CT Imaging Characteristics Correlated with Overall Survival Time in Patients with AIDS-Related non-Hodgkin’s Lymphoma

Background CT can provide useful information for treatment regimens and prognosis prediction of patients with AIDS-related non-Hodgkin’s lymphoma (AR-NHL). It is necessary to investigate the prediction of CT imaging and clinical characteristics for overall survival (OS) in patients with AR-NHL. Methods Data of 121 AR-NHL patients [median age: 41 (range 22-78), 112 male] between July 2012 and November 2019 were retrospectively reviewed. Patients were divided into two groups by median OS time and data were compared between two groups. K-M survival analysis and Cox proportional hazards regression analysis were used to determine the prognostic risk factors for OS.


Background
Acquired immune de ciency syndrome (AIDS)-related non-Hodgkin's lymphoma (AR-NHL), is a high-risk factor for morbidity and mortality in patients with AIDS (1,2). Although the incidence of AIDS-related tumors has decreased with the advent of highly active antiretroviral therapy (HAART), the occurrence rate of AR-NHL appears to be stable (3,4). So early diagnosis and evaluation are extremely important for treatment and prognosis.
HAART, chemotherapy, CD4 count, HIV RNA levels, Ann Arbor stage, lactate dehydrogenase (LDH) levels, international prognostic index (IPI) score, and age are key predictors of survival in AR-NHL patients in previous studies (5)(6)(7). However, few studies investigated the signi cance of imaging characteristics for the prediction of prognosis and survival, Novel imaging modalities for assessing lymphoma can provide useful information for treatment regimens and prediction of patients' prognoses. With the advancement of imaging technology such as CT, MR and PET/CT, radiologic techniques play an increasingly essential role in detecting lesion and evaluating disease (8)(9)(10). CT can nd the enlargement of lymph nodes and in ltration of extranodal organs, guide biopsy and observe early relapse through follow-up (11,12). MR has similar e cacy of detecting the space occupying effect with CT. Although MR is superior to diagnosis primary central nervous system lymphoma and lymphoma in skeletal muscular system, cases involved there are relatively rare (13). PET/CT can assess lymphoma stage, grade of malignancy and evaluate treatment response (14), but sometimes PET/CT is di cult to differentiate lymphoma from active in ammatory lesion or other lesions in brain and its utilization is limited due to various economic and social factors in some developing countries (15). So it is very pragmatic to assess the AR-NHL and determine the predictive factors via CT.
We hypothesized that there would be imaging differences in CT between patients with different overall survival (OS). The rst purpose of the study was to nd imaging and clinical factors indicated shorter OS. Second, we aimed to build a predictive model for the OS of AR-NHLbased on clinical and imaging parameters .

Patients
The study was conducted under an approval by the Institutional Review Board. In this multi-center retrospective study, information of 181 patients with AIDS-related lymphoma from three tertiary infectious disease hospitals were reviewed and their clinical and imaging data were analyzed between July 2012 and November 2019. The diagnosis of HIV infection was based on the standards of centers for disease control and prevention of the USA. The diagnosis of lymphoma was based on puncture biopsy (163 patients), endoscopic biopsy (6 patients), and operation specimens (12 patients). All intervention and treatment were processed according to NCCN Clinical Practice Guidelines in Oncology: non-Hodgkin's lymphomas (16)(17)(18). If the patients were in stable conditions, they were followed up once every three months in the rst year, once every six months in the second year and once every year in the third year and beyond. Patients were followed up at any time if disease progression and deterioration occurred. OS was chosen as the end point. OS was measured from the lymphoma diagnosis until last follow-up or death from any cause. Follow-up was continued until November, 2019.

Inclusion and exclusion criteria
Patients eligible for this study were (a) Age > 18 years. (b) With history of HIV-infection (c) With de nite pathologically con rmed diffuse large B-cell lymphoma (DLBCL) or Burkitt's lymphoma (BL) [19] (d) With available clinical and CT imaging data before any clinical intervention. Patients with Hodgkin's lymphoma (HL, n = 5), indolent B-cell NHL (n = 4), and T-cell NHL (n = 7) or lacking a speci c pathological type (n = 13) were excluded. One patient younger than 18 years of age and two patients with severe artifacts in the CT images were also excluded. 28 patients who were lost to follow-up were also excluded from the study. Through follow-up, the median OS time was 17 months (range, 0.5-60 months). 121 patients [median age: 41 (range 22-78), 112 male] were nally included in our research and they were subdivided into two groups in accordance with OS time. Group1 included patients with OS time ≤ median OS time (61 patients) and group 2 included those with OS time median OS time (60 patients). Figure 1 shows ow chart of patient inclusion and exclusion criteria.

