Prevalence of non Helicobacter pylori species in patients presenting with dyspepsia.

BACKGROUND: Helicobacter species associated with human infection include Helicobacter pylori, Helicobacter heilmannii and Helicobacter felis among others. In this study we determined the prevalence of H. pylori and non-Helicobacter pylori organisms H. felis and H. heilmannii and analyzed the association between coinfection with these organisms and gastric pathology in patients presenting with dyspepsia. Biopsy specimens were obtained from patients with dyspepsia on esophagogastroduodenoscopy (EGD) for rapid urease test, histology and PCR examination for Helicobacter genus specific 16S rDNA, H. pylori phosphoglucosamine mutase (glmM) and urease B (ureB) gene of H. heilmannii and H. felis. Sequencing of PCR products of H. heilmannii and H. felis was done.
RESULTS: Two hundred-fifty patients with dyspepsia were enrolled in the study. The mean age was 39 ± 12 years with males 162(65%). Twenty-six percent (66 out of 250) were exposed to cats or dogs. PCR for Helicobacter genus specific 16S rDNA was positive in 167/250 (67%), H. pylori glmM in 142/250 (57%), H. heilmannii in 17/250 (6%) and H. felis in 10/250 (4%), respectively. All the H. heilmannii and H. felis PCR positive patients were also positive for H. pylori PCR amplification. The occurrence of coinfection of H. pylori and H. heilmannii was 17(6%) and with H. felis was 10(4%), respectively. Only one out of 66 exposed to pets were positive for H. heilmannii and two for H. felis. Histopathology was carried out in 160(64%) of 250 cases. Chronic active inflammation was observed in 53(56%) (p = 0.001) of the patients with H. pylori infection alone as compared to 3(37%) (p = 0.73) coinfected with H. heilmannii and H. pylori and 3(60%) coinfected with H. felis and H. pylori (p = 0.66). Intestinal metaplasia was observed in 3(3%)(p = 1.0) of the patients with H. pylori infection alone as compared to 2(25%) (p = 0.02) coinfected with H. heilmannii and H. pylori and 1(20%) coinfected with H. felis and H. pylori (p = 0.15).
CONCLUSION: The prevalence of H. heilmannii and H. felis was low in our patients with dyspepsia. Exposure to pets did not increase the risk of H. heilmannii or H. felis infection. The coinfection of H. pylori with H. heilmannii was seen associated with intestinal metaplasia, however this need further confirmation.

Background

Helicobacter species infect the gastrointestinal tracts of many animals from birds through humans. Some of these have been linked to a range of human diseases [1,2] including chronic gastritis, peptic ulcer disease, mucosa-associated lymphoid tissue lymphoma, and gastric adenocarcinoma [1,3]. The principal Helicobacter infection in humans is Helicobacter pylori, with infection rates in developing countries reaching 50% to 90% [2,4]. Human gastric biopsy samples, however, have shown to harbor bacteria which were morphologically different from H. pylori [5,6]. These include Helicobacter heilmannii and Helicobacter felis which are primarily pathogens of domestic animals and were later found to infect humans as well [7-9].

Gastric non-Helicobacter pylori helicobacters constitute a diverse group of bacterial species that are known to colonize the gastric mucosa of several animals [10]. These include morphologically distinct, typically long spiral shaped bacteria originally referred to as Gastrospirillum hominis and later as H. heilmannii. The latter was further subdivided in two taxa, types 1 and 2 [10]. H. heilmannii type 1 are identical to H. suis which colonizes the stomachs of pigs. The former H. heilmannii type 2 represent a group of species, known to colonize the gastric mucosa of dogs and cats and include H. felis, H. bizzozeronii, H. salomonis, H. cynogastricus, H. baculiformis and a bacterium provisionally named in 2004 as ''Candidatus H. heilmannii'' because at that time, it could not be cultured in vitro [10,11]. However, recently, in vitro cultures have been obtained resulting in description of H. heilmannii, as a novel Helicobacter species [12]. Sequencing of the 16S or 23S rRNA-encoding genes allows differentiation of H. suis from the other gastric non-H. pylori helicobacters species, but it cannot distinguish between H. felis, H. bizzozeronii, H. salomonis, H. cynogastricus, H. baculiformis and Candidatus H. heilmannii [10]. For differentiation between these species, sequencing of the heat shock protein 60 (hsp60) or gyrase B (gyrB) gene is used while sequencing of the urease A and B genes is considered to be the most suitable method since sequences of these genes are available [10,11,13,14].

Dyspepsia describes a variety of symptoms, including abdominal pain, bloating, nausea, and vomiting. In these patients, endoscopy is considered to rule out gastroesophageal reflux disease, peptic or duodenal ulcer and gastric cancer. The role of H. pylori infection in dyspepsia remains controversial. This study aims to identify the prevalence of H. pylori and non-H. pylori helicobacters, H. felis and H. heilmannii and to analyze the gastric pathology associated with coinfection of these organisms in patients presenting with dyspepsia.

