ArticlePDF Available

Validación del peso e índice de masa corporal auto-declarados de los participantes de una cohorte de graduados universitarios [in Spanish]

Authors:

Abstract

Resumen Objetivos: Valorar la validez del peso y del índice de masa corporal (IMC) auto-declara- dos por los participantes de un estudio prospectivo de cohortes multipropósito. Métodos: El estudio SUN (Seguimiento Universidad de Navarra) es una cohorte formada por gra- duados universitarios reclutados y seguidos mediante cuestionarios enviados por correo. Se comparó el peso y la talla auto-declarados en el cuestionario basal con el peso y talla medidos en una consulta médica después de su declaración en el cuestionario. Los parti- cipantes no conocían en el momento de completar el cuestionario que sus pesos y tallas auto-declarados se compararían con los medidos objetivamente. Resultados: El error re- lativo medio fue de -1,45% (IC 95%: -2,03% a -0,86%) para el peso y de -2,64% (IC 95%: -3,70% a -1,60%) para el IMC. Para la detección de sobrepeso/obesidad (IMC≥ 25 kg/m2) la auto-declaración tuvo una sensibilidad del 90%, una especificidad del 100%, valores pre- dictivos positivo y negativo del 100% y del 93%, respectivamente. El índice kappa de con- cordancia entre sobrepeso/obesidad directamente medido y el auto-declarado fue de 0,91 (IC 95%: 0,81-0,99). El coeficiente de correlación entre el peso medido y declarado fue de 0,991 (IC 95%: 0,986-0,994) y para el IMC de 0,944 (IC 95%: 0,911-0,965). La validez del IMC auto-declarado disminuía al aumentar el peso (p= 0,021) y al disminuir la talla (p= 0,034). Conclusiones: La validez del IMC y del peso auto-declarado por los graduados universitarios de esta cohorte es suficientemente adecuada como para utilizarla en estu- dios epidemiológicos, bien sea como variable cuantitativa de resultado o para ajustar por ella como posible confusor. Summary Objectives: To assess the validity of self-reported weight and Body Mass Index (BMI) in the participants of a multi-purpose prospective cohort. Methods: The SUN Study is a cohort of university graduates, recruited and followed-up using mailed questionnaires. We compared
TRABAJO INÉDITO
352
Maira Bes-Rastrollo
1
, José Ramón Pérez Valdivieso
2
, Almudena Sánchez-Villegas
1,3
, Álvaro Alonso
1
,
Miguel Ángel Martínez-González
1
1
Departamento de Medicina Preventiva y Salud Pública, Facultad de Medicina, Universidad de Navarra;
2
Departamento
de Anestesiología y Reanimación, Clínica Universitaria de Navarra, Facultad de Medicina, Universidad de Navarra;
3
Departamento de Ciencias Clínicas, Facultad de Medicina, Universidad de las Palmas de Gran Canaria.
Correspondencia: Dr. Miguel Ángel Martínez-González. Departamento de Medicina Preventiva y Salud Pública.
Facultad de Medicina. Universidad de Navarra. C/ Irunlarrea, s/n. 31080. Pamplona. Tfno.: +34 948425600, Ext. 6463.
E-mail: mamartinez@unav.es
Validación del peso e índice de masa corporal
auto-declarados de los participantes de una cohorte
de graduados universitarios
Resumen
Objetivos: Valorar la validez del peso y del índice de masa corporal (IMC) auto-declara-
dos por los participantes de un estudio prospectivo de cohortes multipropósito. Métodos:
El estudio SUN (Seguimiento Universidad de Navarra) es una cohorte formada por gra-
duados universitarios reclutados y seguidos mediante cuestionarios enviados por correo.
Se comparó el peso y la talla auto-declarados en el cuestionario basal con el peso y talla
medidos en una consulta médica después de su declaración en el cuestionario. Los parti-
cipantes no conocían en el momento de completar el cuestionario que sus pesos y tallas
auto-declarados se compararían con los medidos objetivamente. Resultados: El error re-
lativo medio fue de –1,45% (IC 95%: –2,03% a –0,86%) para el peso y de –2,64% (IC 95%:
–3,70% a –1,60%) para el IMC. Para la detección de sobrepeso/obesidad (IMC
25 kg/m
2
)
la auto-declaración tuvo una sensibilidad del 90%, una especificidad del 100%, valores pre-
dictivos positivo y negativo del 100% y del 93%, respectivamente. El índice kappa de con-
cordancia entre sobrepeso/obesidad directamente medido y el auto-declarado fue de 0,91
(IC 95%: 0,81-0,99). El coeficiente de correlación entre el peso medido y declarado fue de
0,991 (IC 95%: 0,986-0,994) y para el IMC de 0,944 (IC 95%: 0,911-0,965). La validez del
IMC auto-declarado disminuía al aumentar el peso (p= 0,021) y al disminuir la talla (p=
0,034). Conclusiones: La validez del IMC y del peso auto-declarado por los graduados
universitarios de esta cohorte es suficientemente adecuada como para utilizarla en estu-
dios epidemiológicos, bien sea como variable cuantitativa de resultado o para ajustar por
ella como posible confusor.
Summary
Objectives: To assess the validity of self-reported weight and Body Mass Index (BMI) in the
participants of a multi-purpose prospective cohort. Methods: The SUN Study is a cohort of
university graduates, recruited and followed-up using mailed questionnaires. We compared
04 trabajo inedito 3-6 23/11/05 18:00 Página 352
Validación del peso e índice de masa corporal auto-declarados de los participantes de una cohorte de graduados universitarios 353
Rev Esp Obes 2005; 3 (6): 352-358
Introducción
La creciente epidemia de obesidad y sobrepeso
es el problema de nutrición más importante ac-
tualmente para la salud pública en países desarro-
llados.
1,2
La evidencia epidemiológica y experi-
mental confirma que la obesidad es un importan-
te factor de riesgo de las principales enfermedades
crónicas de gran prevalencia y trascendencia en los
países desarrollados, como son la cardiopatía is-
quémica, la hipertensión arterial, la diabetes me-
llitus tipo 2, entre otras enfermedades.
3-6
Además, en la actualidad se considera que el so-
brepeso y la obesidad han alcanzado una magnitud
epidémica incluso en países en vías de desarrollo
7
y que estarán entre los principales problemas de sa-
lud pública en el siglo XXI.
8
Hoy en día, en nues-
tro país más del 13% de los varones y cerca del 16%
de las mujeres en edades comprendidas entre 25 y
60 años presentan obesidad (índice de masa cor-
poral [IMC] ≥ 30). Además, el 38% de la población
española presenta sobrepeso (IMC entre 25 y 29,99
kg/m
2
).
9
Por otra parte, hay evidencias de una ten-
dencia al aumento.
10
Debido a problemas de factibilidad y limitacio-
nes de recursos, muchos estudios epidemiológicos
intentan evaluar la presencia de sobrepeso u obe-
sidad basándose en información declarada direc-
tamente por el participante, incluso mediante en-
cuestas telefónicas, para disminuir al máximo los
costes.
11
El problema para usar estos datos es que
pueden existir dudas sobre su fiabilidad y validez,
aunque algunos estudios han hallado en nuestro
país buenas correlaciones entre el peso corporal
medido y el peso declarado.
12,13
Al igual que en es-
tudios de cohortes de gran tamaño desarrollados
en EE.UU. y otros países, se están usando habi-
tualmente los datos auto-declarados de los partici-
pantes en estos estudios para valorar el sobrepeso
y la obesidad. De todas formas, sobre este asunto
existe todavía en nuestro país cierto desacuerdo o
debate. Actualmente están poniéndose en marcha
diversas cohortes de gran envergadura en España
y resultaría muy pertinente poder disponer de in-
formación para valorar si el peso auto-declarado (y
el IMC) es suficientemente válido al menos para al-
gunos de los objetivos de estas cohortes, ya que me-
dirlo con carácter longitudinal encarecería todavía
mucho más estos estudios, haciéndolos menos efi-
cientes. Así pues, el objetivo de este trabajo fue va-
lorar si en una cohorte que incluye exclusivamen-
te a graduados universitarios, como es el caso del
estudio SUN (Seguimiento Universidad de Nava-
rra), resultan válidos los datos auto-declarados por
los propios participantes acerca del peso corporal
y del IMC para poder hacer inferencias acerca del
sobrepeso y obesidad.
