Joint Latent Class Research Articles

Patterns of intrinsic capacity trajectory and onset of activities of daily living disability among community-dwelling older adults.

Global population ageing has brought about new challenges for elderly care. Exploring intrinsic capacity (IC) over time, which is designed as a composite measure of an individual's physical and mental capabilities, is essential for promoting healthy ageing and preventing dependency, such as that emerging from disability in activities of daily living (ADL). We aimed to identify and examine the differences between classes of IC trajectory and onset of ADL disability. We conducted an observational study using data from three waves (2011-15) of the China Health and Retirement Longitudinal Study, comprising 2609 participants with 6034 observations. IC was measured by five domains, including locomotion, cognition, psychological, sensory capacities, and vitality. We used joint latent class modelling to identify distinct classes with similar patterns of IC trajectory and onset of ADL disability, as well as to explore the variation in IC trajectory and predict five-year risks of ADL disability considering the heterogeneity in the elderly population. The average baseline IC score was 7.15 (range: 0-15). We observed that IC scores slowly decreased with age, with 17.25% of participants developing ADL disability. We identified three classes of IC, which could be described as moderate health (class 1: n = 1634, 62.63%), at-risk (class 2: n = 716, 27.44%; had the highest risk of ADL disability), and optimal health (class 3: n = 259, 9.93%; had the lowest baseline risk of ADL disability). The probability of being in the moderate health class was decreased the most by emotional problems (odds ratio (OR) = 0.219; P < 0.001). Having a self-rated poor standard of living substantially reduced the chances of moderate (OR = 0.308; P = 0.001) and optimal health (OR = 0.110; P < 0.001). Observing IC trajectories and the onset of ADL disability can stratify the elderly into heterogeneous groups, as well as provide data for implementing person-centred care plans to reverse the trend and delay the adverse outcomes in clinical practice.

Trajectory of haemoglobin A1c and incidence of cardiovascular disease in patients with type 2 diabetes mellitus.

To evaluate the association between changes in haemoglobin A1c (HbA1c) and the concurrent incidence of cardiovascular disease (CVD) in type 2 diabetes mellitus (T2DM) patients. We conducted a retrospective cohort study among T2DM patients with HbA1c measurement after T2DM diagnosis between August 2009 and September 2010. The patients were classified into six subgroups based on baseline HbA1c (<7%; 7%-7.9%; ≥8%) and age (<65; ≥65 years), and then clustered into classes by HbA1c trajectory and CVD incidence over the 12-year follow-up period using joint latent class mixture models. We explored the HbA1c trajectories and CVD incidences in each latent class. Multinomial logistic regression was used to compare the baseline characteristics among different latent classes. A total of 128 843 T2DM patients were included with a median follow-up period of 11.7 years. Ten latent classes were identified in patients with baseline HbA1c ≥ 8% and age <65 years, while seven classes were identified in the other five groups. Among all the identified latent classes, patients with fluctuating HbA1c trajectories, characterized by alternating periods of increase and decrease, had higher CVD incidences. Male patients, and patients with higher baseline HbA1c and use of antidiabetic drugs were more likely to have a fluctuating HbA1c trajectory. More specifically, patients aged < 65 years with younger age or a smoking habit, and patients aged ≥ 65 years with a longer duration of T2DM were more likely to have a fluctuating HbA1c trajectory. We found that T2DM patients with fluctuating HbA1c trajectories could have a higher CVD risk. Different trajectory-associated characteristics in age subgroups highlight the need for individualized management of T2DM patients.

Longitudinal trajectories of frailty are associated with short-term mortality in older people: a joint latent class models analysis using 2 UK primary care databases

