Reading Response Discussions

Defining the time dimension is a fundamental challenge in longitudinal data analysis.  The most common choice is age or time since study enrollment, however, for many questions other time dimensions are relevant, for example grade in school, or time before or after stroke incidence, or time since release from prison.  Identify an article in the applied literature using a longitudinal analysis based on age as the time dimension, time since study enrollment as the time dimension, and one other possible time dimension (ie, not age or time since study enrollment).  For each study, briefly describe the research question, the study sample, the longitudinal design, and the analysis approach. Please post links to the studies.

Time Dimension: Age; Time since diagnosis

Article: Helzner EP, Scarmeas N, Cosentino S, Tang MX, Schupf N, Stern Y. Survival in Alzheimer disease: a multiethnic, population-based study of incident cases. Neurology. 2008 Nov 4;71(19):1489-95. doi: 10.1212/01.wnl.0000334278.11022.42.

Research Question: What factors are associated with survival in Alzheimer Disease (AD)?

Study Sample: Washington Heights Inwood Columbia Aging Project, a multi-ethnic cohort of people living in Northern Manhattan. This study included 323 people from this cohort diagnosed with incident AD during study follow-up.  

Study Design/Approach: Authors aimed to assess the effect of patients’ characteristics at time of diagnosis on survival. These characteristics included sex, race, education, comorbidities (heart disease, stroke, hypertension, diabetes). Cox proportional hazards models were employed to estimate effects of baseline characteristics on survival following AD diagnosis. Log-rank tests were used to compared cumulative survival between groups. Time scale was age, calculated from date of birth through date of death or last follow-up.

In an attempt to make their results comparable to prior work, the authors also compared post-diagnostic survival, using time since diagnosis as the time scale.  Again, these patients were enrolled as healthy individuals and incident AD diagnosis occurred while in the study.  As above, only patients with AD diagnosis were included in this analysis. Note: patients were assessed for AD every 18 months, so date of diagnosis was interval censored. In this analysis, they calculated time-to-event as the time of documented AD diagnosis to date of death or last follow-up.  

 Time Dimension: Time since study enrollment

 Article: Wilson KM, Kasperzyk JL, Rider JR, Kenfield S, van Dam RM, Stampfer MJ, Giovannucci E, Mucci LA. Coffee consumption and prostate cancer risk and progression in the health professionals follow-up study. JNCI. 2011 June 8;103(11):876-84

 Research Question: What is the relationship between coffee consumption and risk of incident (overall and aggressive) prostate cancer?

 Study Sample: Health Professionals Follow-Up Study (HPFS). This is a prospective cohort of over 51,000 male health professionals between the ages of 40-75 at time of study enrollment in 1986. Follow-up questionnaires on health outcomes are completed every two years and diet assessments occur every 4 years. This study included 47,911 men from HPFS without baseline cancer diagnosis and who completed a baseline food frequency question (FFQ).

 Study Design/Approach: Authors used data from FFQ to calculate total coffee consumption (regular and decaf). The authors performed various sensitivity analyses to assess for reverse causation (e.g., lagging the exposure). Self-reported prostate cancer diagnosis was confirmed by medical record review; deaths were identified via the National Death Index. Multivariable Cox proportional hazards regression models were adjusted for race, height, body mass index, physical activity, smoking status, diabetes status, family history of prostate cancer, PSA testing, and various dietary factors. Time since enrollment was used as the time scale and thus time-to-event was calculated as the date of baseline questionnaire to date of prostate cancer diagnosis, death, or end of follow-up period.

 

 

Age as the time dimension

Weyde, Kjell Vegard & Krog, Norun & Oftedal, Bente & Magnus, Per & White, Richard & Stansfeld, Stephen & Øverland, Simon & Aasvang, Gunn Marit. A Longitudinal Study of Road Traffic Noise and Body Mass Index Trajectories from Birth to 8 Years. Epidemiology. 2018 Sep;29(5):729-738

Research question: Are noise and adiposity associated? More specifically, (1) is road traffic noise during pregnancy associated with different BMI trajectories from birth to 8 years of age? And (2) is early childhood road traffic noise associated with different BMI trajectories from 18 months to 8 years of age?

Study sample: The Norwegian Mother and Child Cohort Study, MoBa. This is a pregnancy cohort including 95,200 mothers and 114,500 children born between 2000 and 2009 in Oslo. The authors considered two (overlapping) study samples: the pregnancy sample (6,963 children) and the childhood sample (6,403 children).

Longitudinal design: Children were followed from pregnancy to 8 years of age. Parents filled out questionnaires at 6 and 18 months and 3, 5, 7, and 8 years to report BMI. Noise was measured at the location of the household.

Analysis approach: The authors specified linear mixed models (with an outcome of BMI for individual i at time j) with random subject intercepts, random slopes for age, and an unstructured covariance matrix. The authors included an interaction term between age (modeled non-linearly using splines) and noise to evaluate the effect of noise on BMI trajectories.

 

Time since study enrollment as the time dimension

Ickovics JR, Hamburger ME, Vlahov D, et al. Mortality, CD4 Cell Count Decline, and Depressive Symptoms Among HIV-Seropositive Women: Longitudinal Analysis From the HIV Epidemiology Research Study. JAMA. 2001;285(11):1466–1474. doi:10.1001/jama.285.11.1466

Research question: Are depressive symptoms associated with mortality and CD4 count among women living with HIV?

Study sample: The HIV Epidemiologic Research Study. This is a large cohort of women living with HIV followed for 7 years selected from 4 study sites (Baltimore, Providence, Bronx, and Detroit). At enrollment women were 16-55 years of age and tested positive for HIV (although they excluded women in late stages of AIDS-related illness).

Longitudinal design: Women were interviewed and underwent physical examinations every 6 months. Loss to follow-up was relatively low (11.9%). The primary outcome was HIV-related mortality measured through the National Death Index. The primary exposure was depression as measured through the CES-D, categorized to chronic depression, intermittent depression, and limited/no depressive symptoms.

Analysis approach: To assess the relationship between depression and mortality, the authors used a Cox proportional hazards approach with time since enrollment as the time scale, adjusting for clinical features, substance abuse, and sociodemographic characteristics. To assess the relationship between depression and CD4 count, the authors specified a "hierarchical linear" model with an interaction term between depressive symptoms and time. While I am assuming this is a mixed model, the authors do not mention random effects.

 

Other time dimension

Assefa T, Haile Mariam D, Mekonnen W, Derbew M. Survival analysis to measure turnover of the medical education workforce in Ethiopia. Hum Resour Health. 2017;15(1):23. Published 2017 Mar 14. doi:10.1186/s12960-017-0197-0

Research question: What is the magnitude of physician turnover in medical schools in Ethiopia, and what factors predict physician turnover?

Study sample: The authors used physician workforce data form seven government run medical schools across Ethiopia, obtained from human resources. The dataset represented approximately six years of workforce information, with 1258 faculty physicians included in the sample.