Sujetos y métodos
El proyecto SUN es un estudio prospectivo de co-
hortes dinámicas (con reclutamiento permanente-
mente abierto) que incluye sólo a personas con es-
tudios universitarios. Son personas con un nivel
educativo alto que se comprometieron a contestar
cuestionarios enviados cada dos años. La cohorte
SUN se puso en marcha a finales de 1999 con el
the self-reported weight and height in the baseline questionnaire with the objectively measu-
red weight and height in a medical visit after the participants filled the questionnaire. Partici-
pants did not know that the self-reported weight and height would be compared with the me-
asured weight and height. Results: The average relative error was –1.45% (95% CI: –2.03%
to –0.86%) for the weight and –2.64% (95% CI: –3.70% to –1.60%) for the BMI. Self-re-
ported BMI showed values for sensitivity of 90%, for specificity of 100%, for positive and ne-
gative predictive values of 100% and 93% respectively, to detect overweight/obesity (BMIε
25 kg/m
2
). The kappa index was 0.91 (95% CI: 0.81 to 0.99). The correlation coefficient bet-
ween self-reported and measured weight was 0.991 (95% CI: 0.986 to 0.994) and it was 0.944
(95% CI: 0.911 to 0.965) for the BMI.Validity of self-reported weight was poorer among par-
ticipants with higher weight (p= 0.021) and lower height (p= 0.034). Conclusions: Self-re-
ported weight and BMI in university graduates is good enough to be used in epidemiologi-
cal studies as a continuous outcome variable or to adjust for it as a potential confounder.
04 trabajo inedito 3-6 23/11/05 18:00 Página 353
354 Bes-Rastrollo
et al.
Rev Esp Obes 2005; 3 (6): 352-358
objetivo de examinar la asociación entre diferentes
estilos de vida, incluyendo factores nutricionales,
y la incidencia de las principales enfermedades cró-
nicas. Algunas publicaciones recientes explican
con mayor detalle los aspectos principales del
SUN.
14-17
Hasta diciembre de 2004, aproximada-
mente 17.000 participantes habían completado la
evaluación basal y 10.000 el primer cuestionario de
seguimiento a dos años.
Es lógico pensar que, si a los participantes en
un estudio de validación se les informa de que su
auto-declaración va a ser después comprobada,
objetivamente, serán más sinceros. Por eso, dise-
ñamos el estudio de validación incluyendo sólo
a participantes que no sabían en el momento de
cumplimentar el cuestionario que se iba a validar
su peso y talla, pero de los que disponíamos de
datos para hacerlo. El presente análisis incluye,
por tanto, a los 70 participantes de la cohorte (eda-
des entre 24 y 81 años), que fueron atendidos en
alguna consulta médica de la Clínica Universita-
ria de Navarra como máximo tres meses después
de contestar el cuestionario del estudio. Estos par-
ticipantes en el momento de completar el cues-
tionario no sabían que se iba a usar su peso y ta-
lla medida en la Clínica Universitaria de Navarra
como validación del peso y el IMC auto-declara-
dos. Por motivos éticos, se usó un procedimiento
para enmascarar la identidad de los participantes
ante quienes comparaban sus datos medidos con
sus datos declarados. Este proceso se desarrolló
de la siguiente forma: una primera persona (A)
preparó un listado con la identificación numéri-
ca y los nombres y apellidos de los participantes,
una segunda persona (B) se encargó de buscar el
peso y la talla medida en consulta y devolver la
lista con los pesos y tallas medidos sin los nom-
bres y apellidos a una tercera persona (C), que fue
la encargada de realizar los análisis de ambos fi-
cheros (datos medidos y datos auto-declarados)
emparejándolos sólo mediante un número de
identificación anónimo.
El IMC fue calculado como el cociente entre pe-
so/talla
2
(kg/m
2
). En el caso del IMC declarado se
sustituyó la ecuación por los valores auto-declara-
dos por los participantes y, para calcular el IMC me-
dido, se usaron los valores determinados objetiva-
mente en la consulta médica (balanza calibrada y
tallímetro estandarizado de pared).
Análisis de datos
Se estimaron los siguientes parámetros: la dife-
rencia entre el peso declarado y el peso medido de
los participantes, el error relativo medio del peso y
del IMC en porcentaje calculado a partir de un co-
ciente, en el que el numerador fue la diferencia en-
tre las variables auto-declaradas y las medidas y el
denominador fue la variable medida. Como indi-
cadores de validez para los índices auto-declarados,
se estimaron la sensibilidad, especificidad y valo-
res predictivos de la clasificación del sobrepeso/obe-
sidad (IMC≥ 25 kg/m
2
) basada en datos declarados.
Se calculó también el índice kappa de concordan-
cia entre la clasificación de sobrepeso/obesidad ba-
sada en datos declarados y el diagnóstico basado
en datos medidos, el índice kappa ponderado con
pesos cuadráticos entre la clasificación de normo-
peso, sobrepeso y obesidad basada en datos decla-
rados y el diagnóstico basado en datos medidos, así
como los coeficientes de correlación de Pearson en-
tre el peso y el IMC declarados frente al peso e IMC
medidos. Se ajustó una regresión lineal múltiple
usando como variable dependiente el error relati-
vo del IMC (%) y las variables edad, sexo, peso y ta-
lla medida como variables independientes.
Se ha representado gráficamente el error relativo
del IMC (%) frente a la media del IMC declarado y
el medido, tal y como propusieron Bland y Altman.
18
Resultados
Al comparar las características de los 70 partici-
pantes incluidos en el estudio con el resto de la co-
horte se puede observar que no existen diferencias
sustanciales que reseñar entre un grupo y otro, sal-
vo una mayor edad en la submuestra de validación
(Tabla 1).
El 54,3% de los participantes eran mujeres, la me-
dia de edad fue de 49,4 años (IC 95%: 46,1 a 52,5).
En la Tabla 2 podemos observar que el peso medio
declarado, 69,3 kg (IC 95%: 66,3 a 72,3), fue infe-
rior al peso medio medido, 70,3 kg (IC 95%: 67,3 a
73,4). El IMC declarado era ligeramente inferior:
24,0 kg/m
2
(IC 95%: 23,2 a 24,8) al IMC medido:
24,7 kg/m
2
(IC 95%: 23,8 a 25,5). Por el contrario,
la talla declarada era 1,08 cm superior a la talla me-
dida en consulta (IC 95%: 0,42 a 1,74). El porcen-
04 trabajo inedito 3-6 23/11/05 18:00 Página 354
taje de sobrepeso/obesidad (IMC≥ 25 kg/m
2
) de la
muestra usando los datos auto-declarados fue del
38,6% (IC 95%: 27,2 a 51,0), mientras que usando
los datos medidos en la consulta médica fue del
42,9% (IC 95%: 31,1 a 55,3).
Se ha observado un mayor error relativo medio
del peso y del IMC en las mujeres (–1,71% y –2,73%,
respectivamente) que en los varones (–1,13% y
–2,53%). De la misma forma, las mujeres infraesti-
maron más su peso (–1,12 kg) que los hombres
(–0,97 kg). En cambio, los hombres sobreestima-
ron más su talla (hombres: +1,32 cm/mujeres: +0,88
cm) e infraestimaron más su IMC (hombres: –0,72
kg/m
2
/mujeres: –0,68 kg/m
2
). No obstante, estas di-
ferencias no fueron estadísticamente significativas
en el análisis univariante.