ObjectivesFrailty is a dynamic health state that changes over time. Our hypothesis was that there are identifiable subgroups of the older population that have specific patterns of deterioration. The objective of this study was to evaluate the application of joint latent class model in identifying trajectories of frailty progression over time and their group-specific risk of death in older people. Study Design and SettingThe primary care records of UK patients, aged over 65 as of January 1, 2010, included in the Clinical Practice Research Datalink: GOLD and AURUM databases, were analyzed and linked to mortality data. The electronic frailty index (eFI) scores were calculated at baseline and annually in subsequent years (2010-2013). Joint latent class model was used to divide the population into clusters with different trajectories and associated mortality hazard ratios. The model was built in GOLD and validated in AURUM. ResultsFive trajectory clusters were identified and characterized based on baseline and speed of progression: low–slow, low–moderate, low–rapid, high–slow, and high–rapid. The high–rapid cluster had the highest average starting eFI score; 7.9, while the low–rapid cluster had the steepest rate of eFI progression; 1.7. Taking the low–slow cluster as reference, low–rapid and high–rapid had the highest hazard ratios: 3.73 (95% CI 3.71, 3.76) and 3.63 (3.57-3.69), respectively. Good validation was found in the AURUM population. ConclusionOur research found that there are vulnerable subgroups of the older population who are currently frail or have rapid frailty progression. Such groups may be targeted for greater healthcare monitoring.

Identifying novel clinical phenotypes of acute respiratory distress syndrome using trajectories of daily fluid balance: a secondary analysis of randomized controlled trials

BackgroundPreviously identified phenotypes of acute respiratory distress syndrome (ARDS) could not reveal the dynamic change of phenotypes over time. We aimed to identify novel clinical phenotypes in ARDS using trajectories of fluid balance, to test whether phenotypes respond differently to different treatment, and to develop a simplified model for phenotype identification.MethodsFACTT (conservative vs liberal fluid management) trial was classified as a development cohort, joint latent class mixed models (JLCMMs) were employed to identify trajectories of fluid balance. Heterogeneity of treatment effect (HTE) for fluid management strategy across phenotypes was investigated. We also constructed a parsimonious probabilistic model using baseline data to predict the fluid trajectories in the development cohort. The trajectory groups and the probabilistic model were externally validated in EDEN (initial trophic vs full enteral feeding) trial.ResultsUsing JLCMM, we identified two trajectory groups in the development cohort: Class 1 (n = 758, 76.4% of the cohort) had an early positive fluid balance, but achieved negative fluid balance rapidly, and Class 2 (n = 234, 24.6% of the cohort) was characterized by persistent positive fluid balance. Compared to Class 1 patients, patients in Class 2 had significantly higher 60-day mortality (53.5% vs. 17.8%, p < 0.001), and fewer ventilator-free days (0 vs. 20, p < 0.001). A significant HTE between phenotypes and fluid management strategies was observed in the FACTT. An 8-variables model was derived for phenotype assignment.ConclusionsWe identified and validated two novel clinical trajectories for ARDS patients, with both prognostic and predictive enrichment. The trajectories of ARDS can be identified with simple classifier models.

20-year depressive symptoms, dementia, and structural neuropathology in older women.

Thecourse of depressive symptoms and dementia risk is unclear, as are potential structural neuropathological common causes. Utilizing joint latent class mixture models, we identified longitudinal trajectories of annually assessed depressive symptoms and dementia risk over 21 years in 957 older women (baseline age 72.7 years old) from the Women's Health Initiative Memory Study. In a subsample of 569 women who underwent structural magnetic resonance imaging, we examined whether estimates of cerebrovascular disease and Alzheimer's disease (AD)-related neurodegeneration were associated with identified trajectories. Five trajectories of depressive symptoms and dementia risk were identified. Compared to women with minimal symptoms, women who reported mild and stable and emerging depressive symptoms were at the highest risk of developing dementia and had more cerebrovascular disease and AD-related neurodegeneration. There are heterogeneous profiles of depressive symptoms and dementia risk. Common neuropathological factors may contribute to both depression and dementia. Highlights The progression of depressive symptoms and concurrent dementia risk is heterogeneous. Emerging depressive symptoms may be a prodromal symptom of dementia. Cerebrovascular disease and AD are potentially shared neuropathological factors.

