Implementing cohort research in Portugal. The past, current and future challenges

Implementing cohort research in Portugal. The past, current and future challenges

Abstract: .

The Lisbon Cohort of Men who have Sex with Men: main results and challenges

The Lisbon Cohort of Men who have Sex with Men: main results and challenges

Abstract: The Lisbon Cohort of Men Who Have Sex with Men (MSM) is an ongoing observational study designed as an open, prospective, and non-interval cohort. It is a major research structure for life-sciences and social-sciences aiming at quantifying the frequency of the disease by estimating the incidence of HIV infection among MSM, monitoring trends in primary and secondary prevention, and identifying strategies to improve the provision of HIV testing and linkage to care. It is a joint project of GAT and the ISPUP, that takes CheckpointLX - a community-based voluntary HIV and other sexually transmitted infections (STI) counselling and testing center devoted to MSM - as the recruitment base and following its privileged anchorage in the community. The cohort started recruiting the day the center opened its activity in April 2011.
All men, who present for testing and have a negative test at first visit, aged 15 or older (18 until 2021) who report having had sex with other men are invited to participate and follow-up visits scheduled according to their convenience and needs. At each evaluation a structured questionnaire is administered and an HIV rapid testing are performed by peer community health workers at CheckpointLX. Men are also offered testing for viral Hepatitis B and C, and Syphilis according to predefined criteria. In case of a reactive test a referral for an HIV/STI clinic is offered.
The establishment of the Lisbon Cohort of MSM has proved to be an important decision for a comprehensive approach into the HIV epidemic among this key population in Portugal. The first estimates of HIV incidence, covering the period from April 2011 to February 2014, of 2.80/100 person-years (95% CI: 1.89-4.14) were higher compared to other European settings and were likely driven by short-term contextual and behavioral changes during follow-up such as partner disclosure of HIV status, newly adopted condomless anal intercourse (CAI) with a steady partner, and being newly diagnosed with syphilis during follow-up. A recent work also showed that participants in the cohort were likely to transition between being eligible or ineligible for HIV pre-exposure prophylaxis (PrEP) throughout the follow-up and that once an individual met any of the eligibility criteria for PrEP, he was at a much higher risk of seroconversion. These results showed that a longitudinal approach to HIV risk factors and dynamics can identify critical periods for preventive interventions and unveil the changing nature of human behavior.
The Lisbon Cohort of MSM has a unique design and strengths that offer excellent conditions to pursue our objectives. It has engaged over 4000 men in active follow-up (more than two visits) corresponding to over 16,000 persons-years of observation. It has no fixed times for follow-up evaluations. This poses enormous challenges in data handling and analysis, nevertheless, it offers the rare opportunity to measure patterns in risk-taking behaviors, HIV prevention, and testing uptake as they occur in participants' lives.

Improvement of COVID-19 symptoms: a survival analysis study from a Portuguese cohort

Improvement of COVID-19 symptoms: a survival analysis study from a Portuguese cohort

Abstract: COVID-19 is caused by the SARS-COV-2 virus and affects individuals with varying severities. It has a high capacity for rapid spread and presents numerous genetic mutations, resulting in the emergence of several variants. Although many studies have investigated risk factors for COVID-19-related mortality, little is known about the duration of symptom resolution. We conducted a prospective study using data from 3481 patients diagnosed with COVID-19 at the Centro Hospitalar Universitário de São João (CHUSJ) between March 2020 and January 2021. We used survival and descriptive analysis techniques on a cohort basis. The estimated survival curves were used to compare the improvement of COVID-19 symptoms for categorical predictors, and formal hypothesis tests were used. Simple and multiple regression models were used to estimate the effect of potential predictors on the improvement of COVID-19 symptoms.

EPIPorto - A population-based cohort

EPIPorto - A population-based cohort

Abstract: EPIPorto is a population-based cohort representative of the adult population residing in Porto. Participants (n=2485) were recruited in 1999 and have been assessed periodically: 1999-2003 (baseline), 2005-2008, 2013-2015, 2017-2018, 2020, and 2021-2022. It allows for a social, demographic, behavioral, and clinical characterization of the participants over time. The last two assessments were conducted following the COVID-19 pandemic and included, exceptionally, the cohabitants of the EPIPorto participants, aiming to characterize the SARS-CoV-2 infection in the city and households. In the 2020 evaluation, 869 participants (59.5% of the eligible) were evaluated together with 671 cohabitants. Thirteen participants/cohabitants (0.8%) reported a previous diagnosis, and 50 (3.2%) had serological evidence of infection. In the last evaluation (2021- 2022), 836 (63.6% of the eligible) participants were evaluated together with 449 cohabitants. Of those, 348 (27.1%) reported a previous diagnosis, and 315 (24.5%) had serological evidence of infection - with vaccination rollout, humoral immunity might be misattributed to vaccination only, despite being likely induced by both infection and vaccination in some cases. Around one-third of those who reported a previous diagnosis were identified with the post COVID-19 condition.

On the special nature of survival data

On the special nature of survival data

Abstract: The estimation of a survival function is most of the times a non-trivial issue due to the special nature of the sampling information. This also applies to the fitting of regression models like, for instance, the proportional hazards model or the accelerated failure time model. Survival data typically suffer from random censoring and/or truncation, as recognized in most textbooks on the topic. In this talk I will revisit these issues and discuss the difficulties that appear when handling censored and truncated survival data. Special attention will be paid to data recruitment procedures that are often used in clinical and epidemiological studies and that induce particular censoring and truncation patterns. This includes, but is not limited to, cross-sectional sampling of prevalent cases, studies with delayed entries, periodical follow-ups with scheduled visits, or interval sampling. In all these scenarios a sampling bias is expected, and proper corrections of the ordinary estimation procedures are needed. Illustrative examples, simulation results and real data applications will be given.

A joint model for multiple longitudinal biomarkers and multiple event times using patient registry data

A joint model for multiple longitudinal biomarkers and multiple event times using patient registry data

Abstract: Joint models have been widely used in clinical and public health research to study the relationships between longitudinal biomarkers and times-to-events of interest. The basic framework for a single event time and a continuous longitudinal biomarker has been extended to accommodate multivariate longitudinal and multiple event data. These advancements have improved statistical inference by better capturing the complex underlying disease dynamics.
In this work, motivated by a clinical study in cystic fibrosis, we propose a Bayesian shared-parameter joint model that accommodates multiple longitudinal markers, a recurrent event process, and multiple terminal events (competing risks). The model is available in the R package JMbayes2. The longitudinal outcomes can follow various distributions: beta, beta-binomial, binomial, censored normal, gamma, negative binomial, normal, Poisson, Student's t, and unit-Lindley distributions. Regarding the recurrent event process, our model accommodates discontinuous risk intervals and the gap and calendar timescales. Our implementation provides great flexibility for modeling the association structure of the longitudinal markers within the relative-risk models, supporting forms such as underlying value, slope, standardized cumulative effect, and various combinations of these, enabling researchers to capture complex relationships in their analyses. .