En la Fig. 1 se representa el gráfico de dispersión
del error relativo del IMC situado en el eje de or-
denadas y la media del IMC medido e IMC decla-
rado en el eje de las abscisas, tal y como propusie-
ron Altman y Bland. Puede observarse que no pre-
senta la forma de embudo, sino que fundamental-
mente recoge variabilidad aleatoria.
En la Tabla 3 se muestra la sensibilidad, especi-
ficidad, el valor predictivo positivo, el valor pre-
dictivo negativo, el índice kappa como medida de
acuerdo del índice de masa corporal declarado pa-
ra la clasificación del sobrepeso/obesidad (IMC≥
25 kg/m
2
) y el índice kappa ponderado con pesos
cuadráticos para la clasificación de normopeso
(IMC< 25 kg/m
2
), sobrepeso (30 kg/m
2
>IMC≥ 25
kg/m
2
) y obesidad (IMC≥ 30 kg/m
2
). La sensibili-
dad obtenida fue del 90% (IC 95%: 74% a 98%), la
especificidad del 100% (IC 95%: 91% al 100%), el
valor predictivo positivo del 100% (IC 95%: 87% al
100%), el valor predictivo negativo del 93% (IC 95%:
81% a 99%), el índice kappa de 0,91 (IC 95%: 0,81 a
0,99) y el índice kappa ponderado con pesos cua-
dráticos de 0,98 (IC
95%: 0,84 a 0,99).
La Tabla 3 también
presenta los índices de
correlación paramétrica
de Pearson del peso y
del IMC para examinar
la asociación entre las
variables declaradas y
las variables medidas.
Para la variable peso, el
coeficiente de correlación encontrado fue de 0,991
(IC 95%: 0,986-0,994) y para la variable IMC fue de
0,944 (IC 95%: 0,911-0,965).
En la Tabla 4 se observan los valores de los coe-
ficientes de la regresión lineal múltiple usando el
error relativo medio del IMC como variable de-
pendiente y la edad, el sexo, el peso medido y la ta-
lla medida como variables independientes. Se ob-
servó que el error relativo medio de las mujeres fue
–1,4% superior al de los hombres, aunque de forma
no significativa (p= 0,40), y este error aumentaba
en un 0,1% por cada kilogramo más de peso medi-
do en los participantes a igualdad de edad, sexo y
talla (p= 0,021). Por el contrario, el error disminuía
0,2% por cada centímetro más de talla de los parti-
cipantes a igualdad de edad, sexo y peso (p= 0,034).
La edad se asociaba a un aumento del error relativo
Tabla 1. Comparación de los valores de las variables de los participantes de la cohorte SUN con los valores de
las variables en la submuestra de participantes del estudio de validación
Variables Cohorte SUN (n= 11.177) Submuestra estudio de validación (n= 70)
Mujeres (%) 56,3 54,3
Edad (años, media, DE) 40,69 (12,73) 49,42 (13,39)
Peso declarado (kg, media, DE) 67,46 (13,62) 69,3 (12,75)
Talla declarada (cm, media, DE) 168,72 (8,65) 169,67 (8,49)
IMC declarado (kg/m
2
, media, DE) 23,54 (3,53) 23,97 (3,31)
No fumadores (%) 42,9 37,1
IMC: índice de masa corporal.
*IMC 25 kg/m
2
.
Figura 1. Error relativo* (%) del peso frente a la media del peso medi-
do y el peso declarado. Método de Altman y Bland.
18
*[(Declarado-medido)/medido] x 100
Error relativo* (%)
Validación del peso e índice de masa corporal auto-declarados de los participantes de una cohorte de graduados universitarios 355
Rev Esp Obes 2005; 3 (6): 352-358
04 trabajo inedito 3-6 23/11/05 18:00 Página 355
356 Bes-Rastrollo
et al.
Rev Esp Obes 2005; 3 (6): 352-358
pero no resultó estadísticamente significativa en el
análisis.
Discusión
Estos resultados sugieren que la va-
lidez del peso auto-declarado en una
cohorte de graduados universitarios
puede ser suficientemente adecuada,
aunque se podría correr el riesgo de in-
fraestimar ligeramente la prevalencia
de sobrepeso/obesidad al dicotomizar
el IMC. Debe tenerse en cuenta que,
al igual que otras cohortes, el estudio
SUN no está concebido para estimar
prevalencias, sino asociaciones pros-
pectivas, al tratarse de un diseño lon-
gitudinal.
19
El peso y el IMC declara-
dos de los participantes serán utiliza-
dos en primer lugar para medir los
cambios que se produzcan en dicha
variable como consecuencia de deter-
minados estilos de vida. Para este objetivo, puede
aceptarse su validez, ya que están altamente corre-
lacionados con el peso e IMC reales de los partici-
pantes, es decir, si el peso o el IMC real aumenta o
disminuye, también se producirá un cambio simi-
Tabla 3. Estimadores de acuerdo, concordancia y correlación del peso y del IMC declarado
(n= 70)
Estimación puntual IC 95%*
Sensibilidad
a
0,90 0,74-0,98
Especificidad
a
1,00 0,91-1,00
Valor predictivo positivo (VPP)
a
1,00 0,87-1,00
Valor predictivo negativo (VPN)
a
0,93 0,81-0,99
Indice kappa
a
0,91 0,81-0,99
Indice kappa ponderado cuadráticamente
b
0,98 0,84-0,99
Coeficiente de correlación (peso)
c
0,991 0,986-0,994
Coeficiente de correlación (IMC)
c
0,944 0,911-0,965
IC 95%: intervalo de confianza al 95%.
IMC: índice de masa corporal (kg/m
2
).
*Para las proporciones se usó el método binomial exacto.
a
Se consideró dicotómicamente el sobrepeso/obesidad (IMC 25 kg/m
2
).
b
Se consideraron 3 categorías:
IMC< 25 (kg/m
2
)
25 (kg/m
2
) IMC < 30 (kg/m
2
)
IMC 30 (kg/m
2
)
c
Se consideró la variable como continua.
Tabla 2. Comparación entre los valores de las variables declaradas con los valores de las variables medidas
Total Hombres Mujeres
n= 70 n= 32 n=38
Declarado Medido Declarado Medido Declarado Medido
(IC 95%) (IC 95%) (IC 95%) (IC 95%) (IC 95%) (IC 95%) p
b
Peso medio (kg) 69,3 70,3 79,1 80,1 61,1 62,2
(66,3 a 72,3) (67,3 a 73,4) (75,2 a 83,0) (76,0 a 84,2) (58,7 a 63,4) (59,7 a 64,6)
Talla media (cm) 169,7 168,6 175,3 174,0 165,0 164,1
(167,7 a 171,7) (166,5 a 170,6) (172,8 a 177,8) (171,4 a 176,6) (162,8 a 167,1) (161,8 a 166,4)
IMC medio (kg/m
2
) 24,0 24,7 25,7 26,4 22,5 23,2
(23,2 a 71,9) (23,8 a 25,5) (24,7 a 26,8) (25,3 a 27,6) (21,6 a 23,5) (22,1 a 24,2)
% Sobrepeso/Obesidad 38,6 42,9 62,5 65,6 18,4 23,7
(IMC 25 kg/m
2
) (27,2 a 51,0) (31,1 a 55,3) (43,7 a 78,9) (46,8 a 81,4) (7,7 a 34,3) (11,5 a 40,2)
ERM del peso (%) –1,45 –1,13 –1,71 0,33
(–2,03 a –0,86) (–1,90 a –0,36) (–2,59 a –0,83)
ERM del IMC (%) –2,64 –2,53 –2,73 0,85
(–3,70 a –1,60) (–4,27 a –0,78) (–4,04 a –1,41)
Diferencia
a
en el peso (kg) –1,05 –0,97 –1,12 0,90
(–1,47 a –0,63) (–0,32 a –1,61) (–1,69 a –0,54)
Diferencia
a
en la talla (cm) +1,08 +1,32 +0,88 0,52
(+0,42 a +1,74) (+0,13 a +2,49) (+0,14 a +1,63)
Diferencia
a
en el IMC (kg/m
2
) –0,70 –0,72 –0,68 0,73
(–0,97 a –0,42) (–0,23 a –1,20) (–0,34 a –1,00)
IC 95%: intervalo de confianza al 95%.