Exploring transdiagnostic stress and trauma-related symptoms across the world: a latent class analysis

ABSTRACT Background: Although trauma exposure is universally prevalent, the ways in which individuals respond to potentially traumatic events vary. Between-country differences have been identified as affecting the development and manifestation of transdiagnostic psychological symptoms, but it remains unclear how stress and trauma-related transdiagnostic symptoms and risk patterns differ based on geographic region. Objective: To explore whether there are distinct classes of stress and trauma-related transdiagnostic symptoms and to determine predictors of class membership in a global sample. Method: Participants (N = 8675) from 115 different countries were recruited online between 2020–2022 and completed the Global Psychotrauma Screen, which assesses stress and trauma exposure, related symptoms, and risk factors. A latent class analysis (LCA) was used to identify classes of stress and trauma-related symptoms per world region (African States, Asia-Pacific States, Eastern European States, Latin American and Caribbean States, Western European and Other States, and North America) and the total sample. Likelihood of class membership was assessed based on demographics, characteristics of the potentially traumatic event, and potential risk factors across the world regions. Results: Similar class compositions were observed across regions. A joint latent class analysis identified three classes that differed by symptom severity (i.e. high, moderate, low). Multinomial logistic regression analyses revealed several factors that conferred greater risk for experiencing higher levels of symptoms, including geographic region, gender, and lack of social support, among others. Conclusions: Stress and trauma-related symptoms seem to be similarly transdiagnostic across the world, supporting the value of a transdiagnostic assessment.

Integrating joint latent class mixed models and Bayesian network for uncovering clinical subgroups of COVID-19 patients

When modelling the dynamics of biomarkers in biomedical studies, it is essential to identify homogeneous clusters of patients and analyse them from a precision medicine perspective. This need has emerged as crucial and urgent during the COVID-19 pandemic: early understanding of symptoms and patient heterogeneity has significant implications for prevention, early diagnosis, effective management, and treatment. Additionally, biomarker progression may be associated with clinically relevant time-toevent data. Therefore, statistical models are necessary to gain insight into complex disease mechanisms by properly accounting for unobservable heterogeneity in patients while jointly modelling longitudinal and time-to-event data. In this study, we leverage the key features of Latent Class modelling and Bayesian Network approaches and propose a unified framework to (a) uncover homogeneous subgroups of patients concerning their longitudinal and survival data and (b) describe patient subgroups within a multivariate framework.

Serial platelet count as a dynamic prediction marker of hospital mortality among septic patients.

Platelets play a critical role in hemostasis and inflammatory diseases. Low platelet count and activity have been reported to be associated with unfavorable prognosis. This study aims to explore the relationship between dynamics in platelet count and in-hospital morality among septic patients and to provide real-time updates on mortality risk to achieve dynamic prediction. We conducted a multi-cohort, retrospective, observational study that encompasses data on septic patients in the eICU Collaborative Research Database (eICU-CRD) and the Medical Information Mart for Intensive Care IV (MIMIC-IV) database. The joint latent class model (JLCM) was utilized to identify heterogenous platelet count trajectories over time among septic patients. We assessed the association between different trajectory patterns and 28-day in-hospital mortality using a piecewise Cox hazard model within each trajectory. We evaluated the performance of our dynamic prediction model through area under the receiver operating characteristic curve, concordance index (C-index), accuracy, sensitivity, and specificity calculated at predefined time points. Four subgroups of platelet count trajectories were identified that correspond to distinct in-hospital mortality risk. Including platelet count did not significantly enhance prediction accuracy at early stages (day 1 C-indexDynamic vs C-indexWeibull: 0.713 vs 0.714). However, our model showed superior performance to the static survival model over time (day 14 C-indexDynamic vs C-indexWeibull: 0.644 vs 0.617). For septic patients in an intensive care unit, the rapid decline in platelet counts is a critical prognostic factor, and serial platelet measures are associated with prognosis.

Comorbid Prolonged Grief, PTSD, and Depression Trajectories for Bereaved Family Surrogates

