The effects of anti-inflammatory agents as host-directed adjunct treatment of tuberculosis in humans: a systematic review and meta-analysis

In this meta-analysis, two authors searched Medline (Additional file 2), Cochrane Central Register of Controlled Trials, embase, Web of Science and Google scholar for trials, using the keywords “pulmonary tuberculosis” or “tuberculoses”, “anti-inflammatory agents” and “host-directed-therapeutics”, with no limitations in the publication type, language or period. The detailed search string was as follows: Tuberculosis [MeSH Terms] OR tuberculoses [MeSH Terms] OR Mycobacterium tuberculosis [MeSH Terms] AND Mycobacterium tuberculoses [MeSH Terms] OR pulmonary tuberculosis [MeSH Terms] OR pulmonary tuberculoses [MeSH Terms] OR adjunctive immunotherapy in tuberculosis [MeSH Terms] OR adjunctive immunotherapy in tuberculoses [MeSH Terms] OR adjunct tuberculosis treatment [MeSH Terms] AND adjunct tuberculoses treatment [MeSH Terms] OR host-directed-therapeutics in tuberculosis [MeSH Terms] OR host-directed-therapeutics in tuberculoses [MeSH Terms] OR anti-tubercular agent [MeSH Terms] OR anti-inflammatory agents [MeSH Terms] AND sputum smear conversion [MeSH Terms] OR sputum culture conversion [MeSH Terms] OR immune inflammatory response [MeSH Terms] OR anti-inflammatory response [MeSH Terms] OR inflammatory response [MeSH Terms] AND immunomodulatory regulators [MeSH Terms] OR immuno modulators [MeSH Terms] OR human [MeSH Terms] OR human subjects [MeSH Terms] OR patients [MeSH Terms] AND adults [MeSH Terms] AND randomised control trials [MeSH Terms]. The database search was from inception up to and including November 2019 and was regularly updated. The reference or citation listed in each identified article was also reviewed for potentially relevant studies. These searches were supplemented by searches of review articles and reference lists of trial publications. When full-length articles were not available from the databases, these were requested from the authors. Four authors then determined which studies met the eligibility criteria.

Four authors were involved in data extraction and quality assessment. Two authors independently performed the initial extraction of data.

Data relating to the following study characteristics were collected: authors, publication year, trial registration number, study setting, mode of anti-TB therapy administration, details of adjunctive therapies being evaluated in each trial (type, dose and route of treatment) and primary study outcome. Baseline characteristics of trial participant data were extracted for the following variables: age, sex ratio, body mass index (BMI), mid-upper arm circumference (MUAC), number of participants randomised and those included, intervention treatment duration and follow-up period, HIV status, proportion with MDR-TB, extent of disease on baseline chest radiograph (as measured by the proportion of zones involved and the presence or absence of cavitation), TB score, C-reactive protein (CRP), erythrocyte sedimentation rate (ESR), other blood indices, adverse effects and mortality in a standard form as recommended by Cochrane [37, 38]. Follow-up data after 4–8 weeks of initiating antimicrobial and adjunct treatment were also extracted, namely sputum smear and culture status to estimate conversion rates, time from initiation of antibiotic treatment to stable sputum smear/culture conversion (the date of stable smear/culture conversion being estimated as the mid-point between the date of the last positive sputum smear/culture and the date of the first negative sputum smear/culture thereafter); TB score, chest radiography, weight gain, BMI, MUAC, CRP, ESR and other blood indices, as well as adverse effects and mortality. For any missing information, corresponding authors were contacted by email. Three items were used to assess the methodological quality of each included study, based on the Jadad scale [39], which includes criteria of randomisation, blinding, and addressing the problem of incomplete outcome data (withdrawals and dropouts). The other two authors were consulted to solve any disagreement on study selection, data extraction, or quality assessment.

