Off-label drug use during the COVID-19 pandemic in Africa: topic modelling and sentiment analysis of ivermectin in South Africa and Nigeria as a case study

Although rejected by the World Health Organization, the human and even veterinary formulation of ivermectin has widely been used for prevention and treatment of COVID-19. In this work we leverage Twitter to understand the reasons for the drug use from ivermectin supporters, their source of information, their emotions, their gender demographics, and location information, in Nigeria and South Africa. Topic modelling is performed on a Twitter dataset gathered using keywords 'ivermectin' and 'ivm'. A model is fine-tuned on RoBERTa to find the stance of the tweets. Statistical analysis is performed to compare the stance and emotions. Most ivermectin supporters either redistribute conspiracy theories posted by influencers, or refer to flawed studies confirming ivermectin efficacy in vitro. Three emotions have the highest intensity, optimism, joy and disgust. The number of anti-ivermectin tweets has a significant positive correlation with vaccination rate. All the provinces in South Africa and most of the provinces of Nigeria are pro-ivermectin and have higher disgust polarity. This work makes the effort to understand public discussions regarding ivermectin during the COVID-19 pandemic to help policy-makers understand the rationale behind its popularity, and inform more targeted policies to discourage self-administration of ivermectin. Moreover, it is a lesson to future outbreaks.

Twitter-based gender recognition using transformers

Social media contains useful information about people and society that could help advance research in many different areas of health (e.g. by applying opinion mining, emotion/sentiment analysis and statistical analysis) such as mental health, health surveillance, socio-economic inequality and gender vulnerability. User demographics provide rich information that could help study the subject further. However, user demographics such as gender are considered private and are not freely available. In this study, we propose a model based on transformers to predict the user's gender from their images and tweets. The image-based classification model is trained in two different methods: using the profile image of the user and using various image contents posted by the user on Twitter. For the first method a Twitter gender recognition dataset, publicly available on Kaggle and for the second method the PAN-18 dataset is used. Several transformer models, i.e. vision transformers (ViT), LeViT and Swin Transformer are fine-tuned for both of the image datasets and then compared. Next, different transformer models, namely, bidirectional encoders representations from transformers (BERT), RoBERTa and ELECTRA are fine-tuned to recognize the user's gender by their tweets. This is highly beneficial, because not all users provide an image that indicates their gender. The gender of such users could be detected from their tweets. The significance of the image and text classification models were evaluated using the Mann-Whitney U test. Finally, the combination model improved the accuracy of image and text classification models by 11.73 and 5.26% for the Kaggle dataset and by 8.55 and 9.8% for the PAN-18 dataset, respectively. This shows that the image and text classification models are capable of complementing each other by providing additional information to one another. Our overall multimodal method has an accuracy of 88.11% for the Kaggle and 89.24% for the PAN-18 dataset and outperforms state-of-the-art models. Our work benefits research that critically require user demographic information such as gender to further analyze and study social media content for health-related issues.

Big data- and artificial intelligence-based hot-spot analysis of COVID-19: Gauteng, South Africa, as a case study

The coronavirus disease 2019 (COVID-19) has developed into a pandemic. Data-driven techniques can be used to inform and guide public health decision- and policy-makers. In generalizing the spread of a virus over a large area, such as a province, it must be assumed that the transmission occurs as a stochastic process. It is therefore very difficult for policy and decision makers to understand and visualize the location specific dynamics of the virus on a more granular level. A primary concern is exposing local virus hot-spots, in order to inform and implement non-pharmaceutical interventions. A hot-spot is defined as an area experiencing exponential growth relative to the generalised growth of the pandemic. This paper uses the first and second waves of the COVID-19 epidemic in Gauteng Province, South Africa, as a case study. The study aims provide a data-driven methodology and comprehensive case study to expose location specific virus dynamics within a given area. The methodology uses an unsupervised Gaussian Mixture model to cluster cases at a desired granularity. This is combined with an epidemiological analysis to quantify each cluster's severity, progression and whether it can be defined as a hot-spot.

