART initiation) exposure variables. IMD score deciles represent relative levels of deprivation of income, employment, health, education, housing and services, crime and living environment for small areas in England and Wales, where 1 = most deprived and 10 = least deprived [22,23].
Composite variables were created combining ethnicity and country of birth or sex and infection route due to mutually exclusive combinations (e.g. being a woman and a MSM is impossible) and known associations. As a proxy TB exposure, countries of birth outside the United Kingdom were grouped by TB incidence; ‘high incidence’ was defined as more than 40 cases/100 000 adult population in 2013. The most recent IMD data for each country between 2000 and 2014 were used, 2010 for England and 2014 for Wales.
Statistical analysis
Data were analysed in Stata version 13.1 (StataCorp LP, College Station, Texas, USA). Descriptive analyses of the cohort were undertaken. To investigate risk factors for developing TB, we calculated incidence rates of TB per 100 000 person-years follow-up and assessed TB incidence over time using Nelson–Aalen cumulative hazard plots. We estimated incidence rate ratios (IRRs) using univariable and multivariable Poisson regression models, offset by follow-up time, Cox regression was precluded as our data did not satisfy the proportional hazards assumption for key variables such as route of HIV infection. Individuals diagnosed with TB 91 days or less after HIV diagnosis were excluded to investigate subsequent TB. Follow-up began 92 days from date of HIV diagnosis or first presentation to UK health services and ended on the date of TB diagnosis, death or 31 December 2014, whichever was earliest. CD4+ cell count and ART initiation were included as time-updated covariates. Incidence rates for different CD4+ strata were calculated using the number of days from each CD4+ cell count to the date of the next CD4+ cell count for each patient. To compare incidence between ART-naive patients and patients who had initiated ART, we split each patient’s follow-up period at the date they first initiated ART to calculate the duration of ART-naive person-time, and person-time having initiated ART.
Potential confounders and effect modifiers were prospectively identified [24]. Our causal framework determined that viral load should be excluded from the multivariable model because of the potential for causal loops between viral load and CD4+ cell count, which could not be adequately accounted for in the data available. We excluded patients missing data on one or more variables. Linearity (of age, CD4+ cell count and year of HIV diagnosis) and statistical interactions (between ART status and CD4+ cell count) were assessed using likelihood ratio tests. As we were not investigating a single ‘main’ exposure variable, there were no confounders in the traditional sense, and therefore, the multivariable model was informed by a causal inference framework defined a priori. To assess the likely impact of missing data, we compared the distributions of age, sex, route of HIV infection, CD4+ cell count and ethnicity/country of birth for cases with missing versus complete data on infection route, CD4+ cell count, IMD score and country of birth. Statistical interactions were considered significant at P less than 0.05. All stated confidence intervals (CIs) are two-sided 95% CIs.
Planned sensitivity analyses investigated the impact of using a 6-month threshold (182 days) for simultaneous diagnosis, excluding weaker matches between HARS and ETS, and excluding people who acquired HIV infection through mother-to-child transmission, as the dataset only contained adults and so individuals infected through this route could be missing 15-year follow-up.
Ethics, consent and permissions
This analysis was approved by the University College London (UCL) student Research Ethics Committee (5683/001). PHE has authority under the Health and Social Care Act 2012 to hold and analyse national surveillance data for public health and research purposes.
Role of the funding source
The funding source had no involvement in the study design; the collection, analysis and interpretation of the data; the writing of the report or the decision to submit the article for publication.
Results
Description of coinfected patients
Between 2000 and 2014, 102 202 adults were newly diagnosed with HIV, among whom 5649 (6%) had TB, 3103 (55%) were simultaneously diagnosed with TB and HIV, 2187 (39%) developed TB after more than 91 days and 359 (6%) were diagnosed with TB first (Table 1).
Table 1. Tuberculosis diagnoses in people notified with HIV from 2000 to 2014 in England, Wales and Northern Ireland, and the incidence rates of tuberculosis in people who were diagnosed with tuberculosis more than 91 days following HIV diagnosis.
| HIV cases | TB cases Total | Prior to HIV diagnosis | Simultaneous with HIV diagnosis | Following HIV diagnosis | ||||
|---|---|---|---|---|---|---|---|---|
| n (column %) | n (row %) | n (row %) | n (row %) | n (row %) | PY follow-up | Incidence ratea (95% CI) | Incidence rate after 1 year from HIV diagnosisa (95% CI) | |
| Total | 102 202 | 5649 (5.5) | 359 (6) | 3103 (55) | 2187 (39) | 635 591 | 344 (330–359) | 247 (234–260) |
| Route of HIV infection | ||||||||
| MSM | 35 879 (35.1) | 462 (1.3) | 31 (7) | 195 (42) | 236 (51) | 212 844 | 111 (98–126) | 86 (74–100) |
| Heterosexual men | 18 738 (18.3) | 2013 (10.7) | 127 (6) | 1205 (60) | 681 (34) | 113 802 | 598 (555–645) | 402 (365–443) |
| Heterosexual women | 30 489 (29.8) | 2815 (9.2) | 167 (6) | 1520 (54) | 1128 (40) | 201 644 | 559 (528–593) | 404 (376–434) |