IMC: índice de masa corporal (kg/m
2
).
ERM: error relativo medio ([(declarado-medido)/medido] x 100).
a
Declarado – medido.
b
Significación estadística al comparar las medias en función del sexo.
04 trabajo inedito 3-6 23/11/05 18:00 Página 356
lar en el peso o IMC declarado y no se perderá ca-
si ninguna información relevante al usar datos au-
to-declarados. Nuestros resultados indican que en
análisis longitudinales será preferible usar como
variable dependiente el cambio de peso autode-
clarado, en vez del cambio del IMC.
El peso y el IMC también podrán ser considera-
dos en ocasiones como factores de confusión y ajus-
tar por ellos las estimaciones de riesgos relativos u
otras medidas de asociación. Del mismo modo, se-
rá más apropiado utilizar el IMC como variable
cuantitativa continua (y no dicotomizarla, por ejem-
plo, en obesidad: sí/no) en el momento de ajustar
para limitar la confusión residual.
Como en la mayoría de los estudios publicados
hasta el momento,
12,19-24
las mujeres presentan una
mayor tendencia a infraestimar su verdadero peso,
aunque en este caso de forma no significativa. Del
mismo modo, las personas con valores en el peso
más elevados y/o valores en la talla inferiores fue-
ron las que más imprecisión aportaron en los valo-
res auto-declarados. Este hecho puede ser debido
a un sesgo de deseabilidad social.
El error relativo medio del peso es en términos
absolutos ligeramente inferior al encontrado en
otros estudios.
12,25
Esto puede deberse a que los par-
ticipantes del SUN son voluntarios movidos por un
sentido de altruismo (la participación no es remu-
nerada ni incentivada) y muchos de ellos son pro-
fesionales sanitarios que, además de tener un me-
jor conocimiento de variables sanitarias, proba-
blemente sean más conscientes de la importancia
de la exactitud de sus datos declarados.
La separación en el tiempo de los datos compa-
rados podría ser una limitación del estudio, pues la
consulta médica se realizó como mucho 3 meses
después de la contestación al cuestionario. Pero,
en todo caso, esto llevaría a que parte de la falta de
validez sea sólo aparente, debido a un cambio real
de peso producido en ese tiempo.
La ventaja del diseño que hemos seguido se basa
en que los participantes no fueron avisados de que
se iban a usar sus datos medidos como referente pa-
ra validar sus datos declarados. Esto hace suponer
que la validez de la información no estará artifi-
cialmente elevada. Por el contrario, si a un subgru-
po de la cohorte se le dice después de dar sus datos
de peso declarado que se le va a pesar, será muy po-
sible que quienes sean conscientes de haber in-
fraestimado más su peso sean los que más motivos
tengan para no querer participar en el estudio de
validación. Por otro lado, si los participantes cono-
cen de antemano que después de rellenar el cues-
tionario se les pesará y tallará, muy probablemen-
te serán artificialmente más sinceros y precisos en
sus respuestas que lo que se espera que ocurra ru-
tinariamente en la cohorte. Es por este motivo que
recurrimos al procedimiento de validación que pre-
sentamos, aun a riesgo de reducir el tamaño mues-
tral. No obstante, la limitación inherente a usar un
tamaño muestral de 70 participantes no represen-
ta a nuestro juicio un problema importante, te-
niendo en cuenta que los límites de confianza de
los parámetros estimados son suficientemente es-
trechos. También podría criticarse que la selección
de la submuestra para la validación no ha sido alea-
toria. Sin embargo, hemos recogido el 100% de los
participantes del SUN que han sido atendidos, pe-
sados y tallados en la Clínica Universitaria de Na-
varra en un plazo inferior o igual a 3 meses después
de cumplimentar nuestros cuestionarios. Si hu-
biéramos elegido una submuestra aleatoria de to-
da la cohorte, a la que invitáramos a participar en
el estudio de validación, habría surgido el proble-
ma antes mencionado (participación selectiva en
función de la validez), que introduciría un sesgo
muy preocupante.
A pesar de estas limitaciones, el óptimo valor de
los índices kappa junto con la alta sensibilidad, es-
pecificidad, valores predictivos y las excelentes co-
rrelaciones constatan que puede asumirse una ade-
cuada validez para el peso auto-declarado por los
Tabla 4. Análisis por regresión lineal múltiple usando el error relativo
a
del índice de masa corporal (%) como variable dependiente e introdu-
ciendo simultáneamente las 4 variables independientes presentadas
en la tabla
Coeficiente
de regresión Significación
Variables B (IC 95%) estadística (p)
Edad (años) –0,03 0,52
(–0,12 a –0,06)
Sexo femenino –1,36 0,41
(–4,68 a –1,95)
Peso medido (kg) –0,140 0,021
(–0,26 a –0,02)
Talla medida (cm) +0,180 0,034
(+0,01 a +0,35)
a
[(IMC declarado – IMC medido)/IMC medido] x 100.
IC 95%: intervalo de confianza al 95%.
Validación del peso e índice de masa corporal auto-declarados de los participantes de una cohorte de graduados universitarios 357
Rev Esp Obes 2005; 3 (6): 352-358
04 trabajo inedito 3-6 23/11/05 18:00 Página 357
358 Bes-Rastrollo
et al.
Rev Esp Obes 2005; 3 (6): 352-358
participantes de una cohorte de universitarios es-
pañoles. Esta información directamente propor-
cionada por el participante puede usarse con con-
fianza para los objetivos que antes hemos mencio-
nado: valorar longitudinalmente cambios en el pe-
so y en el IMC, y ajustar por estas mismas variables
las estimaciones epidemiológicas de medidas de
asociación y efecto.
Agradecimientos
Expresamos nuestros agradecimientos a los par-
ticipantes del estudio SUN por su continuada coo-
peración y participación. Al Ministerio de Sanidad
y Consumo (Fondo de Investigaciones Sanitarias,
proyectos PI040233 y G03/140, Red Temática de
Dieta y Enfermedad Cardiovascular). Agradece-
mos a la Fundación “Grupo Eroski” y a la Asocia-
ción “Amigos de la Universidad de Navarra” su apo-
yo con una beca de posgrado.
Otros miembros del estudio SUN son: M. Seguí-
Gómez, J. de Irala, J.A. Martínez-Hernández, R.M.
Pajares, C. de la Fuente, M. Delgado-Rodríguez,
M. Serrano-Martínez, F. Guillén-Grima, I. Agui-
naga, A. Martí, M. Marqués y M. Muñoz.
Bibliografía
1. James PT, Leach R, Kalamara E, Shayeghi M. The worldwide obe-
sity epidemic. Obes Res 2001; 9 (Suppl): S228-33.
2. OMS. Obesity. Preventing and managing the global epidemic. Gi-
nebra: World Health Organization, 1998.
3. Pi-Sunyer FX. Medical hazards of obesity. Ann Intern Med 1993;
119: 655-60.
4. Must A, Spadano J, Coakley EH, Filed AE, Colditz GA, Dietx
WH. The disease burden associated with overweight and obesity.
JAMA 1999; 282: 1523-9.
5. Ernst ND, Obarzanek E, Clark MB, Brieffel RR, Brown CD, Do-
nato K. Cardiovascular health risks related to overweight. J Am
Diet Assoc 1997; 97 (Suppl 7): 47-51.
6. Colditz GA, Willet WC, Rotnitsky A, Manson JE. Weight gain as
a risk factor for clinical diabetes mellitus in women. Ann Intern
Med 1995; 122: 481-6.
