Statistics in Health Sciences PDF

B.Sc. Degree in Applied Statistics Statistics in Health Sciences 3. Classification of studies Jose Barreraab [email protected] https://sites.google.com/view/josebarrera a ISGlobal Barcelona Institute for Global Health - Campus MAR b Department of Mathematics (UAB) This work is licensed under...

B.Sc. Degree in Applied Statistics Statistics in Health Sciences 3. Classification of studies Jose Barreraab [email protected] https://sites.google.com/view/josebarrera a ISGlobal Barcelona Institute for Global Health - Campus MAR b Department of Mathematics (UAB) This work is licensed under a Creative Commons “Attribution-NonCommercial-ShareAlike 4.0 International” license. Statistics in Health Sciences 1 Epidemiological studies Notation Classification criteria Epidemiological study designs Randomized epidemiological trials Cohorts Case-control Cross-sectional Ecological Time series analysis 2 (Main) studies classification diagram Jose Barrera (ISGlobal & UAB) Statistics in Health Sciences, 2023/2024 2 / 20 Classification of epidemiological studies Exposure, outcome and study design • One of the main aims of epidemiology is to analize the relationship between an exposure and a health indicador (outcome) of interest. For instance: • Working in a coal mine and suffering a respiratory disease • Suffering melanoma (i.e. skin cancer) and being exposed to UV radiation (e.g. sunbathing) • Excess of mortality attributable to heat waves  In each of the previous examples, identify the exposure and the outcome. • To analyze such scientific questions, several study designs have been developed. Next, we will introduce the main ones. Notation Onwards, when working with binary both exposure and outcome, we will use the following notation: D D̄ E Ē : : : : The disease (or health indicator) of interest is present The disease (or health indicator) of interest is absent The exposure of interest is present The exposure of interest is absent Jose Barrera (ISGlobal & UAB) Statistics in Health Sciences, 2023/2024 3 / 20 Classification of epidemiological studies Classification criteria There is a number of characteristics of the study that are related with its design: • Allocation of study participants to the different study groups: random allocation or allocation under predetermined criteria. • Number of measurements in each participant: one (cross-sectional study) or more than one (longitudinal study). • Time direction of the study: prospective of retrospective study. • Selection criteria for the participants: none, exposure based o disease status based. • Study unit: individual or population. Jose Barrera (ISGlobal & UAB) Statistics in Health Sciences, 2023/2024 4 / 20 Classification of epidemiological studies Study designs The main epidemiological study designs are: • Randomized epidemiological trials • Cohorts • Case-control • Cross-sectional • Ecological • Time series analysis Jose Barrera (ISGlobal & UAB) Statistics in Health Sciences, 2023/2024 5 / 20 Randomized epidemiological trials (1/2) Characteristics • Random sample of individuals from the population of interest • Random allocation regarding the exposure • Prospective longitudinal study • The units of the study disease free individuals Advantages • Prospectivity and selection criteria regarding the disease imply exposure precedes disease • We can compare risks P(D|E) and P(D|Ē), and we can estimate incidence rates • Random allocation minimizes potential bias (groups only differ in exposure and outcome) • High external validation • The study design allows to analyse several diseases Disadvantages • The study could be long and expensive • Ethical problems (most of studies analyze potentially harmful exposures) Jose Barrera (ISGlobal & UAB) Statistics in Health Sciences, 2023/2024 6 / 20 Randomized epidemiological trials (2/2) Examples 1 Study of long-term efficacy of an exercise intervention to increase physical activity in patients with COPD, by Arbillaga-Etxarri et al. [1] .  