| Men who inject drugs | 1453 (1.4) | 132 (9.1) | 5 (4) | 55 (42) | 72 (55) | 8216 | 876 (696–1104) | 660 (499–873) |
| Women who inject drugs | 532 (0.5) | 35 (6.6) | 1 (3) | 15 (43) | 19 (54) | 3138 | 605 (365–945) | 526 (295–868) |
| Blood/tissue transfer | 505 (0.5) | 58 (11.5) | 6 (10) | 31 (53) | 21 (36) | 2928 | 717 (468–1100) | 527 (288–883) |
| Mother-to-child | 253 (0.2) | 15 (5.9) | 1 (7) | 4 (27) | 10 (67) | 863 | 1159 (556–2131) | 836 (307–1819) |
| Unknownb | 14 353 (14.0) | 119 (0.8) | 21 (18) | 78 (66) | 20 (17) | 92 155 | 22 (14–34) | 13 (7–24) |
| Ethnicity/country of birth | ||||||||
| White, UK-born | 27 320 (26.7) | 359 (1.3) | 24 (7) | 161 (45) | 174 (48) | 160 488 | 108 (93–126) | 84 (70–100) |
| Black African, UK-born | 947 (0.9) | 51 (5.4) | 6 (12) | 25 (49) | 20 (39) | 5556 | 360 (232–558) | 260 (151–448) |
| Other ethnicity, UK-born | 2687 (2.6) | 72 (2.7) | 6 (8) | 26 (36) | 40 (56) | 14 948 | 268 (196–365) | 217 (151–313) |
| Ethnicity unknown, UK-born | 403 (0.4) | 3 (0.7) | 0 (0) | 3 (100) | 0 (0) | 544 | 0 (0–678)c | 0 (0–876)c |
| Born in low-TB incidence country | 11 551 (11.3) | 245 (2.1) | 11 (4) | 118 (48) | 116 (47) | 65 376 | 177 (148–213) | 125 (99–157) |
| White, born in high-TB incidence country | 7461 (7.3) | 126 (1.7) | 4 (3) | 71 (56) | 51 (40) | 47 593 | 107 (81–141) | 84 (61–116) |
| Black African, born in high-TB incidence country | 35 035 (34.3) | 3877 (11.1) | 223 (6) | 2142 (55) | 1512 (39) | 234 853 | 644 (612–677) | 454 (426–483) |
| Other ethnicity, born in high-TB incidence country | 6756 (6.6) | 518 (7.7) | 52 (10) | 311 (60) | 155 (30) | 35 614 | 435 (372–509) | 290 (236–356) |
| Ethnicity unknown, born in high-TB incidence country | 1140 (1.1) | 13 (1.1) | 2 (15) | 10 (77) | 1 (8) | 7556 | 13 (0–74)c | 15 (0–81) |
| White, country of birth unknown | 3065 (3.0) | 52 (1.7) | 4 (8) | 31 (60) | 17 (33) | 23 968 | 71 (41–114) | 54 (28–95) |
| Other ethnicity, country of birth unknown | 4226 (4.1) | 300 (7.1) | 23 (8) | 181 (60) | 96 (32) | 33 093 | 290 (237–354) | 210 (164–268) |
| Both unknownb | 1611 (1.6) | 33 (2.0) | 4 (12) | 24 (73) | 5 (15) | 6002 | 83 (27–194) | 39 (5–141) |
| Age at HIV diagnosis (years) | ||||||||
| 15–24 | 11 513 (11.3) | 437 (3.8) | 25 (6) | 173 (40) | 239 (55) | 73 647 | 325 (286–368) | 260 (224–302) |
| 25–34 | 38 910 (38.1) | 2227 (5.7) | 129 (6) | 1121 (50) | 977 (44) | 261 955 | 373 (350–397) | 280 (260–302) |
| 35–44 | 31 894 (31.2) | 1944 (6.1) | 133 (7) | 1147 (59) | 664 (34) | 199 946 | 332 (308–358) | 232 (211–255) |
| 45–64 | 18 357 (18.0) | 973 (5.3) | 64 (7) | 619 (64) | 290 (30) | 93 708 | 309 (276–347) | 183 (156–214) |
| 65+ | 1479 (1.4) | 68 (4.6) | 8 (12) | 43 (63) | 17 (25) | 5764 | 295 (172–472) | 99 (172–472) |
| CD4+ cell count at HIV diagnosisd (incidence rates are calculated for time-updated CD4+) | ||||||||
| ≥500 | 20 153 (19.7) | 381 (1.9) | 30 (8) | 88 (23) | 263 (69) | 187 994 | 139 (123–157) | 122 (106–139) |
| 350–499 | 14 801 (14.5) | 455 (3.1) | 34 (7) | 133 (29) | 288 (63) | 114 505 | 259 (231–290) | 270 (241–304) |
| 200–349 | 16 282 (15.9) | 861 (5.3) | 61 (7) | 388 (45) | 412 (48) | 81 579 | 527 (480–579) | 454 (407–506) |
| 100–199 | 9514 (9.3) | 1039 (10.9) | 79 (8) | 613 (59) | 347 (33) | 24 933 | 1356 (1219–1508) | 785 (673–916) |
| 50–99 | 5039 (4.9) | 718 (14.2) | 35 (5) | 525 (73) | 158 (22) | 6247 | 2209 (1870–2610) | 1072 (817–1407) |
| 0–49 | 8731 (8.5) | 1241 (14.2) | 63 (5) | 956 (77) | 222 (18) | 5166 | 2788 (2368–3282) | 891 (648–1224) |
| Unknownb | 27 682 (27.1) | 954 (3.4) | 57 (6) | 400 (42) | 497 (52) | – | – | – |
| Viral load at diagnosis (copies/ml) | ||||||||
| ≤200 | 13 951 (13.7) | 580 (4.2) | 51 (9) | 311 (54) | 218 (38) | 63 098 | 345 (303–395) | 227 (190–270) |
| >200 | 58 824 (57.6) | 3735 (6.3) | 229 (6) | 2050 (55) | 1456 (39) | 339 621 | 428 (407–451) | 305 (286–325) |
| Unknownb | 29 427 (28.8) | 1334 (4.5) | 79 (6) | 742 (56) | 513 (38) | 232 872 | 221 (202–241) | 170 (153–188) |
| Ever started ART (incidence rates calculated for time-updated ART)e | ||||||||
| No | 32 207 (31.5) | 809 (2.5) | – | – | 1336e | 261 662 | 511 (484–539) | 337 (314–362) |
| Yes | 69 995 (68.5) | 4840 (6.9) | – | – | 851e | 373 929 | 228 (213–243) | 188 (174–203) |
| IMD decile | ||||||||
| 1 | 13 498 (13.2) | 900 (6.7) | 64 (7) | 470 (52) | 366 (41) | 75 516 | 485 (437–537) | 343 (301–390) |
| 2 | 15 075 (14.8) | 920 (6.1) | 66 (7) | 510 (55) | 344 (37) | 86 339 | 398 (358–443) | 286 (251–327) |
| 3 | 12 746 (12.5) | 688 (5.4) | 53 (8) | 385 (56) | 250 (36) | 72 760 | 344 (304–389) | 247 (212–288) |
| 4 | 9150 (9.0) | 474 (5.2) | 29 (6) | 273 (58) | 172 (36) | 52 758 | 326 (281–379) | 222 (183–268) |
| 5 | 6732 (6.6) | 336 (5.0) | 22 (7) | 191 (57) | 123 (37) | 37 961 | 324 (272–387) | 235 (189–293) |
| 6 | 5233 (5.1) | 253 (4.8) | 18 (7) | 134 (53) | 101 (40) | 29 630 | 341 (280–414) | 238 (186–304) |
| 7 | 3870 (3.8) | 164 (4.2) | 10 (6) | 89 (54) | 65 (40) | 21 596 | 301 (236–384) | 233 (174–312) |
| 8 | 3304 (3.2) | 140 (4.2) | 6 (4) | 83 (59) | 51 (36) | 17 934 | 290 (221–381) | 207 (147–291) |
| 9 | 2809 (2.7) | 110 (3.9) | 7 (6) | 64 (58) | 39 (35) | 15 846 | 246 (180–337) | 163 (108–245) |
| 10 | 2217 (2.2) | 97 (4.4) | 3 (3) | 52 (54) | 42 (43) | 11 925 | 352 (260–477) | 274 (190–394) |
| Unknownb | 27 568 (27.0) | 1567 (5.7) | 81 (5) | 852 (54) | 634 (40) | 213 326 | 297 (274–321) | 217 (197–238) |
ART, antiretroviral therapy; CI, confidence interval; IMD, index of multiple deprivation; PLHIV, people living with HIV; PWID, people who inject drugs; PY, person-years; TB, tuberculosis.
