Predicting the risk of pulmonary tuberculosis based on the neutrophil-to-lymphocyte ratio at TB screening in HIV-infected individuals

In Thailand, day care centers (DCC) for HIV-infected individuals were established at the district hospitals in 1995 to support HIV treatment according to the national HIV/AIDS clinical guideline [13]. These centers provide social and psychological care [14]. Mae Chan hospital is a one of the large district hospitals in Chiang Rai province, which is located in the northernmost region of Thailand. In this hospital, the DCC works closely with the HIV and TB clinic to provide medical services. HIV-infected individuals are routinely introduced to the DCC at the hospital and are screened for TB based on clinical history, chest radiography, and sputum smear microscopic examination for acid fast bacilli (AFB) by trained laboratory technicians from the TB clinic. Culture and drug susceptibility testing is primarily performed in patients with positive sputum smear results after the samples are transferred to the microbiology laboratory at Chiang Rai hospital. In Chiang Rai province, isoniazid preventive therapy (IPT), as recommended in the national guideline, has been provided since 2001 to HIV-infected individuals who have positive tuberculin skin test (TST) results. However, only 9% of HIV-infected individuals received the IPT during this study period, because the initiation of IPT prophylaxis was dependent on the accessibility of isoniazid for preventive therapy in the hospitals and based on the decision of physicians.

Patients eligible for this prospective cohort study were HIV-infected individuals older than 18 years of age who had not previously received TB treatment and were registered at the DCC between 2002 and 2015. Patients who started IPT before the registration at the DCC and who started IPT after DCC registration were excluded from this analysis to avoid the intervention effect. Patients registered at the DCC received their TB screening at their first visit to the DCC. People who were diagnosed with active TB and were prescribed anti-TB treatment at the first TB screening were excluded from the analysis.

Baseline TB screening data were collected by trained research nurses at the DCC. A questionnaire was used to collect demographic information (age, sex, and ethnicity), recent TB symptoms, current and prior anti-retroviral treatment (ART), and TB treatment history (Additional file 1). Recent TB symptoms included: cough, night sweats, fever, weight loss, and hemoptysis within the last 4 weeks. All patients received 3-day sputum smear tests for AFB, chest X-ray, and blood laboratory tests at the hospital. When a sputum AFB smear result was positive, or if it was negative but from a patients with a high suspicion of TB, sputum samples were processed for Mycobacterium tuberculosis culture tests after being transferred to the microbiology laboratory in Chiang Rai hospital. The blood laboratory tests included CD4 cell counts, white blood cell counts, and differential counts. NL ratio was calculated from the absolute neutrophil counts divided by the absolute lymphocyte counts. Chest X-rays were interpreted for abnormal signs such as infiltrate, cavity, and pleural effusion by physicians who treated TB patients. TST was also performed during TB screening to identify latent TB infection. The skin test reaction was read 48–72 h after administration. A diameter of 5 or more millimeters in the indurated area was considered positive for latent TB infection [15]. The provincial-wide TB registry database was used to ascertain and validate the TB diagnosis and the treatment outcome, including mortality data. The provincial-wide TB registry database is a reporting system for TB, which records socio-demographic information, medication, treatment outcome, and laboratory examinations for all TB patients diagnosed in the TB clinic and the public hospital. TB cases were diagnosed based on clinical symptoms; and smear AFB results and TB culture results used for TB diagnosis could not be accessed during the follow-up.

Baseline characteristics at the first TB screening were summarized as median and interquartile range (IQR) for continuous variables and as frequency for categorical variables. Comparisons between positive TB screening and negative TB screening groups were tested using the chi-square test or Wilcoxon rank-sum test for categorical and continuous variables, respectively.