Family surrogates of patients who die in an intensive care unit (ICU) are at risk of cooccurring prolonged grief disorder (PGD), posttraumatic stress disorder (PTSD), and depressive disorder during bereavement, but symptom trajectories are often explored individually. To simultaneously examine and determine co-occurrence of PGD, PTSD, and depressive symptom trajectories. This cohort study was conducted in ICUs of 2 Taiwanese medical centers from January 2018 to March 2020, with follow-up through July 2022. Participants included surrogates responsible for decision-making who provided data 6 to 24 months after the death of their loved one. Data were analyzed from August to December 2022. PGD was assessed with the 11 grief symptom items of the PG-13; PTSD, the Impact of Event Scale-Revised; and depressive symptoms, the depression subscale of the Hospital Anxiety and Depression Scale at 6, 13, 18, and 24 months after the death. Latent growth mixture modeling was conducted to identify distinct trajectories, and joint latent class analysis was used to assess joint patterns of trajectories. A total of 303 participants were included, with most younger than 56 years (207 participants [68.3%]), female (177 participants [58.4%]), and married (228 participants [75.2%]), and their relationship with the patient was mostly spouse (88 participants [29.0%]) or adult child (166 participants [54.8%]). Three trajectories were identified each for PGD, PTSD, and depressive symptoms. A resilience trajectory was predominant across PGD (253 participants [83.5%]), PTSD (250 participants [82.5%]), and depressive (200 participants [66.0%]) symptoms. Second most common was a recovery trajectory identified for PGD (36 participants [11.9%]) and PTSD (41 participants [13.5%]) symptoms, while for depressive symptoms, a moderate trajectory (72 participants [23.8%]) signified persistent moderate distress. A chronic trajectory characterized by persistently high distress was identified for PGD (14 participants [4.6%]) and depressive (31 participants [10.2%]) symptoms, whereas a unique delayed-onset trajectory (12 participants [4.0%]) was identified for PTSD symptoms. Most family surrogates (228 participants [75.2%]) experienced cooccurring PGD, PTSD, and depressive symptom trajectories, but multiple patterns were discordant. Symptom trajectories cooccurred in joint patterns: resilient (247 participants [81.5%]), recovered (43 participants [14.1%]), and distressed (14 participants [4.5%]). These patterns were characterized by high conditional probabilities for the resilience (PGD, 0.999; PTSD, 0.999; depressive, 0.804), recovery (PGD, 0.854; PTSD, 0.890; depressive, 0.588), and chronic (PGD, 0.921; PTSD, 0.789; depressive, 0.980) symptom trajectories. In this cohort study, grief-related psychological symptoms evolved in complex ways during ICU bereavement, as characterized by heterogeneous trajectories. Some ICU bereaved surrogates experienced persistent elevated PGD, PTSD, and depressive symptoms individually or conjointly, underscoring the importance of early screening to identify this population at high risk of comorbid psychological distress trajectories.

Haemoglobin trajectories in chronic kidney disease and risk of major adverse cardiovascular events.

The trajectories of haemoglobin in patients with chronic kidney disease (CKD) have been poorly described. In such patients, we aimed to identify typical haemoglobin trajectory profiles and estimate their risks of major adverse cardiovascular events (MACE). We used 5-year longitudinal data from the CKD-REIN cohort patients with moderate to severe CKD enrolled from 40 nationally representative nephrology clinics in France. A joint latent class model was used to estimate, in different classes of haemoglobin trajectory, the competing risks of (i) MACE+defined as the first event among cardiovascular death, non-fatal myocardial infarction, stroke or hospitalization for acute heart failure, (ii) initiation of kidney replacement therapy (KRT) and (iii) non-cardiovascular death. During the follow-up, we gathered 33874 haemoglobin measurements from 3011 subjects (median, 10 per patient). We identified five distinct haemoglobin trajectory profiles. The predominant profile (n=1885, 62.6%) showed an overall stable trajectory and low risks of events. The four other profiles had nonlinear declining trajectories: early strong decline (n=257, 8.5%), late strong decline (n=75, 2.5%), early moderate decline (n=356, 11.8%) and late moderate decline (n=438, 14.6%). The four profiles had different risks of MACE, while the risks of KRT and non-cardiovascular death consistently increased from the haemoglobin decline. In this study, we observed that two-thirds of patients had a stable haemoglobin trajectory and low risks of adverse events. The other third had a nonlinear trajectory declining at different rates, with increased risks of events. Better attention should be paid to dynamic changes of haemoglobin in CKD.

Describing complex disease progression using joint latent class models for multivariate longitudinal markers and clinical endpoints.