Longitudinal design: Data were continually collected from the seven medical schools, including inflow data (when new physicians were hired as faculty) and outflow data (when the physician left their position). The authors had information on reasons why the physicians left (i.e., transferred, retired, death, etc.) in addition to sociodemographic characteristics on the physicians. Time zero for each subject was defined as date of employment.

Analysis approach: For their descriptive objective (measuring the magnitude of physician turnover), the authors used Kaplan-Meier survival curves to describe the length of stay of physicians in medical school, stratified by academic rank. To determine what factors predict turnover, the authors used Cox proportional hazards models, including gender, academic rank, age category, and a categorical variable representing the school as covariates.

Study 1

Reference: Huang C, Singh K, Handa S, Halpern C, Pettifor A, Thirumurthy H. Investments in children's health and the Kenyan cash transfer for orphans and vulnerable children: evidence from an unconditional cash transfer scheme. Health policy and planning 2017; 32(7): 943-55. PMC5886102

https://academic.oup.com/heapol/article/32/7/943/3739868

Research question: Does the Kenya Cash Transfer for Orphans and Vulnerable Children program, targeting HIV/AIDS-orphans, affect the incidence of upper respiratory illness in children?

Study sample: Researchers evaluated children aged 0-7 years and under-5 years of age in multiple locations throughout Kenya. The sample frame was from a list of all eligible households (N=1,542) provided by Government of Kenya’s Department of Children’s Services and the control sample frame, eligible households with delayed program entry, from households (N=755) within randomly sampled census tracts. Baseline data were collected from March-August, 2007 and follow-up occurred from March-July, 2009.

Longitudinal design: This is a cluster-randomized longitudinal study where the time dimension is time since study enrollment. Two health surveys were administered in the years 2007 and 2009.

Analysis approach: A three-level generalized linear latent and mixed model (GLLAMM) was used to estimate the average treatment effect of the program. This type of model allowed for nesting of hierarchical data as levels were thought to impact outcomes. The GLLAMM estimated the effect on an individual i living in household j in k location and time t’s likelihood of experiencing the outcome (incidence of malaria or pneumonia, or whether health-seeking occurred), while considering the correlation that existd between individuals in the same household and households in the same area, clustered at the location level.

 

Study 2

Reference: Holmes MJ, Robertson FC, Little F, Randall SR, Cotton MF, van der Kouwe AJW, et al. (2017) Longitudinal increases of brain metabolite levels in 5-10 year old children. PLoS ONE 12(7): e0180973.

https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0180973

Research question: Are there regional (mid-frontal gray, basal ganglia, and white matter), age-related changes in brain metabolite levels in HIV-negative, South African children?

Study sample: The study sample was from a subset of controls from a longitudinal neuroimaging study investigating the central nervous system effect of HIV throughout childhood. HIV-negative children aged 5-9 years, with a birth weight > 2000g, and no history or diagnosis of central nervous system problems or dysmorphic syndromes were included in this secondary analysis.

Longitudinal design: This is a secondary analysis of a prospective cohort study. Children received an MRI every two years, at age 5, 7 and 9 years, depending on timing of enrollment. The time dimension was age of the patient.

Analysis approach: The authors used a linear mixed effects regression model to estimate the association between age and regional brain metabolite levels and to account for repeated measures in some children. Additional covariates were included (sex, ethnicity, and HIV-exposure) in the model, as were standard errors of the brain metabolites to adjust for differences in measurement accuracy, and the gray + white matter tissue percentage of the total tissue to account for differences in tissue composition. To adjust for multiple comparisons, the authors relied on the false discovery rate method.

 

Study 3

Reference: Guesh G, Degu G, Abay M, Beyene B, Brhane E, Brhane K. Survival status and predictors of mortality among children with severe acute malnutrition admitted to general hospitals of Tigray, North Ethiopia: a retrospective cohort study. BMC research notes 2018; 11(1): 832. PMC6257969

https://bmcresnotes.biomedcentral.com/articles/10.1186/s13104-018-3937-x

Research question: What are predictors of mortality and the survival status in Ethiopian children with severe acute malnutrition admitted to malnutrition stabilization centers in general hospitals in the Tigray region?

Study sample: The medical records of 569 children were sampled from a source population of under age five children with severe acute malnutrition admitted to malnutrition stabilization centers in general hospitals in Tigray from January 01, 2013 to December 30, 2015. The study sample was allocated proportionally for randomly selected hospitals and a simple random sampling technique was used to select individual records from the entire medical record.

Longitudinal design: This was a 24-month, retrospective, longitudinal study where the time dimension was time since admission up to 45 days maximum.

Analysis approach: The authors constructed a life table to estimate the probabilities of death over time and fit a Kaplan–Meier survival curve with a log-rank test to test for mortality differences among categorical variables. The authors used a bi-variable Cox regression to determine the effect of each independent variable on the outcome; those with a p-value<0.25 were included in the multivariable Cox regression model to determine predictors of mortality. The proportionality of the hazard assumption was tested with a global test on Schoenfeld residuals, and the authors summarized associations with an adjusted hazard ratio.

Time Dimension: Time since study enrollment
Chockalingam, L., Pence, B., Frangakis, C. E., Ha, T. V., Latkin, C. A., Sripaipan, T., ... & Go, V. F. (2019). The relationship between health-related variables and increases in smoking among recently diagnosed HIV+ people who inject drugs in Vietnam. Addictive behaviors, 95, 118-124.

Research Question: What is the relationship between recent HIV diagnosis and increase in smoking and health-related behaviors among people who inject drugs (PWID) in Vietnam?
Study Sample: 323 HIV infected PWID who participated in a four-arm randomized control trial from 2009 to 2013 in the Thai Nguyen province in Vietnam. The trial evaluated the effect of community and individual level interventions on injection and sexual risk behavior. Inclusion criteria included: confirmed HIV diagnosis, ability to bring an injected network partner for screening, 18+, male, had sex in past 6 months, injected drugs in past 6 months, and planed to live in Thai Nguyen for the next 2 years.
Longitudinal Design: Participants completed a 1-hour face-to-face interview using a structured questionnaire at baseline, and at month 6, 12, 18, and 24.
Analysis Approach: Descriptive statistics were used to describe baseline characteristics and smoking trends over time. The authors used generalized estimating equations for repeated measures to estimate the bivariate and multivariable associations between participant characteristics and increases in smoking, with binomial distribution and logit link function. All variables were initially included in the multivariable model, and were dropped if dropping them improved goodness of fit (evaluated with Quasi Akaike’s information criterion – QIC).

Time Dimension: Age
Hoogendijk, E. O., Rockwood, K., Theou, O., Armstrong, J. J., Onwuteaka-Philipsen, B. D., Deeg, D. J., & Huisman, M. (2018). Tracking changes in frailty throughout later life: results from a 17-year longitudinal study in the Netherlands. Age and ageing, 47(5), 727-733.