Results and discussion

Majority of the patients with H. pylori infection were in the age range of 18-39 years, while H. felis and H. heilmannii positive patients did not show this distribution. (Table 1). There was no difference in the gender, ethnicity of patients, crowding index (CI) and source of water distribution among the patients with H. pylori and non-H. pylori infections (Table 1). All patients had abdominal pain with endoscopic gastritis as the predominant finding. The false positive and false negative results obtained with RUT were 15(36%) and 6(12%), respectively while with histology the false positive and false negative results obtained were 20(30%) and 10(11%), respectively (Table 1-2).

Table 1

Demography and clinical features of patients enrolled

	PCR for H. pylori			PCR for H. heilmannii			PCR for H. felis
	Positive n = 142	Negative n = 108	P value	Positive n = 17	Negative n = 233	P value	Positive n = 10	Negative n = 240	P value
Age
18-39 years	81(57)	55(51)		8(47)	128(55)		6(60)	130(55)
40-55 years	53(37)	35(32)	0.02	7(41)	81(35)	0.82	3(30)	85(35)	0.93
56-75 years	8(6)	18(17)		2(12)	24(10)		1(10)	25(10)
Gender
Male	98(69)	64(59)	0.11	11(65)	151(65)	0.99	5(50)	157(65)	0.33
Female	44(31)	44(41)		6(35)	82(35)		5(50)	83(35)
Ethnicity
Karachiite	36(25)	36(33)		4(24)	68(29)		2(20)	70(29)
Quetta resident	47(36)	31(25)	0.15	5(29)	73(31)	0.81	3(30)	75(31)	0.75
Afghan	55(39)	45(42)		8(47)	92(40)		5(50)	95(40)
Crowding Index (CI)
0-1(low)	59(41)	39(36)		5(29)	93(40)		5(50)	93(39)
2-4 (moderate)	82(58)	62(57)	0.03	12(71)	132(57)	0.35	5(50)	139(58)	0.59
> 4 (crowding)	1(1)	7(7)		0(0)	8(3)		0(0)	8(3)
Water supply
Municipal	86(61)	59(55)	0.34	8(47)	137(59)	0.34	7(70)	138(58)	0.52
Boring Water	56(39)	49(45)		9(53)	96(41)		3(30)	102(42)
EGD
Gastritis	136(96)	106(96)	0.47	17(100)	225(97)	1	10(100)	232(97)	1
Duodenal ulcer	6(4)	2(4)		0(0)	8(3)		0(0)	8(3)
Rapid Urease test (n = 90)
Positive	42(88)	15(36)	> 0.001	4(44)	53(63)	0.28	2(40)	55(63)	0.35
Negative	6(12)	27(64)		5(56)	28(37)		3(60)	30(37)
Histopathology (n = 160)
Bacterial density
Occasional	10(11)	40(61)		3(37)	47(31)		0(0)	50(32)
Few in some fields	59(63)	21(32)	> 0.001	2(25)	78(51)	0.27	4(80)	76(49)	0.14
Only 1/2 small clusters	25(27)	5(7)		3(38)	27(18)		1(20)	29(19)
Inflammation type
Chronic	41(44)	47(71)	0.001	5(63)	83(55)	0.73	2(40)	86(55)	0.66
Chronic active inflammation	53(56)	19(29)		3(37)	69(45)		3(60)	69(45)
Lymphoid follicles
Positive	14(15)	8(12)	0.73	0(0)	22(14)	0.30	0(0)	22(14)	0.47
Negative	80(85)	58(88)		8(100)	130(86)		5(100)	133(86)
Intestinal metaplasia
Positive	3(3)	2(3)	1.0	2(25)	3(2)	0.02	1(20)	4(3)	0.15
Negative	91(97)	64(97)		6(75)	149(98)		4(80)	151(97)

Univariate analysis was performed by using the independent sample t-test, Pearson Chi-square or Fisher Exact test where appropriate. A P-value of < 0.05 was considered as statistically significant. *All the H. heilmannii and H. felis PCR positive patients were also positive for H. pylori PCR amplification.

Table 2

PCR results for Helicobacter species

	PCR for Helicobacter genus specific 16SrRNA
	Positive n = 167	Negative n = 83	P value
H. pylori glmM
Positive	133(80)	9(11)	< 0.001
Negative	34(20)	74(89)
H. heilmannii ureB
Positive	17(10)	0(0)	0.003
Negative	150(90)	83(100)
H. felis ure A and B
Positive	10(6)	0(0)	0.03
Negative	157(94)	83(100)

Univariate analysis was performed by using the independent sample t-test, Pearson Chi-square test or Fisher Exact test where appropriate. A P-value of < 0.05 was considered as statistically significant. *All the H. heilmannii and H. felis PCR positive patients were also positive for H. pylori PCR amplification.

PCR for Helicobacter genus specific 16S rDNA was positive in 167/250 (67%), glmM (H. pylori) in 142/250 (57%), H. heilmanii in 17/250 (6%) and H. felis in 10/250 (4%), respectively (Table 2).

PCR was positive for both H. pylori and H. heilmannii in 17(6%) and for H. pylori and H. felis in 10(4%), respectively (Table 2). All the H. heilmannii and H. felis positive patients were also positive for H. pylori glmM PCR amplification (Table 2).