Nowcasting unemployment rate during the COVID-19 pandemic using Twitter data: The case of South Africa

The global economy has been hard hit by the COVID-19 pandemic. Many countries are experiencing a severe and destructive recession. A significant number of firms and businesses have gone bankrupt or been scaled down, and many individuals have lost their jobs. The main goal of this study is to support policy- and decision-makers with additional and real-time information about the labor market flow using Twitter data. We leverage the data to trace and nowcast the unemployment rate of South Africa during the COVID-19 pandemic. First, we create a dataset of unemployment-related tweets using certain keywords. Principal Component Regression (PCR) is then applied to nowcast the unemployment rate using the gathered tweets and their sentiment scores. Numerical results indicate that the volume of the tweets has a positive correlation, and the sentiments of the tweets have a negative correlation with the unemployment rate during and before the COVID-19 pandemic. Moreover, the now-casted unemployment rate using PCR has an outstanding evaluation result with a low Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), Symmetric MAPE (SMAPE) of 0.921, 0.018, 0.018, respectively and a high R²-score of 0.929.

Planetary sleep medicine: Studying sleep at the individual, population, and planetary level

Circadian rhythms are a series of endogenous autonomous oscillators that are generated by the molecular circadian clock which coordinates and synchronizes internal time with the external environment in a 24-h daily cycle (that can also be shorter or longer than 24 h). Besides daily rhythms, there exist as well other biological rhythms that have different time scales, including seasonal and annual rhythms. Circadian and other biological rhythms deeply permeate human life, at any level, spanning from the molecular, subcellular, cellular, tissue, and organismal level to environmental exposures, and behavioral lifestyles. Humans are immersed in what has been called the "circadian landscape," with circadian rhythms being highly pervasive and ubiquitous, and affecting every ecosystem on the planet, from plants to insects, fishes, birds, mammals, and other animals. Anthropogenic behaviors have been producing a cascading and compounding series of effects, including detrimental impacts on human health. However, the effects of climate change on sleep have been relatively overlooked. In the present narrative review paper, we wanted to offer a way to re-read/re-think sleep medicine from a planetary health perspective. Climate change, through a complex series of either direct or indirect mechanisms, including (i) pollution- and poor air quality-induced oxygen saturation variability/hypoxia, (ii) changes in light conditions and increases in the nighttime, (iii) fluctuating temperatures, warmer values, and heat due to extreme weather, and (iv) psychological distress imposed by disasters (like floods, wildfires, droughts, hurricanes, and infectious outbreaks by emerging and reemerging pathogens) may contribute to inducing mismatches between internal time and external environment, and disrupting sleep, causing poor sleep quantity and quality and sleep disorders, such as insomnia, and sleep-related breathing issues, among others. Climate change will generate relevant costs and impact more vulnerable populations in underserved areas, thus widening already existing global geographic, age-, sex-, and gender-related inequalities.

Leveraging artificial intelligence to optimize COVID-19 robust spread and vaccination roll-out strategies in Southern Africa

The outbreak of coronavirus in the year 2019 (COVID-19), caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) prompted widespread illness, death, and extended economic devastation worldwide. In response, numerous countries, including Botswana and South Africa, instituted various clinical public health (CPH) strategies to mitigate and control the disease. However, the emergence of variants of concern (VOC), vaccine hesitancy, morbidity, inadequate and inequitable vaccine supply, and ineffective vaccine roll-out strategies caused continuous disruption of essential services. Based on Botswana and South Africa hospitalization and mortality data, we studied the impact of age and gender on disease severity. Comparative analysis was performed between the two countries to establish a vaccination strategy that could complement the existing CPH strategies. To optimize the vaccination roll-out strategy, artificial intelligence was used to identify the population groups in need of insufficient vaccines. We found that COVID-19 was associated with several comorbidities. However, hypertension and diabetes were more severe and common in both countries. The elderly population aged ≥60 years had 70% of major COVID-19 comorbidities; thus, they should be prioritized for vaccination. Moreover, we found that the Botswana and South Africa populations had similar COVID-19 mortality rates. Hence, our findings should be extended to the rest of Southern African countries since the population in this region have similar demographic and disease characteristics.