Neurodegenerative diseases are characterized by numerous markers of progression and clinical endpoints. For instance, multiple system atrophy (MSA), a rare neurodegenerative synucleinopathy, is characterized by various combinations of progressive autonomic failure and motor dysfunction, and a very poor prognosis. Describing the progression of such complex and multi-dimensional diseases is particularly difficult. One has to simultaneously account for the assessment of multivariate markers over time, the occurrence of clinical endpoints, and a highly suspected heterogeneity between patients. Yet, such description is crucial for understanding the natural history of the disease, staging patients diagnosed with the disease, unravelling subphenotypes, and predicting the prognosis. Through the example of MSA progression, we show how a latent class approach modeling multiple repeated markers and clinical endpoints can help describe complex disease progression and identify subphenotypes for exploring new pathological hypotheses. The proposed joint latent class model includes class-specific multivariate mixed models to handle multivariate repeated biomarkers possibly summarized into latent dimensions and class-and-cause-specific proportional hazard models to handle time-to-event data. Maximum likelihood estimation procedure, validated through simulations is available in the lcmm R package. In the French MSA cohort comprising data of 598 patients during up to 13 years, five subphenotypes of MSA were identified that differ by the sequence and shape of biomarkers degradation, and the associated risk of death. In posterior analyses, the five subphenotypes were used to explore the association between clinical progression and external imaging and fluid biomarkers, while properly accounting for the uncertainty in the subphenotypes membership.

#3381 PROFILES OF HEMOGLOBIN TRAJECTORY IN CKD PATIENTS AND ASSOCIATED RISKS OF MAJOR ADVERSE CARDIOVASCULAR EVENTS: THE CKD-REIN COHORT STUDY

Abstract Background and Aims Anemia in chronic kidney disease (CKD) has been repeatedly associated with increased cardiovascular and all-cause morbidity and mortality. Most observational studies have focused on one single measurement of hemoglobin, but very few have investigated the trajectory of hemoglobin over time. Our aim was to identify, in moderate to severe CKD, typical profiles of hemoglobin trajectories, and to estimate their associated risk of major adverse cardiovascular event (MACE). Method We used data from the CKD-REIN cohort, which included 3033 patients with moderate to severe CKD from 40 nationally representative nephrology clinics in France between 2013 and 2016. A minimum of one hemoglobin measurement was necessary to be included. The primary endpoint was MACE defined as the first event among cardio-vascular death, myocardial infarction, stroke or hospitalization for acute heart failure. Secondary events included initiation of kidney replacement therapy (KRT) and death. A joint latent class mixed model was used to identify classes of hemoglobin trajectory and estimate the risk of each event. Once the model had identified classes and assigned participants in their class, we described the characteristics of patients in each class a posteriori. We also performed a posterior analysis to describe the course of estimated glomerular filtration rate (eGFR) in each class. Results A total of 3011 subjects were included in the analysis: 66% were men, median age at inclusion was 69 years. A total of 33874 hemoglobin measurements have been analyzed (median of 10 per patient). A total of 460 MACE, 522 KRT before MACE and 216 deaths before MACE or KRT, were recorded over a median follow-up of 4.3 years (Interquartile Range IQR 2.3-5.0). Five distinct classes were identified with a predominant one, Class-1 (‘constant’) including 63% (n = 1885) of the studied population with an overall stable trajectory surrounding a hemoglobin concentration of 13 g/dL (See figure below). For this class, risks of events were very low throughout the follow-up. Patients in this class had fewer cardiovascular risk factors, lower eGFR, and less often increased proteinuria at baseline than in other classes. Class-2 trajectory (‘late strong decline’, n = 75, 2.5%) had an abrupt decline at 2 years of follow-up; concomitantly risks of KRT and death before MACE emerged but the risk of MACE remained low over the follow-up. Patients classified in this class had more often tubulo-interstitial or unknown nephropathy than in other classes. Classes 3 ‘(late moderate decline, n = 438, 14.6%) and 4 (‘early moderate decline, n = 356, 11.8%) exhibited more moderate decline of hemoglobin level occurring on average at 3 and 1 year of follow-up, respectively. For these two classes, the risk of KRT before MACE increased from the start of these declines but their risks of MACE increased from the cohort entry. Class-5 (‘early strong decline’, n = 257, 8.5%) had a collapsing trajectory right from the entry into the cohort, and had a risk of MACE increasing rapidly in the very first year. Patients from Class-3 and Class-5, who had the highest risk of MACE, also had a worse cardiovascular profile at baseline. The average eGFR trajectories estimated in each of the 5 identified classes of hemoglobin showed similar patterns as those for hemoglobin. Conclusion In patients with CKD under nephrology care, most of patients have normal and stable hemoglobin values over time. In about one third of patients, 4 profiles of hemoglobin decline were observed, likely matching those of eGFR decline. These 4 classes with declining profiles had different levels of MACE risk and a strong increased risk of KRT after hemoglobin decline. This study suggests that more attention should be paid to dynamic changes of hemoglobin in the management of CKD.