Read the paper, identify the exposure and the outcome of interest and summarize the main results. 2 Study of short-term effect of air pollution on attention function in adolescents in high schools in Barcelona, Gignac et al. [2] .  Read the paper, identify the exposure and the outcomes. How was the exposure randomized? How were the outcomes measured? Jose Barrera (ISGlobal & UAB) Statistics in Health Sciences, 2023/2024 7 / 20 Cohorts (1/2) Characteristics • Same design than a randomized epidemiological trial, except that. . . • Exposure status is observed instead of randomly assigned (observational study) • Usually prospective (rarely retrospective) Advantages • Prospectivity and selection criteria regarding the disease imply exposure precedes disease • We can compare risks P(D|E) and P(D|Ē), and we can estimate incidence rates • Allows for the study of rare exposure • High external validation (if groups are comparable) • The study design allows to analyze several diseases Disadvantages • The study could be long and expensive • Requires high sample sizes in the case of rare diseases • The exposure status can vary during the study • Lost of follow-up risk Jose Barrera (ISGlobal & UAB) Statistics in Health Sciences, 2023/2024 8 / 20 Cohorts (2/2) Example Study of long-term association between television viewing during childhood and adolescent neuropsychological outcomes, by O’Connor et al. [3] .  Read the paper, identify the exposure and the outcome of interest and summarize the main results. Exercise: randomized epidemiological trial vs cohort  Think of a situation where a randomized epidemiological trial would not be viable but a cohort study would.  What is the main advantage of randomized epidemiological trial over cohort study would? Jose Barrera (ISGlobal & UAB) Statistics in Health Sciences, 2023/2024 9 / 20 Case-control (1/2) Characteristics • Selection criteria based on the disease of interest: individuals are randomly selected among cases (affected by the disease) and controls (disease free) • Sometimes, matching case-control. [4] • Retrospective study Advantages • Appropriate to study rare or high period of latency diseases • The study design allows to analyse several exposures • Require lower economical and time costs than cohorts studies Disadvantages • Unable to estimate P(D|E) • Unable to guarantee that exposure precedes the disease • Assessing the exposure with precision could be difficult • Potential selection bias Jose Barrera (ISGlobal & UAB) Statistics in Health Sciences, 2023/2024 10 / 20 Case-control (2/2) Example Study of factors associated to cancer as, for instance, Papantoniou et al. [5] , Garcia-Saenz et al. [6] or a matching design by Castaño-Vinyals et al. [4] .  Read one of the papers, identify the exposure and the outcome of interest and summarize the main results.  In the case of Castaño-Vinyals et al. [4] , how were the matching strata defined? Exercise: cancer and case-control design  Most of the epidemiological studies related to cancer use the case-control design. Why? Jose Barrera (ISGlobal & UAB) Statistics in Health Sciences, 2023/2024 11 / 20 Cross-sectional (1/2) Characteristics • Random sample from the population of interest • Only one measurement in each individual • In each individual, the exposure status and the disease status are assessed posterior to the individual selection Advantages • We can estimate P(D), P(E) and P(D ∩ E) • Require lower economical and time costs than cohorts studies • Useful for hypothesis generation Disadvantages • Unable to guarantee that exposure precedes the disease • Unable to estimate incidence • Proportions of exposed or cases cannot be set Jose Barrera (ISGlobal & UAB) Statistics in Health Sciences, 2023/2024 12 / 20 Cross-sectional (2/2) Example Study of association between traffic-related air pollution and noise at school, and behavioral problems in Barcelona schoolchildren, by Forns et al. [7] .  