Knowing the unknown: The underestimation of monkeypox cases. Insights and implications from an integrative review of the literature

Monkeypox is an emerging zoonotic disease caused by the monkeypox virus, which is an infectious agent belonging to the genus Orthopoxvirus. Currently, commencing from the end of April 2022, an outbreak of monkeypox is ongoing, with more than 43,000 cases reported as of 23 August 2022, involving 99 countries and territories across all the six World Health Organization (WHO) regions. On 23 July 2022, the Director-General of the WHO declared monkeypox a global public health emergency of international concern (PHEIC), since the outbreak represents an extraordinary, unusual, and unexpected event that poses a significant risk for international spread, requiring an immediate, coordinated international response. However, the real magnitude of the burden of disease could be masked by failures in ascertainment and under-detection. As such, underestimation affects the efficiency and reliability of surveillance and notification systems and compromises the possibility of making informed and evidence-based policy decisions in terms of the adoption and implementation of ad hoc adequate preventive measures. In this review, synthesizing 53 papers, we summarize the determinants of the underestimation of sexually transmitted diseases, in general, and, in particular, monkeypox, in terms of all their various components and dimensions (under-ascertainment, underreporting, under-detection, under-diagnosis, misdiagnosis/misclassification, and under-notification).

Public sentiments toward COVID-19 vaccines in South African cities: An analysis of Twitter posts

Amidst the COVID-19 vaccination, Twitter is one of the most popular platforms for discussions about the COVID-19 vaccination. These types of discussions most times lead to a compromise of public confidence toward the vaccine. The text-based data generated by these discussions are used by researchers to extract topics and perform sentiment analysis at the provincial, country, or continent level without considering the local communities. The aim of this study is to use clustered geo-tagged Twitter posts to inform city-level variations in sentiments toward COVID-19 vaccine-related topics in the three largest South African cities (Cape Town, Durban, and Johannesburg). VADER, an NLP pre-trained model was used to label the Twitter posts according to their sentiments with their associated intensity scores. The outputs were validated using NB (0.68), LR (0.75), SVMs (0.70), DT (0.62), and KNN (0.56) machine learning classification algorithms. The number of new COVID-19 cases significantly positively correlated with the number of Tweets in South Africa (Corr = 0.462, P < 0.001). Out of the 10 topics identified from the tweets using the LDA model, two were about the COVID-19 vaccines: uptake and supply, respectively. The intensity of the sentiment score for the two topics was associated with the total number of vaccines administered in South Africa (P < 0.001). Discussions regarding the two topics showed higher intensity scores for the neutral sentiment class (P = 0.015) than for other sentiment classes. Additionally, the intensity of the discussions on the two topics was associated with the total number of vaccines administered, new cases, deaths, and recoveries across the three cities (P < 0.001). The sentiment score for the most discussed topic, vaccine uptake, differed across the three cities, with (P = 0.003), (P = 0.002), and (P < 0.001) for positive, negative, and neutral sentiments classes, respectively. The outcome of this research showed that clustered geo-tagged Twitter posts can be used to better analyse the dynamics in sentiments toward community-based infectious diseases-related discussions, such as COVID-19, Malaria, or Monkeypox. This can provide additional city-level information to health policy in planning and decision-making regarding vaccine hesitancy for future outbreaks.

A call for citizen science in pandemic preparedness and response: beyond data collection

The COVID-19 pandemic has underlined the need to partner with the community in pandemic preparedness and response in order to enable trust-building among stakeholders, which is key in pandemic management. Citizen science, defined here as a practice of public participation and collaboration in all aspects of scientific research to increase knowledge and build trust with governments and researchers, is a crucial approach to promoting community engagement. By harnessing the potential of digitally enabled citizen science, one could translate data into accessible, comprehensible and actionable outputs at the population level. The application of citizen science in health has grown over the years, but most of these approaches remain at the level of participatory data collection. This narrative review examines citizen science approaches in participatory data generation, modelling and visualisation, and calls for truly participatory and co-creation approaches across all domains of pandemic preparedness and response. Further research is needed to identify approaches that optimally generate short-term and long-term value for communities participating in population health. Feasible, sustainable and contextualised citizen science approaches that meaningfully engage affected communities for the long-term will need to be inclusive of all populations and their cultures, comprehensive of all domains, digitally enabled and viewed as a key component to allow trust-building among the stakeholders. The impact of COVID-19 on people’s lives has created an opportune time to advance people’s agency in science, particularly in pandemic preparedness and response.