Dynamic personalized prediction of the individual liver-related risk after sustained viral response in HCV patients.

Sustained viral response (SVR) significantly improves the prognosis in patients with hepatitis C virus (HCV) chronic infection but does not totally alleviate the risk of liver-related complications (LRC). We aimed to evaluate whether the dynamics of multiple measurements of simple parameters after SVR enable the development of a personalized prediction of prognosis in HCV patients. HCV mono-infected patients who experienced SVR in two prospective cohorts (ANRS CO12 CirVir cohort: derivation set; ANRS CO22 HEPATHER cohort: validation set) were included. The study outcome was LRC, a composite criterion including decompensation of cirrhosis and/or hepatocellular carcinoma. Joint latent class modelling accounting for both biomarker trajectory and event occurrence during follow-up was developed in the derivation set to compute individual dynamic predictions, with further evaluation in the validation set. In the derivation set (n=695; 50 LRC during the median 3.8 [1.6-7.5] years follow-up), FIB4 was identified as a biomarker associated with LRC occurrence after SVR. Joint modelling used sex and the dynamics of FIB4 and diabetes status to develop a personalized prediction of LRC. In the validation set (n=7064; 273 LRC during the median 3.6 [2.5-4.9] years follow-up), individual dynamic predictions from the model accurately stratified the risk of LRC. Time-dependent Brier Score showed good calibration that improved with the accumulation of visits, justifying our modelling approach considering both baseline and follow-up measurements. Dynamic modelling using repeated measurements of simple parameters predicts the individual residual risk of LRC and improves personalized medicine after SVR in HCV patients.

Circulating thrombospondin-2 level for identifying individuals with rapidly declining kidney function trajectory in type 2 diabetes: a prospective study of the Hong Kong West Diabetes Registry.

Thrombospondin-2 (TSP2) is a matricellular protein with tissue expression induced by hyperglycaemia. TSP2 has been implicated in non-diabetic renal injury in preclinical studies and high circulating levels were associated with worse kidney function in cross-sectional clinical studies. Therefore, we investigated the prospective associations of circulating TSP2 level with kidney function decline and the trajectories of estimated glomerular filtration rate (eGFR) in type 2 diabetes. Baseline serum TSP2 level was measured in 5471 patients with type 2 diabetes to evaluate its association with incident eGFR decline, defined as≥40% sustained eGFR decline, using multivariable Cox regression analysis. Among participants with relatively preserved kidney function (Baseline eGFR≥60ml/min/1.73m2), joint latent class modelling was employed to identify three different eGFR trajectories. Their associations with baseline serum TSP2 was evaluated using multinomial logistic regression analysis. The predictive performance of serum TSP2 level was examined using time-dependent c-statistics and calibration statistics. Over a median follow-up of 8.8 years, 1083 patients (19.8%) developed eGFR decline. Baseline serum TSP2 level was independently associated with incident eGFR decline (HR 1.21, 95%CI 1.07-1.37, P=0.002). With internal validation, incorporating serum TSP2 to a model of clinical risk factors including albuminuria led to significant improvement in c-statistics from 83.9 to 84.4 (P<0.001). Among patients with eGFR≥60ml/min/1.73m2, baseline serum TSP2 level was independently associated with a rapidly declining eGFR trajectory (HR 1.63, 95%CI 1.26-2.10, P<0.001). Serum TSP2 level was independently associated with incident eGFR decline, particularly a rapidly declining trajectory, in type 2 diabetes.

