Use of a scoring strategy to determine clinical risk of progression and risk group-specific treatment adherence in subjects with latent tuberculosis infection

Background Annual incidence of active tuberculosis (TB) cases has plateaued in the US from 2013-2015. Most cases are from reactivation of latent tuberculosis infection (LTBI). A likely contributor is suboptimal LTBI treatment completion rates in subjects at high risk of developing active TB. It is unknown whether these patients are adequately identified and treated under current standard of care. Methods In this study, we sought to retrospectively assess the utility of an online risk calculator (tstin3d.com) in determining probability of LTBI and defining the characteristics and treatment outcomes of Low: 0-<10%, Intermediate: 10-<50% and High: 50-100% risk groups of asymptomatic subjects with LTBI seen between 2010-2015. Results 51(41%), 46 (37%) and 28 (22%) subjects were in Low, Intermediate and High risk groups respectively. Tstin3d.com was useful in determining the probability of LTBI in tuberculin skin test positive US born subjects. Of 114 subjects with available treatment information, overall completion rate was 61% and rates of completion in Low (60%), Intermediate (63%) and High (57%) risk groups were equivalent. 75% subjects in the 3HP group completed treatment compared to 58% in the INH group. Provider documentation of important clinical risk factors was often incomplete. Logistic regression analysis showed no clear trends of treatment completion being associated with assessment of a risk factor. Conclusion These findings suggest tstin3d.com could be utilized in the US setting for risk stratification of patients with LTBI and select treatment based on risk. Current standard of care practice leads to subjects in all groups finishing treatment at equivalent rates.


56
One third of the world population is estimated to have LTBI [1], a state of infection caused by 57 Mycobacterium tuberculosis (Mtb) characterized by temporary immune containment of the 58 bacteria and lack of any clinical or microbiologic evidence of active tuberculosis (TB). As these 59 subjects have no evidence of disease, they can currently be diagnosed only by measuring the 60 cutaneous Delayed Type Hypersensitivity reaction (tuberculin skin test, TST) or by measuring 61 production of Interferon-γ in the blood using interferon-γ release assays (IGRAs) [2]. It is 62 estimated that 5-10% of all subjects with LTBI will have a lifetime risk of progression to active 63 TB disease. Treatment of LTBI decreases the overall burden of active TB by 60-90% [3]. In low 64 burden TB countries like the US where the overall active TB rates are <10/1000 population, the Variables generated by the calculator 117 Using the above data, the calculator was used to generate a positive predictive value (PPV) for 118 TST performed to detect TB infection. As IGRAs have higher specificity for assessment of Mtb [9]. In this algorithm the highest annual risk is assigned (in descending order) to transplantation 127 requiring immunosuppressant therapy (7.4%), HIV (5%), pulmonary silicosis (3%), chronic renal 128 failure (requiring hemodialysis) (2.5%), carcinoma of the head and neck(1.6%), close contact of 129 person with active TB (1.5%) and recent TST/IGRA conversion (≤ 2years) (1.5%) . The 130 cumulative risk refers to the annual risk of TB reactivation multiplied by the number of years 131 before the patient reaches an age of 80 years. In addition, provider awareness of risk was 132 assessed by identifying how many of the above variables were documented by the treating 133 physician at the initial clinical encounter.

135
Concordance between TST and IGRA in subjects that received both tests was analyzed by 136 McNemar's test of corrected proportions. Descriptive statistics were analyzed to determine the 137 characteristics of subjects with a PPV of greater than 50%. Patients were also stratified into Low 138 (<10%), Intermediate (10% to <50%), and High (50%-100%) cumulative risk categories.
Furthermore, percentages of patients in each cumulative risk group completing treatment were 140 calculated. Fisher's exact test was used to determine the association between the types of drug 141 regimen with treatment completion rates. Logistic regression analysis was performed to examine 142 the association between provider assessment of an individual clinical risk factor and the patient's 143 likelihood of completing the treatment. All statistical analyses were conducted using SAS 9.4 144 (SAS Institute, Cary, NC), two-sided with a significance level of 0.05.

Characteristics of the study subjects
148 Table 1 shows the baseline demographic information for the study subjects. The median age was 149 49 years and a little less than half were female (43.31%). Of the 125 patients included, 94 were 150 from the US and US territories, 32 were non-US born and 1 subject had no documentation of 151 country of origin. Of the US-born individuals most were Caucasian (52%) or African American 152 (43%). The mean age at immigration for non-US born subjects was 29.22 years. TST data were 153 available on 69 and IGRA data on 91 subjects. Of all 35 TST positive subjects who also had 154 IGRA results available, 19 were IGRA positive, 13 were IGRA negative and 1 indeterminate .

