Biol Trace Elem Res DOI 10.1007/s12011-016-0712-5

Integrative Multivariate Logistic Regression Analysis of Risk Factors for Kashin-Beck disease Fang-fang Yu 1 & Huan Liu 1 & Xiong Guo 1

Received: 19 February 2016 / Accepted: 18 April 2016 # Springer Science+Business Media New York 2016

Abstract To determine the current evidence on risk factors for Kashin-Beck disease (KBD) using an integrative metaanalysis. We searched five English and three Chinese databases from inception to September 2015, to identify casecontrol studies that examined risk factors for KBD using multivariate logistic analysis. DerSimonian and Laird effective models are applied in processing data using pooled odds ratios (ORs) and 95 % confidence intervals (CI). Seven studies were identified with 3087 cases and 6402 controls. The main risk factors found to be significantly associated with the onset of KBD were age (OR 1.19, 95 % CI 1.10–1.28), parents prevalence (OR 5.16, 2.51–7.80), family hygiene (OR 1.68, 1.42– 1.93), food source (OR 3.29, 2.38–4.19), wheat (OR 1.12, 1.08–1.16), wheat germ necrosis rate (OR 6.03, 1.87–12.92), total volatile basic nitrogen (OR 6.85, 1.01–28.67), low selenium in hair (OR 2.29, 1.08–3.50) were found to be significant risks factors. The pooled ORs (95 % CI) of protein intake and rice were 0.79 (0.66–0.93) and 0.90 (0.86–0.95), respectively, indicating that the two factors may be protective for KBD. We found that the combination of low protein intake, polluted grain, and selenium deficiency may contribute to be onset of KBD together.

Fang-fang Yu and Huan Liu contributed equally to this work and should be considered co-first authors. * Xiong Guo [email protected] 1

Institute of Endemic Diseases, School of Public Health of Health Science Center, Key Laboratory of Trace Elements and Endemic Diseases, National Health and Family Planning Commission, Xian Jiaotong University, No. 76 Yan Ta West Road, Xi’an, Shaanxi 710061, People’s Republic of China

Keywords Kashin-Beck disease . Selenium deficiency . Protein intake . Polluted grain

Introduction Kashin-Beck disease (KBD) was an endemic, progressive, and degenerative osteoarthropathy of unknown etiology. KBD is distributed in specific geographic areas ranging from northeastern Siberia to southwest of China [1–3]. Up to 2012, there were more than 642,000 patients with KBD in 378 endemic counties and a population of 37,917,000 individuals at risk for KBD in China according to the BChina Health and Family Planning Statistical Yearbook 2013^ [4]. The environmental risk factors that result in the deep cartilage cell death of KBD have been known for a long time, and more than 50 environmental risk factors have been proposed [4, 5]. Some risk factors have been phased out with increased understanding of the etiology of KBD. After extensive study of environment factors hypothesis, we concluded that the risk factors of KBD mainly concentrated in three environmental factors. One factor is environmental geochemistry, represented by selenium and iodine deficiency [6–9]. Another factor is poisoning water by organic matter and humic acid [10, 11]. The third factor is poisoning food by mycotoxins, represented by T-2 toxin [12–14]. Currently, a multifactorial model considering the interactions of the multiple environmental and genetic factors has been developed for KBD. It is likely that KBD has a multifactorial origin. We did this meta-analysis to determine the validity of evidence on environmental risk factors for the occurrence of KBD. This review focused on non-clinical risk factors, such as general demographic characteristics, diet, and the three hypotheses. We summarized the risk factors to provide clues for

Yu et al.

further study of the pathogenesis of KBD and to suggest strategies for clinical prevention and treatment.

Methods Search Strategy and Study Selection We followed the guidelines of the Meta-analysis of Observational Studies in Epidemiology (MOOSE) [15]. Databases searched included Medline, ISI Web of Science, Google scholar, Elsevier, Springer Link, Chinese National Knowledge Infrastructure (CNKI), Wan Fang data, and Chinese Science and Technique Journals Database (VIP) from inception to September, 2015. The search strategy used the keywords BKashin-Beck disease^ or BKBD^ or Blogistic regression^ or Brisk factor^ and Bcase-control study.^ Publications of logistic regression and case-control studies both in English and Chinese were reviewed. In addition, we also searched the references of reviews to identify additional articles. Two authors screened all the records independently using the aforementioned search strategy and then assessed the final results by discussion. The study selection flowchart is shown in Fig. 1. Eligibility Criteria The eligibility criteria were as follows: (i) The subjects were KBD patients in endemic areas who met the diagnostic standard. (ii) The objective was to explore the external environment relative to KBD with similar exposure. (iii) The experimental design was case-control study using multivariate Fig. 1 Flow diagram of included studies

