Contribuciones del metanálisis a la mejora de la salud y del método científico / Contribution of meta-analysis to scientific method and health care

El metanálisis es fruto del paradigma inductista de investigación: ofrece un análisis del principio de consistencia para una asociación causal (segundo principio de Hill) y es una herramienta intermedia en el análisis de decisiones. No está exento de inconvenientes: la hipótesis que investiga se comprueba generalmente con la misma información que la genera y los estudios que se ponderan no son formalmente independientes, en la medida que la metodología influye en los resultados y los estudios del presente mejoran los del pasado. Ha motivado el desarrollo de cuestionarios y protocolos de evaluación de diferentes tipos de diseños, principalmente ensayos clínicos. Todo esto fue uno de los detonantes de la llamada medicina basada en la ®evidencia¼. Ha impulsado la creación de agencias de evaluación de tecnologías sanitarias y contribuido a la mejora de la salud. En el campo de la metodología de investigación ha profundizado en el estudio del sesgo de publicación y la búsqueda de variables que influyen en la discrepancia entre los estudios individuales. En la actualidad, se metanalizan efectos secundarios y adversos, muchas veces objetivos secundarios de los estudios originales, lo que ocasiona riesgos al combinar estudios con insuficiente potencia estadística y que comunican irregularmente esos resultados.

Meta-analysis can be considered a result of the application of the induction model of thinking given that it allows athorough analysis of the consistency of an association(2nd principle of Hill). Besides, its usefulness as anintermediate tool for decision analysis has also beenproved. However, meta-analysis has some drawbacks:(a) the hypothesis under analysis is usually proved usingsome of the data that generated it; (b) the studies to be assessed are not formally independent as research methods influence on results and the methodology of current studies is better than that of previous ones. Meta analysishas prompted the development of evaluationprotocols for different designs, mainly clinical trials. It has favoured the emergence and spreading of evidence based medicine, thus contributing to the developmentof agencies for technology assessment in health care. As regards to methodology, it has improved the knowledgeon publication bias and determinants of lack ofconsistency among primary studies. Currently, meta analysisis also applied to the study of secondary outcomes(e.g., adverse effects). This practice is risky due to the combination of studies with poor statistical data of these effects, apart from the fact that they report results inadequately.

Alcohol drinking as a predictor of intensive care and hospital mortality in general surgery: a prospective study.

AIMS: To analyse whether alcohol drinking increases admission to intensive care and in-hospital mortality in general surgery. DESIGN AND PARTICIPANTS: A prospective cohort study on a consecutive series of 1505 hospitalized patients in a Service of General Surgery of a tertiary hospital. MEASUREMENTS: Drinking pattern was defined by quantity, frequency and volume of drinking. Information on relevant confounders was obtained: smoking, body mass index, nutritional status (measured by serum albumin), cholesterol and its fractions, severity of the underlying disease and all therapeutic measures. Multivariate logistic regression was applied to assess the relationship between drinking and both admission to intensive care and in-hospital death. RESULTS: Twenty-nine (1.9%) patients died and 33 (2.1%) were admitted to the intensive care unit (ICU). Drinking was heavier in men, patients without antecedents of cancer, with lower preoperative risk assessment scores, number of co-morbidities and age and higher serum albumin levels. After adjusting for age, severity of underlying disease, smoking and serum albumin, male drinkers of 72+ g/day had an increased risk of being admitted to ICU, the effect being stronger for week-day drinking (odds ratio, OR = 8.48; 95% confidence interval, CI = 1.68-42.8). A significant association was also seen between week-day drinking (72+ g/day) and death in men (OR = 7.19, 95% CI = 1.43-36.1). Numbers for women were too small to evaluate. CONCLUSION: Heavy drinking increases admission to intensive care and in-hospital mortality in hospitalized male patients undergoing general surgery procedures.

A prospective study of tobacco smoking as a predictor of complications in general surgery.