The primary and secondary outcome measures of the meta-analysis comprised of efficacy assessment and safety evaluation, respectively. The primary outcome was sputum smear conversion rate, estimated as the percentage of smear-positive PTB (PTB+) cases registered in a specified period that converted to smear-negative status [40]. Secondary outcomes included sputum smear and culture time to conversion, sputum culture conversion rate, TB score, chest radiography, weight gain, BMI, MUAC, CRP, ESR, other blood indices associated with infectivity and inflammation (total white blood cells, lymphocytes, neutrophils, monocytes, interferon gamma (IFN-γ), and TNF-α), incidence of potential adverse reactions attributable to the study intervention agents and all-cause mortality. These outcome measures are considered very relevant in TB disease progression and treatment success.

Risk of bias assessment was conducted using the Cochrane Collaboration Risk of Bias tool [37], considering the following parameters: sequence generation, allocation concealment, blinding of participants, personnel and outcome assessors, completeness of outcome data, evidence of selective outcome reporting, and other potential threats to validity. The Cochrane risk of bias tool was used for the assessment of risk of bias in estimating the study outcomes. The selective reporting within studies was assessed by answering whether the results were fully reported, as the study was pre-specified (for example, if all the results were reported at all follow-up time points). For the primary analysis, the likelihood of publication bias was investigated through the construction of a funnel plot and Egger’s test [41, 42]. Study quality, in this case, was assessed independently by three authors. Discrepancies were resolved by a fourth author.

Meta-analysis of sputum smear conversion rate was conducted using both random and fixed effect methods, resulting in conversion hazard of the treatment group with respect to the control group. Study weights reported on the forest plots were derived from the random effect analysis and were computed as w_i = 1/(s_i² + t²) with s_i² being the variance estimate from the i-th study and t² representing the overall variance. Standard errors were computed as di/1.96 where di = max [(log (upper 95% confidence interval bound)-log (RRi); (log (RRi)-log (lower 95% bound)].

Heterogeneity between studies was reported by means of Cochrane Q test and I² statistic [43]. When relevant heterogeneity was detected (significant Cochrane Q test and/or I² > 50%), stratification and study exclusion were conducted to identify its source. Publication bias was assessed by means of funnel plot visual inspection and Egger’s test [42]. Influence analysis, excluding one study at a time, was conducted to assess robustness of results.

The initial search yielded more than 1000 records (Additional file 1). After the initial screening and duplicates were removed, the remaining 55 unique studies identified were further screened for eligibility. Out of the 18 full-text articles identified and assessed for eligibility, 15 studies met the inclusion criteria of this meta-analysis. Some reasons for excluding articles, were that the adjunct treatments used could not be isolated and some of the articles were sub-studies of an article that was already included. Fifteen articles were included in the qualitative, whereas only 11 in the final quantitative synthesis due to the unavailability of data from authors. The detailed literature search and selection process is shown in Fig. 1. The clinical trials were conducted in 12 countries and most of them were registered as described in Table 1. Ten studies investigated the effect of vitamin D [14‐16, 18, 20, 21, 23, 44‐46] as adjunct HDT agent, while five studies used either recombinant human interleukin [25], prednisolone [47], pentoxifylline [48], N-acetylcysteine [49] or recombinant human granulocyte-macrophage colony-stimulating factor [50]. Out of the fifteen studies, thirteen studies used two-arm parallel designs to investigate the effects of only one HDT treatment [14‐16, 21, 23, 25, 44‐50]. Two studies included L-arginine [20] and phenylbutyrate [18] concurrently with vitamin D as HDT in a factorial design such that they could be isolated. The treatments were administered either by oral [15, 16, 18, 20, 21, 44‐49], intramuscular [14, 23], intradermal [25] or subcutaneous [50] routes. Apart from three studies [15, 16, 21], all others followed the standard anti-TB therapy regime of 2 months of isoniazid, rifampicin, pyrazinamide, ethambutol (HRZE), then 4 months of rifampicin and isoniazid (RH) (Table 1). All studies included sputum smear and/or culture conversion as a primary (or co-primary) or secondary outcome measure. The following studies reported these as primary outcomes: sputum culture conversion rate [15, 16, 18, 20, 21, 25, 46], sputum smear conversion rate [23, 44, 49], TB clinical scores [18, 20, 23, 45], chest radiograph resolution [14], weight gain [14], treatment safety [47], treatment tolerance [50] and treatment effect [48] (Table 2). According to the Jadad scale, all the trials included in the meta-analysis were considered high-quality studies as shown in supplementary Table 1 (Jadad score 3–5). All trials achieved the maximum quality score of 5, except two studies which scored 3 [44] and 4 [50] on the scale. Details of the quality assessment are provided in supplementary Table 1 (Additional file 2).