Incidence is given per 100 000 population aged at least 15 years, per year.
Unknown strata includes both unknown and missing data.
One-sided, 97.5% CI.
Incidence rates are calculated for time-updated CD4+ cell count.
Of the 5649 PLHIV who got TB, 809 never initiated ART. However, of the 2187 who got TB > 91 days after their HIV infection, 1336 had not initiated TB at the time of their HIV diagnosis.
Of people with TB who acquired HIV infection through heterosexual sex, over half were diagnosed simultaneously with TB and HIV, 60% for men and 54% for women. In contrast, more TB cases in MSM and PWID were diagnosed more than 91 days after diagnosis of HIV infection (51 and 54%, respectively). The proportion of TB cases occurring after HIV diagnosis was highest in white, UK-born individuals (179/359, 48%) and those born in low TB incidence countries (116/245, 47%); these two groups comprise 38% of the cohort.
Incidence of tuberculosis following HIV diagnosis
A total of 95 003 adults were TB-free 92 days after presenting for HIV care, with a total of 635 591 person-years follow-up. Median age at HIV diagnosis was 34 years [interquartile range (IQR) 28–42] and median CD4+ cell count was 340 cells/µl (IQR 170–527). A total of 95% of patients had more than one CD4+ cell count (median 14).
Overall TB incidence was 344/100 000 person-years (95% CI: 330–359, Table 1). The probability of developing TB was highest in the year following HIV diagnosis and then decreased (Fig. 1a). Incidence was high in PWID [men 876/100 000 (696–1104/100 000); women 605/100 000 (386–949/100 000)] and heterosexuals [men 598/100 000 (555–645/100 000); women 559 (528–593/100 000)], particularly compared with MSM [111/100 000 (98–126/100 000)]. The largest differences in cumulative probability of TB diagnosis between PWID, black Africans from high-TB incidence countries and MSM were in the first 2 years following HIV diagnosis; the rate of diagnosis remained relatively constant across all groups thereafter (Fig. 1b).
Fig. 1. Cumulative hazard plot of the probability of developing tuberculosis from more than 91 days following HIV diagnosis.
TB incidence increased with decreasing time-updated CD4+ cell count, from 139/100 000 (123–157/100 000) for those with CD4+ cell count at least 500 cells/ml to 2788/100 000 (2368–3282/100 000) for those with CD4+ cell count less than 50 cells/µl. TB incidence was 511/100 000 (484–539/100 000) in people who had never received ART (26% of all person-years) compared with 228/100 000 (213–243/100 000) in people who had (74% of person-years). TB incidence was higher for PWID who had never initiated ART [1478/100 000 (95% CI 1157–1888/100 000)] than for black Africans from high-TB incidence countries who had never initiated ART [991/100 000 (929–1058/100 000)], although incidence rates following ART initiation were similar in both groups [384/100 000 (264–560/100 000) for PWID versus 421 (389–456/100 000) for black Africans]. TB incidence was highest in those living in areas of England and Wales with the lowest decile of IMD score [485/100 000 (437–537/100 000)].
Factors associated with developing tuberculosis disease
A total of 62 684 individuals with complete case data and a TB-free follow-up period of more than 91 days following HIV diagnosis were included in the time-to-event analysis. There were a total of 414 714 person-years of follow-up (median follow-up 7.1 years, IQR 3.6–10.4), during which there were 1591 TB diagnoses (Table 2). The median duration of follow-up was 7.3 years (IQR 3.9– 10.4) for patients who did not develop TB, whereas patients who did develop TB did so in a median of 0.2 years (IQR 0.1–0.5). Black African patients born in high-TB countries had a slightly higher median follow-up period of 8.2 (4.7–11.0) years, compared with 6.3 (3.1–9.8) for MSM and 6.5 (3.4–9.8) for PWID, as black Africans were more likely to be diagnosed earlier in the study period than PWID or MSM.
Table 2. Univariable and multivariable incidence rate ratios from Poisson regression of factors associated with incident tuberculosis disease (>91 days after HIV diagnosis) among people living with HIV in England, Wales and Northern Ireland from 2000 to 2014.