Research Question: The purpose of the study was to investigate changes in the degree of frailty during later life, and the extent to which changes are determined by socio-demographic characteristics.
Study Sample: Investigators used data from the Longitudinal Aging Study Amsterdam (LASA), which is a nationally representative, ongoing study on physical, emotional, cognitive and social functioning of older adults in the Netherlands. Data from the second LASA measurement wave (1995-1996) were used as the baseline observation. Data from six consecutive measurement waves over a period of 17 years were used. The sample consisted of 1659 respondents aged 65 years and older. 399 respondents who were missing follow up data were older, lower educated, more often male, more often without a partner, and had a higher degree of frailty.
Longitudinal Design: The investigators used data from six waves of the Longitudinal Aging Study Amsterdam (LASA) to examine frailty trajectories over 17 years.
Analysis Approach: Generalized estimating equations were used to study longitudinal frailty trajectories over a period of 17 years. The GEE analysis was executed with a 5-dependent correlation matrix. Because the Frailty Index score skewed slightly to the right, the natural log of the FI score was used in the GEE analysis; estimates were back-transformed for descriptive purposes. Three models were tested: 1) changes over time in FI score adjusted for age at baseline and sex; 2) educational level and partner status were additionally included as covariates; 3) interaction effects between socio-demographic factors and time were included in the third model. Non-linear change was tested by adding a quadratic term for time, but it was excluded from the final models as it was not statistically significant.

Time Dimension: Time since polytrauma event
da Costa, L. G. V., Carmona, M. J. C., Malbouisson, L. M., Rizoli, S., Rocha-Filho, J. A., Cardoso, R. G., & Auler-Junior, J. O. C. (2017). Independent early predictors of mortality in polytrauma patients: a prospective, observational, longitudinal study. Clinics, 72(8), 461-468.

Research Question: This study identifies early predictors of mortality in severely injured polytrauma patients across all stages of care, from the earliest stage of care in the pre-hospital setting to admission to the ICU and hospital discharge. The demographic profile of the study population was also ascertained.
Study Sample: Screening included identification of general trauma patients (>18 years old) submitted to high-energy trauma (severe bleeding, TBI, significant damages resulting from high-velocity car crashes, falls, gunshots, penetrating torso/abdominal injuries, pedestrian car accidents and traumatic limb amputations) who were attended to and screened by rescue system medical teams and taken to Hospital das Clínicas - University of São Paulo, School of Medicine, Teaching Hospital for treatment. Patients with an injury severity score (ISS) <16 were excluded. Patients were also excluded if they did not provide written informed consent, in situations where data collection could compromise victim care, technical problems during data collection, insufficient blood samples or data, or if they died prior to arrival. Missing data were treated as missing at random (MAR).
Longitudinal Design: For all patients, data were recorded 1) at the trauma scene, 2) in the ER, 3) 3 hours after hospital admission, and 4) 24 hours after hospitalization. Data were collected during the first 24 hours of treatment, and patients were clinically followed up for 30 days.
Analysis Approach: Sample power analyses were conducted and 200 patients was conservatively defined, with a margin of error included to account for the possibility of death. Data analysis was divided into three interconnecting parts: 1) utilized descriptive data analysis and tests of association between independent variables and death; 2) addressed the profiles of time-dependent measures and their relationships with death through analysis of nonparametric variance for repeated measures; 3) evaluated results of all previous analysis, and constructed a generalized estimating equation.

Age as the time dimension

Renal injury in children with a solitary functioning kidney—the KIMONO study

https://academic-oup-com.ucsf.idm.oclc.org/ndt/article/26/5/1533/1893727

The research question of this study was whether children with a solitary functioning kidney have an increased risk of chronic kidney disease, hypertension and albuminuria.

The study sample (n=206) included all patients with a single functioning kidney diagnosed under the age of 18 years and followed at the Pediatric Renal Centre of the VU University Medical Centre. Diagnosis was confirmed by unilateral absence of (functional) renal tissue on ultrasound and/or on renal scintigraphy. Patients were divided into two groups – primary and secondary single functioning kidneys.

The authors collected data by chart review including blood pressure, age, gender and height from clinical visits, urinary albumin, estimated glomerular filtration rate and renal length measured by renal ultrasound. The authors analysed the longitudinal relation of blood pressure course over childhood, albuminuria and estimated glomerular filtration rate by using separate generalized estimated equation (GEE) analyses. All available data was utilized, taking into account that each patient had multiple measurements and irregular time intervals. An exchangeable correlation structure was used, which means that correlations between subsequent measurements are assumed to be the same, irrespective of the time between measurements. Age was used as the independent variable.

Time since study enrollment as the time dimension

Progression of kidney dysfunction in the community-dwelling elderly

https://www.sciencedirect.com/science/article/pii/S0085253815514336

The research aim was to determine the progression of kidney dysfunction among a community-based cohort of elderly subjects.

The study sample included 10,184 subjects 66 years of age or older, who had one or more outpatient serum creatinine measurements during each of two time periods: July 1 to December 31, 2001 and July 1 to December 31, 2003. To avoid episodes of acute renal failure laboratory measurements associated with a hospital admission were excluded. Subjects who were already receiving dialysis at study entry were excluded. Estimated glomerular filtration rate (eGFR) was calculated using the MDRD equation.

Using the unique provincial health care number for each subject, laboratory data from study subjects was linked to provincial administrative data to obtain information on prescription drug use in the year prior to their index eGFR. Prescription drug data were used to determine the presence of diabetes mellitus, and to derive a measure of comorbidity based on the Chronic Disease Score.

For the primary assessment of the progression of kidney function, the authors determined the rate of decline in eGFR in ml/min/1.73 m² per year, using a mixed effects model with random intercepts and random slopes. This model was used to estimate rate of change in eGFR over time, taking into account the varying number and spacing of measurements of eGFR, as well as the variable follow-up for each subject. Covariates in the model included age, gender, diabetes mellitus and comorbidity score. Age was entered as a continuous variable and the comorbidity score was divided into quartiles. Results were presented as the rate of change in eGFR per year.

Other time dimension

Urine Neutrophil Gelatinase-Associated Lipocalin (uNGAL) as a Marker for Acute Kidney Injury in Kidney Surgery Patients

https://www-ncbi-nlm-nih-gov.ucsf.idm.oclc.org/pmc/articles/PMC4167359/

The research aim was to determine if urine neutrophil gelatinase-associated lipocalin (uNGAL) is a marker for acute kidney injury (AKI) in patients undergoing partial or radical nephrectomy.

Between April 2010 and April 2012, 220 consecutive patients were enrolled in the prospective study. 162 patients (partial nephrectomy: 88, radical nephrectomy: 32; thoracic surgery: 42) had adequate specimens to be included in the final analysis. The authors prospectively collected and analyzed urine and serum of partial nephrectomy, radical nephrectomy, and thoracic surgery patients between April 2010 and April 2012. Urine was collected preoperatively and at multiple time points postoperatively. The urine specimens were analyzed for NGAL and creatinine levels using a commercially available ELISA assay.

