Central
Medical Journal of Obstetrics and Gynecology
Cite this article: Ohkuchi A (2013) Biological Interaction as a Useful Tool to Understand Complicated Relationship among Multiple Risk Factors. Med J
Obstet Gynecol 1(1): 1001.
*Corresponding author
Akihid e O hkuc hi, De p a rtm e nt o f O b ste tric s a nd
G yne c o lo g y, Jic hi Me d ic a l Unive rsity Sc ho o l o f
Me d ic ine , 3311-1 Ya kushiji, Shim o tsuke -shi, To c hig i
329-0498, Japan, Tel: 81-285-58-7376; Fax: 81-285-44-
8505; Ema il: o kuc hi@ jic hi.a c .jp
Submitte d: 23 August 2013
Accepted: 24 August 2013
Publishe d: 26 August 2013
Copyright
© 2013 Ohkuc hi
OPEN ACCESS
Editorial
Biological Interaction as a
Useful Tool to Understand
Complicated Relationship
among Multiple Risk Factors
Akihide Ohkuchi*
Department of Obstetrics and Gynecology, Jichi Medical University School of
Medicine, Japan
IntroductIon
Preeclampsia
Preeclampsia (PE) is a multifactorial disease [1]. A systematic
review elucidated that there are several risk factors for PE [2].
In 52 selected studies, antiphospholipid antibodies, pre-existing
diabetes, previous PE, family history of PE, nulliparity, multiple
pregnancies, raised body mass index (BMI) at booking or before
pregnancy, systolic blood pressure (SBP) ≥130 mmHg at booking,
diastolic blood pressure (DBP) ≥80 mmHg at booking, and age ≥40
were significant risks for PE [2]. Recent multivariate analysis of
a prospective multicenter cohort of nulliparous women revealed
that clinical risk factors at 14-16 weeks of gestation were age,
mean arterial blood pressure (MAP), BMI, family history of PE,
family history of coronary heart disease, maternal body weight,
and vaginal bleeding for at least five days; and factors associated
with reduced risk were a previous single miscarriage with the
same partner, taking at least 12 months to conceive, high intake
of fruit, cigarette smoking, and alcohol use in the first trimester;
however, the area under the receiver operation characteristics
curve (AUC) was only 0.71; moreover, the addition of uterine
artery Doppler (UAD) indices did not improve performance [3].
Thus, a multivariate model for predicting all PE using maternal
factors and UAD may not be promising for clinical use.
Biological interaction
High BP and abnormal UAD findings in the second trimester
are well-known risk factors for the subsequent occurrence of PE,
[4,5] especially earlier onset of PE [6-8]. We recently found that
high BP and abnormal UAD had a strong biological interaction
for the occurrence of early-onset PE (EO-PE) with onset at <34
weeks of gestation [9]. In 2410 pregnant women, the adjusted
odds ratio (95% CI) in women with abnormal UAD alone, high
BP alone, and both high BP and abnormal UAD for predicting EO-
PE was 4.3 (0.37-49), 12 (2.6-55) and 85 (17-422), respectively;
and that for predicting late-onset PE (LO-PE) was 6.3 (1.5-27),
6.1 (2.1-17) and 15 (3.6-61), respectively. The relative excess risk
due to biological interaction (RERI) between the two risk factors
for EO-PE in the logistic regression model was calculated as 70;
therefore, we judged that there was a very strong synergistic
interaction between high BP and abnormal UAD for EO-PE risk;
and the RERI between the two risk factors for LO-PE in the
logistic regression model was calculated as 3.3; therefore, there
was a weak synergistic interaction for LO-PE risk. These results
suggested that high BP and abnormal UAD synergistically interact
for the genesis of EO-PE.
Biological interaction is a relatively unfamiliar, but clinically
important phenomenon. The degree of biological interaction is
measured between risk factors as the deviation from additivity
by the corresponding disease rates, not as deviation from
multiplicity [10]. In order to specify the models, let i = 1 when
the first risk factor is present and 0 otherwise, and let j = 1 when
the second risk factor is present and 0 otherwise; furthermore,
let RRij be relative risk in exposure category i, j. Thus, RR
11
, RR
10
,
RR
01
, and RR
00
are the relative risks for each of the four categories.
We also define those unexposed to both the first and second risk
factors as the reference category, i.e., RR
00
= 1. The three relative
risk estimates other than RR
00
could obtained from a logistic
regression model or Cox regression model [10].
The relative excess risk due to biological interaction (RERI) is
defined as follows: [10]
RERI = RR
11
− RR
10
− RR
01
+ 1
If there is no biological interaction, RERI is equal to 0; if
there is positive biological interaction, RERI is >0; and if there
is negative biological interaction, RERI is <0 [10]. In our recent
study, weak, moderate, and strong biological interactions were
defined when the absolute values of RERI were 2.0-4.9, 5.0-9.9,
and ≥10 [9].
To calculate biological interaction using multivariate analyses,
we should perform one invention to enter variables into the
models. RR
11
, RR
10
, RR
01
, and RR
00
are created by the combination
of two dichotomous variables as a series of dummy variables. For
example, using SPSS, dummy variables can be easily generated
using a built-in function.