Hypertension is a clinical condition characterized by high blood pressure (BP). The reduction of high blood pressure is an effective strategy to decrease the CV risk. Usually a drug therapy with antihypertensives (AHs) is necessary to reduce BP levels and the main techniques to measure BP in clinical practice are i) Office and ii) 24-hour ambulatory blood pressure monitoring (ABPM). The office measurement consists in carrying out a single BP measurement in the clinical setting. This technique could not show the true BP level due to the white coat effect (WCE). For the ABPM method, the patient wears blood pressure equipment continuously for 24 hours. ABPM offers several advantages, for example, it is not modified by the WCE. In the last years many randomized clinical studies aimed to evaluate the effects of AH drugs in reducing BP used indifferently these techniques, while the choice of which tool to use has relevant implications. The effectiveness of AH treatment in the prevention of CV events can be evaluated only after a long follow-up period. In order to avoid long trials, it has been suggested to assess the treatment effectiveness considering organ damages as surrogate endpoint. The organs most frequent damaged by hypertension are i) brain, ii) heart and iii) kidney. The objectives of this thesis were to quantify the BP changes (i.e. the difference of the measure of BP between baseline and end of follow-up) measured with both Office and ABPM techniques on the same patients treated with AH and to investigate the association between AH use and risk of dementia and left ventricular hypertrophy (CV surrogate endpoints). In particular, in the study on BP measurements a meta-analysis of randomized clinical trials with correlated measures was carried out and the summary estimates were calculated through linear mixed models. In one of the studies on organ damage the adjustment for unmeasured confounder was taken into account because meta-analysis based on observational studies. Regarding the LMM, a macro SAS was created to calculate meta-analytic estimates, choosing between the fixed and random effects model for both correlated and non-correlated data. Instead, the problem of unmeasured confounder was addressed with two approach: i) Bayesian approach through a method proposed by McCandless and ii) Monte Carlo Sensitivity Analysis (MCSA). Another macro SAS is made for the application of MCSA method. Finally, the issue of the optimization of the process of identifying the papers to include in the meta-analysis was faced. This process usually is very expensive in terms of time and the probability to mistake something is not low. The meta-analysis on BP reduction was based on 52 RCTs and the estimate of interest was the difference between Office and ABPM reduction. The results shows significant difference in the changes of BP levels when measured with Office or ABPM techniques. This difference seems to be at least partly accountable by a placebo-effect and to largely result from the regression of the mean phenomenon, though a WCE cannot be excluded. Regarding the meta-analysis on the use of ACE-inhibitors and risk of dementia, the results seem to show that the treatment helps to preserve cognitive function. It is possible that ACE drug’s effects is nonspecific and related more generally to lowering arterial hypertension, and thereby reducing silent vascular events and cerebrovascular events. Instead, the meta-analysis on reduction of left ventricular mass due to antihypertensive drugs shows that all antihypertensive drug classes have the same effect on the reduction of the left ventricular mass index even if there are actually few observational studies available on this issue. The problem of the optimization of papers’ search was addressed by designing a prototype based on machine-learning method.

L'ipertensione è una condizione clinica caratterizzata da un’elevata pressione arteriosa (PA). Solitamente è necessaria una terapia farmacologica con antipertensivi (AI) per ridurre i livelli di PA. Le principali tecniche utilizzate per misurare la PA sono la tecnica Office e il monitoraggio ambulatoriale della pressione arteriosa nelle 24 ore (ABPM). La tecnica Office prevede di effettuare una singola misurazione in un ambiente clinico. Questa tecnica potrebbe però non mostrare il vero livello di PA a causa dell'effetto del camice bianco. L'ABPM è una tecnica in cui la misurazione della PA viene registrata continuamente per 24 ore. Questo strumento, al contrario del precedente, non è influenzato dall’ “effetto camice bianco”. Negli ultimi anni sono stati condotti differenti studi clinici (RCT) volti a valutare gli effetti dei farmaci AI nel ridurre la PA. Questi studi hanno usato indifferentemente le due tecniche anche se la scelta di quale strumento utilizzare può avere rilevanti implicazioni sui risultati. L'efficacia del trattamento con AI nella prevenzione degli eventi cardiovascolari può essere valutata solo dopo un lungo periodo di follow-up. Per evitare lunghi studi, è stato suggerito di valutare l'efficacia del trattamento considerando il danno d'organo come evento surrogato. Questa tesi ha l’obiettivo di quantificare, in pazienti trattati con AI, i cambiamenti di PA (fine del follow-up - baseline) misurati con le due tecniche sugli stessi pazienti e di indagare l'associazione tra utilizzo di antipertensivi e il rischio di due importanti danni d’organo: demenza e ipertrofia ventricolare sinistra. Tutti gli obiettivi di questa tesi sono stati affrontati attraverso un approccio meta-analitico. In particolare, nello studio sulle misurazioni della PA è stata condotta una meta-analisi di RCT con misure correlate e le stime riassuntive sono state calcolate mediante modelli lineari misti (MLM). In uno degli studi sul danno d’organo, l'aggiustamento per il fattore di confondimento non misurato è stato preso in considerazione in quanto la meta-analisi era basata su studi osservazionali. Per quanto riguarda il MLM, è stata creata una macro SAS che consente di calcolare le stime meta-analitiche scegliendo tra il modello ad effetti fissi o casuali per dati correlati e non. Invece, il problema del confondente non misurato è stato affrontato con due approcci: i) approccio bayesiano e ii) Monte Carlo Sensitivity Analysis (MCSA). Un'altra macro SAS è creata per l’applicazione del metodo MCSA. Infine, è stato affrontato il tema dell'ottimizzazione del processo di identificazione degli articoli da includere nella meta-analisi. La meta-analisi sulla riduzione della PA era basata su 52 RCT e la stima dell'interesse era la differenza tra la riduzione di Office e ABPM. I risultati mostrano una differenza significativa tra cambiamenti dei livelli di PA misurati con le tecniche di Office e ABPM. Questa differenza sembra essere almeno in parte dovuta all’effetto placebo e in gran parte alla regressione verso la media, sebbene non sia possibile escludere una componente dell’effetto camice bianco. Per quanto riguarda la meta-analisi sull’utilizzo di ACE inibitori e il rischio di demenza, i risultati supportano la probabilità che il trattamento potrebbe aiutare a preservare la funzione cognitiva. È possibile che l’effetto degli ACE non sia specifico e correlati con l'abbassamento della PA e quindi alla riduzione degli eventi vascolari silenti ed eventi cerebrovascolari. Invece, la meta-analisi sulla riduzione dell’indice di massa ventricolare sinistra dovuta a all’uso di antipertensivi mostra che le diverse classi di farmaci sembrano avere medesimi effetti anche se sono disponibili pochi studi osservazionali. Il problema dell'ottimizzazione della ricerca degli articoli è stato affrontato progettando un prototipo basato sul metodo di machine-learning.