| TB cases | PY | Univariable IRR (95% CI) |
Multivariable IRR (95% CI) |
|
|---|---|---|---|---|
| Route of HIV infection | ||||
| MSM | 184 | 172 708 | 1.00 (P < 0.001) | 1.00 (P < 0.001) |
| Male heterosexual | 474 | 82 460 | 5.40 (4.55–6.40) | 1.70 (1.38–2.10) |
| Female heterosexual | 837 | 148 391 | 5.29 (4.51–6.21) | 1.86 (1.51–2.29) |
| Male PWID | 61 | 5895 | 9.71 (7.27–12.97) | 5.47 (4.07–7.35) |
| Female PWID | 16 | 2514 | 5.97 (3.58–9.95) | 4.59 (2.75–7.67) |
| Blood/tissue transfer | 14 | 2251 | 5.84 (3.39–10.05) | 2.70 (1.55–4.71) |
| Mother-to-child | 5 | 494 | 9.51 (3.91–23.11) | 2.80 (1.13–6.97) |
| Ethnicity/country of birth | ||||
| White, UK-born | 134 | 127 453 | 1.00 (P < 0.001) | 1.00 (P < 0.001) |
| Black African, UK-born | 13 | 4317 | 2.86 (1.62–5.06) | 1.97 (1.10–3.51) |
| Other ethnicity, UK-born | 31 | 12 040 | 2.45 (1.66–3.62) | 1.92 (1.29–2.84) |
| Ethnicity unknown, UK-born | 0 | 252 | a | a |
| Born in low-TB incidence country | 98 | 53 647 | 1.74 (1.34–2.25) | 1.33 (1.02–1.73) |
| White, born in high-TB incidence country | 38 | 12 606 | 2.87 (2.00–4.11) | 2.19 (1.53–3.15) |
| Black African, born in high-TB incidence country | 1093 | 148 017 | 7.02 (5.87–8.40) | 4.27 (3.42–5.33) |
| Other ethnicity, born in high-TB incidence country | 105 | 22 219 | 4.50 (3.48–5.80) | 3.36 (2.57–4.39) |
| Ethnicity unknown, born in high-TB incidence country | 1 | 323 | 2.95 (0.41–21.07) | 1.35 (0.19–9.71) |
| White, country of birth unknown | 12 | 15 491 | 0.74 (0.41–1.33) | 0.52 (0.29–0.94) |
| Other ethnicity, country of birth unknown | 66 | 18 348 | 3.42 (2.55–4.59) | 1.60 (1.17–2.20) |
| CD4+ cell count (cells/µl) | ||||
| ≥500 | 259 | 185719 | 1.00 (P <0.001) | b |
| 350–499 | 293 | 113185 | 1.86 (1.57–2.19) | |
| 200–349 | 427 | 80 443 | 3.81 (3.26–4.44) | |
| 100–199 | 332 | 24 367 | 9.77 (8.30–11.49) | |
| 50–99 | 137 | 6093 | 16.12 (13.11–19.83) | |
| 0–49 | 143 | 4905 | 20.90 (17.04–25.64) | |
| Ever on ART | ||||
| No | 928 | 107477 | 1.00 (P <0.001) | b |
| Yes | 663 | 307 237 | 0.25 (0.23–0.28) | |
| Viral load at diagnosis (copies/ml) | ||||
| ≤200 | 154 | 43347 | 1.00 (P = 0.006) | – |
| >200 | 1063 | 261249 | 1.15 (0.97–1.36) | |
| Age at HIV diagnosis (years) | ||||
| 15–24 | 169 | 48 805 | 0.95 (0.79–1.13) | 0.92 (0.77–1.10) |
| 25–34 | 714 | 170 957 | 1.14 (1.02–1.28) | 1.06 (0.94–1.19) |
| 35–44 | 477 | 130441 | 1.00 (P < 0.001) | 1.00 (P = 0.332) |
| 45–64 | 220 | 61 028 | 0.99 (0.84–1.16) | 1.11 (0.95–1.31) |
| ≥65 | 11 | 3484 | 0.86 (0.47–1.57) | 0.92 (0.51–1.68) |
| Year of HIV diagnosis (for each year increase from 2000) | 1591 | 414 714 | 0.98 (0.97–1.00) P = 0.036 |
1.02 (1.00–1.04) P = 0.014 |
| IMD decile (England and Wales only) | ||||
| 1 | 264 | 51 685 | 1.00 (P < 0.001) | – |
| 2 | 269 | 63 391 | 0.83 (0.70–0.98) | |
| 3 | 193 | 54 955 | 0.69 (0.57–0.83) | |
| 4 | 127 | 38 159 | 0.65 (0.53–0.81) | |
| 5 | 83 | 26 725 | 0.61 (0.48–0.78) | |
| 6 | 78 | 20 986 | 0.73 (0.57–0.94) | |
| 7 | 47 | 15 254 | 0.60 (0.44–0.82) | |
| 8 | 38 | 12 644 | 0.59 (0.42–0.83) | |
| 9 | 24 | 10 743 | 0.44 (0.29–0.66) | |
| 10 | 32 | 8326 | 0.75 (0.52–1.09) | |
62 684 PLHIV were included in this analysis; 32 319 were excluded from the model due to missing data on ethnicity and country of birth, route of HIV infection, CD4+ cell count or age at HIV diagnosis. Viral load was not included in the multivariable model due to collinearity with CD4+ cell count and ART status. ART, antiretroviral therapy; CI, confidence interval; IMD, index of multiple deprivation; IRR, incidence rate ratio; PWID, people who inject drugs; PY, person years; TB, tuberculosis.
Not calculated as numerator was zero.
Interaction present between time-updated CD4+ cell count and time-updated ART status, see Tables 3 and 4.
All exposures were included in the multivariable Poisson regression model (Table 2), except viral load and IMD decile. IMD decile was excluded as there was a high degree of missing data and no association with the outcome in a multivariable model (Supplementary Tables 1–3, http://links.lww.com/QAD/B155). CD4+ cell count and age at HIV diagnosis were treated as categorical variables (tests for linearity P < 0.001, P 0.005, respectively), year of HIV diagnosis was treated as a linear variable (P > 0.05). There was a statistically significant interaction between time-updated CD4+ cell count and time-updated ART status (P < 0.001).
Compared with MSM, PWID had increased rates of TB [IRR for men 5.47 (95% CI 4.07–7.35); women 4.59 (2.75–7.67)]. Rates were also higher in those infected through heterosexual sex [men 1.70 (1.38–2.10); women 1.86 (1.51–2.29)]. UK-born black Africans [1.97 (1.10–3.51)] and people of other ethnicities [1.92 (1.29–2.84)] were associated with increased incidence rates versus white UK-born individuals, as were those born in high TB incidence countries [black African 4.27 (3.42–5.33); white 2.19 (1.53–3.15); other ethnicities 3.36 (2.57–4.39)].
Overall, and within each stratum of CD4+ cell count, TB rates were greatly reduced in individuals who had received ART compared with those who had not (Table 3). When stratifying by ART initiation status, lower time-updated CD4+ cell count was strongly associated with increased TB rates (Table 4). For individuals who had never initiated ART, the IRR for TB increased with decreasing CD4+ cell count to 6.42 (4.87–8.46) for 0–49 cells/µl cf. at least 500 cells/ml. The increased risk at low CD4+ cell count was higher in individuals who had initiated ART, with an IRR of 44.21 (30.90–63.24) for 0–49 cells/µl, cf. at least 500 cells/µl.