Harnessing Artificial Intelligence to assess the impact of nonpharmaceutical interventions on the second wave of the Coronavirus Disease 2019 pandemic across the world

In the present paper, we aimed to determine the influence of various non-pharmaceutical interventions (NPIs) enforced during the first wave of COVID-19 across countries on the spreading rate of COVID-19 during the second wave. For this purpose, we took into account national-level climatic, environmental, clinical, health, economic, pollution, social, and demographic factors. We estimated the growth of the first and second wave across countries by fitting a logistic model to daily-reported case numbers, up to the first and second epidemic peaks. We estimated the basic and effective (second wave) reproduction numbers across countries. Next, we used a random forest algorithm to study the association between the growth rate of the second wave and NPIs as well as pre-existing country-specific characteristics. Lastly, we compared the growth rate of the first and second waves of COVID-19. The top three factors associated with the growth of the second wave were body mass index, the number of days that the government sets restrictions on requiring facial coverings outside the home at all times, and restrictions on gatherings of 10 people or less. Artificial intelligence techniques can help scholars as well as decision and policy-makers estimate the effectiveness of public health policies, and implement "smart" interventions, which are as efficacious as stringent ones.

Management of hospital beds and ventilators in the Gauteng province, South Africa, during the COVID-19 pandemic

We conducted an observational retrospective study on patients hospitalized with COVID-19, during March 05, 2020, to October 28, 2021, and developed an agent-based model to evaluate effectiveness of recommended healthcare resources (hospital beds and ventilators) management strategies during the COVID-19 pandemic in Gauteng, South Africa. We measured the effectiveness of these strategies by calculating the number of deaths prevented by implementing them. We observed differ ences between the epidemic waves. The length of hospital stay (LOS) during the third wave was lower than the first two waves. The median of the LOS was 6.73 days, 6.63 days and 6.78 days for the first, second and third wave, respectively. A combination of public and private sector provided hospital care to COVID-19 patients requiring ward and Intensive Care Units (ICU) beds. The private sector provided 88.4% of High care (HC)/ICU beds and 49.4% of ward beds, 73.9% and 51.4%, 71.8% and 58.3% during the first, second and third wave, respectively. Our simulation results showed that with a high maximum capacity, i.e., 10,000 general and isolation ward beds, 4,000 high care and ICU beds and 1,200 ventilators, increasing the resource capacity allocated to COVID- 19 patients by 25% was enough to maintain bed availability throughout the epidemic waves. With a medium resource capacity (8,500 general and isolation ward beds, 3,000 high care and ICU beds and 1,000 ventilators) a combination of resource management strategies and their timing and criteria were very effective in maintaining bed availability and therefore preventing excess deaths. With a low number of maximum available resources (7,000 general and isolation ward beds, 2,000 high care and ICU beds and 800 ventilators) and a severe epidemic wave, these strategies were effective in maintaining the bed availability and minimizing the number of excess deaths throughout the epidemic wave.