A total of 2728 randomised participants from 15 studies fulfilled the eligibility criteria, of which 2540 were included in the final analysis. One thousand two hundred and sixty-nine patients received treatment, whilst 1271 were on identical placebo in addition to TB treatment. Two thousand and fifty-seven participants contributed data for the vitamin D treatment group whilst 483 participants contributed data for the other anti-inflammatory agents group, for analysis of various study outcomes. The treatment outcome endpoint measured, ranged from 4 to 8 weeks, with follow-up duration up to 32 weeks. The age range of the 2540 participants included in the meta-analysis was from 18 to 70 years in both the intervention and control groups, and 1898 (74.7%) were male. Administration of treatment to participants in the intervention arm was as follows: treatment was administered at baseline only [23], daily [18, 25, 44, 47‐49], twice a week [50], weekly then 2-weekly [46], 2-weekly [15, 16, 21], 4-weekly [20], 3 times over a period of 32 weeks (at baseline, 20 weeks and 32 weeks) [45] and at baseline and 4 weeks [14]. Four hundred and three (28.6%) of the 1411 participants in 11 studies [16, 20, 21, 23, 25, 44‐48, 50] tested positive for HIV status at baseline. The proportion of participants with MDR-TB (i.e. their Mtb complex isolate was resistant to at least isoniazid and rifampicin) was 3.9%, which is 57 out of 1460 reported in 7 studies [15, 16, 18, 20, 21, 25, 46] (Table 3).

Twelve studies reported the sputum smear conversion rate [14‐16, 18, 20, 21, 23, 44‐50], whilst ten studies reported the sputum culture conversion rate [15, 16, 18, 20, 21, 25, 46‐48, 50]. Eleven reported the time to sputum smear conversion [15, 18, 20, 21, 23, 25, 45, 47‐50] and eight provided the time to sputum culture conversion [15, 16, 21, 25, 46‐48, 50]. Six reported changes in BMI [14‐16, 21, 23, 44], four presented changes in mean MUAC [14, 15, 23, 45], ten provided the data on chest radiograph change [14, 15, 18, 20, 21, 25, 44, 46‐49] and five presented results on TB clinical score changes [14, 18, 20, 23, 45]. Eleven presented the incidence of adverse events [15, 16, 18, 20, 21, 25, 45‐48, 50], nine indicated the changes in weight gain [14, 18, 20, 25, 45, 47‐50], eight provided the incidence of all-cause mortality [15, 16, 20, 21, 45‐47, 50], four exhibited changes in CRP and ESR [15, 18, 21, 23], and eight studies showed changes in other blood parameters such as total white cells, lymphocytes, neutrophils, monocytes and TNF-α [15, 18, 21, 23, 25, 44, 47, 48, 50]. Details on study outcome assessments are summarised in supplementary Tables 3 and 4 (Additional file 2).