Table 3. Multivariable Poisson regression of the association between time-updated antiretroviral therapy status and tuberculosis disease, stratified by CD4+ cell count, among people living with HIV in England, Wales and Northern Ireland from 2000 to 2014.
| Ever on ART | CD4+ cell count (cells/µl) | |||||
|---|---|---|---|---|---|---|
| ≥500 IRR (95% CI) |
350–499 IRR (95% CI) |
200–349 IRR (95% CI) |
100–199 IRR (95% CI) |
50–99 IRR (95% CI) |
0–49 IRR (95% CI) |
|
| No | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| Yes | 0.07 (0.05–0.10) | 0.14 (0.11–0.18) | 0.21 (0.17–0.25) | 0.32 (0.26–0.40) | 0.35 (0.25–0.49) | 0.49 (0.35–0.69) |
Incidence rate ratios derived from multivariable Poisson regression of the association between time-updated ART status and TB disease, stratified by CD4+ cell count. Model adjusted for the variables in the multivariable model in Table 2. 62 684 PLHIV were included in this analysis; 32 319 were excluded from the model due to missing data on ethnicity and country of birth, route of HIV infection, CD4+ cell count or age at HIV diagnosis. ART, antiretroviral therapy; CI, confidence interval; IRR, incidence rate ratio; TB, tuberculosis.
Table 4. Multivariable Poisson regression of the association between time-updated CD4+ cell count and tuberculosis disease, stratified by antiretroviral therapy status, among people living with HIV in England, Wales and Northern Ireland from 2000 to 2014.
| CD4+ cell count (cells/µl) | Ever on ART | |
|---|---|---|
| No IRR (95% CI) |
Yes IRR (95% CI) |
|
| ≥500 | 1.00 | 1.00 |
| 350–499 | 1.28 (1.06–1.55) | 2.51 (1.77–3.56) |
| 200–349 | 2.22 (1.84–2.66) | 6.37 (4.66–8.72) |
| 100–199 | 4.74 (3.79–5.93) | 21.21 (15.59–28.85) |
| 50–99 | 7.07 (5.26–9.51) | 34.29 (24.10–48.77) |
| 0–49 | 6.42 (4.87–8.46) | 44.21 (30.90–63.24) |
Incidence rate ratios derived from multivariable Poisson regression of the association between time-updated CD4+ cell count and TB disease, stratified by ART status. 62 684 PLHIV were included in this analysis; 32 319 were excluded from the model due to missing data on ethnicity and country of birth, route of HIV infection, CD4+ cell count or age at HIV diagnosis. Model adjusted for the variables in the multivariable model in Table 2. ART, antiretroviral therapy; CI, confidence interval; IRR, incidence rate ratio; TB, tuberculosis.
In a post-hoc analysis of patients who had initiated ART, we found that those who developed TB were more likely to have discontinued ARTat their last clinic visit (27 versus 6% of those without TB, P < 0.001, Supplementary Table 4, http://links.lww.com/QAD/B155). ART initiation rates and time from the most recent clinic visit to the end of the study were similar for MSM, heterosexuals and PWID.
There was no substantial difference in the age, sex, ethnicity/country of birth, route of HIV infection or CD4+ cell count of patients with missing data on any of the following variables: route of HIV infection, CD4+ cell count, IMD decile and country of birth. Patients with missing route of infection were less likely to be diagnosed with TB; however, there were no substantial differences for patients missing data on any other variable.
Sensitivity analysis
Sensitivity analyses were conducted as follows: excluding 241 individuals who acquired HIV infection through mother-to-child transmission, excluding 595 individuals with TB whose probabilistic matching scores (linking to their HIV record) were in the lowest quartile, excluding 137 individuals with TB who were matched to their HIV record using the three lowest ranked deterministic criteria, excluding 424 individuals diagnosed with TB 92–182 days after HIV diagnosis, including IMD score and excluding data on 12 432 individuals missing IMD score. All analyses provided consistent results with the main model (Supplementary Tables 1–3, http://links.lww.com/QAD/B155).
Discussion
People who acquired HIV infection through IDU (largely UK-born patients) had a high risk of TB following their HIV diagnosis, with incidence rates comparable with those in black Africans born in high TB incidence countries; almost five-fold more than MSM after accounting for other factors including starting ART. Consistent with previous research [3,6], declining CD4+ cell count was associated with higher TB rates.
The current study benefits from the very large national HIV-positive cohort, providing comprehensive results for England, Wales and Northern Ireland. The algorithm linking patients with TB and HIV utilizes ethnicity, year and country of birth; all variables with very high completeness: 97.3, 99.9 and 90.5%, respectively.
We found no substantial differences in the demographics or proportion of TB in patients missing data on each of these variables; however, patients missing data on one variable were more likely to have other missing data. In addition, patients missing data for multiple variables were less likely to be linked to a TB notification, and therefore, we may have under-estimated TB incidence rates; it is likely that the low incidence of TB in patients with ‘unknown’ route of HIV infection is a symptom of this and patients with extensive missing data may be more likely to be from populations at high risk for TB. In addition, the record linkage algorithm is less sensitive to non-English names [20]; therefore, we may have under estimated TB incidence in foreign populations.
One limitation was missing CD4+ cell count data for approximately a third of patients, who were therefore excluded from the risk factor analysis. This is partly due to difficulties linking data, and partly because some large hospitals do not supply CD4+ cell count data to HARS. However, we found no evidence that patients with missing CD4+ cell count data were systematically different to our analysis cohort. As our sample size remained very large, and there was no evidence that patients missing data were systematically different, we chose not to use multiple imputation due to the complexity of the dataset as a result of using time-updated CD4+ cell count and ART initiation. Data were available on ART discontinuation but were of poor quality and could not be included in the model. Consequently, we may have underestimated the association between starting ART and lower TB incidence by assuming all individuals remained on treatment for the duration of our study.