A Bibliometric Network Analysis of Recent Publications on Digital Agriculture to Depict Strategic Themes and Evolution Structure

The agriculture sector is one of the backbones of many countries’ economies. Its processes have been changing to enable technology adoption to increase productivity, quality, and sustainable development. In this research, we present a scientific mapping of the adoption of precision techniques and breakthrough technologies in agriculture, so-called Digital Agriculture. To do this, we used 4694 documents from the Web of Science database to perform a Bibliometric Performance and Network Analysis of the literature using SciMAT software with the support of the PICOC protocol. Our findings presented 22 strategic themes related to Digital Agriculture, such as Internet of Things (IoT), Unmanned Aerial Vehicles (UAV) and Climate-smart Agriculture (CSA), among others. The thematic network structure of the nine most important clusters (motor themes) was presented and an in-depth discussion was performed. The thematic evolution map provides a broad perspective of how the field has evolved over time from 1994 to 2020. In addition, our results discuss the main challenges and opportunities for research and practice in the field of study. Our findings provide a comprehensive overview of the main themes related to Digital Agriculture. These results show the main subjects analyzed on this topic and provide a basis for insights for future research.

Leveraging Artificial Intelligence and Big Data to Optimize COVID-19 Clinical Public Health and Vaccination Roll-Out Strategies in Africa

COVID-19 is imposing massive health, social and economic costs. While many developed countries have started vaccinating, most African nations are waiting for vaccine stocks to be allocated and are using clinical public health (CPH) strategies to control the pandemic. The emergence of variants of concern (VOC), unequal access to the vaccine supply and locally specific logistical and vaccine delivery parameters, add complexity to national CPH strategies and amplify the urgent need for effective CPH policies. Big data and artificial intelligence machine learning techniques and collaborations can be instrumental in an accurate, timely, locally nuanced analysis of multiple data sources to inform CPH decision-making, vaccination strategies and their staged roll-out. The Africa-Canada Artificial Intelligence and Data Innovation Consortium (ACADIC) has been established to develop and employ machine learning techniques to design CPH strategies in Africa, which requires ongoing collaboration, testing and development to maximize the equity and effectiveness of COVID-19-related CPH interventions.

Diagnostic utility and therapeutic impact of PET/CT [18F]F-Fluoromethylcholine -Choline in the biochemical recurrence of prostate cancer

OBJECTIVE

To assess the diagnostic capability of PET/CT with [¹⁸F]F-Fluoromethylcholine in prostate cancer (PC) with biochemical recurrence and its therapeutic impact.

MATERIAL AND METHODS

We included 108 patients, diagnosed with PC with biochemical criteria for recurrence. A PET/CT Choline scan was performed by dynamic pelvic and whole body study at 60min post-tracer injection. The relationship between the positive studies and the PSA value was analysed by classifying patients into three groups (<1.2/1.2-2/>2ng/ml), and the diagnostic capacity was assessed with respect to pelvic MRI and the impact on the therapeutic decision.

RESULTS

The location of recurrence was identified in 85 of 108 patients (78.7%): 34 local, 47 pelvic lymph nodes and 58 distant lesions, including retroperitoneal, mediastinal lymph nodes and distant organ lesions (bone and lung). Second tumors were diagnosed in 4 patients. No significant differences were found in the percentage of positive studies depending on primary treatment. Patients with PSA>2ng/ml showed a higher percentage of disease detection than patients with a lower PSA level, with significant differences (p<0.0001). PET/CT [¹⁸F]F-Choline was able to detect local disease, not previously known from MRI, in 29.41% of patients. PET/CT Choline had an impact on therapeutic management in 67 of 108 patients (62%).

CONCLUSIONS

PET/CT with [¹⁸F]F-Fluoromethylcholine is a useful tool in the detection of locoregional and disseminated disease of PC treated with suspicion of recurrence, providing a change in therapeutic management in 62% of patients.

Association between PD1 mRNA and response to anti-PD1 monotherapy across multiple cancer types

Background

We hypothesized that the abundance of PD1 mRNA in tumor samples might explain the differences in overall response rates (ORR) observed following anti-PD1 monotherapy across cancer types.