CT image acquisition and interpretation
All examinations were imaged with Philips iCT 256 (Philips; Amsterdam, Netherlands), 39 patients accepted contract-enhanced CT scan via intravenous contrast materials. The CT protocols was as follows: tube voltage, 120 kV; automatic tube current, 30-300 mA; rotation time, 0.75 s; collimation, 0.625 mm; pitch, 0.945; matrix, 512*512; section thickness, 5 mm; breath hold at full inspiration. The images were transmitted to the workstation and picture achieving and communication systems (PACS) for multiplanar reconstruction and post-processing. All images (both axial CT images and multiplanar reconstruction images) were reviewed by three radiologists (Doctor A with 22 years' experience, B with 7 years' experience and C with 10 years' experience) blinded to clinical and laboratory data. Three estimators assessed the CT features independently. After separate evaluations, any divergences were resolved by discussion or consultation from a specialist in infectious imaging (Doctor D with 33 year' experience), eventually reviewed by Doctor E for consistency analysis.

Data Analysis
The baseline data were recorded, including age, sex, pathological types, clinical manifestations, time from detecting positive HIV-antibodies to admission, time from initial mass discovery to admission, Ann Arbor stages, laboratory test results, treatment situations for each group.
Regarding the different sites of intranodal involvement, we subdivided the lesions into groups of axillary lymph nodes, cervical lymph nodes, mediastinal or hilar lymph nodes, abdominal pelvic and peritoneal lymph nodes, retroperitoneal lymph nodes, and inguinal lymph nodes. For the extranodal sites, involvement of the gastrointestinal tract, urinary system organs, liver, lung, pancreas or spleen were evaluated.
A detailed evaluation of the individual maximal lesion were analyzed including the following characteristics: diameters greater than or less than 5 cm; shape of the lesion (circular/irregular); fusion tendency; presence of extracapsular in ltration; evidence of necrosis; attenuation (hyper/iso/hypo); and texture (homogeneous/heterogeneous) in plain scans and contrast-enhanced behavior (including texture and the degree of enhancement, the CT value increased by 0-20 Hu for poor enhancement, 20-50 Hu for moderate enhancement, greater than 50 Hu for severe enhancement) in enhanced scans.

Statistical analysis
The imaging ndings were tested for agreement using the Kappa test. If the Kappa value was less than 0.4, the consistency of the diagnostic ndings was poor. If the Kappa value was greater than 0.75, then the diagnostic ndings were considered to be su ciently consistent.
Continuous variables of parameters were tested for normal distribution using the Kolmogorov-Smirnov method. If the data tted a normal distribution, mean ± standard deviation (SD) and the t-test were used to check for differences between the two group. If the data were not normally distributed, then the median (IQR) and Mann-Whitney U test were used. The chi-square test and Fisher's exact test were used to compare categorical variables.
The univariate analysis of a Kaplan-Meier analysis model was tted to determine the signi cant prognostic factors for OS in all patients. If P values of prognostic factors were less than 0.1, they were tested in a multivariate Cox proportional hazard model for independence of association and factors showing signi cant impact in the multivariate analysis were expressed via forest graph. Proportional hazards assumption was assessed through visual inspection of (log-log) plots of log cumulative hazard against time. A predictive model was developed for AR-NHL using Cox regression and illustrated by nomogram. The accuracy of predictions was assessed by estimating the model's discrimination measured by the Harrell's Concordance index (C-index). The C-index is the probability that for two patients chosen at random, the patient who had the event rst had a higher probability of having the event according to the model. C-index = 0.50 represents agreement by chance; C-index = 1.0 represents perfect discrimination (20). The calibration of the nomogram was evaluated by the Hosmer-Lemeshow test. All statistical analyses were performed using SPSS version 22.0 (IBM Corp., Armonk, NY, USA). The gures were created using GraphPad Prism 7 (GraphPad Software; San Diego, CA, USA) and R software (version 4.0.1; http://www.r-project.org, with package of "rms"). The signi cance level was set at P < 0.05. Table 1 shows the demographic data of patients in each group. The median time from diagnosis of HIV to admission in group 1 was 9(2-36) months, which was statistically longer than that of group 2 [1.5(0.5-12) months, p = 0.007; Table 1]. 80% (48/60) of the patients with OS time > median OS time accepted chemotherapy vs. only 57.4% (35/61) of the patients with OS ≤ median OS time, which was a signi cant difference (P = 0.007). The CD4 count of patients with OS time ≤ median OS time was lower than the other group [157 (55-280) v 212 (130-375), p = 0.050]. Other laboratory results, such as WBC, NEUT, and lymphocytes, revealed no statistical differences between two groups. There was signi cant difference in Ann Arbor stage between two groups (p = 0.007). there were 13 patients in group 1 and 11 patients in group 2 who did not find the mass themselves; ¶ there were 2 missing data in group 1 and 1 missing data in group 2.