7. Anónimo. The catastrophic failures of public health [editorial].
Lancet 2004; 363: 745.
8. OMS. World Health Report, 2003. 1: Global Health: today´s
challenges. [publicado 18-12-2003, citado 24-01-04]. Disponi-
ble en: http://www.who.int/whr/2003/chapter1/en/ (accedido
25-04-05).
9. Aranceta J, Pérez Rodrigo C, Serra Majem L, Ribas Barba L, Qui-
les Izquierdo J, Vioque J, et al. Prevalencia de la obesidad en Es-
paña: resultados del estudio SEEDO 2000. Med Clin (Barc) 2003;
120: 608-12.
10. Gutiérrez-Fisac JL, Banegas Banegas JR, Rodríguez Artalejos F,
Regidor E. Increasing prevalence of overweight and obesity among
Spanish adults, 1987-1997. Int J Obes 2000; 24: 1677-82.
11. Galán I, Rodríguez-Artalejo F, Zorrilla B. Comparación entre en-
cuestas telefónicas y encuestas “cara a cara” domiciliarias en la
estimación de hábitos de salud y prácticas preventivas. Gac Sanit
2004; 18: 440-50.
12. Quiles Izquierdo J, Vioque J. Validez de los datos antropométri-
cos declarados para la determinación de la prevalencia de obesi-
dad. Med Clin (Barc) 1996; 106: 725-9.
13. Galán I, Gandarillas A, Febrel C, Meseguer CM. Validación del
peso y la talla autodeclarados en población adolescente. Gac Sa-
nit 2001; 15: 490-7.
14. Martínez-González MA, Sánchez-Villegas A, De Irala J, Martí A,
Martínez JA. Mediterranean diet and stroke: objectives and de-
sign of the SUN Project. Nutr Neurosci 2002; 5: 65-73.
15. Alonso A, de la Fuente C, Martín-Arnau AM, de Irla J, Martínez
JA, Martínez-González MA. Fruit and vegetable consumption is
inversely associated with blood pressure in a Mediterranean po-
pulation with a high vegetable-fat intake: the Seguimiento Uni-
versidad de Navarra (SUN) Study. Br J Nutr 2004; 92: 311-9.
16. Sánchez-Villegas A, Delgado-Rodríguez M, Martínez-González
MA, de Irala J, for the SUN group. Gender, age, socio-demogra-
phic and life-style factors associated with major dietary patterns
in the Spanish Project SUN (Seguimiento Universidad de Nava-
rra). Eur J Clin Nutr 2003; 57: 285-92.
17. García de Albéniz XA, Guerra-Gutiérrez F, Ortega-Martínez R,
Sánchez-Villegas A, Martínez-González MA. Smoking among a
cohort of Spanish university graduates. The SUN Project. Gac Sa-
nit 2004; 18: 108-17.
18. Bland JM, Altman DG. Comparing methods of measurement:
why plotting difference against standard method is misleading.
Lancet 1995; 346: 1085-7.
19. Spencer EA, Appleby PN, Davey Gk, Key TJ. Validity of self-re-
ported height and weight in 4808 EPIC-Oxford participants. Pu-
blic Health Nutr 2002; 5: 561-5.
20. Rowland LM. Self-reported weight and height. Am J Clin Nutr 1990;
52: 1125-33.
21. Nieto-García FJ, Bush TL, Keyl PM. Body Mass Definitions of
Obesity: Sensitivity and Specificity using self-reported weight and
height. Epidemiology 1990; 1: 146-52.
22. Stewart AL. The reliability and validity of self-reported weight and
height. J Chron Dis 1982; 35: 295-309.
23. Pirie P, Jacobs D, Jeffery R, Hannan P. Distortion in self-reported
height and weight data. J Am Diet Assoc 1981; 78: 601-6.
24. Palta M, Prineas RJ, Berman R, Hannan P. Comparision of self-
reported and measured height and weight. Am J Epidemiol 1982;
115: 223-30.
25. Shapiro JR, Anderson DA. The effects of restraint, gender, and
body mass index on the accuracy of self-reported weight. Int J Eat
Disord 2003; 34: 177-80.
04 trabajo inedito 3-6 23/11/05 18:00 Página 358
... The baseline questionnaire included information on sociodemographic, lifestyle and medical variables. BMI was calculated as weight (in kilograms)/square of height (in meters), using self-reported information, whose accuracy to estimate the BMI has been previously validated in this cohort [23]. Information on physical activity during leisuretime was collected at baseline with a specific questionnaire previously validated in Spain [24]. ...
Article
Full-text available
Purpose The quality, rather than the quantity, of carbohydrate intake may play a major role in the etiology of obesity-related cancers (ORCs). We assessed the association between a previously defined carbohydrate quality index (CQI) and the risk of developing ORCs in the “Seguimiento Universidad de Navarra” (SUN) cohort. Methods A total of 18,446 Spanish university graduates [mean age 38 years (SD 12 years), 61% women, mean BMI 23.5 kg/m² (SD 3.5 kg/m²)], with no personal history of cancer, were followed-up. Baseline CQI was assessed summing quintiles of four previously defined criteria: high dietary fiber intake, low glycemic index (GI), high whole-grain: total-grain carbohydrates ratio and high solid carbohydrates: total carbohydrates ratio. Participants were classified into tertiles of their total CQI. Incident ORCs were confirmed by an oncologist using medical records and by querying the National Death Index blindly to dietary exposures. Results During a median follow-up of 13.7 years, 269 incident cases of ORC were confirmed. A higher CQI was inversely associated with ORC incidence [multivariable-adjusted hazard ratio (HR) for the upper (T3) versus the lowest tertile (T1) of 0.68 (95% CI: 0.47–0.96), p for trend = 0.047]. Particularly, higher dietary fiber intake was inversely associated with ORC, HRT3 vs. T1=0.57 (95% CI 0.37–0.88 p for trend = 0.013). Conclusion In this prospective Mediterranean cohort, an inverse association between a better global quality of carbohydrate intake and the risk of ORCs was found. Strategies for cancer prevention should promote a higher quality of carbohydrate intake.
... The baseline questionnaire collected information on sex, age, sociodemographic characteristics, weight and height, smoking, physical activity, and medical history. Selfreported weight and height were previously validated in a subgroup of our cohort [25]. Physical activity was assessed using the Spanish version of the 17-item Harvard Nurses' Health Study physical activity questionnaire [26]. ...
Article
Full-text available
Purpose Consumption of ultra-processed foods (UPF) has increased despite potential adverse health effects. Recent studies showed an association between UPF consumption and some gastrointestinal disorders. We evaluated the association between UPF consumption and peptic ulcer disease (PUD) in a large Spanish cohort. Methods We conducted a prospective analysis of 18,066 participants in the SUN cohort, followed every two years. UPF was assessed at baseline and 10 years after. Cases of PUD were identified among participants reporting a physician-made diagnosis of PUD during follow-ups. Cases were only partially validated against medical records. Cox regression was used to assess the association between baseline UPF consumption and PUD risk. Based on previous findings and biological plausibility, socio-demographic and lifestyle variables, BMI, energy intake, Helicobacter pylori infection, gastrointestinal disorders, aspirin and analgesic use, and alcohol and coffee consumption were included as confounders.We fitted GEE with repeated dietary measurements at baseline and after 10 years of follow-up. Vanderweele’s proposed E value was calculated to assess the sensitivity of observed associations to uncontrolled confounding. Results During a median follow-up of 12.2 years, we recorded 322 new PUD cases (1.56 cases/1000 person-years). Participants in the highest baseline tertile of UPF consumption had an increased PUD risk compared to participants in the lowest tertile (HR = 1.52, 95% CI: 1.15, 2.00, Ptrend=0.002). The E-values for the point estimate supported the observed association. The OR using repeated measurements of UPF intake was 1.39 (95% CI: 1.03, 1.87) when comparing extreme tertiles. Conclusion The consumption of UPF is associated with an increased PUD risk.