Trajectories of Change in Acute Dynamic Risk Ratings and Associated Risk for Recidivism in Paroled New Zealanders: A Joint Latent Class Modelling Approach

ObjectivesPrior studies indicate risk for recidivism declines with time spent in the community post-incarceration. The current study tested whether declines in risk scores occurred uniformly for all individuals in a community corrections sample or whether distinct groups could be identified on the basis of similar trajectories of change in acute risk and time to recidivism. We additionally tested whether accounting for group heterogeneity improved prospective prediction of recidivism.MethodsThis study used longitudinal, multiple-reassessment data gathered from 3,421 individuals supervised on parole in New Zealand (N = 92,104 assessments of theoretically dynamic risk factors conducted by community corrections supervision officers). We applied joint latent class modelling (JLCM) to model group trajectories of change in acute risk following re-entry while accounting for data missing due to recidivism (i.e., missing not at random). We compared accuracy of dynamic predictions based on the selected joint latent class model to an equivalent joint model with no latent class structure.ResultsWe identified four trajectory groups of acute dynamic risk. Groups were consistently estimated across a split sample. Trajectories differed in direction and degree of change but using the latent class structure did not improve discrimination when predicting recidivism.ConclusionsThere may be significant heterogeneity in how individuals’ assessed level of acute risk changes following re-entry, but determining risk for recidivism should not be based on probable group membership. JLCM revealed heterogeneity in early re-entry unlikely to be observed using traditional analytic approaches.

AFFECTIVE TRAJECTORIES: RISK OF DEMENTIA AND UNDERLYING STRUCTURAL BRAIN VARIABLES IN OLDER WOMEN

Abstract Understanding how trajectories of positive and negative affect relate to dementia risk and underlying structural brain variables is important for dementia prevention. We examined associations between annually assessed Positive and Negative Affect Scale subscales and dementia risk (2000-18) among cognitively-intact community-dwelling women (N=948; aged 72.9±3.7) from the Women’s Health Initiative Study of Cognitive Aging (years 2000-2010) and Magnetic Resonance Imaging Study (2005-2006). Joint latent class mixture models were constructed to identify latent classes of women with similar trajectories of affect and dementia risk over time. Multinomial and logistic regressions examined whether structural MRI measures predicted latent class membership (adjusted for sociodemographic, lifestyle, clinical characteristics, and intracranial volume). Two latent classes of positive affect (high stable:88% and decreasing:12%) and four classes of negative affect (Minimal stable:75%; high stable:4%; emerging:12%; moderate decreasing:9%) were identified. With the high stable trajectory as referent, women with decreasing positive affect were more likely to develop dementia (HR=4.33;p&amp;lt;.001). The odds of being classified as this high-dementia risk group were increased among women with more (per SD) global small vessel ischemic disease (SVID;OR=1.42;p&amp;lt;.001), deep white matter SVID (OR=1.93;p&amp;lt;.001), and smaller parahippocampal volumes (OR=1.41;p=.016). For negative affect, with minimal stable negative affect as referent, women with smaller hippocampal volumes were more likely to be classified as having moderate decreasing negative affect (OR=1.45;p=.024) while emerging negative affect was associated with higher dementia risk (HR=2.00;p=.014). These findings highlight the importance of changes in affect in later-life with dementia risk and potential underlying role of cerebrovascular disease and medial temporal lobe structures.

Joint models for dynamic prediction in localised prostate cancer: a literature review

BackgroundProstate cancer is a very prevalent disease in men. Patients are monitored regularly during and after treatment with repeated assessment of prostate-specific antigen (PSA) levels. Prognosis of localised prostate cancer is generally good after treatment, and the risk of having a recurrence is usually estimated based on factors measured at diagnosis. Incorporating PSA measurements over time in a dynamic prediction joint model enables updates of patients’ risk as new information becomes available. We review joint model strategies that have been applied to model time-dependent PSA trajectories to predict time-to-event outcomes in localised prostate cancer.MethodsWe identify articles that developed joint models for prediction of localised prostate cancer recurrence over the last two decades. We report, compare, and summarise the methodological approaches and applications that use joint modelling accounting for two processes: the longitudinal model (PSA), and the time-to-event process (clinical failure). The methods explored differ in how they specify the association between these two processes.ResultsTwelve relevant articles were identified. A range of methodological frameworks were found, and we describe in detail shared-parameter joint models (9 of 12, 75%) and joint latent class models (3 of 12, 25%). Within each framework, these articles presented model development, estimation of dynamic predictions and model validations.ConclusionsEach framework has its unique principles with corresponding advantages and differing interpretations. Regardless of the framework used, dynamic prediction models enable real-time prediction of individual patient prognosis. They utilise all available longitudinal information, in addition to baseline prognostic risk factors, and are superior to traditional baseline-only prediction models.