3.2.Comparison of CT imaging characteristics
There was good agreement among the three doctors in evaluating the CT imaging characteristics (Kappa value 0.778-0.998, Table S1). There were signi cant differences in the involvement situations of inguinal  Fig. 2f] were more common seen in patients of group 1 (Table 3).

3.3.Prognostic factors for OS
In the univariate K-M survival analysis ( Figure S1; Table S2- The predictive models based on Cox regression and illustrated by nomogram (Fig. 4)

Discussion
There were three important ndings in current study. First, presence of extracapsular in ltration, necrosis, CD4 ≤ 100 cells/µL, period from nding mass to admission>1 month, without chemotherapy, liver involved, gastrointestinal tract involved and mediastinal or hilar lymph nodes involved were associated with shorter OS. Second, liver involved, mediastinal or hilar lymph nodes involved, necrosis in lesion, CD4 ≤ 100 cells/µL and treatment without chemotherapy were independently risk factors for shorter OS. Third, chemotherapy weights most in the new predictive model for AR-NHL, then followed by liver involved, CD4 counts>100 cells/µL, mediastinal or hilar lymph nodes involved and necrosis. If the total points<295, three-year overall survival probability might<50%.
Imaging plays an important role in the detection and evaluation of AR-NHL lesions (8,10,12). CT of the head and neck, chest, abdomen, and pelvis is a critical staging modality recommended by the National Comprehensive Cancer Network guidelines (18). Necrosis shows hypoattenuation without enhancement in CT images. The potential mechanism of necrosis in lymphoma is the occlusion of the supplying hilar artery by the tumor (compression or invasion) in addition to lymphatic ow obstruction (21,22). Previous study indicated HIV(-) NHL patients with necrosis had signi cantly higher Ann Arbor stages, greater IPI, and higher serum LDH levels than those without necrosis but in Kaplan-Meier survival analysis, no statistically signi cant difference was noted for necrosis (23). Our study focused on AR-NHL patients and necrosis was an independent risk factor. Although the pathogenesis of AR-NHL necrosis remains unclear, it can be speculated that it indicates aggressive tumor growth with an apparent tendency for treatment resistance. Extracapsular in ltration was common in Kikuchi's disease (24). On CT it corresponds to periadenitis pathologically. It is caused by in ltration of in ammatory cells and karyorrhectic debris, which is the destructive fragmentation of the nucleus of dying histiocytes and plasmacytoid monocytes around the lymph node. So lymph node capsule may be broken by in ammation (24). However, it could also be observed in malignant lymphadenopathy with extracapsular extension such as lymphoma and metastasis (25). The invasion of tumor cells may be a potential mechanism of extracapsular in ltration and may correlate with poor prognosis. Natural killer (NK) cells play an important role in growth and in ltration of lymphoma cells, and activated NK cells could be a promising immunotherapeutic tool against lymphoma cells either alone or in combination with conventional therapy (26).
AR-NHLs are usually B-cell, high-grade, and poorly differentiated lymphomas (27). Extranodal sites involvement are common, the liver is the second most common site of abdominal involvement after the gastrointestinal tract with an incidence ranging from 26-45%. HIV (+) patients have a higher relative incidence of NHL than HIV (-) patients (28). HIV (+) patients are also more likely to have HCV, and vice versa. Persons with HCV have an almost two-fold greater risk of NHL, in addition, higher prevalence of HCV is associated with hepatocellular carcinoma (29 Intensive chemotherapy is relative safe and effective in AIDS non-Hodgkin's lymphoma (35).
Chemotherapy and concomitant HAART for AIDS-related NHL does not cause prolonged suppression of lymphocyte subsets. On the contrary, chemotherapy can increase the counts of CD4, CD8, CD19 and CD56 cell populations, which provide reassurance regarding the long-term consequences of chemotherapy in these individuals (36). Even for advanced AR-BL 70% patients achieved complete response by intensive chemotherapy regimen (LMB86), less than 10% patients occurred treatment-related deaths severe bone marrow toxicity should be concerned (37). Despite the chemotherapy treatmentrelated toxicity and mortality, standard chemotherapy remains the rst line treatment for patients with AR-NHL .
Our study has several limitations. First, it is a retrospective study with a limited number of patients. The small number of cases in some categories may in uence the stability of risk estimation. Second, we analyzed clinical and imaging data of AR-DLBCL and AR-BL together although no signi cant difference in pathology classi cations between two groups was found in current study. Other limitations include shorter follow-up time and potential selection bias. Prospective investigations with larger samples, focus on certain pathological type should be designed in order to nd more predictive factors for prognoses of AR-NHL patients.

Conclusion
CT is essential and pragmatic to assess AR-NHL patient's condition. Intensive chemotherapy regimens and more frequent follow-up should be considered for patients with characteristics including lower CD4 counts, necrosis and extracapsular in ltration in lesion, vital sites like mediastinal or hilar lymph nodes, liver, gastrointestinal tract involved.

Abbreviations
CT=computed tomography