... The baseline questionnaire gathered information about socio-demographics (marital status, educational level), validated anthropometric measures [31], lifestyle habits (smoking status, physical activity or time spent watching television), and medical variables (prevalence of cancer, hypertension, hypertriglyceridemia and hypercholesterolemia, among others). Information about the total amount of energy intake (kcal/day) was also collected. ...
Article
Full-text available
Purpose Vitamin D deficiency has been associated with multiple chronic diseases, including metabolic disorders such as insulin resistance and type 2 diabetes (T2D). The aim of the study was to analyze the association between validated predicted serum vitamin D status and the risk of developing T2D in a large prospective cohort based on a Mediterranean population. Methods The SUN project is a prospective and dynamic Spanish cohort that gathers university graduates who have answered lifestyle questionnaires, including a validated Food Frequency Questionnaire. The association between predicted serum vitamin D and the risk of T2D was assessed through Cox regression models according to quartiles (Q) of predicted vitamin D at baseline. The models were adjusted for potential confounders and sensitivity analyses were performed to ensure the robustness of our findings. Results Our study included a total of 18,594 participants and after a total follow-up of 238,078 person-years (median follow-up of 13.5 years), 209 individuals were diagnosed with incident T2D. We found a significant inverse association between predicted levels of serum vitamin D and the risk of developing T2D, after adjusting for potential confounders and performing different sensitivity analyses (hazard ratio Q4 vs. Q1: 0.48, 95% CI 0.26–0.88; p for trend = 0.032). Conclusion The outcomes suggest that higher levels of vitamin D at baseline may be associated with a reduced risk of developing T2D.
... Information about non-dietary lifestyle factors (i.e., smoking, alcohol intake, and physical activity) was collected from the baseline questionnaire. A previously validated questionnaire was used to evaluate physical activity [27] and selfreported weight and height measurements were validated in a cohort subgroup [28]. BMI was calculated by dividing weight by height squared (kg/m 2 ). ...
Article
Full-text available
Purpose We aimed to prospectively investigate the association of an overall oxidative balance score (OBS) with all-cause death and cause-specific mortality among participants in the Seguimiento Universidad de Navarra (SUN) Study, a Mediterranean cohort of Spanish graduates. Methods Using baseline information on 12 a priori selected dietary and non-dietary lifestyle pro- and antioxidants exposures—vitamins C and E, β-carotenes, selenium, zinc, heme iron, polyphenols, total antioxidant capacity, body mass index, alcohol, smoking, and physical activity—we constructed an equally weighted OBS categorized into quartiles, with higher scores representing greater antioxidant balance. Cox proportional hazards models were fitted to evaluate the association between the OBS and mortality. Results A total of 18,561 participants (mean [SD] age, 38.5 [12.4] years; 40.8% males) were included in the analysis. During a median follow-up of 12.2 years (interquartile range 8.3–14.9), 421 deaths were identified, including 80 deaths from cardiovascular disease (CVD), 215 from cancer, and 126 from other causes. After adjustment for potential confounders, the hazard ratios and 95% confidence interval (CIs) between the highest quartile (predominance of antioxidants) vs. the lowest quartile (reference category) were 0.35 (95% CI 0.22–0.54, P-trend < 0.001) for all-cause mortality, 0.18 (95% CI 0.06–0.51, P-trend = 0.001) for CVD mortality, 0.35 (95% CI 0.19–0.65, P-trend = 0.002) for cancer mortality, and 0.45 (95% CI 0.20–1.02, P-trend = 0.054) for other-cause mortality. Conclusion Our findings suggest a strong inverse association between the OBS and all-cause, CVD, and cancer mortality. Individuals exposed to both antioxidant dietary and lifestyle factors may potentially experience the lowest mortality risk. Study registry number Dynamic Mediterranean Prospective Cohort: the SUN Project; NCT02669602. https://clinicaltrials.gov/ct2/show/NCT02669602. https://proyectosun.es
... At baseline, information from demographic and lifestyle factors, including age, sex, educational level, smoking, physical activity, alcohol intake, dietary intakes, body weight, height, and health status -such as use and time of hormonal therapy for menopause, menopausal status, or family history of breast cancer-were collected and updated throughout follow-up [25,26]. Age at menopause was updated in the questionnaire after 18 years of follow-up. ...
Article
Objective Dietary patterns may have a greater influence on human health than individual foods or nutrients and they are also of substantial interest in the field of breast cancer prevention. Beyond the adequate balance of macronutrients, evidence shows that the quality of macronutrient sources may play an important role in health outcomes. We sought to examine the relationship between a healthful and an unhealthful low-fat dietary pattern in relation to breast cancer. Methods We used an observational data from a Mediterranean cohort study (the SUN project). We prospectively followed 10,930 middle-aged women initially free of breast cancer during a median follow-up of 12.1 years. We calculated an overall, an unhealthful, and a healthful low-fat-diet score proposed by Shan et al., based on a previously validated 136-item FFQ and grouped participants into tertiles. Incident breast cancer –overall and stratified by menopausal status– was the primary outcome. It was self-reported by participants and confirmed based on medical reports or consultation of the National death index. We used multivariable Cox regression models adjusted for potential confounders. Results During 123,297 person-years of follow-up, 150 cases of incident breast cancer were confirmed. No significant associations were observed for overall or premenopausal breast cancer. For postmenopausal women, we observed a significant association for moderate adherence to the unhealthful LFD score and postmenopausal breast cancer (comparing tertile 2 vs. tertile 1, HR= 2.18; 95% confidence interval, 1.15–4.13) Conclusion : In conclusion, no clear associations were observed although more research is needed to address the association between an unhealthful dietary pattern and postmenopausal breast cancer risk.
... NCDs such as diabetes, cancer, and cardiovascular events were confirmed by medical reports. Self-reported measurements and diagnosis at baseline and during the follow-up were previously validated in a subsample of the cohort and have been found reliable (18)(19)(20)(21). ...
Article
Full-text available
Background and aims Front-of-Pack (FoP) nutrition labelling has been established as a policy, empowering consumers to choose healthy food options for preventing diet-related non-communicable diseases. This study aimed to evaluate the association between the nutrient profile underlying the Chilean warning label score and all-cause mortality and to conduct a calibration with the Nutri-Score in a large cohort of Spanish university graduates. Materials and methods This prospective cohort study analysed 20,666 participants (8,068 men and 12,598 women) with a mean (standard deviation) age of 38 years (±12.4) from the SUN cohort. Dietary food intake was assessed by a validated semi-quantitative food-frequency questionnaire at baseline and after 10 years of follow-up. The warning label score was calculated by considering the threshold of nutrients (sugar, saturated fat, and sodium) and energy density per 100 g/ml of product, as established by Chilean Legislation. Participants were classified according to quartiles of consumption of daily label score: Q1 (≤5.0), Q2 (>5.0–7.1), Q3 (>7.1–9.8), and Q4 (>9.8). Time-dependent, multivariable-adjusted Cox models were applied. To compare the performance of the warning label score and Nutri-Score to predict mortality, we used the Akaike information criterion (AIC) and Bayesian information criterion (BIC) methods. Results During a median of 12.2 years of follow-up, 467 deaths were identified. A higher score in the warning label values (lower nutritional quality) was associated with an increased risk of all-cause mortality [HR (95% CI) Q4 vs. Q1: 1.51 (1.07–2.13); p -trend = 0.010] and cancer mortality [HR (95% CI) Q4 vs. Q1: 1.91 (1.18–3.10); p -trend = 0.006]. However, no statistically significant association was found for cardiovascular mortality. Furthermore, the warning label score and Nutri-Score exhibited comparable AIC and BIC values, showing similar power of prediction for mortality. Conclusion A diet with a higher warning label score (>9.8 per day) was a good predictor of all cases and cancer mortality in a large Spanish cohort of university graduates. Also, the warning label score was capable to predict mortality as well as the Nutri-Score. Our findings support the validity of the warning label score as a FoP nutrition labelling policy since it can highlight less healthy food products.