(2018). STUDY OF SEVERAL ISSUES RELATED TO HYPERTENSION THROUGH CLASSICAL AND MODERN META-ANALYTICAL METHODS. (Tesi di dottorato, Università degli Studi di Milano-Bicocca, 2018).

STUDY OF SEVERAL ISSUES RELATED TO HYPERTENSION THROUGH CLASSICAL AND MODERN META-ANALYTICAL METHODS

SORANNA, DAVIDE
2018

Abstract

Hypertension is a clinical condition characterized by high blood pressure (BP). The reduction of high blood pressure is an effective strategy to decrease the CV risk. Usually a drug therapy with antihypertensives (AHs) is necessary to reduce BP levels and the main techniques to measure BP in clinical practice are i) Office and ii) 24-hour ambulatory blood pressure monitoring (ABPM). The office measurement consists in carrying out a single BP measurement in the clinical setting. This technique could not show the true BP level due to the white coat effect (WCE). For the ABPM method, the patient wears blood pressure equipment continuously for 24 hours. ABPM offers several advantages, for example, it is not modified by the WCE. In the last years many randomized clinical studies aimed to evaluate the effects of AH drugs in reducing BP used indifferently these techniques, while the choice of which tool to use has relevant implications. The effectiveness of AH treatment in the prevention of CV events can be evaluated only after a long follow-up period. In order to avoid long trials, it has been suggested to assess the treatment effectiveness considering organ damages as surrogate endpoint. The organs most frequent damaged by hypertension are i) brain, ii) heart and iii) kidney. The objectives of this thesis were to quantify the BP changes (i.e. the difference of the measure of BP between baseline and end of follow-up) measured with both Office and ABPM techniques on the same patients treated with AH and to investigate the association between AH use and risk of dementia and left ventricular hypertrophy (CV surrogate endpoints). In particular, in the study on BP measurements a meta-analysis of randomized clinical trials with correlated measures was carried out and the summary estimates were calculated through linear mixed models. In one of the studies on organ damage the adjustment for unmeasured confounder was taken into account because meta-analysis based on observational studies. Regarding the LMM, a macro SAS was created to calculate meta-analytic estimates, choosing between the fixed and random effects model for both correlated and non-correlated data. Instead, the problem of unmeasured confounder was addressed with two approach: i) Bayesian approach through a method proposed by McCandless and ii) Monte Carlo Sensitivity Analysis (MCSA). Another macro SAS is made for the application of MCSA method. Finally, the issue of the optimization of the process of identifying the papers to include in the meta-analysis was faced. This process usually is very expensive in terms of time and the probability to mistake something is not low. The meta-analysis on BP reduction was based on 52 RCTs and the estimate of interest was the difference between Office and ABPM reduction. The results shows significant difference in the changes of BP levels when measured with Office or ABPM techniques. This difference seems to be at least partly accountable by a placebo-effect and to largely result from the regression of the mean phenomenon, though a WCE cannot be excluded. Regarding the meta-analysis on the use of ACE-inhibitors and risk of dementia, the results seem to show that the treatment helps to preserve cognitive function. It is possible that ACE drug’s effects is nonspecific and related more generally to lowering arterial hypertension, and thereby reducing silent vascular events and cerebrovascular events. Instead, the meta-analysis on reduction of left ventricular mass due to antihypertensive drugs shows that all antihypertensive drug classes have the same effect on the reduction of the left ventricular mass index even if there are actually few observational studies available on this issue. The problem of the optimization of papers’ search was addressed by designing a prototype based on machine-learning method.
ZAMBON, ANTONELLA
antihypertensive; macroSAS; dementia; LVM; meta-analysis
antihypertensive; macroSAS; dementia; LVM; meta-analysis
MED/01 - STATISTICA MEDICA
English
7-mar-2018
SANITA' PUBBLICA - 78R
30
2016/2017
open
(2018). STUDY OF SEVERAL ISSUES RELATED TO HYPERTENSION THROUGH CLASSICAL AND MODERN META-ANALYTICAL METHODS. (Tesi di dottorato, Università degli Studi di Milano-Bicocca, 2018).
File in questo prodotto:
File Dimensione Formato  
phd_unimib_713384.pdf

Accesso Aperto

Descrizione: tesi di dottorato
Tipologia di allegato: Doctoral thesis
Dimensione 2.93 MB
Formato Adobe PDF
2.93 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/10281/199089
Citazioni
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
Social impact