Dynamics of circulating calprotectin accurately predict the outcome of moderate COVID-19 patients.

SummaryBackgroundSevere COVID-19 is associated with a high circulating level of calprotectin, the S100A8/S100A9 alarmin heterodimer. Baseline calprotectin amount measured in peripheral blood at diagnosis correlates with disease severity. The optimal use of this biomarker along COVID-19 course remains to be delineated.MethodsWe focused on patients with a WHO-defined moderate COVID-19 requiring hospitalization in a medical ward. We collected plasma and serum from three independent cohorts (N = 626 patients) and measured calprotectin amount at admission. We performed longitudinal measures of calprotectin in 457 of these patients (1461 samples) and used a joint latent class mixture model in which classes were defined by age, body mass index and comorbidities to identify calprotectin trajectories predicting the risk of transfer into an intensive care unit or death.FindingsAfter adjustment for age, sex, body mass index and comorbidities, the predictive value of baseline calprotectin in patients with moderate COVID19 could be refined by serial monitoring of the biomarker. We discriminated three calprotectin trajectories associated with low, moderate, and high risk of poor outcome, and we designed an algorithm available as online software (https://calpla.gustaveroussy.fr:8443/) to monitor the probability of a poor outcome in individual patients with moderate COVID-19.InterpretationThese results emphasize the clinical interest of serial monitoring of calprotectin amount in the peripheral blood to anticipate the risk of poor outcomes in patients with moderate COVID-19 hospitalized in a standard care unit.FundingThe study received support (research grants) from ThermoFisher immunodiagnostics (France) and Gustave Roussy Foundation.

Comorbidities in osteoarthritis (ComOA): a combined cross-sectional, case–control and cohort study using large electronic health records in four European countries

IntroductionOsteoarthritis (OA) is one of the leading chronic conditions in the older population. People with OA are more likely to have one or more other chronic conditions than those without....

Using real-world data to dynamically predict flares during tapering of biological DMARDs in rheumatoid arthritis: development, validation, and potential impact of prediction-aided decisions

BackgroundBiological disease-modifying antirheumatic drugs (bDMARDs) are effective in the treatment of rheumatoid arthritis. However, as bDMARDs may also lead to adverse events and are expensive, tapering them is of great clinical interest. Tapering according to disease activity-guided dose optimization (DGDO) does not seem to affect long term remission rates, but flares are frequent during this process. Our objective was to develop a model for the prediction of flares during bDMARD tapering using data from routine care and to evaluate its potential clinical impact.MethodsWe used a joint latent class model to repeatedly predict the probability of a flare occurring within the next 3 months. The model was developed using longitudinal data on disease activity (DAS28) and other routine care data from two clinics. Predictive accuracy was assessed in cross-validation and external validation was performed with data from the DRESS (Dose REduction Strategy of Subcutaneous tumor necrosis factor inhibitors) trial. Additionally, we simulated the reduction in number of flares and bDMARD dose when implementing the model as a decision aid during bDMARD tapering in the DRESS trial.ResultsData from 279 bDMARD courses were used for model development. The final model included two latent DAS28-trajectories, bDMARD type and dose, disease duration, and seropositivity. The area under the curve of the final model was 0.76 (0.69–0.83) in cross-validation and 0.68 (0.62–0.73) in external validation. In simulation of prediction-aided decisions, the mean number of flares over 18 months decreased from 1.21 (0.99–1.43) to 0.75 (0.54–0.96). The reduction in he bDMARD dose was mostly maintained, increasing from 54 to 64% of full dose.ConclusionsWe developed a dynamic flare prediction model, exclusively based on data typically available in routine care. Our results show that using this model to aid decisions during bDMARD tapering may significantly reduce the number of flares while maintaining most of the bDMARD dose reduction.Trial registrationThe clinical impact of the prediction model is currently under investigation in the PATIO randomized controlled trial (Dutch Trial Register number NL9798).