... The baseline questionnaire collected information on sociodemographic, lifestyle, and medical variables. The self-reported accuracy of height and weight to estimate the BMI has been previously validated in this cohort [24]. Physical activity was also assessed through a validated questionnaire [25]. ...
Article
Full-text available
Obesity is associated with a higher risk of several types of cancer, grouped as obesity-related cancers (ORC). Vitamin D deficiency is more prevalent in obese subjects, and it has been suggested to play a role in the association between obesity and cancer risk. The aim of the study was to analyze the association between vitamin D intake and the subsequent risk of ORC in a prospective Spanish cohort of university graduates. The SUN Project, initiated in 1999, is a prospective dynamic multipurpose cohort. Participants answered a 556-item lifestyle baseline questionnaire that included a validated food-frequency questionnaire. We performed Cox regression models to estimate the hazard ratios (HRs) of ORC according to quartiles of energy-adjusted vitamin D intake (diet and supplements). We included 18,017 participants (mean age = 38 years, SD = 12 years), with a median follow-up of 12 years. Among 206,783 person-years of follow-up, we identified 225 cases of ORC. We found no significant associations between vitamin D intake and ORC risk after adjusting for potential confounders: HRQ2vsQ1 = 1.19 (95% CI 0.81-1.75), HRQ3vsQ1 = 1.20 (95% CI 0.81-1.78), and HRQ4vsQ1 = 1.02 (95% CI 0.69-1.51). Dietary and supplemented vitamin D do not seem to be associated with ORC prevention in the middle-aged Spanish population.
... Level of adiposity was calculated using self-reported weight and height, which were then used to calculate body mass index (BMI) using the standard formula and reported in kg/m 2 . In order to demonstrate the reliability of self-reported weight and height data, a validation study was conducted in 2005, in which it was concluded that the data were of sufficient reliability [28]. Smoking status was included as a categorical variable identifying current, former and non-smokers; missing data were imputed and accounted for <1% of the sample. ...
Article
Full-text available
Healthy and unhealthy lifestyles are tightly linked to general health and well-being. However, measurements of well-being have failed to include elements of health and easy to interpret information for patients seeking to improve lifestyles. Therefore, this study aimed to create an index for the assessment of general health and well-being along with two cut-off points: the lifestyle and well-being index (LWB-I). This was a cross-sectional analysis of 15,168 individuals. Internally valid multivariate linear models were constructed using key lifestyle features predicting a modified Short Form 36 questionnaire (SF-36) and used to score the LWB-I. Categorization of the LWB-I was based on self-perceived health (SPH) and analyzed using receiver operating characteristic curve analysis. Optimal cut-points identified individuals with poor and excellent SPH. Lifestyle and well-being were adequately accounted for using 12 lifestyle items. SPH groups had increasingly healthier lifestyle features and LWB-I scores; optimal cut-point for poor SPH were scores below 80 points (AUC: 0.80 (0.79, 0.82); sensitivity 75.7%, specificity 72.3%)) and above 86 points for excellent SPH (AUC: 0.67 (0.66, 0.69); sensitivity 61.4%, specificity 63.3%). Lifestyle and well-being were quantitatively scored based on their associations with a general health measure in order to create the LWB-I along with two cut points.
Article
Tobacco and alcohol co‐use are two major lifestyle modifiable risk factors. Understanding the determinants of both behaviors helps to develop interventions to prevent these exposures. However, previous studies have focused on predictors of individual tobacco or alcohol use. This study aims to explore the potential predictors of tobacco and alcohol co‐use among Spanish university graduates from the “Seguimiento Universidad de Navarra” (SUN) cohort study. A total of 7175 participants who were co‐users of tobacco and alcohol were selected for this cross‐sectional analysis. Their mean age was 39.1 years (12.04 SD) and 57.3% were women. Univariate regression models were used to select the potential predictors of tobacco and alcohol co‐use, and the areas under the ROC curves (AUC) were calculated. Multivariable logistic regression models were used to create a predictive model. Baseline potential predictors included sociodemographic factors, lifestyle habits, and perceived personality aspects. In the multivariable model, the main significant potential predictors of tobacco and alcohol co‐use were driving under the influence of alcohol (odds ratio [OR] = 1.65 [1.43–1.90]), drinking 1–2 cups of coffee daily (OR = 1.50 [1.24–1.84]), and drinking three or more cups of coffee daily (OR = 1.61 [1.35–1.91]) when compared with the reference group. Conversely, those who were married (OR = 0.87 [0.75–0.99], ate at home 7 days a week (OR = 0.69 [0.60–0.80]), did more physical activity than recommended (OR = 1.18 [1.02–1.34]), or had a high perceived level of competitiveness (OR = 0.83 [0.72–0.95]) had a lower risk of co‐use (AUC 0.61 [confidence interval 95% 0.59–0.63]), compared to the reference group. These results could be used by healthcare professionals, especially nurses, to effectively assess patients at higher risk of tobacco and alcohol co‐use.
Article
Several studies have tried to analyse the association between all-cause mortality and different risk factors, (especially those which are modifiable, such as smoking, diet or exercise), to develop public health preventive strategies. However, a specific analysis of predictors of premature and late mortality is needed to give more precise recommendations. Considering that there are risk factors which exert an influence on some diseases and not on others, we expect that, similarly, they may have a different impact depending on the timing of mortality, separating premature (≤65 years) from late mortality (>65 years). Thus, we prospectively followed-up during a median of 12 years a cohort of 20,272 university graduates comprising an ample range of ages at inception. Time-dependent, covariate-adjusted Cox models were used to estimate adjusted hazard ratios (HR) and their 95 % confidence intervals (CI) for each predictor. The strongest independent predictor of mortality at any age was physical activity which was associated with reduced risk of total, premature and late mortality (range of HRs when comparing the highest vs. the lowest level: 0.24 to 0.48). Specific strong predictors for premature mortality were smoking, HR: 4.22 (95 % CI: 2.42–7.38), and the concurrence of ≥2 metabolic conditions at baseline, HR: 1.97 (1.10–3.51). The habit of sleeping a long nap (≥30 min/d), with HR: 2.53 (1.30–4.91), and poor adherence to the Mediterranean Diet (≤3 points in a 0 to 8 score vs. ≥6 points), with HR: 2.27 (1.08–4.76), were the strongest specific predictors for late mortality. Smoking, diet quality or lifestyles, probably should be differentially assessed as specific predictors for early and late mortality. In the era of precision medicine, this approach will allow tailored recommendations appropriate to each person's age and baseline condition.
Article
Full-text available
Objective: To examine the relation between adult weight change and the risk for clinical diabetes mellitus among middle-aged women. • Design: Prospective cohort study with follow-up from 1976 to 1990. • Setting: 11 U.S. states. • Participants: I'll 281 female registered nurses aged 30 to 55 years who did not have diagnosed diabetes mellitus, coronary heart disease, stroke, or cancer in 1976. • Outcome Measures: Non-insulin-dependent diabetes mellitus. • Results: 2204 cases of diabetes were diagnosed during 1.49 million person-years of follow-up. After adjustment for age, body mass index was the dominant predictor of risk for diabetes mellitus. Risk increased with greater body mass index, and even women with average weight (body mass index, 24.0 kg/m 2) had an elevated risk. Compared with women with stable weight (those who gained or lost less than 5 kg between age 18 years and 1976) and after adjustment for age and body mass index at age 18 years, the relative risk for diabetes mellitus among women who had a weight gain of 5.0 to 7.9 kg was 1.9 (95% CI, 1.5 to 2.3). The corresponding relative risk for women who gained 8.0 to 10.9 kg was 2.7 (CI, 2.1 to 3.3). In contrast, women who lost more than 5.0 kg reduced their risk for diabetes mellitus by 50% or more. These results were independent of family history of diabetes. • Conclusion: The excess risk for diabetes with even modest and typical adult weight gain is substantial. These findings support the importance of maintaining a constant body weight throughout adult life and suggest that the 1990 U.S. Department of Agriculture guidelines that allow a substantial weight gain after 35 years of age are misleading.