155
The Only 2 subjects with positive IGRA had a negative TST result (Figure 1). There was 156 significant discordance between the two tests ( p = 0.0045,Mcnemar's test) in subjects who 157 underwent both TST and IGRA for diagnosis .There were 90 subjects who had a single test

174
The overall distribution of risk factors in the study patients comparing US born with immigrants 175 is shown in Figure 2. Subjects with HIV, diabetes, history of smoking, taking a TNF-α inhibitor 176 and on transplant immunosuppression were more common in US born subjects compared to 177 immigrants. The calculator allowed us to divide subjects with LTBI into "Low (<10% risk)," 178 "Intermediate (10-<50% risk)," and "High (50-100% risk)" cumulative risk groups. All patients 179 with AIDS had the highest annual (>10%) and cumulative (>50%) risk. The calculator assigns a 180 slightly increased risk to African Americans compared to age and sex matched Caucasians and a larger proportion of African Americans (20% and 43%) were in the highest annual (>10%) and 182 cumulative risk category (>50%) compared to Caucasians (2% and 16% ). However, this was 183 primarily because of the significantly increased prevalence of HIV/AIDS noted in this population 184 (15/42, 37.5%) compared to the Caucasian population (4/49, 8.2%) (p=0.001 Fisher exact test).
185 Table 2 shows characteristics of all the patients in the High cumulative risk group. Furthermore, 186 for all patients who tested positive on TST but negative on IGRA, the median cumulative risk 187 was 19% suggesting that they were not in the high risk group.   The calculated risk of INH induced hepatitis correlates with the age of the patient more than any 217 of the other risk factors. Of the five patients that developed elevated LFTs, all were above 38 218 years of age with three being above 65 years. Only one was on INH and four were on rifampin.

219
The calculator estimated a risk of hepatitis of 1.2% for two of these subjects, 5% for two and 220 2.3% for one subject. Each patient was concomitantly using at least one other potentially 221 hepatotoxic medication. treated for LTBI were the best for 3HP and worst for Rifabutin, 75% and 50% subjects 229 respectively ( Table 3). Amongst the documented reasons for stopping treatment early ( Figure   230 3), loss to follow-up accounted for a majority of incomplete treatments (23 subjects  factor was not associated with improved treatment completion but this could be because providers often select regimens based on side effect profiles of drugs rather than the patient's 272 risk of progression to active TB. We speculate that provider awareness of a numerical risk for 273 patients with LTBI can allow them to use short course LTBI treatment regimens more cost-274 effectively, ensuring completion in the highest risk group, an approach that needs to be tested in have shown that false positive IGRAs remain a concern among specific groups assessed for 281 LTBI [10].We found significant discordance between PPD and IGRA positivity in our study in 282 keeping with results obtained by others [11,12]. We found the PPV calculated by the calculator to improve treatment adherence [16][17][18]. This approach also minimizes the loss to follow up 306 which was the major reason for incomplete treatment in our cohort.

307
In agreement with the published literature on adherence with 3HP based regimens, we found the 308 highest rates of treatment completion in subjects prescribed 3HP compared to those receiving 309 INH or Rifampin. Surprisingly, we saw low rates of treatment completion seen with Rifampin or 310 Rifabutin only regimens. Cytochrome P (CYP) 450 isoenzyme induction by Rifamycin based 311 regimens remains a concern in subjects with HIV on ARVs. This is especially important as more 312 than 85% of subjects with HIV/AIDS were prescribed INH and only 30% were able to complete 313 treatment in our study. Current US guidelines recommend use of 3HP only with efavirenz (EFV)

319
Subjects on TNF-α blockers are another high risk group at risk for TB disease reactivation. The 320 majority of subjects with LTBI in our study were referred because of recent TST/IGRA 321 conversion (within ≤ 2years) which put them at an even higher risk of TB reactivation disease.

322
Rifamycin based regimens have recently been shown to be effective with minimal side effects in 323 this group of patients [19]. Although only 25% of patients on TNF-α blockers were prescribed 324 3HP, 70% were able to successfully complete LTBI therapy. This is likely due to close clinic 325 follow-up that these patients receive for their underlying autoimmune disease.

326
Two important limitations of our study are the retrospective nature of the study and that subjects 327 already receiving care for co-morbid conditions at different clinics were referred to our 328 Infectious Diseases clinic, making them more likely to seek healthcare. These subjects are 329 therefore more likely to have higher rates of LTBI treatment completion compared to the overall 330 population of subjects with LTBI. Furthermore, as patients were often referred from community 331 and other specialty clinics, ID clinic providers were often aware of medical comorbidities for e.g.

332
HIV/AIDS, Diabetes, renal failure on dialysis, and ongoing use of TNF-α blocker use at initial 333 clinic assessment. Therefore we could not properly assess the actual frequency with which these 334 risk factors would be assessed had the provider not been made aware beforehand. Nevertheless, 335 we used relatively strict criteria for inclusion of subjects with previously untreated LTBI in our 336 study and had follow up data on the majority of our study cohort. Our study suggests that 337 tstin3d.com could be used in future prospective studies for determining a numerical risk of TB 338 progression in patients with LTBI for improved provider awareness of those at high risk.
can be improved by "risk score targeted" treatment (i.e. selecting a 3HP based regimen for all 341 subjects at high cumulative risk). The utility of "risk score targeted" treatment as a strategy for 342 decreasing the community burden of TB in the US needs to be validated in future prospective 343 studies.