Patients and methods

RNASeqv2 data from 10 078 tumor samples representing 34 different cancer types was analyzed from TCGA. Eighteen immune-related gene signatures and 547 immune-related genes, including PD1, were explored. Correlations between each gene/signature and ORRs reported in the literature following anti-PD1 monotherapy were calculated. To translate the in silico findings to the clinical setting, we analyzed the expression of PD1 mRNA using the nCounter platform in 773 formalin-fixed paraffin embedded (FFPE) tumor samples across 17 cancer types. To test the direct relationship between PD1 mRNA, PDL1 immunohistochemistry (IHC), stromal tumor-infiltrating lymphocytes (sTILs) and ORR, we evaluated an independent FFPE-based dataset of 117 patients with advanced disease treated with anti-PD1 monotherapy.

Results

In pan-cancer TCGA, PD1 mRNA expression was found strongly correlated (r > 0.80) with CD8 T-cell genes and signatures and the proportion of PD1 mRNA-high tumors (80th percentile) within a given cancer type was variable (0%-84%). Strikingly, the PD1-high proportions across cancer types were found strongly correlated (r = 0.91) with the ORR following anti-PD1 monotherapy reported in the literature. Lower correlations were found with other immune-related genes/signatures, including PDL1. Using the same population-based cutoff (80th percentile), similar proportions of PD1-high disease in a given cancer type were identified in our in-house 773 tumor dataset as compared with TCGA. Finally, the pre-established PD1 mRNA FFPE-based cutoff was found significantly associated with anti-PD1 response in 117 patients with advanced disease (PD1-high 51.5%, PD1-intermediate 26.6% and PD1-low 15.0%; odds ratio between PD1-high and PD1-intermediate/low = 8.31; P < 0.001). In this same dataset, PDL1 tumor expression by IHC or percentage of sTILs was not found associated with response.

Conclusions

Our study provides a clinically applicable assay that links PD1 mRNA abundance, activated CD8 T-cells and anti-PD1 efficacy.

Fierce-21: Phase II study of vofatmab (B-701), a selective inhibitor of FGFR3, as salvage therapy in metastatic urothelial carcinoma (mUC)

409

Background: Patients (pts) with mUC who have failed platinum-based chemotherapy have a poor prognosis. About 20% of them usually respond to immune checkpoint inhibitors (ICI). Also, 20% of pts with mUC harbor FGFR3 mutations or fusions (M/F), and this feature may be associated with lower sensitivity to ICI. Vofatamab (B-701) is a fully human monoclonal antibody against FGFR3 that blocks activation of the wildtype and genetically activated receptor. FIERCE-21 is a Phase 1b/2 study designed to evaluate vofatamab monotherapy (VM) or in combination with docetaxel (VD). Methods: The study consists of a P 1b lead-in (P1b with VD), previously reported followed by P2 expansion cohort in FGFR3 M/F+ pts (identified with the FoundationONE CDx™ assay on archival samples). The study enrolled mUC pts with failure to ≥ 1 line prior chemotherapy (including prior taxane treatment) or ≤12 months of (neo)adjuvant chemotherapy, measurable disease and ECOG ≤ 1. Treatment consisted of vofatamab at 25 mg/kg alone (VM) and in combination with D (VD) at 75 mg/m2 q3w. Efficacy was assessed by investigators (RECIST 1.1). Primary objectives were safety and activity (objective response-rate [ORR]). Results: 55 pts have received treatment. In the Ph 2 study, 21 pts received VM, 15 pts received VD. 35% of pts were included as 2nd line therapy, 65% in 3rd or later line of treatment. For pts receiving VD, safety was consistent with Ph 1B data. For pts receiving VM, median age was 67 yrs, ECOG 1 = 71.4%, Hgb < 10 gm/dL 13%, liver metastases 19%, ≥ 2 prior regimens 57%, (best response to prior therapy PD 31%). TEAEs occurring in ≥5% patients were asthenia (19%), diarrhea (9.5%), flushing 14%, chills (9.5%), hypotension (9.5%), decreased appetite (19%) and creatinine increased (9.5%). The majority of TEAS were grade 1 and 2. Only 1 patient on MF had a grade 3 TEAE and only 1 patient discontinued treatment due to an AE. The ORR have been seen in 7 pts to date including those receiving both VM and VD. Conclusions: Vofatamab both alone and combined with D in an every 3-week schedule are well tolerated with a low frequency of grade 3 TEAEs. Both VM and VD have demonstrated efficacy in terms of ORR, and mature results with PFS data will be presented. Clinical trial information: NCT02401542.