Article
Full-text available
Objective This study examines the influence of the interview method (telephone or face-to-face in households) on the assessment of health behaviors and preventive practices. Material and method The same questionnaire was completed by two independent samples of the population aged 18-64 years living in the municipality of Madrid. One sample (n = 1,391 subjects) completed the questionnaire by telephone interview and the other (n = 739) by face-to-face interview in households. The results of the two samples for 28 variables related to anthropometry, physical activity, food consumption, tobacco and alcohol use, preventive practices and injuries were compared. Results The telephone sample had a higher rate of failed contact (31.8% vs. 22.2%) but a greater degree of cooperation than the sample for the face-to-face interview (83.0% vs. 74.0%). In total, 19 of the 28 variables showed a relative variation of less than 10% between the two surveys; the differences found were between 10 and 20% for eight variables and were higher than 20% for one variable. Differences were statistically significant for only four variables (sedentary leisure time, consumption of vegetables, giving up smoking and cholesterol measurement), with a relative variation of 6.1% (p < 0.01), 10% (p < 0.001), 36.7% (p < 0.01) and 8.6% (p < 0.01), respectively. The total cost of the telephone interview was half that of the face-to-face household interview. Conclusions The results of both surveys were very similar. Because of its lower cost, the telephone interview is a good option in public health research when data collection by interview is required.
Article
Introduction The aim of this study was to examine the relationship between self-reported population in order to evaluate the validity of self-reported measures for the purpose ofestimating the prevalence of low (≤ 15th percentile) and high (≥ 85th percentile) body mass index (BMI) in the study population. Subjects and method Information on self-reported and objective weight and height was obtained from a representative sample of 3,244 adolescents, aged 15-18 years, in secondary education schools in the Autonomous Community of Madrid. We calculated the mean relative error; the correlation between subjective and objective parameters, sensitivity, specificity and predictive value positive of low and high BMIs. Results The mean relative errors were as follows: weight: +0.07% for males versus and –0.79% for females; height, +0.51% for males versus +0.98% for females; BMI: –0.88% for males versus –2.63% for females. The correlation between self-reported and objective BMI was 0.87 for males and 0.90 for females. The prevalence of high BMI was underestimated by 34.1% and 34.4% of females white that of low BMI was overestimated by 10.7% of males and 14.8% of females. Conclusion Analysis of BMI as a continuous variable, based on self-reported weight and height measurement data, entails a small margin of error. However, its use as a categorical variable involves a considerable underestimate of the prevalence of high BMI, and an smaller overestimate of the prevalence of low BMI.
Article
. The protocol used in each survey was in accordance to the recommen- dations of the Spanish Society for the study of Obesity (SEEDO) to estimate the prevalence of obesity in population studies. RESULTS: The prevalence of obesity in Spanish adult population was 14.5% (95% CI, 13.93-15.07%), significantly higher among women 15.75% (95% CI, 14.89-16.61%), than men 13,39% (95% CI, 11.84-14.94%) (χ2 = 12.470; p = 0.000). Prevalence of obesity significantly increased with age in men and women. The highest rates were estimated for the age group older than 55 years, both among males and females, 21.58% (95% CI, 18.68-24.48%) and 33.9% (95% CI, 32.73-35.07%), res- pectively. CONCLUSION: Obesity is a health problem which affects an important proportion of the Spanish adult
Article
Purpose: The Mediterranean diet has been postulated as a protective factor against different diseases including stroke. Thus, an epidemiological study in a Mediterra- nean country, such as Spain, focused on diet may offer new insights of the potential benefits of this nutritional pattern to prevent the onset of cerebrovascular diseases. Methods: The SUN ("Seguimiento Universidad de Navarra") project is a prospective study among Spanish university alumni, aimed to identify the dietary determinants of stroke, coronary disease and other disorders. Two pilot studies have been developed. The first pilot study was focused on the understanding of the questionnaire. The second study used a random sample to assess the response proportion and the feasibility of using a mailing system for following-up the cohort. The first informative results are expected to be available after the first four years of following-up the cohort (2005). Here, we report the description of the baseline diet of the first participants in the cohort using data from 1587 men and 2260 women. Results: The outcome of our pilot studies ensure the feasibility of a mail-based cohort. In the baseline assessment, we found a high consumption of olive oil (18.5 g/person/day), red wine (28.8 g/person/day), legumes (102.5 g/person/day), vegetables (507.8 g/person/ day) and fruits (316.7 g/person/day), with a great between-subject variability. Also, the values for cereals (170.4 g/person/day), dairy products (239.3 g/person/day) and meat and meat products consumption (186 g/person/ day) in this cohort were estimated. The coefficients of variation ranged in women from 56 (for vegetables) to 240% (for red wine) and in men from 62 to 180% (for these same two items), reflecting a wide heterogeneity in the diet of participants. Conclusions: Although the participation was not high (22% according to the estimates of the pilot study), it was comparable to the proportion found in large previous cohorts such as the Nurses-II Health Study (24%). The sharp contrast in dietary habits between the US and Spain together with the high between-subjects varia- bility we have found in our Spanish cohort provides an exceptional opportunity to assess the aspects of the Mediterranean diet, which may be protective against stroke and other neurological disorders.
Article
A high correlation between continuous measures of self-reported and measured weight and height has led investigators to infer that self-report of these variables is appropriate in epidemiologic studies. We evaluated the sensitivity and specificity of categorical definitions of body mass defined using self-reported height and weight on 7,455 adult participants of the Lipid Research Clinics Family Study (1975-1978) on whom both self-reported and measured height and weight were available. The categorical definition of obesity used here was a body mass index of at least 30 kg/m2. Overall, the sensitivity of the obese category when defined with self-reported weight and height was 74% (95% CI = 72%-76%), and the specificity was 99%. The sensitivities of the categories defined using self-reported measures varied considerably by sex, age, and educational level. Overall, the sensitivities were higher for women compared with men, as men were less accurate in reporting height. The sensitivity of the categorical definitions of obesity decreased with increasing age and education in both men and women. The sensitivity for elderly obese men was below 50%. These results suggest that substantial misclassification can occur when self-reported information is used to define body mass categories.
Article
The error in self-reported weight and height compared with measured weight and height was evaluated in a nationally representative sample of 11,284 adults aged 20-74 y from the second National Health and Nutrition Examination Survey of 1976-1980. Although weight and height were reported, on the average, with small errors, self-reported weight and height are unreliable in important population subgroups. Errors in self-reporting weight were directly related to a person's overweight status--bias and unreliability in self-report increased directly with the magnitude of overweight. Errors in self-reported weight were greater in overweight females than in overweight males. Race, age, and end-digit preference were ancillary predictors of reporting error in weight. Errors in self-reporting height were related to a person's age--bias and unreliability in self-reporting increased directly with age after age 45 y. Overweight status was also a predictor of reporting error in height.
Article
Screening data from the Hypertension Detection and Follow-up Program in Minneapolis, MN, 1973-1974, provided an opportunity to evaluate the accuracy of self-report of height and weight. It was found that both were reported, on the average, with small but systematic errors. Large errors were found in certain population subgroups. Also, men and women differed somewhat in their pattern of misreporting. Weight was understated by 1.6% by men and 3.1% by women, whereas height was overstated by 1.3% by men and 0.6% by women. As in previous studies, it was found that the most important correlates of the amount of error were the actual measurements of height and weight. An interesting finding was that misreporting of both height and weight in men was correlated with both aspects of body size, whereas for women, it was related mainly to the characteristic in question. Certain other demographic variables, such as age and educational level, were also found to have some importance as factors influencing misreporting.