Individuals entered the study cohort 92 days after HIV diagnosis or first presentation to UK health services; therefore, we may have underestimated TB incidence in people diagnosed abroad who were at risk prior to entering the United Kingdom, as we would have missed TB cases diagnosed during the initial period following HIV diagnosis when TB incidence is highest. A recent study of PLHIV had 18% loss to follow-up over 4 years, and 14% of TB cases diagnosed more than 91 days after HIV diagnosis were amongst these patients [4]. As TB and HIV are sometimes treated (and usually reported) separately in the United Kingdom, dropping out of HIV care does not prevent notification of a TB diagnosis. We therefore used passive censoring, continuing follow-up until 31 December 2014 rather than the date last seen for HIV care. Consequently, migration out of the United Kingdom may mean we underestimated TB incidence.
A limitation of the Poisson regression model was censoring due to competing risks, specifically deaths from non-TB causes. However, few patients died (3%) and median time to death was 3.4 years, substantially longer than median time to TB diagnosis (1.8 years); therefore, any impact of censoring is likely to be minimal.
Although PWID represented less than 2% of PLHIV, they accounted for 3% of TB cases in this population and more than 4% of cases diagnosed more than 91 days after HIV diagnosis. TB incidence in PWID in our study (876/ 100 000 person-years in men and 605/100 000 in women) was substantially higher than that in a cohort of German PLHIV [25], possibly because this cohort utilized active rather than passive follow-up and excluded patients who did not present to care for 6 months or more, who may be more likely to develop TB disease than patients who remain engaged with care. PWID are typically diagnosed with HIV late [18], have slower rates of linkage to care and lower rates of viral suppression [26], all of which may contribute to increased risk of TB. We found ART initiation and the time from the last clinic visit to the end of the study were comparable for MSM, heterosexuals and PWID, and that PWID did not have higher rates of ART discontinuation at their last clinic visit prior to study end (Supplementary Table 4, http://links.lww.com/QAD/B155). Consequently, it seems high rates of TB among PWID are caused by difficulties in linking to care and not lack of engagement with health services once linked. Many PWID have other comorbidities which may cause immunosuppression, make HIV care more challenging, or be associated with increased risk of TB [27]. In addition, there are high rates of alcoholism and homelessness, and living in hostels is common [28]. These, in addition to injecting drugs in shared social settings, may drive close mixing of people with similar risk factors for TB disease, driving transmission. High rates of smoking may also have impacts on both local lung immunity and TB transmission. Further studies are needed to explore the impact of these factors and to design effective interventions. The British HIV Association (BHIVA) guidelines currently recommend testing and treating latent TB infection (LTBI) among PLHIV using criteria based on CD4+ cell count, time on ART and country of birth [29]. As the incidence of TB among PWID was comparable with that of black African patients born in countries with high TB incidence, we suggest that additionally screening and treating PWID for LTBI should be considered.
The majority of PWID were white (51%) and born in the United Kingdom or low TB incidence countries (72%). It is therefore likely that most TB in this group was acquired in the United Kingdom, meaning these cases may be preventable by diagnosing HIV sooner and ensuring prompt ART initiation. We could also do more to diagnose TB cases sooner; the impact of active case finding in PLHIV should be evaluated. In contrast, heterosexuals were typically black African (61%) and born in high TB incidence countries (69%), both populations which also have high rates of TB among HIV-negative people. Consequently, they are likely to have acquired TB abroad, limiting our ability to prevent these TB infections if they present with clinical TB at the time of HIV diagnosis [30]. As more than 60% of heterosexuals were diagnosed with TB simultaneously or prior to HIV diagnosis, greater efforts to diagnose these HIV infections and initiate ART would reduce TB in this population. A greater focus on screening and treating LTBI could also prevent these cases [31]. There is little data available on the prevalence of LTBI and the use of preventive therapy among PLHIV in the United Kingdom. Rates of LTBI screening and uptake of preventive therapy vary substantially between HIV clinics [32,33], and a survey of UK HIV healthcare providers providing care to 90% of PLHIV in the United Kingdom found that only 54% offered LTBI screening and preventive therapy [34]. Health economics evaluations would be useful to determine the most effective screening measures for these populations.
Over half of all TB cases (55%) were diagnosed simultaneously with HIV infection, and of the 39% diagnosed later, the probability of a TB diagnosis was highest in the first year following HIV diagnosis (Fig. 1). This suggests that TB disease is largely the result of TB infection acquired prior to HIV diagnosis. This could result from late diagnosis of existing active TB, particularly in migrants who have recently moved to the United Kingdom from high-burden countries and whose TB is largely attributable to reactivation of remotely acquired infection [35]. In addition, the incidence of TB amongst migrants decreases with time since entry to the United Kingdom, as new TB infection is less likely in the United Kingdom than their country of origin. Other factors which could explain this trend are increased surveillance for opportunistic infections following HIV diagnosis, or ‘unmasking-type’ immune reconstitution inflammatory syndrome as a consequence of ART. Although TB incidence was lower after the first year since HIV diagnosis (Table 1), 25% of all TB cases occurred more than 1 year after HIV diagnosis. These cases can certainly be attributed to reactivation of LTBI and could be preventable with LTBI treatment.
Patients who had initiated ART had greatly reduced rates of TB compared with those who had not (Table 3); however, time-updated CD4+ cell count and ART initiation status interacted within our model. Higher rate ratios for TB at low CD4+ cell count in people on ART may be attributable to late ART start (i.e. long periods of low CD4+ cell count prior to initiating ART and then little time on ART prior to TB diagnosis), or due to ART discontinuation. The Strategies for Management of Antiretroviral Therapy trial demonstrated an association between stopping ART and increased risk of opportunistic disease and death [36]. Our post-hoc analysis of patients who had started ART demonstrated that patients who went on to develop TB were more likely to have discontinued ART at their last study visit than individuals who remained TB-free (Supplementary Table 4, http://links.lww.com/QAD/B155). This suggests ART discontinuation could leave patients at risk of new TB disease.
In England, Wales and Northern Ireland, PLHIV who acquired HIV by injecting drugs had higher rates of TB after their HIV diagnosis than MSM, comparable with black Africans born in countries with high TB incidence. High rates of TB in PWID are likely to result from transmission within the United Kingdom. ART is highly protective against TB, but the majority of TB diagnoses were in people who have never started ART. ART discontinuation rates were much higher in people who subsequently developed TB than those who did not. Quicker initiation of ART, as per the recently updated BHIVA guidelines [37], and improving retention in care and ART continuation should decrease incident TB in PLHIV.