The authors evaluated the relationship between AKI and postoperative uNGAL. The relationship was assessed using a linear GEE model with an autoregressive correlation structure. Since multiple uNGAL measurements from the same patient are not independent, a GEE model was used to incorporate the correlation between different measures from the same patient. Time from surgery (hours) was used as the independent variable.

Assignment: 

Defining the time dimension is a fundamental challenge in longitudinal data analysis.  The most common choice is age or time since study enrollment, however, for many questions other time dimensions are relevant, for example grade in school, or time before or after stroke incidence, or time since release from prison.  Identify an article in the applied literature using a longitudinal analysis based on age as the time dimension, time since study enrollment as the time dimension, and one other possible time dimension (ie, not age or time since study enrollment).  For each study, briefly describe the research question, the study sample, the longitudinal design, and the analysis approach. Please post links to the studies. 

Age as longitudinal measure

https://doi-org.ucsf.idm.oclc.org/10.1111/cdev.12043

McAlister AR, Peterson CC. Siblings, Theory of Mind, and Executive Functioning in Children Aged 3–6 Years; New Longitudinal Evidence. Child Development. 2013;84:1442-1458.

Research Questions: 1. Does demonstration of executive functioning in early life predict later theory of mind development?  2. What is the effect of having siblings at home (vs. being an only-child) have on the development of theory of mind and executive functioning?

Study Sample: 157 children ages 3 years 3 months to 5 years 6 months were recruited from Australian kindergartens and preschools.  (information on eligibility criteria is limited)

Longitudinal Design: Participants were enrolled and underwent a battery of 4 theory of mind tasks, 2 executive functioning tasks, and a language measure.  This was repeated when the children were ~1 year older.  The first timepoint was at preschool for subjects, at the second time point some participants moved on to primary school.

Analysis: One-way analysis of covariance was used with age and test scores were used to examine the presence of siblings as a predictor of theory of mind development.  Another ANCOVA was used for the follow-up time of 1 year older as the dependent variable.

 

Time since enrollment

https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0064392

Maskew M, Fox MP, van Cutsem G, Chu K, MacPhail P, Boulle A, et al. (2013) Treatment Response and Mortality among Patients Starting Antiretroviral Therapy with and without Kaposi Sarcoma: A Cohort Study. PLoS ONE 8(6): e64392.

 

Research Question: What is the effect of KS on survival (as well as loss to follow-up and immunologic and virologic responses) in a South African population undergoing ART.

Study Sample: Cohort data from two large urban HIV clinics were combined: the Themba Lethu Clinic and Khayelitsha sites in Johannesburg and Cape Town, respectively.  Enrollment for ART within these cohorts included either a CD4 count < 200 or a WHO stage 4 illness (eg. Kaposi sarcoma).  Within these cohorts, 13,847 adults (>=18 years of age) who initiated ART between 1/1/2001 – 12/31/2007 with standard public sector first-line ART regimens were included for analysis.  Notably they included any KS diagnosis within 6 months (pre- or post-) of enrollment as prevalent KS cases, also important: KS was largely identified on a clinical basis.

Longitudinal Design: Authors assessed the above 3 outcomes at 6- and 12-month on ART (study enrollment) based on follow-up at respective clinics or active tracing of patients who did not return to the clinic (including data for those lost compared to the South African National Vital Registration system, though this system likely does not cover all study subjects).  Notably mortality and losses to follow-up were considered the effect of KS if they occurred any time after initiation of ART.

Analysis Approach: Authors compared baseline characteristics stratified by KS status.  Cox proportional hazard models were used to estimate the effect of KS on mortality and loss to follow-up on ART at 6- and 12-months.  Person-time was calculated from ART initiation to censoring (mortality, loss to follow up, or transfer to another facility).  Linear mixed models for repeated measures were used with an unstructured correlation matrix to estimate CD4 trajectories over time with separate models based on KS status.  Multivariate GEE was used to estimate association of KS with change in CD4 count; covariates used in their model included age, gender, baseline CD4, TB treatment status, year of ART initiation, and site).  Log-binomial regression was used to estimate impact of KS status on CD4.  Authors reported a sensitivity analysis where they attributed the outcomes to all missing CD4 counts, and that such analysis did not change the results at either time points.  More specific information regarding the process was not reported.

 

Time since release from prison

https://www-sciencedirect-com.ucsf.idm.oclc.org/science/article/pii/S0277953607003589

Seal DW, Eldrige GD, Kacanek D, Binson D, MacGowan RJ, Project START Study Group. A longitudinal, qualitative analysis of the context of substance use and sexual behavior among 18- to 29-year-old men after their release from prison. Social Science & Medicine. 2007;65:2394-2406.

Research Question:

Study Sample: 89 men between 18-29 years of age were recruited within 60 days of their scheduled release from one of five state prisons (California, Mississippi, Rhode Island, and Wisconsin).

Longitudinal Design: 5 Qualitative interviews were conducted on a range of topics.  The current study reports on only the 6-month follow-up on topics related to substance use, sexual behavior, and reincarceration.

Analysis: In addition to qualitative data, a series of ratings were developed to assess participants regarding their reintegration to life outside prison.  A “global reintegration” code ranging from 1 – 3 was developed as well as 4 themed categories identified from qualitative interviews as being relevant to successful reentry and a risk behavior trajectory.  These include “social consistency”, “Social extensiveness”, “social support”, and “structural stability”.  Additional quantitative measures include substance use, sexual risk behavior, and reincarceration.

Age as time dimension

Lamar, M., Resnick, S. M., & Zonderman, A. B. (2003). Longitudinal changes in verbal memory in older adults: Distinguishing the effects of age from repeat testing.Neurology, 60(1), 82-86. doi:10.1212/WNL.60.1.82

https://www-ncbi-nlm-nih-gov.ucsf.idm.oclc.org/pubmed/12525723

1. The research question - The purpose of the current research is to characterize the nature of age-related changes on a measure of new learning and recall, to evaluate the influence of repeat testing on performance, and to examine sex, education, and baseline performance as modulators of longitudinal age change in memory in nondemented older adults. 

2. The study sample - Subjects were >= 55 years old from the Baltimore Longitudinal Study of Aging (BLSA). THey were included if they had completed the California Verbal Learning Test (CVLT)--a cognitive performance test. The sample was comprised of community-dwelling, generally healthy group of volunteers; cognitively normal for age at baseline and all follow-up evaluations; n = 385

3. The longitudinal design - In approximately 2 to 7 years of longitudinal follow-up, subjects received repeated CVLT tests every 2 years

4. The analysis approach- series of mixed-effects regression analyses to investigate age-related changes in CVLT test performance and the possible modulating effects of repeat testing on longitudinal change; Separate analyses were conducted for each CVLT variable of interest.