A funnel plot and Egger’s test for the outcome of rate of sputum smear conversion across studies did suggest a publication bias in trials involving vitamin D but not for those involving other anti-inflammatory treatments, in relation to this outcome (Additional file 2:Supplementary Figs. 1 & 2). An Egger’s test revealed an uneven spread of results of RCTs skewed at one side of the overall adjusted hazard ratio as shown in Supplementary Fig. 1. Seven RCTs [14, 15, 21, 25, 46‐48] were at low risk of bias for all aspects analysed according to the Cochrane Collaboration Risk of Bias tool. The study led by Nursyam and colleagues [44] had an unclear risk bias due to selection bias (adequate sequence generation) and detection bias (blinding). Methods used in random sequence generation, blinding of participants, personnel and outcome assessments were not clearly or adequately described. Even though this could have affected the outcome, no evidence was found that it did. Selection bias was also observed in the trial by Pedral-Sampaio and colleagues, as it did not adequately describe a clear method used in random sequence generation [50]. RCTs conducted by Daley, Mahakalkar, Mily, Wejse and Ralph had an unclear risk of attrition bias due to incomplete outcome data [16, 18, 20, 45, 49] as a result of relatively high rates of loss to follow-up (LFU) of participants (> 20%). In addition, no evidence was found that it affected the overall final outcome of these studies, since the LFU was similar amongst groups. Some reasons for the LFU common to these studies included defaulting, death or transfer of patients to other locations of residence or for treatment. Details on the risk of bias assessment are provided in supplementary Table 2 (Additional file 2). There was a low statistical heterogeneity observed between/within treatments using other anti-inflammatory agents (I² = 11.2%, p = 0.342), compared to the vitamin D supplementation group (I² = 69.5%, p = 0.006) in the proportion of the rate of sputum smear conversion (Fig. 2 & Fig. 3). The high heterogeneity observed in the vitamin D supplementation group could be as a result of the different adjunct treatment doses administered, timing (Table 1) or duration of treatment (Table 3). The exclusion of the study from Salahuddin et al. [14] reduced the between-study heterogeneity to 53.3%. Overall, influence analysis did not detect any relevant difference in the relative risk (RR) estimate when excluding single studies. Likewise, no study was excluded due to dubious decisions as a result of the sensitivity analysis conducted to exclude any study.

The results represent the sputum smear conversion rate upon HDT treatment measured after a period of 4–8 weeks. It was observed that there was a 38% increased rate of sputum smear negativity among patients administered with vitamin D, compared to those receiving only the TB treatment (RR 1.38, 95% CI 1.03 to 1.84). There was moderate to high between-study heterogeneity observed in this group (p = 0.006 and I² = 69.5%) (Fig. 2). Similarly, patients treated with other anti-inflammatory HDT agents had a 29% increased sputum smear conversion rate (RR 1.29, 95% CI 1.09 to 1.53). No heterogeneity was observed between studies in this group (p = 0.342 and I² = 11.2%) (Fig. 3).

Sputum smear and culture time to conversion tended to be faster in the treatment group than in placebo in both vitamin D and other anti-inflammatory adjunct therapies in general. However, the sputum culture conversion rate at 8 weeks did not differ significantly and was similar between groups in the majority of trials. Administration of vitamin D did not influence the proportion of participants with sputum culture conversion at 8 weeks (treatment vs placebo, 150 of 190 vs. 148 of 200, respectively; adjusted odds ratio, 1.47; 95% CI, 0.88–2.45; P = 0.14). However, vitamin D did significantly accelerate sputum smear conversion (adjusted HR, 1.47; 95% CI,1.09–1.98) more in the treatment arm according to Ganmaa et al. [15] (Supplementary Table 4a).

Similarly, higher sputum clearance in the treatment group were reported by Farazi et al. (OR 0.20, 95% CI 0.04–1.02; p = 0.037) [23], Nursyam et al. (p = 0.002) [44] and Mayanja-Kizza et al. [47]. The proportion of participants with negative sputum culture at 4 weeks was significantly different for those randomised to vitamin D versus placebo (62% vs 37%, p = 0.001) but not at week 8 [47]. Mily et al. reported similar observations (OR 3.37, 95% CI 1.0–10.8; p = 0.041) [18]. According to Mahakalkar et al., there was a significant clearing of infiltration and reduction in cavity size in treatment compared to placebo (87.5% vs 33.33%, p < 0.05) and (p < 0.01) at 8 weeks, respectively [49]. Ganmaa et al. also reported vitamin D accelerated radiographic resolution (mean number of zones affected on chest radiograph at 8 weeks, treatment vs placebo, 5.48 vs 5.69 respectively; 95% CI for difference, − 0.06–0.77 zones; p = 0.02) [15]. There was also a 50% reduction in cavity size observed in the vitamin D arm compared to the placebo arm (p = 0.035) [14]. Most trials, however, did not observe significant clearing of infiltration or reduction in cavity size in treatment compared to placebo groups (Supplementary Table 4a).