Supplementary Material
Acknowledgements
J.R.W. is funded by a UCL IMPACT studentship. This report is independent research supported by the National Institute for Health Research (Postdoctoral Fellowship, H.R.S., PDF-2014-07-008). I.A. acknowledges funding from NIHR (NF-SI-0616-10037 and SRF-2011-04-001), MRC and the Wellcome Trust. The views expressed in this publication are those of the authors and not necessarily those of the NHS, the National Institute for Health Research or the Department of Health.
Footnotes
Author contributions: J.R.W. designed the study, linked the TB and HIV surveillance datasets, conducted the analysis and drafted the article. H.R.S. and C.J.S. designed the study, analysed and interpreted the data and critically revised the article. A.E.B., M.K.L., A.S., H.L.T., Z.Y. and P.K. gave input on the study design, collected the data, linked the datasets, interpreted the results and critically revised the article. V.D. and I.A. designed the study, collected, linked, analysed and interpreted the data and revised the article. M.L. and A.P. interpreted the results and critically revised the article. All authors approved the final version of the article for publication.
The article utilized two surveillance datasets collected by the respiratory (Tuberculosis section) and HIV departments in the National Infections Service at Public Health England. In light of the work involved in collecting and linking these two datasets, and designing a study utilizing both of them, we have listed 13 authors for this article.
Conflicts of interest
J.R.W., A.E.B., M.K.L., M.L., A.S., P.K., Z.Y., H.L.T., V.D. and I.A. have no conflicts of interests to declare. H.R.S. declares funding from the National Institute for Health Research, UK during the conduct of the study; and, outside of the submitted work, grants and personal fees from Otsuka Pharmaceutical, nonfinancial support from Sanofi, and other support from the WHO. Outside the submitted work, C.J.S. reports personal fees from Gilead Sciences and ViiV Healthcare. A.P. is chair of the BHIVA TB guidelines committee. J.R.W. had full access to all the data in the study and had final responsibility for the decision to submit for publication.
References
- 1.WHO. Tuberculosis and HIV 2017. [Accessed 1 July 2016]. Available from: http://www.who.int/hiv/topics/tb/about_tb/en/
- 2.Health Protection Agency. HIV in the United Kingdom: 2011 report. Health Protection Services, Colindale; London: 2011. [Google Scholar]
- 3.Grant AD, Bansi L, Ainsworth J, Anderson J, Delpech V, East-erbrook P, et al. United Kingdom Collaborative HIV Cohort Study Group Tuberculosis among people with HIV infection in the United Kingdom: opportunities for prevention? AIDS. 2009;23:2507–2515. doi: 10.1097/QAD.0b013e3283320dfd. [DOI] [PubMed] [Google Scholar]
- 4.Gupta RK, Rice B, Brown AE, Thomas HL, Zenner D, Anderson L, et al. Does antiretroviral therapy reduce HIV-associated tuberculosis incidence to background rates? A national observational cohort study from England, Wales, and Northern Ireland. Lancet HIV. 2015;2:e243–e251. doi: 10.1016/S2352-3018(15)00063-6. [DOI] [PubMed] [Google Scholar]
- 5.WHO. WHO TB burden estimates 2017. [Accessed 23 July 2017]. Available from: http://www.who.int/tb/country/data/download/en/
- 6.Rice B, Elford J, Yin Z, Kruijshaar M, Abubakar I, Lipman M, et al. Decreasing incidence of tuberculosis among heterosexuals living with diagnosed HIV in England and Wales. AIDS. 2013;27:1151–1157. doi: 10.1097/QAD.0b013e32835e2cb1. [DOI] [PubMed] [Google Scholar]
- 7.Selwyn PA, Hartel D, Lewis VA, Schoenbaum EE, Vermund SH, Klein RS, et al. A prospective study of the risk of tuberculosis among intravenous drug users with human immunodeficiency virus infection. N Engl J Med. 1989;320:545–550. doi: 10.1056/NEJM198903023200901. [DOI] [PubMed] [Google Scholar]
- 8.Monge S, Diez M, Pulido F, Iribarren JA, Campins AA, Arazo P, et al. Tuberculosis in a cohort of HIV-positive patients: epidemiology, clinical practice and treatment outcomes. Int J Tuberc Lung Dis. 2014;18:700–708. doi: 10.5588/ijtld.13.0778. [DOI] [PubMed] [Google Scholar]
- 9.Girardi E, Sabin CA, d’Arminio Monforte A, Hogg B, Phillips AN, Gill MJ, et al. Incidence of Tuberculosis among HIV-infected patients receiving highly active antiretroviral therapy in Europe and North America. Clin Infect Dis. 2005;41:1772–1782. doi: 10.1086/498315. [DOI] [PubMed] [Google Scholar]
- 10.Sterling TR, Lau B, Zhang J, Freeman A, Bosch RJ, Brooks JT, et al. Risk factors for tuberculosis after highly active antiretroviral therapy initiation in the United States and Canada: implications for tuberculosis screening. J Infect Dis. 2011;204:893–901. doi: 10.1093/infdis/jir421. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Taarnhoj GA, Engsig FN, Ravn P, Johansen IS, Larsen CS, Roge B, et al. Incidence, risk factors and mortality of tuberculosis in Danish HIV patients 1995–2007. BMC Pulm Med. 2011;11:26. doi: 10.1186/1471-2466-11-26. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Mor Z, Lidji M, Cedar N, Grotto I, Chemtob D. Tuberculosis incidence in HIV/AIDS patients in Israel, 1983–2010. PLoS One. 2013;8:e79691. doi: 10.1371/journal.pone.0079691. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Moreno S, Jarrin I, Iribarren JA, Perez-Elias MJ, Viciana P, Parra-Ruiz J, et al. Incidence and risk factors for tuberculosis in HIV-positive subjects by HAART status. Int J Tuberc Lung Dis. 2008;12:1393–1400. [PubMed] [Google Scholar]
- 14.Zhang Y, Yu L, Tang ZR, Huang SB, Zheng YJ, Meng ZH, et al. Diagnosis of pulmonary tuberculosis among asymptomatic HIVR patients in Guangxi, China. Chin Med J. 2010;123:3400–3405. [PubMed] [Google Scholar]
- 15.Skingsley A, Yin Z, Kirwan P, Croxford S, Chau C, Conti S, et al. HIV in the UK – situation report 2015: data to end 2014. Public Health England; London: 2015. [Google Scholar]
- 16.Yin Z, Brown AE, Hughes G, Nardone A, Gill ON, Delpech VC, et al. HIV in the United Kingdom 2014 report: data to end 2013. Public Health England; London: 2014. [Google Scholar]
- 17.Zenner D, Abubakar I, Conti S, Gupta RK, Yin Z, Kall M, et al. Impact of TB on the survival of people living with HIV infection in England, Wales and Northern Ireland. Thorax. 2015;70:566–573. doi: 10.1136/thoraxjnl-2014-206452. [DOI] [PubMed] [Google Scholar]
- 18.Skingsley A, Kirwan P, Yin Z, Nardone A, Hughes G, Tosswill J, et al. HIV new diagnosis, treatment and care in the UK 2015 report: data to end 2014. Public Health England; London: 2015. [Google Scholar]
- 19.Gupta RK, Brown AE, Zenner D, Rice B, Yin Z, Thomas HL, et al. CD4R cell count responses to antiretroviral therapy are not impaired in HIV-infected individuals with tuberculosis co-infection. AIDS. 2015;29:1363–1368. doi: 10.1097/QAD.0000000000000685. [DOI] [PubMed] [Google Scholar]
- 20.Aldridge RW, Shaji K, Hayward AC, Abubakar I. Accuracy of probabilistic linkage using the enhanced matching system for public health and epidemiological studies. PLoS One. 2015;10:e0136179. doi: 10.1371/journal.pone.0136179. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Winter JR, Delpech V, Kirwan P, Stagg HR, Venugopalan S, Skingsley A, et al. Linkage of UK HIV and tuberculosis data using probabilistic and deterministic methods; Conference on Retroviruses and Opportunistic Infections; Boston, USA. 2016. [Google Scholar]
- 22.Department for Communities and Local Government. English indices of deprivation 2015. [Accessed 20 May 2016]. Available from: https://www.gov.uk/government/collections/english-indices-of-deprivation.