Time since study enrollment as the time dimension

Buchman, A. S., Boyle, P. A., Wilson, R. S., James, B. D., Leurgans, S. E., Arnold, S. E., & Bennett, D. A. (2010). Loneliness and the rate of motor decline in old age: The rush memory and aging project, a community-based cohort study. BMC Geriatrics, 10(1), 77-77. doi:10.1186/1471-2318-10-77

1. The research question - Hypothesis H1: A higher level of loneliness at study entry is associated with a more rapid rate of motor decline during the course of the study. 

2. The study sample - Subjects were recruited from forty retirement facilities and subsidized housing facilities in northeastern Illinois. Eligibility: The absence of baseline clinical dementia, assessment of loneliness at baseline, and baseline motor evaluation with at least one follow-up evaluation. n=985.  Mean age 79.67 years (SD 7.36). 

3. The longitudinal design -  Enrollees had detailed examinations each year for up to 12 years. Mean follow-up of 5.0 years (SD 2.44 years). 

4. The analysis approach - Mixed-effect models to examine the relationship of loneliness with baseline global motor function and its annual rate of change. The core model included terms for time in years since baseline as well as terms for loneliness at baseline and a term for its interaction with time since baseline. Models included terms for age, sex, and education and their interaction with time.

Time since cART initiation as time dimension

Cysique, L. A., Vaida, F., Letendre, S., Gibson, S., Cherner, M., Woods, S. P., . . . Ellis, R. J. (2009). Dynamics of cognitive change in impaired HIV-positive patients initiating antiretroviral therapy. Neurology, 73(5), 342-348. doi:10.1212/WNL.0b013e3181ab2b3b

1. The research question/Study aims: estimate the rate and nature of neuropsychiatric (NP) test improvement in people living with HIV initiating cART (combined antiretroviral therapy); Determine which demographic, clinical, labs, and treatment factors are associated with NP improvement. 

2. The study sample - HIV+ people enrolled into the CIT cohort study at the HIV Neurobehavioral Research Center. Study participants included 37 HIV+ individuals with mild to moderate NP impairment who initiated cART.

3. The longitudinal design -  This study examines cognitive performance from four assessments over 48 weeks following cART initiation. Participants underwent NP testing at 12, 24, 36, and 48 weeks after starting cART. 

   4. The analysis approach - Mixed-effect regression models were used to evaluate the time course of cognitive change and its association with baseline and time-varying NP predictors. (NP change was assessed using a regression-based change score that was normed on a separate NP-stable group). A mixed effects model was fitted for the longitudinal NP score (at weeks 12, 24, 36, and 48), with a linear time effect and random, subject-specific intercepts and slopes. 

Article: Newby et al. Lung function decline and variable airway inflammatory pattern: Longitudinal analysis of severe asthma. 2014 PMID: 24928647 link to paper

The research question: they asked 2 main questions:

• to determine if eosinophilic inflammation in sputum is associated with FEV1 decrease in patients with severe asthma and

• whether they could identify subgroups of asthma severity based on the variation of eosinophilic airway inflammation over time

the study sample: 908 adults with asthma from Leicester Difficult Asthma Database cohort at Glenfield Hospital in the U.K.

the longitudinal design: patients with severe asthma who had at least 4 visits with sputum and postbronchodilator lung function measurements over 5 or more years in addition to clinical data related to their asthma history.

the analysis approach. The authors used linear mixed-effects modeling to identify the relationship between sputum eosinophilia and decrease in postbronchodilator FEV1. The final model included the fixed effects of time (in years); height; (3) age of onset of asthma symptoms (in years); (4) exacerbations; mean log10 sputum eosinophil percentage over time; and sex. The random effects of time (in years) and log10 sputum eosinophil percentage were modeled to allow us to model patient-specific lung function decline. Their supplemental data show that they tested 30 different models and picked the one using Bayesian information criterion (I think, but not sure from methods and results). They then used unsupervised cluster analysis to cluster patients based on their mean (representing the amplitude of eosinophilic airway inflammation) and SD log10 sputum eosinophil counts (representing the variation of eosinophilic airway inflammation) over time to identify whether we could generate clusters of differential lung function decline within their cohort of severe asthma. They used some type of BIC metric to decide which number of clusters best captured the variability in the data.

 

 

Time Dimension: Enrollment, procedure

Article: https://www.ncbi.nlm.nih.gov/pubmed/30946773

Research question: Can sociodemographic or medical predictors, predict patient urinary and sexual dysfunction after Robot-Assisted Radical Prostatectomy?

Study sample: Italian men with localized prostate cancer who had undergone RARP. The men must be native Italian speakers, no neurological or psychopathological problems and referred to value based project. 478 men were selected in the end with localized pc and undergone rarp. 

Longitudinal design: T0 baseline of measurements pre hospitalization and t1 45 days’ post RARP  with follow ups after being 3 month intervals until 12 months. Patients were put into two categories using baseline measurements for membership and latent class growth analysis on EPIC-26 survey a wellbeing urology survey. 

Analysis approach: T0 would be used as a way to categorize groups into trajectories as baseline assessments had low outcome variability and T1 would be the start of the analysis. EPIC-26 was used to create longitudinal trajectories, urinary and sexual dysfunction using non-liner latent class growth analysis. The authors used chi-square test to assess the association between the trajectories urinary and sexual dysfunction and membership.

1.     Age as the time dimension- Kari JT, Pehkonen J, Hutri-Kähönen N, Raitakari OT, Tammelin TH. Longitudinal Associations Between Physical Activity and Educational Outcomes. Medicine & Science in Sports & Exercise. 2017;49(11):2158-2166.  https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5647118/pdf/mss-49-2158.pdf

In this study researchers wanted to evaluate whether physical activity in adolescence was associated with academic achievement at the end of compulsory education and with education later in life. They used data from the Cardiovascular Risk in Young Finns Study, and analyzed 2445 subjects that were 12-15 years when information about physical activity was measured. Physical activity (exposure) was self-reported, and researchers used several questions to come up with an index. Participants were followed until 2010; mean age of the group at that time was 40 years. Education information (outcomes) was also self-reported at age 15 (ie, GPA) and by registry for the years of educational attainment as adults. The analysis consisted of ordinary least squares models stratified by sex. Models were adjusted by birth cohort, birth month, chronic conditions, body fat, family income, parents’ education and family size. Additionally, authors conducted instrumental variable analyses and included as instruments, in three different models: individual’s body height, physical activity of the father and physical activity of  father and mother.  

 2.     Time since study enrollment as the time dimension- Rebeiro PF, Howe CJ, Rogers WB, et al. The Relationship Between Adverse Neighborhood Socioeconomic Context and HIV Continuum of Care Outcomes in a Diverse HIV Clinic Cohort in the Southern United States. AIDS Care. 2018;30(11):1426-1434. https://www.tandfonline.com/doi/full/10.1080/09540121.2018.1465526

The aim of this study was to assess the relationship between neighborhood socioeconomic context (NSEC) and viral suppression and retention in care in a cohort of HIV positive adults. Viral suppression was defined as <200 copies/mL in last measurement of the year and retention in care as more than 2 visits, 90 days apart per calendar year. The study sample consisted of individuals participating of the Vanderbilt Comprehensive Care Clinic cohort that had more than 1 study visit and resided in zip code areas with available census-derived socioeconomic information for the period of January 1, 2008 and December 31, 2012. Participants (n=2,272 for retention and n=2,541 for viral suppression) were followed from their first study visit until death or end of the study period. Modified Poisson regression was used to calculate relative risks and 95% CIs by quartiles of the NSEC for the two study outcomes. GEE was used to account for correlation within individuals over time and zip code tabulation areas. Models were adjusted for year of birth, race/ethnicity and time since initiating HIV care.       