Read the paper, identify the exposure and the outcome of interest and summarize the main results. Exercise: cross-sectional vs cohort designs  Mention and advantage of cross-sectional design over cohort design.  Mention and advantage of cohort design over cross-sectional design. Jose Barrera (ISGlobal & UAB) Statistics in Health Sciences, 2023/2024 13 / 20 Ecological (1/2) Characteristics • Instead individual data, aggregated data are used (e.g. mortality rates or other indicators at a country level) • Usually, data are available prior to the study (mainly, indicators at country or other regions level) • Usually, cross-sectional design Advantages • Relatively easy to carry out • Usually having high statistical power due to a high number of observations • Allows to compare different populations Disadvantages • Unavailable information at individual level • Confusion is hard (or impossible) to control Jose Barrera (ISGlobal & UAB) Statistics in Health Sciences, 2023/2024 14 / 20 Ecological (2/2) Example Study of worldwide association between UVR exposure and current eczema prevalence in children, by Fuertes et al. [8] .  Read the paper, identify the exposure and the outcome of interest and summarize the main results. Exercise: ecological & longitudinal  Think of a study design being ecological and longitudinal. Jose Barrera (ISGlobal & UAB) Statistics in Health Sciences, 2023/2024 15 / 20 Time series analysis (1/2) Characteristics • Study of short-term effects of an environmental exposure (e.g. ambient temperature or air pollution) on a population health indicator (e.g. mortality or hospitalizations) • Daily time series for the exposure and the outcome are analyzed (ecological study) • Regression model for the mean of the outcome in day i as a function of the exposure in day i • Lagged effects can be considered (e.g. distributed lag models) • Covariates to control for confounding (e.g. time trends, seasonality or relative humidity) Advantages • Efficiently controls for potential confounders • Appropriate to study rare diseases (if conditions above hold) • Data gathering requires low economical and time costs Disadvantages • Aggregated data fails to control for some confounders (e.g. sex and age) • Assessing the exposure with precision could be difficult • Not useful for long-term effects Jose Barrera (ISGlobal & UAB) Statistics in Health Sciences, 2023/2024 16 / 20 Time series analysis (2/2) Example Study of association between high ambient temperatures and risk of motor vehicle crashes in Catalonia, by Basagaña et al. [9] .  Read the paper, identify the exposure and the outcome of interest and summarize the main results. Exercise: PubMed  Use the advanced search in PubMed to find a relatively recent study (2010 to present) using time series analysis to assess the short-term effects of air pollution on some health indicator in some Spanish city. Summarize aims and results. Jose Barrera (ISGlobal & UAB) Statistics in Health Sciences, 2023/2024 17 / 20 (Main) studies classification diagram Assign E? Yes No Experimental study Observational study Reference group? Yes No Controlled trial Non controlled trial Yes No Analytical study Descriptive study Random assignation? Selection criteria No based on E Yes random sample Non randomised trial Cohort study Randomised trial Jose Barrera (ISGlobal & UAB) Case-control study Statistics in Health Sciences, 2023/2024 based on D Cross-sectional study 18 / 20 References I [1] A. Arbillaga-Etxarri, E. Gimeno-Santos, A. Barberan-Garcia, E. Balcells, M. Benet, E. Borrell, N. Celorrio, A. Delgado, C. Jané, A. Marin, C. Martń-Cantera, M. Monteagudo, N. Montellà, L. Muñoz, P. Ortega, DA. Rodríguez, R. Rodríguez-Roisin, P. Simonet, P. Torán-Monserrat, J. Torrent-Pallicer, P. Vall-Casas, J. Vilaró, and J. Garcia-Aymerich. Long-term efficacy and effectiveness of a behavioural and communitybased exercise intervention (Urban TrainingTM ) to increase physical activity in patients with COPD. A randomised controlled trial. European Respiratory Journal, 52(4):1800063, 2018. URL https://doi.org/10.1183/13993003.00063-2018. [2] F. Gignac, J. Barrera-Gómez, C. Persavento, C. Solé, È. Tena, M. López-Vicente, M. Foraster, F. Amato, A. Alastuey, X. Querol, H. Llavador, J. Apesteguia, J. Júlvez, D. Couso, J. Sunyer, and X. Basagaña. Short-term effect of air pollution on attention function in adolescents (ATENC!