- 23.Welsh Government. Welsh index of multiple deprivation (WIMD) 2015. [Accessed 20 May 2016]. Available from: http://gov.wales/statistics-and-research/welsh-index-multiple-deprivation/?lang=en.
- 24.Victora CG, Huttly SR, Fuchs SC, Olinto MTA. The role of conceptual frameworks in epidemiological analysis: a hierarchical approach. Int J Epidemiol. 1997;26:224–227. doi: 10.1093/ije/26.1.224. [DOI] [PubMed] [Google Scholar]
- 25.Karo B, Haas W, Kollan C, Gunsenheimer-Bartmeyer B, Hamouda O, Fiebig L, et al. Tuberculosis among people living with HIV/AIDS in the German ClinSurv HIV Cohort: long-term incidence and risk factors. BMC Infect Dis. 2014;14:148. doi: 10.1186/1471-2334-14-148. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Kirwan PD, Chau C, Brown AE, Gill ON, Delpech VC. Contributors HIV in the UK – 2016 report. Public Health England; London: 2016. [Google Scholar]
- 27.Public Health England; Health Protection Scotland; Public Health Wales; Public Health Agency Northern Ireland. Shooting up: infections among people who inject drugs in the UK, 2015. Public Health England; London: 2016. Nov, [Google Scholar]
- 28.National AIDS Trust. HIV and injecting drug use. London: 2013. [Google Scholar]
- 29.Pozniak AL, Coyne KM, Miller RF, Lipman MCI, Freedman AR, Ormerod LP, et al. British HIV Association guidelines for the treatment of TB/HIV coinfection 2011. HIV Med. 2011;12:517–524. doi: 10.1111/j.1468-1293.2011.00954.x. [DOI] [PubMed] [Google Scholar]
- 30.Rice BD, Elford J, Yin Z, Delpech VC. A new method to assign country of HIV infection among heterosexuals born abroad and diagnosed with HIV. AIDS. 2012;26:1961–1966. doi: 10.1097/QAD.0b013e3283578b80. [DOI] [PubMed] [Google Scholar]
- 31.Danel C, Moh R, Gabillard D, Badje A, Le Carrou J, Ouassa T, et al. TEMPRANO ANRS 12136 Study Group A trial of early antiretrovirals and isoniazid preventive therapy in Africa. N Engl J Med. 2015;373:808–822. doi: 10.1056/NEJMoa1507198. [DOI] [PubMed] [Google Scholar]
- 32.Fox-Lewis A, Brima N, Muniina P, Grant AD, Edwards SG, Miller RF, et al. Tuberculosis screening in patients with HIV: an audit against UK national guidelines to assess current practice and the effectiveness of an electronic tuberculosis-screening prompt. Int J STD AIDS. 2016;27:901–905. doi: 10.1177/0956462416628355. [DOI] [PubMed] [Google Scholar]
- 33.Kall MM, Coyne KM, Garrett NJ, Boyd AE, Ashcroft AT, Reeves I, et al. Latent and subclinical tuberculosis in HIV infected patients: a cross-sectional study. BMC Infect Dis. 2012;12:107. doi: 10.1186/1471-2334-12-107. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.White HA, Miller RF, Pozniak AL, Lipman MC, Stephenson I, Wiselka MJ, et al. Latent tuberculosis infection screening and treatment in HIV: insights from evaluation of UK practice. Thorax. 2017;72:180–182. doi: 10.1136/thoraxjnl-2016-209063. [DOI] [PubMed] [Google Scholar]
- 35.Aldridge RW, Zenner D, White PJ, Williamson EJ, Muzyamba MC, Dhavan P, et al. Tuberculosis in migrants moving from high-incidence to low-incidence countries: a population-based cohort study of 519 955 migrants screened before entry to England, Wales, and Northern Ireland. Lancet. 2016;388:2510–2518. doi: 10.1016/S0140-6736(16)31008-X. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Lundgren JD, Babiker A, El-Sadr W, Emery S, Grund B, Neaton JD, et al. Strategies for Management of Antiretroviral Therapy (SMART) Study Group Inferior clinical outcome of the CD4R cell count-guided antiretroviral treatment interruption strategy in the SMART study: role of CD4R Cell counts and HIV RNA levels during follow-up. J Infect Dis. 2008;197:1145–1155. doi: 10.1086/529523. [DOI] [PubMed] [Google Scholar]
- 37.BHIVA. British HIV Association guidelines for the treatment of HIV-1-positive adults with antiretroviral therapy 2015. BHIVA; London: 2015. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.