 

3.     One other possible time dimension (ie, not age or time since study enrollment)- Haas CB, Phipps AI, Hajat A, Chubak J, Wernli KJ. Time to Fecal Immunochemical Test Completion for Colorectal Cancer Screening. Am J Manag Care. 2019;25(4):174-180. https://ajmc.s3.amazonaws.com/_media/_pdf/AJMC_04_2019_Haas_final.pdf

In this study researchers were interested in estimating time to completion of the fecal immunochemical test (FIT) and in identifying factors associated with it. A total of 63,478 members of Kaiser Permanente Washington who got a FIT order between January 1, 2011 and December 31, 2012 were included in the analysis. Participants were followed from the date recorded of the FIT order until the date recorded of a received FIT. Death, disenrollment from Kaiser, and the day 365 from the FIT order for those not having any of the two described events and did not return the FIT, were considered as censoring events. One year was chosen as the study endpoint based on the FIT is recommended annually. For this time-to-event analysis, authors used the Kaplan-Meier method and compared curves by age, race/ethnicity, BMI and Charlson comorbidity index with the log-rank test. Cox proportional hazards models were also conducted to determine factors independently associated with FIT completion. Predictors included in the models were: gender, age, race/ethnicity, type of insurance, BMI, comorbidities and CRC testing in previous 2 years.      

Age as the time dimension

https://www.ncbi.nlm.nih.gov/pubmed/28499106

Title: Alcohol and Cigarette Use From Ages 23 to 55: Links With Health and Well-Being in the Long-Term National Child Development Study

Research question: They examined investigated how patterns of alcohol and cigarette use from young adulthood (age 23) to midlife (age 55) are associated with health and well-being.

Study sample:  They used a nationally representative British cohort born in one week in 1958. They also included immigrants born the same week.

Longitudinal design: Following initial assessment of 17,415 infants (99% of births), the cohort was assessed at ages 7, 11, 16, 23, 33, 42, 46, 50, and 55 years. They immigrants born the same week were added at ages 7, 11, and 16.

Analytic approach: They used to what they called a “multilevel latent class analysis” to assess how successive combinations of alcohol and cigarette use come together in distinct configurations from ages 23 to 55 (i.e., latent paths). They then assessed how these latent paths relate to health and well-being in young adulthood and midlife.

 

 

Time since enrollment

https://www.ncbi.nlm.nih.gov/pubmed/27017310

Title: A blood RNA signature for tuberculosis disease risk: a prospective cohort study.

Research question: Their main objective was aimed to assess whether global gene expression measured in whole blood of healthy people allowed identification of prospective signatures of risk of active tuberculosis disease.

Study sample: Theyfollowed healthy, South African adolescents aged 12-18 years from the adolescent cohort study (ACS) who were infected with M tuberculosis for 2 years.

Longitudinal design: They then collected blood samples from study participants every 6 months and monitored the adolescents for progression to tuberculosis disease. 

Analytic approach:Although this was a longitudinal study all they really did for analysis was calculate sensitivity/specificity as well as AUROC curves. They probably have additional papers using this particular dataset where they used more family longitudinal-based analyses i.e. survival analysis. 

 

 

Age as the time dimension

https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0092235

Title: Predictors of Full Enteral Feeding (FEF) Achievement in Very Low Birth Weight Infants (VLBW)

Research question: The aim of the present study was to determine the role that some prenatal, neonatal and early postnatal variables play in favoring or delaying the achievement of FEF in VLBW infants and to determine whether neonatal intensive care units differ in this outcome.

Study sample: In this study they used data from an Italian data registry, the “Emilia-Romagna Perinatal Network.” The database included live-born infants with birth weight (BW) between 500 and 1500 grams admitted to the nine Level-3 NICUs of the Region.  For their study they included 2,441 VLBW infants born 2004-2009 .

Longitudinal design: The outcome of interest was time to FEF achievement, measured as the difference in days between the date when FEF was established and the birth date. FEF was defined as enteral tolerance of at least 150 ml/kg/day of milk. How much an infant eats was tracked daily

Analytic approach:A parametric survival analysis was used to estimate the relationship of prenatal, neonatal and early postnatal variables with time to FEF. 

Age as the time dimension

Hardy R, Wadsworth M, Kuh D. The influence of childhood weight and socioeconomic status on change in adult body mass index in a British national cohort. Int J Obesity 2000; 24: 725-34

https://www-nature-com.ucsf.idm.oclc.org/articles/0801238?proof=trueIn

Research question: Overweight and obesity in middle life have been associated with cardiovascular disease, adult-onset diabetes and various types of cancer. The authors investigated the effect of childhood weight and childhood SES on the pattern of change in BMI between 20 and 43 years.

Study Sample: The MRC National Survey of Health and Development is a socially stratified cohort of 2548 women and 2814 men born in one week in 1946 in England, Scotland and Wales.

Longitudinal design: There have been 19 follow-ups, the last follow-up was at age 43 years. The population interviewed at that age was, in most respects, representative of the native population of that age. Drop-outs: 6.8% of the original cohort had died and 12.1% had permanently withdrawn from the study, 11.5% were living abroad and 8.8% had temporarily refused to participate or could not be traced.

Analysis: Hierarchical data with 4 outcomes (adult BMI at age 20, 26, 36 and 43 years) per individual. Age was level 1 (within individual change) and the individual was level 2 (between individuals change) of the hierarchy. Fixed parameters: mean intercept and mean slope. Variation in intercept and slope were the random parameters that allowed each individual had their own intercept and slope. Statistical modeling and testing were carried out using multilevel models fitted using the computer package MLn. Associations between continuous risk factors and BMI were tested for deviations from linearity. Interactions between baseline covariates and age were investigated.

Time since study enrolment as the time dimension

Whitlock E, Diaz-Ramirez, G, Glymour, M. Association Between Persistent Pain and Memory Decline and Dementia in a Longitudinal Cohort of Elders. JAMA Intern Med. 2017;177(8):1146-1153.

https://jamanetwork-com.ucsf.idm.oclc.org/journals/jamainternalmedicine/fullarticle/2629448

Research question: Is persistent pain associated with accelerated cognitive decline in the elderly?

Study sample: This cohort study used data from the Health and Retirement Study. Participants for this analysis answered questions about pain and cognition without proxy in the 1998 and 2000 evaluation waves and were at least 60 years old at the 1998 wave.

