We included all adult patients (≥18 years of age) who presented with miliary nodules, based on computed tomography (CT) of the chest, between November 2001 and April 2007. Patients were selected from Seoul National University Hospital, a tertiary referral hospital in South Korea. Miliary nodules were defined as micronodules (<7 mm in diameter 10 ), regardless of solid or non-solid nature, occupying more than two-thirds of the lung volume on a conventional chest CT or high-resolution chest CT. We excluded patients whose physician had not fully evaluated the cause of the miliary nodules, patients who defaulted before obtaining a final diagnosis, and patients with concurrent discrete macronodules (>7 mm in diameter). However, patients with concurrent consolidations were included.

The study protocol was approved by the institutional review board of Seoul National University Hospital.

We performed a retrospective cohort study to examine the relative frequencies of aetiologies of miliary pulmonary nodules and to assess clinical and radiographic predictors of miliary TB. Enrolled patients were classified into two groups, miliary TB and miliary nodules due to other causes, according to the final diagnosis of their miliary pulmonary nodules. To reveal predictors of miliary TB, clinical and radiographic characteristics were analyzed between the groups.

We retrospectively reviewed the demographic and clinical characteristics, including age, gender, body mass index (BMI), and symptoms, of the selected patients. In addition, we checked for the presence of previous TB, malignancy, diabetes mellitus, immunosuppressive drug therapy, solid organ transplantation, connective tissue disease, chronic kidney disease, chronic liver disease, HIV infection, and alcoholism. We also reviewed the results of laboratory tests, including complete blood cell counts, blood chemistry, and C-reactive protein levels. The results of histopathological examinations and microbiological studies, including acid-fast staining, mycobacterial cultures, and fungal cultures, were also reviewed.

Chest CT scans performed at the time of diagnosis were reviewed by two board-certified radiologists, by consensus. They were blinded to patient clinical data and had not participated in the case selection.

Miliary nodules were assessed for size, margin, profusion, and dominant distribution. Considering previous reports that miliary nodules from cancer usually consist of micronodules of variable size or relatively large micronodules (4–7 mm in diameter) 1 11 , we described the size of the miliary nodules as 'homogeneous' when the sizes of all micronodules were confined to 0–4 mm. The profusion of nodules was estimated by counting all nodules in two contiguous squares of 4 cm² on three levels: just above the aortic arch, the right upper lobar bronchus, and the lower portion of the left atrium 7 . The dominant distribution of miliary nodules within secondary pulmonary lobules was described as centrilobular, perilymphatic, or random 1 2 6 11 .

Poorly defined centrilobular nodules, branching linear lesions, tree-in-bud appearance, and bronchial wall thickening were classified as indicating the bronchogenic spread of lesions 7 . Patients with fibrotic bands, small calcified nodules, or bronchiectasis observable in the upper lobes were regarded as having pre-existing TB sequelae 12 13 . Mediastinal lymph nodes >1 cm in short-axis diameter were considered positive signs of lymphadenopathy. Additionally, we assessed the extent of accompanying ground-glass opacities, consolidations, and pleural effusions.

Miliary TB was confirmed if any of the following was met: (1) positive staining for AFB, the growth of Mycobacterium tuberculosis, or positive PCR for M. tuberculosis DNA in sputum, bronchial lavage fluid, pleural fluid, or tissue; (2) the presence of caseating granulomas in lung tissue and clinical and radiographic improvement with anti-TB medications; or (3) definite clinical and radiographic improvement with empirical anti-TB medications. Miliary metastasis of malignancies was identified when any of the following was met: (1) cytological/histological examination of miliary nodules showing malignant cells; or (2) concurrent changes in the size and number of miliary nodules relative to the size of measurable primary malignant lesions. For example, concurrent decreases in the sizes of both primary lesions and miliary nodules in a patient receiving anti-cancer chemotherapy would be such a case. Pneumoconiosis was diagnosed based on occupational history and compatible clinical and radiographic findings. Disseminated fungal diseases were defined using bacteriological or pathological analyses of miliary nodules. Hypersensitivity pneumonitis was diagnosed by environmental or occupational history and improvement with environmental control or corticosteroid treatment.

To identify factors for differentiating between TB-associated miliary nodules and miliary nodules due to other causes, we performed univariate analyses using the χ² test or Fisher's exact test for categorical variables, and using t tests for continuous variables. To determine predictors of miliary TB, we performed binary logistic regression analysis with significant (P < 0.05) variables from the univariate analysis. Co-linearity was eliminated by stepwise selection (entry, p = 0.05; removal, p = 0.1). Because some predictors were likely to be found only within the miliary TB group, we used the penalized likelihood ratio test to determine significance. For the same reason, odds ratios and 95% confidence intervals were estimated using the penalized maximum likelihood and profile penalized likelihood confidence intervals, respectively. Statistical significance was determined at P < 0.05. All analyses were performed using SPSS (version 12.0, SPSS Inc., Chicago, IL, USA) or SAS (version 8.2, SAS Institute Inc., Cary, NC, USA).

A total of 182 patients diagnosed with and treated for miliary pulmonary nodules during the study period were initially screened for inclusion in this study. Of these, 102 patients were excluded because of concurrent discrete macronodules (>7 mm in diameter), mass lesions suggestive of primary lung cancer, or cavitations suggestive of pulmonary TB. Indeed, the diagnoses of the 102 patients excluded from further analysis were metastatic cancer (65 patients), miliary TB (29 patients), and pneumoconiosis (8 patients). An additional four patients were excluded because they defaulted before obtaining a final diagnosis. Finally, data for 76 patients were included and analyzed. Their median age was 52 years and 38 (50%) of them were males; 18 patients (24%) had a previous or current malignancy and five (7%) had a history of TB (Table 1).

Among the 76 patients, miliary TB accounted for 54% of the miliary pulmonary nodules (41 patients). Of these, 31 patients (76%) were diagnosed by positive staining for AFB, the growth of Mycobacterium tuberculosis, or positive PCR for M. tuberculosis DNA in sputum, bronchial lavage fluid, pleural fluid, cerebrospinal fluid, or tissue. In five patients (12%), diagnoses were made by the presence of caseating granulomas in lung tissue and clinical and radiographic improvement with anti-TB medications. The other five patients (12%) were diagnosed based on definite clinical and radiographic improvement with empirical anti-TB medications. Miliary metastasis of malignancies accounted for 26% of the miliary pulmonary nodules (20 patients). Of these, direct histological confirmation of miliary pulmonary nodules via lung biopsy was achieved in one patient. Three patients were diagnosed with disseminated candidiasis, all with an underlying hematologic malignancy (Table 2).

With univariate analysis, we found that age ≤30 years, HIV infection, corticosteroid use, connective tissue disease, low BMI, body temperature >37.7°C, homogeneous nodule size, high profusion, bronchogenic spread of lesions, and ground-glass opacities occupying >25% of total lung volume were associated with miliary TB. However, a history of solid organ malignancy decreased the probability of miliary TB (Tables 3, 4). Subsequent multivariate analysis supported the association between miliary TB and age ≤30 years, HIV infection, corticosteroid use, bronchogenic spread of lesions, and ground-glass opacities occupying >25% of total lung volume, as well as the association between miliary nodules of non-miliary TB and a history of malignancy (Table 5).