OBJECTIVE: To analyze whether tobacco smoking is related to nosocomial infection, admission to the intensive care unit, in-hospital death, and length of stay. DESIGN: A prospective cohort study. SETTING: The Service of General Surgery of a tertiary-care hospital. PATIENTS: A consecutive series of patients admitted for more than 1 day (N = 2,989). RESULTS: Sixty-two (2.1%) patients died and 503 (16.8%) acquired a nosocomial infection, of which 378 (12.6%) were surgical site and 44 (1.5%) were lower respiratory tract. Smoking (mainly past smoking) was associated with a worse health status (eg, longer preoperative stay and higher American Society of Anesthesiologists score). A long history of smoking (> or = 51 pack-years) increased postoperative admission to the intensive care unit (adjusted odds ratio [OR] = 2.86; 95% confidence interval [CI95], 1.21 to 6.77) and in-hospital mortality (adjusted OR = 2.56; CI95, 1.10 to 5.97). There was no relationship between current smoking and surgical-site infection (adjusted OR = 0.99; CI95, 0.72 to 1.35), whereas a relationship was observed between past smoking and surgical-site infection (adjusted OR = 1.46; CI95, 1.02 to 2.09). Current smoking and, to a lesser degree, past smoking augmented the risk of lower respiratory tract infection (adjusted OR = 3.21; CI95, 1.21 to 8.51). Smokers did not undergo additional surgical procedures more frequently during hospitalization. In the multivariate analysis, length of stay was similar for smokers and nonsmokers. CONCLUSION: Smoking increases in-hospital mortality, admission to the intensive care unit, and lower respiratory tract infection, but not surgical-site infection. Deleterious effects of smoking are also observed in past smokers and they cannot be counteracted by hospital cessation programs.

Cholesterol and serum albumin levels as predictors of cross infection, death, and length of hospital stay.

HYPOTHESIS: The levels of cholesterol, its fractions (high-density lipoprotein cholesterol [HDL-C] and low-density lipoprotein cholesterol [LDL-C]), and serum albumin reflect nutritional status and are related to in-hospital death, nosocomial infection, and length of stay in the hospital. DESIGN: A prospective cohort study of hospitalized patients. SETTING: The Service of General Surgery of a tertiary hospital. PATIENTS: A consecutive series of 2989 patients admitted for more than 1 day. MAIN OUTCOME MEASURES: Nosocomial infection, in-hospital death, and length of stay. RESULTS: During follow-up, 62 (2%) of the patients died, 382 (13%) developed a nosocomial infection, and 257 (9%) developed a surgical site infection. Serum albumin (lowest quintile vs highest quintile: adjusted odds ratio [OR], 1.9; 95% confidence interval, 1.2-2.9) and HDL-C (lowest quintile vs highest quintile: OR, 2.0; 95% confidence interval, 1.3-3.0) levels showed an inverse and highly significant relationship with nosocomial infection (mainly due to surgical site infection) in crude and multivariate analyses (controlling for the Study on the Efficacy of Nosocomial Infection Control [SENIC] index, the American Society of Anesthesiologists' score, cancer, and age). Regarding total and LDL-C levels, only their lowest quintiles increased the risk of nosocomial infection. Serum albumin and HDL-C levels showed an inverse trend (P<.001) with mortality, with high multivariate-adjusted ORs in the lowest quintile (serum albumin: OR, 5.8; 95% confidence interval, 0.8-44.6; HDL-C: OR, 7.2; 95% confidence interval, 0.9-55.0), whereas no trend was appreciated with other cholesterol fractions or ratios. Serum albumin, HDL-C, and LDL-C levels showed independent, significant (P<.001), and inverse relationships with length of stay. CONCLUSION: The levels of serum albumin and cholesterol fractions, mainly HDL-C, which are routinely measured at hospital admission, are predictors of in-hospital death, nosocomial infection, and length of stay.