Longitudinal design: Participants were followed with in-person or telephone interviews approximately every 2 years until death, dropout, or the 2012 evaluation wave.

Analysis: Multivariable linear mixed-effects models were developed to estimate the association of persistent pain and the coprimary outcomes of memory score and dementia probability. Interaction of time since the year 2000 and the persistent pain classification was used to describe differences in the slope of the cognitive trajectory for participants with vs without persistent pain at baseline. Individual participants’ slopes and intercepts for the cognitive trajectory were allowed to vary as random effects. Memory score was modeled without transformation; dementia probability was modeled on the log-odds scale.

Other possible time dimensions: Time before and after AMI

Mendes de Leon C, Bang W, Bienias J et al. Change in disability before and after myocardial infarction in older adults. Arch Intern Med 2005; 165: 763-68

https://www-ncbi-nlm-nih-gov.ucsf.idm.oclc.org/pubmed/15824295

Research question: The authors wanted to characterize changes in disability before and after AMI in older adults and test specifically whether changes in disability during the first year after AMI differ from those occurring in the 3-year period before AMI.

Study sample: Participants from the New Haven study site of the Established Populations for the Epidemiologic Studies of the Elderly (EPESE) project. The sampling frame was a stratified probability sample of the noninstitutionalized New Haven population 65 years and older. A total of 2812 participants.

Longitudinal design: Baseline was 1982. Follow-up interviews were conducted at yearly intervals through 1990/1991, including face-to-face interviews in 1985 and 1988 and telephone interviews in all other years. There was complete information on vital status for the entire follow-up period.

Analysis: yearly collected disability data were linked to the date of AMI. There were up to 3 pre-AMI interviews and 1 post-AMI interview. Multivariable regression models were used to estimate the change in disability outcomes during the pre-AMI and post-AMI period. Owing to their nonnormal distributions, disability scores were considered as the number of tasks a person was unable to perform out of the total number of tasks on each measure (a proportion). They analyzed the data using generalized estimating equations with a logit link function, a binomial error structure, and an exchangeable working correlation matrix. Estimated regression coefficients were expressed as odds ratios, which represent the linear effect of a predictor variable on the odds of the proportion of tasks a person was unable to perform.

Time Dimension	Research Question	Study Design	Analytic Approach	Link
Age	What is the incidence of recovery from stunting and what factors are associated with post-stunting growth in children under-five?	Longitudinal study using repeated measures from a demographic health and surveillance site	Event history analysis of time-to-recover (Cox regression models) and Height-for-Age Z [HAZ] slope modeling (The difference in HAZ for each stunted child is estimated by fitting a line through the points, starting from the first stunting episode, and using the slope for the individual as the measure of change over time)	https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0215488
Time since enrollment	What is the role of hormonal contraception in breast cancer risk?	Cohort study enrolling 100 000 women aged 34 to 49 years in 1991/1992 with follow-up until December 31, 1999	Relative hazards were calculated using Cox proportional hazard models, with use of Oral Contraceptives as the independent variable and breast cancer as the dependent variable. Analyses were stratified by age at enrollment because a significant interaction was found between current use  of  OCs and age at start of follow-up.	http://cebp.aacrjournals.org/content/cebp/11/11/1375.full.pdf
Time to event	What is the effect the effect of Treat All implementation on timely ART initiation and retention in care?	Open observational cohort of patients receiving routine care	Interrupted time series analysis using pre-Treat All and post-Treat All periods.  Baseline characteristics of patients enrolling in the pre‐Treat All and Treat All periods were compared using bivariate logistic regression models that accounted for clustering within health centres. We used the Kaplan–Meier method to estimate median time from enrolment to ART initiation. Segmented linear regression models were used to estimate the predicted probability of initiating ART within 30 days of enrolment and six‐month retention in care among patients entering care in each month.	https://onlinelibrary.wiley.com/doi/full/10.1002/jia2.25279

Age as time dimension

Research question: To determine changes in patient-reported outcomes (PRO) and walking speed (Timed 25-foot walk) over time between those who continued versus discontinued disease modifying therapy among multiple sclerosis patients over age 60.

Study sample: Six hundred Patients over 60 who were seen at three US MS centers (Cleveland, OH; Las Vegas, NV; Weston, FL), who had PRO and T25FW data and who were diagnosed with MS before age 60 were included.

Longitudinal design: PROs and T25FW were collected routinely at clinic visits.  Study participants had a median follow up time of 5.3 years.

Analysis approach: Mixed-effects linear regression models were used. For each model, the dependent variable was either the PRO score or the T25FW speed. The independent variables of interest included time from age 60 (in years) and a treatment group variable comparing DMT continuers and discontinuers consisting of three categories: continuers, discontinuers before discontinuation (DBD), and discontinuers after discontinuation (DAD). To determine if outcome trajectories differed by treatment group over time, authors included an interaction term between the time from age 60 and treatment group.

Link: https://www.sciencedirect.com/science/article/pii/S2211034819301051?via%3Dihub

Time since study enrollment as time dimension

Research question: To compare the effects of a mindfulness yoga program vs stretching and resistance training exercise (SRTE) on psychological distress (measured using HADS score) in patients with mild-to-moderate Parkinson’s disease.

Study sample: A total of 187 adults with a clinical diagnosis of idiopathic PD who were able to stand unaided and walk with or without an assistive device were enrolled via convenience sampling. Participants were recruited through 2 regional neurology outpatient clinics and 4 centers of the Hong Kong Society of Rehabilitation, Hong Kong Parkinson’s Disease Association.

Longitudinal design: This was an RCT conducted at 4 community rehabilitation centers in Hong Kong between December 1, 2016, and May 31, 2017. Assessments were done at baseline, 8 weeks (T1), and 20 weeks (T2).

Analysis approach: GEE models, specifically with a first-order autoregressive structure, were used to assess the differential change in the primary outcome variable (HADS score) between the 2 groups at T1 and T2 compared with T0.

Link: https://jamanetwork-com.ucsf.idm.oclc.org/journals/jamaneurology/fullarticle/2729691

Time since discharge from hospital as time dimension

Research question: Whether in-hospital insomnia (measured at discharge) independently predicts long-term pain after burn injury.

Study sample: Participants were recruited from consecutive admissions to one of the three regional burn centers in the USA contributing data on adults to the Burn Model System database. Study sample included 333 subjects hospitalized for major burn injury.

Longitudinal design: Subjects completed measures of health, function (SF-36), and psychological distress (Brief Symptom Inventory) while in hospital, at 6, 12, and 24 months after discharge. Participants were categorized as either having or not having sleep onset insomnia at discharge.

Analysis approach: Both conditional linear mixed-effects models and GEE were used to model the correlation structure of the repeated measures within each patient. Authors included time from discharge as a predictor. Covariate predictors included in the models were % total body surface area burned, % total body surface area skin grafted, number of days in intensive care unit, time since discharge (measured at 6 months, 1 year and 2 years), SF-36 Mental Health and Bodily Pain scores at the time of discharge, SF-36 General Health, Bodily Pain, and Mental Health Indices before burn injury (measured retrospectively), sex, age in years at the time of burn injury, and ethnicity (white vs. non-white).