There was no clinically significant difference in the change in MUAC between the vitamin D treatment and placebo group according to Salahuddin et al. [14]. However, Ganmaa et al. observed changes in MAUC after 8 weeks of treatment (adjusted mean difference,0.25,95% CI 0.08–0.43; p = 0.04) [15]. No significant effect was observed on concentrations of CRP and ESR at 8 weeks in three reported vitamin D supplemented trials [15, 18, 21]. However, an increase in CRP was seen in the placebo group at the end of 4 weeks (OR 2.98, 95% CI 1.04–8.52; p = 0.038) and 8 weeks (OR 3.33, 95% CI 1–11.14; p = 0.044) of treatment [23]. In addition, the ESR was observed to be higher in the placebo group compared to the Vitamin D group though not significantly (Vitamin D vs placebo, 33.5 ± 22.0 vs 36.3 ± 22.2 respectively) [18]. Similarly, there were no major significant effects of supplementary treatments observed with regards to weight gain and BMI for some of the vitamin D and other anti-inflammatory treatment groups. However, Salahuddin et al., Daley et al., and Mahakalkar et al. [14, 16, 49] reported highly significant weight changes in the vitamin D and N-acetylcysteine treatment groups compared to the placebo groups (3.75 vs 2.61; p < 0.009, 3·1; p < 0·0001, and 2.66 ± 0.18 vs − 0.02 ± 0.03; p < 0.001, respectively). Farazi et al. [23] also found marginal significant changes in BMI in favour of the vitamin D treatment group (24.66 ± 4.3 vs 22.31 ± 3.9, 95% CI 0.23–4.47) in comparison to Daley et al. [16], who observed increased BMI changes in favour of placebo group (0·087 kg/m², p = 0·597) (Supplemetary Table 4b).

Changes in time to clinical improvement of the TB score were also similar in the two groups in three reported vitamin D supplemented trials included in this review. Farazi and colleagues [23], however, showed that the vitamin D group had a low TB score and better health related quality of life in comparison to placebo after 8 weeks of treatment (2.4 ± 1.5 vs 3.7 ± 1.7 95% CI -2.13 -- 0.47; p = 0.003 and 52.3 ± 9.6 vs 46.2 ± 10.1; p = 0.019, respectively). Adverse effects (AEs) were reported in the majority of the studies and there were no differences in the occurrence of most of the AEs between the study arms. However, in the study by Johnson et al., hyperpigmentation (recombinant human Interleukin vs placebo, 55 vs 1; p < 0.0001) and other adverse effects were observed to be higher in the treatment group [25]. No study reported mortality due to study intervention except for the study by Mayanja-Kizza et al. (prednisolone vs placebo, 17 vs 14 respectively; p = 0.28) [47]. Two studies reported that vitamin D administration was associated with reduced total white blood cell counts, neutrophil counts and monocyte counts [15, 21]. The lymphocyte to monocyte ratio was significantly higher in the intervention than in the control group at 8 weeks (3.52 vs 2.70, 95% CI for difference 0.16–1.11; p = 0.009) [21] and adjusted mean difference 0.4, 95% CI 0.2–0.6; p = 0.001 [15]. The median levels of TNF-α decreased by at least 50% from baseline values in the treatment arm compared to the placebo at 4 weeks (p < 0.001) [47]. There was a similar trend toward reduced TNF-α production in the pentoxifylline arm, although the difference between groups was not statistically significant (pentoxifylline vs placebo,0.4 vs 0.6 respectively; p = 0.27) [48] (Supplementary Table 4c)..