logistic regression analysis (Table 1). (iv) The raw data provided the odds ratios (OR) and 95 % confidence intervals (CI), or it could be calculated. We excluded studies that investigated the association between genetic and environmental risk factors of KBD. Additionally, repetitive research, poor-quality studies, and ecological studies with no individual data were excluded. Data Extraction Data extraction was performed independently by two investigators, and in the case of discrepancies, the final decision was discussed and resolved by consensus. The following information was extracted from each eligible article: author, publication year, risk factors, characteristics of study population including mean age, sex, number of cases or controls, OR, and 95 % CI (Table 2). The risk factors included age, gender, education, parents prevalence, family hygiene, family income, residential humidity, storage conditions, protein intake/rice, food source, fresh vegetables, wheat/corn/bean, humic acid, chemical oxygen demand (CON), wheat germ necrosis rate (WGNR), total volatile basic nitrogen (TVBN), selenium in hair. Meta-analysis was performed for the three categories of risk factors where a sufficient number of studies reported findings. The pooled OR and 95 % CI of general demographic characteristics, diets, and the three main hypotheses were calculated. Statistical Analysis Data analysis was performed with StataSE version 11.0 (STATA Corp, College Station, TX, USA). We used OR and

Risk Factors for KBD Table 1

Primary information of multivariate logistic analysis of risk factors associate with KBD

No.

Author(year)

Area

Cases

Controls

Risk factors

1

Xu et al.(2013) [18]

Szechwan

2 3

Zhang et al.(2009) [19] Yu et al.(2009) [20]

Shaanxi Henan

147 133 212 15

813 1292 212 15

Gender, age, family income, parents prevalence, education, protein intake, saple food, fresh vegetables, food source Family income, staple food, food source Family hygiene, protein intake, staple food, food source

4

Ping et al.(2006) [21]

Shaanxi

1959

2977

Family hygiene, protein intake, staple food, food source

5 6

Suetens et al.(2001) [22] Zhou et al.(1990) [23]

Lhasa Gansu

280 136

295 171a/215b

7

Yang et al.(1989) [24]

Heilongjiang

205

204a/208b

Gender, age, family hygiene, staple food, food source Family income, family hygiene, humic acid, COD, WGNR, TVBN, hair selenium Family income, family hygiene, protein intake, staple food, food source, granary humidity, storage conditions, humic acid, COD, WGNR, TVBN, selenium in hair

Abbreviations: COD chemical oxygen demand, WGNR wheat germ necrosis rate, TVBN total volatile basic nitrogen a

Controls in KBD area

b

Controls in non-KBD area

95 % CI as the main effect size. The indicator of risk factors for OR was processing by the use of the DerSimonian and Laird method [16]. The DerSimonian and Laird method for meta-analysis is the random-effect model, and the method is based on the inverse variance approach, making an adjustment to the study weights according to the extent of variation, or heterogeneity, among the varying treatment effects. The principal analysis was the pooling of individual pooled ORs. Summary data were sourced from the trial publication or estimated from the logistic regression. To investigate for statistical heterogeneity between trials, the standard chi-squared Qtest was applied (p < 0.10) [17]. The pooled of OR and 95 %

CI and heterogeneity test were calculated by the Stata11.0. Firstly, data transformation: the different type of data transformed the OR and 95 % CI. If the literature only reported beta (β) and 95 % CI, the effect size (ES) = lnOR = β, the standard error of effect size = [ln (lower limit) –ln (upper limit)] / 3.92. Secondary, heterogeneity test: If the I2 < 50 % was considered as homogeneity between the multiple similar studies, then the fixed-effects model was selected. Conversely, the I2 ≥ 50 % was considered as a larger heterogeneity; therefore, randomeffect model was selected. Thirdly, the pooled OR and 95 % CI were output, and Z test was used to test the pooled statistic test (P < 0.05).

Table 2 Definition of variables Risk factors

Assignment

Age Gender

1 = male, 2 = female

Education Parents prevalence

1 = illiteracy, 2 = primary school, 3 = middle school, 4 = senior school 1 = yes, 2 = no

Family hygiene Family income Residential humidity Storage conditions Protein intake/Rice Food source

1 1 1 1 1 1

Fresh vegetables Wheat/corn/bean Humic acid Chemical oxygen demand

1 = frequently, 2 = occasionally 1 = no, 2 = yes >0.4 mg/l >2.19 mg/l

Wheat germ necrosis rate Total volatile basic nitrogen Selenium in hair

>10 % >0.067 mg/kg