Ó): A randomized controlled trial in high schools in Barcelona, Spain. Environment International, 156:106614, 2021. URL https://doi.org/10.1016/j.envint.2021.106614. [3] G. O’Connor, M. Piñero Casas, X. Basagaña, ML. Vicente, P. Davand, M. Torrent, D. Martínez-Murciano, R. García-Esteban, M. Marinelli, J. Sunyer, and J. Julvez. Television viewing duration during childhood and long- association with adolescent neuropsychological outcomes. Preventive Medicine Reports, 4:447–452, 2016. URL https://doi.org/10.1016/j.pmedr.2016.08.013. [4] G. Castaño-Vinyals, S. Sadetzki, R. Vermeulen, F. Momoli, M. Kundi, F. Merletti, M. Maslanyj, C. Calderon, J. Wiart, AK. Lee, M. Taki, M. Sim, B. Armstrong, G. Benke, R. Schattner, HP. Hutter, D. Krewski, C. Mohipp, P. Ritvo, J. Spinelli, B. Lacour, T. Remen, K. Radon, T. Weinmann, ETh. Petridou, M. Moschovi, A. Pourtsidis, K. Oikonomou, P. Kanavidis, E. Bouka, R. Dikshit, R. Nagrani, A. Chetrit, R. Bruchim, M. Maule, E. Migliore, G. Filippini, L. Miligi, S. Mattioli, N. Kojimahara, N. Yamaguchi, M. Ha, K. Choi, H. Kromhout, G. Goedhart, A. ’t Mannetje, A. Eng, CE. Langer, J. Alguacil, N. Aragonés, M. Morales-Suárez-Varela, F. Badia, A. Albert, G. Carretero, and E. Cardis. Wireless phone use in childhood and adolescence and neuroepithelial brain tumours: Results from the international MOBI-Kids study. Environment International, 160:107069, 2022. URL https://doi.org/10.1016/j.envint.2021.107069. [5] K. Papantoniou, G. Castaño Vinyals, A. Espinosa, N. Aragonés, B. Pérez-Gómez, J. Burgos, I. Gómez-Acebo, J. Llorca, R. Peiró, JJ. Jimenez-Moleón, F. Arredondo, A. Tardón, M. Polan, and M. Kogevinas. Night shift work, chronotype and prostate cancer risk in the MCCSpain case-control study. International Journal of Cancer, 137(5):1147–1157, 2015. URL https://doi.org/10.1002/ijc.29400. Jose Barrera (ISGlobal & UAB) Statistics in Health Sciences, 2023/2024 19 / 20 References II [6] A. Garcia-Saenz, A. Sánchez de Miguel, A. Espinosa, A. Valentin, N. Aragonés, J. Llorca, P. Amiano, V. Martín Sánchez, M. Guevara, R. Capelo, A. Tardón, R. Peiró-Perez, JJ. Jiménez-Moleón, A. Roca-Barceló, B. Pérez-Gómez, T. Dierssen-Sotos, T. Fernández-Villa, C. Moreno-Iribas, V. Moreno, J. García-Pérez, G. Castaño Vinyals, M. Pollán, M. Aubé, and M. Kogevinas. Evaluating the association between artificial light-at-night exposure and breast and prostate cancer risk in Spain (MCC-Spain study). Environmental Health Perspectives, 126(4):047011, 2018. URL https://doi.org/10.1289/EHP1837. [7] J. Forns, P. Dadvand, M. Foraster, M. Alvarez-Pedrerol, I. Rivas, M. López-Vicente, E. Suades-Gonzalez, R. Garcia-Esteban, M. Esnaola, M. Cirach, J. Grellier, X. Basagaña, X. Querol, M. Guxens, MJ. Nieuwenhuijsen, and J. Sunyer. Traffic-related air pollution, noise at school, and behavioral problems in Barcelona schoolchildren: a cross-sectional study. Environmental Health Perspectives, 124(4):529–535, 2016. URL https://doi.org/10.1289/ehp.1409449. [8] E. Fuertes, C. Flohr, JI. Silverberg, M. Standl, DP. Strachan, and ISAAC Phase Three Study Group. Global associations between UVR exposure and current eczema prevalence in children from ISAAC phase three. Journal of Investigative Dermatology, 137(6):1248–1256, 2017. URL https://doi.org/10.1016/j.jid.2017.02.966. [9] X. Basagaña, JP. Escalera-Antenaza, P. Dadvand, Ò. Llatje, J. Barrera-Gómez, J. Cunillera, M. Medina-Ramón, and C. Pérez. High ambient temperatures and risk motor vehicle crashes in Catalonia, Spain (2000–2011): A time-series analysis. Environmental Health Perspectives, 123(12):1309–1316, 2015. URL https://doi.org/10.1289/ehp.1409223. Jose Barrera (ISGlobal & UAB) Statistics in Health Sciences, 2023/2024 20 / 20