Link: https://insights.ovid.com/pubmed?pmid=18362052

Time Dimension: Age of the individual 

Roberts S, Suderman M, Zammit S, et al. Longitudinal investigation of DNA methylation changes preceding adolescent psychotic experiences. Translational Psychiatry. 2019;9:69. (doi: 10.1038/s41398-019-0407-8).

https://www.nature.com/articles/s41398-019-0407-8

Research question: Are changes in or particular patterns of DNA methylation (an essential process in mammal development) at different stages of development predictive of psychotic experiences (PEs) in late-childhood and adolescence?

Study sample: The study sample consists of 901 individuals for whom DNA methylation data were available from all time points, and for whom PE data were available from at least one time point. These individuals are a subset of the ARIES (Accessible Resource for Integrated Epigenomic Studies) cohort, which itself is a subset of the ALSPAC (Avon Longitudinal Study of Parents and Children) cohort. ALSPAC was a large cohort study that recruited over 14,000 pregnant women in Avon, UK, in 1991 and 1992.

Longitudinal design: Blood samples were collected from study participants at birth, age 7, and age 15-17. PE data were obtained via interviews at age 12 and age 18.

Analysis approach: The analyses were carried out in four stages:

1.     The authors performed six “epigenome-wide association studies” of PEs at the two interview time points with DNAm at the three blood sampling time points, controlling for relevant covariates and using a Bonferroni correction for multiple comparisons.

2.     One-way analysis of variance tests were performed to assess the association between DNA methylation at 110 specific regions of DNA (CpG sites) and the continuity of PEs between age 12 and 18 (persistent, remitted, emergent, and none). CpG sites with significant (p<0.00045) associations were then modeled with linear regression to adjust for covariates.

3.     Investigators used multilevel models to determine the association between patterns of DNA methylation over time at the “top” (the authors’ meaning is ambiguous) CpG sites and the continuity of PEs between age 12 and 18.

               a.     A linear spline with one knot at age 7 allowed for different linear relationships from birth to age 7 and from age 7 to age 15-17.

4.     Differentially methylated regions were identified with a particular module in Python, and statistical significance was determined, but I do not understand what analysis was performed nor what the statistical significance reveals. The authors used a Šidák correction for multiple comparisons.

Time Dimension: Time since study enrollment

Elbejjani M, Auer R, Jacobs DR, et al. Cigarette smoking and gray matter brain volumes in middle age adults: the CARDIA Brain MRI sub-study. Translational Psychiatry. 2019;9:78. (doi: 10.1038/s41398-019-0401-1). 

https://www.nature.com/articles/s41398-019-0401-1

Research question: Where and to what extent does cigarette smoking cause atrophy of the brain’s grey matter? And do vascular, respiratory, psychological, or substance-use conditions influence this relationship?

Study sample: The study sample includes 698 individuals who were recruited at the 25-year follow-up visit of the CARDIA (Coronary Artery Risk Development in Young Adults) study. The CARDIA cohort included over 5000 individuals—balanced by race, sex, and education—who were followed for 25 years starting in the mid-1980s. From among the participants in the 25-year follow-up, a subsample of 719 people was enrolled in this sub-study, 698 of whom had complete data for smoking and other covariates.  

Longitudinal design: Brain MRIs were conducted for the 698 individuals recruited for this study (the 25^th year since enrollment in the CARDIA study). Exposure data (never, former, or current smoker) were collected consistently at CARDIA’s baseline and at each study examination (years 2, 5, 7, 10, 15, 20, and 25 after baseline). Analyses adjusted for vascular risk factors, respiratory risk factors, and substance-use/psychological risk factors. The values of these covariates used in the analyses were collected at years 20 (FEV1) and 25 (all other covariates).

Analysis approach: Study investigators used multivariable linear regression models to examine the relationship between historical smoking status and measures of total and regional grey matter volume (including the frontal, temporal, occipital, parietal, amygdala, insula, entorhinal cortex, and cingulate regions). Analyses adjusted for age, sex, race, educational attainment, intracranial volume, and study center in all analyses. Three models that separately controlled for vascular factors, respiratory factors, and substance-use/psychological factors were also evaluated. Interaction terms were used to determine whether any of these three mechanisms modified the relationship between smoking status and grey matter volumes. Finally, the investigators performed a sensitivity analysis to assess the impact of the high prevalence of a self-reported history of cannabis use in the sample.

Time Dimension: Time since treatment 

Sato M, Yamato M, Mitani G, et al. Combined surgery and chondrocyte cell-sheet transplantation improves clinical and structural outcomes in knee osteoarthritis. Regenerative Medicine. 2019;4:1-11. (doi: 10.1038/s41536-019-0069-4).

https://www.nature.com/articles/s41536-019-0069-4

Research question: Do transplanted chondrocyte cell sheets, in combination with conventional surgical treatment, improve clinical and structural outcomes for patients with osteoarthritis of the knee?

Study sample: Ten patients with osteoarthritis of the knee and symptomatic cartilage lesions were enrolled in this study in Japan. Two individuals were later excluded, and eight received the experimental treatment (conventional surgery in combination with chondrocyte cell-sheet transplantation).

Longitudinal design: Clinical examinations of each individual were performed at the following nine time points: 6 weeks before treatment, 4-to-3 weeks before treatment, day zero (the day of treatment), and 1, 3, 6, 12, 24, and 36 months after treatment.

Analysis approach: The investigators used one-way analysis of variance tests (comparing pre-treatment values to post-treatment values for the eight study participants, not accounting for correlation between measurements from a given individual) with post-hoc Holm-Bonferroni corrections for multiple comparisons.

Outcomes:

1.     Pre- and post-op imaging assessment of x-rays and MRIs

               a.     X-ray: Pre- and post-op progression of osteoarthritis evaluated using Kellgren-Lawrence grading scale

               b.     MRI: MOCART method for evaluation: magnetic resonance observation of the cartilage repair tissue

2.     Pre- and post-op arthroscopy

               a.     evaluation of cartilage defects for condition and size

               b.     Outerbridge grade of joint cartilage breakdown

3.     Pre- and post-op evaluation of cartilage viscoelasticity

               a.     LIPA = laser-induced photoacoustic evaluation

4.     Pre- and post-op (1, 3, 6, 12, 24, 36 month) clinical outcomes

               a.     KOOS = Knee injury and Osteoarthritis Outcome Score

               b.     LKS = Lysholm Knee Score

5.     Histological outcomes

               a.     12-month arthroscopic biopsy of regenerated cartilage

                                    i.     OARSI (Osteoarthritis Research Society International) histological score assessed independently by 3 trained ortho surgeons

                                    ii.     Mankin score assessed independently by 3 trained ortho surgeons