Univariate Cox-proportional hazards models were used to estimate the relationship between potential predictive variables and time-to-event outcome among individuals who screened negative for TB. The multivariate Cox proportional hazards model for the risk of TB diagnosis (including the variables (age, sex, ethnicity, year of DCC registration, ART status, TB symptoms, TST results, and chest X-ray results) assessed the effect of the NL ratio on the incidence of TB. Cubic-spline curves were used to assess possible nonlinear relationships between NL ratio and the outcome, and to determine the threshold of the NL ratio. The analysis was conducted using multivariate-adjusted cox-regression models and used 5 knots. The threshold level of NL ratio was identified on the cubic spline curve, which showed a statistically significant effect on TB development (adjusted hazard ratio > 1). The NL ratio threshold was used to categorize NL ratio into low or high. The effects of the categorized NL ratio (based on the threshold level) were analyzed in sub-groups (normal chest X-ray, TB symptoms and each CD4 count category). To determine the effect of NL ratio combined with TB symptoms and chest X-ray at the TB screening, we categorized participants into four groups (no-TB symptoms and normal chest X-ray, TB symptoms and normal chest X-ray, no-TB symptoms and abnormal chest X ray, and TB symptoms and abnormal chest X-ray) and assessed the effect of NL ratio on TB development. P-values were calculated using the likelihood ratio test. The threshold of statistical significance was a P-value of < 0.05. All analyses were conducted using STATA 14.0 (STATA Corp., College Station, Texas, USA).

A total of 1118 HIV-infected individuals (median age 35 years; 46.4% male) underwent TB screening with at least 3 sputum assays at the DCC between 2002 and 2015. Fifty-four patients diagnosed with active TB and who started TB medication were excluded from the subsequent analysis, based on the detection of at least 1 AFB positive smear (N = 8, 14.8%), TB culture positive (N = 40, 74.1%), or clinical symptoms highly suspected to be TB (N = 41, 75.9%). Patients that were TST positive, had abnormal chest X-rays, or reported TB symptoms were strongly associated with active TB compared to patients who were TB screening negative (Table 1). Baseline median NL ratio (median 3.27; IQR: 1.99–5.41) in active TB at the first TB screening were significantly higher than that in patients who were TB screening negative (median NL ratio 1.66; IQR: 1.14–2.54; p < 0.001).

Of 1064 individuals who were TB screening negative (916.23 person-years follow-up), 5.6% (n = 60) of patients developed TB (incidence rate: 65.5 per 1000 person-year, 95% CI: 50.9–84.3) and 78.3% (n = 47) of them were diagnosed within 6 months after the first TB screening (Fig. 1). Among the 60 patients who were diagnosed with TB after the first TB screening, 26 died within 6 months of TB diagnosis (median 31 days, IQR: 10–68 days). After adjusting for age, sex, ethnicity, year of registration, CD4 counts, TB symptoms, abnormal chest x-ray, TST results and NL ratio, low CD4 counts (< 200/mm³), positive TST, TB symptoms and male sex were found to increase the risk of TB development. TST positive patients had a 3.5-fold higher risk of becoming a new TB case. However, the proportion of TST positive patients attributable to TB cases was only 10.1%. A total of 330 (31.0%) patients commenced ART before TB diagnosis, while 123 (11.6%) patients started ART after or at the time of TB diagnosis. After adjusting for age, sex, ethnicity, year of DCC registration, CD4 counts, ART, TST results, chest X-ray and TB symptoms (Table 2), each unit difference of NL ratio corresponded to a 1.06-fold higher risk of TB development. Multivariable-adjusted restricted cubic spline analysis (Fig. 2) suggested that the aHR of TB diagnosis gradually increased corresponding to increasing NL ratio. NL ratio above 2 was the threshold level for a significant effect on TB development.

Categorized NL ratios (high: NL ratio > 2 and low: NL ratio ≦2) were analyzed to examine whether the NL ratio was associated with outcomes in different conditions (TB symptoms, chest X-ray and CD4 count categories) (Fig. 1). A high NL ratio corresponded to a 2.2-folds higher risk of progression to active TB compared to low NL ratio among all HIV-infected individuals with negative TB screening. Even among the normal chest X-ray group, high NL ratio was a high risk for active TB. Moreover, among the normal chest X-ray group, combined TB symptoms with a high NL ratio could predict a 2.5-times higher risk of TB development compared to a low NL ratio combined with TB symptoms. When stratified by CD4 counts, the association between high NL ratio and TB disease was stronger among the lowest CD4 count group (CD4 counts < 50/mm³) compared to the group with higher CD4 counts (≧50/mm³), (Table 3).