Acute tubulointerstitial nephritis associated with atezolizumab, an anti-programmed death-ligand 1 (pd-l1) antibody therapy

Direct stimulation of the antitumor activity of immune system through checkpoint inhibitors (ICIs) has demonstrated efficacy in the treatment of different cancer types. The activity of these antibodies takes place in the immunological synapse blocking the binding of the negative immunoregulatory proteins, thus leading to the finalization of the immune response. Despite having a favorable toxicity profile, its mechanism of action impedes the negative regulation of the immune activity which can potentially favor autoimmune attacks to normal tissues. Renal toxicity has been described in several ICI but not with atezolizumab, an IgG1 monoclonal antibody targeting PD-L1 (programmed death ligand 1), approved by FDA as a second-line therapy for advanced urothelial carcinoma. Here we present a patient with a single kidney and metastatic renal cell carcinoma treated with atezolizumab and bevacizumab combination, with biopsy-proven acute interstitial nephritis, who had a complete resolution of renal dysfunction after steroid therapy.

Androgen receptor gene status in plasma DNA associates with worse outcome on enzalutamide or abiraterone for castration-resistant prostate cancer: a multi-institution correlative biomarker study

BACKGROUND

There is an urgent need to identify biomarkers to guide personalized therapy in castration-resistant prostate cancer (CRPC). We aimed to clinically qualify androgen receptor (AR) gene status measurement in plasma DNA using multiplex droplet digital PCR (ddPCR) in pre- and post-chemotherapy CRPC.

METHODS

We optimized ddPCR assays for AR copy number and mutations and retrospectively analyzed plasma DNA from patients recruited to one of the three biomarker protocols with prospectively collected clinical data. We evaluated associations between plasma AR and overall survival (OS) and progression-free survival (PFS) in 73 chemotherapy-naïve and 98 post-docetaxel CRPC patients treated with enzalutamide or abiraterone (Primary cohort) and 94 chemotherapy-naïve patients treated with enzalutamide (Secondary cohort; PREMIERE trial).

RESULTS

In the primary cohort, AR gain was observed in 10 (14%) chemotherapy-naïve and 33 (34%) post-docetaxel patients and associated with worse OS [hazard ratio (HR), 3.98; 95% CI 1.74-9.10; P < 0.001 and HR 3.81; 95% CI 2.28-6.37; P < 0.001, respectively], PFS (HR 2.18; 95% CI 1.08-4.39; P = 0.03, and HR 1.95; 95% CI 1.23-3.11; P = 0.01, respectively) and rate of PSA decline ≥50% [odds ratio (OR), 4.7; 95% CI 1.17-19.17; P = 0.035 and OR, 5.0; 95% CI 1.70-14.91; P = 0.003, respectively]. AR mutations [2105T>A (p.L702H) and 2632A>G (p.T878A)] were observed in eight (11%) post-docetaxel but no chemotherapy-naïve abiraterone-treated patients and were also associated with worse OS (HR 3.26; 95% CI 1.47-not reached; P = 0.004). There was no interaction between AR and docetaxel status (P = 0.83 for OS, P = 0.99 for PFS). In the PREMIERE trial, 11 patients (12%) with AR gain had worse PSA-PFS (sPFS) (HR 4.33; 95% CI 1.94-9.68; P < 0.001), radiographic-PFS (rPFS) (HR 8.06; 95% CI 3.26-19.93; P < 0.001) and OS (HR 11.08; 95% CI 2.16-56.95; P = 0.004). Plasma AR was an independent predictor of outcome on multivariable analyses in both cohorts.

CONCLUSION

Plasma AR status assessment using ddPCR identifies CRPC with worse outcome to enzalutamide or abiraterone. Prospective evaluation of treatment decisions based on plasma AR is now required.

CLINICAL TRIAL NUMBER

NCT02288936 (PREMIERE trial).