26% (66 out of 250) were exposed to pets either cats or dogs. Most H. heilmannii positive patients did not have pet contact. Only one out of 66 exposed to pets was positive for H. heilmannii and two for H. felis (Table 3).

Table 3

Association of Helicobacter species with pets

	PCR for H. pylori			PCR for H. heilmannii			PCR for H. felis
	Positive n = 142	Negative n = 108	P value	Positive n = 17	Negative n = 233	P value	Positive n = 10	Negative n = 240	P value
Pets
Yes	42(30)	24(22)	0.19	1(6)	65(28)	0.05	2(20)	64(27)	1
No	100(70)	84(78)		16(94)	168(72)		8(80)	176(73)

Univariate analysis was performed by using the independent sample t-test, Pearson Chi-square test or Fisher Exact test where appropriate. A P-value of < 0.05 was considered as statistically significant. *All the H. heilmannii and H. felis PCR positive patients were also positive for H. pylori PCR amplification.

A higher degree of bacterial density was associated with H. pylori infection alone (p < 0.001) (Table 1). Chronic active inflammation was observed in 53(56%) cases with H. pylori alone infection (p = 0.001) compared to 3(37%) in H. heilmannii (p = 0.73) and 3(60%) in H. felis positive patients coinfected with H. pylori (p = 0.66) (Table 1). Intestinal metaplasia (IM) was present in 3(3%) out of 94 cases with H. pylori infection alone compared to 2(25%) out of 8 cases of H. Heilmannii and H. pylori coinfection, and 1(20%) out of 5 cases of H. felis and H. pylori coinfection in which histology has been performed.

PCR product sequences were compared to the sequences of ureaseB of different H. heilmannii and H. felis strains. The H. heilmannii sequences had 100% similarity to 'Candidatus Helicobacter heilmannii' strains GenBank: AF508012 and L25079; while it was 99% to GenBank: AY139170, AF507996, AY139172, AY139173 and 98% to GenBank: AY139171, respectively. The H. felis sequences had 100% similarity to H. felis strains GenBank: FQ670179 and X69080; while it was 99% to H. felis GenBank: AY368267, AY368261 and 98% to GenBank: DQ865138, respectively.

Among our patients, the cohort exposed to pet animals was limited to 26%. There were more patients with H. pylori infection who were in the 18-39 years age range. Such age distribution was not seen in cases with H. felis and H. heilmannii infection. There was no difference in the gender, ethnicity of patients, crowding index (CI) and source of water distribution among the patients with H. pylori and non-H. pylori helicobacter species infections. There were no statistically significant differences in the endoscopic findings in patients with H. pylori infection alone or with coinfection of H. pylori and non-H. pylori Helicobacter species. Chronic active inflammation was associated with H. pylori infection compared to H. heilmannii or H. felis coinfections with H. pylori (Table 1). However, the histology was not obtained in all the cases that showed H. heilmannii and H. felis infection. Intestinal metaplasia was present in 2(25%) out of 8 cases of H. heilmannii coinfection with H. pylori and in 1(20%) out of 5 cases of H. felis coinfection with H. pylori as compared to 3(3%) of 94 cases with H. pylori infection alone who had undergone the histological study. Although it was not possible to draw a conclusion that IM was significantly associated with the coinfection of either of the species and H. pylori, a tendency in that way would be likely, as it has also been reported by other authors.

PCR positives at the species level were also positive for the Helicobacter genus specific 16S rDNA and all the H. heilmannii and H. felis positive patients were also positive for H. pylori glmM PCR (Table 2). PCR product sequences of ureaseB gene of H. heilmannii had shown 100% similarity to 'Candidatus H. heilmannii strains' GenBank: AF508012 and L25079; while H. felis sequences had shown 100% similarity to strains GenBank: FQ670179 and X69080.

In this study, we used urease gene-based PCR method developed by Nieger et al that detected only 'Candidatus H. heilmannii' DNA from pure in vitro cultures of other non-H. pylori helicobacter species [14]. This method was also used by other investigators to demonstrate the presence of Candidatus H. heilmannii DNA in gastric biopsies from patients with dyspepsia [11,15,16]. The limitations of our study include the small number of patients who had non-H. pylori helicobacter infection and the presence of H. pylori co-infection which precluded assessment of the histological effect of these species under consideration. Also, the significance of coinfection in terms of disease development could not be determined. We could have identified few more cases of non-H. pylori helicobacter species by other reported methods used to study non H. pylori helicobacter species including fluorescent in situ hybridization (FISH), transmission electron microscopy (TEN) and partial 16S ribosomal sequencing for analyses of the amplified products [12,17].

The implications of this study are that non-H. pylori helicobacter species infection occurs in patients with abdominal pain or discomfort similar to H. pylori infection. Most of our H. heilmannii infections were not associated with contact with animals. This is in contrast to a previous analysis of 125 patients with confirmed H. heilmannii infection that showed some 70.3% of the 111 patients had a history of contact with one or more animals [17,18]. All of our patients with non-H. pylori infection had endoscopic gastritis though their association with peptic ulcer is well known [19,20]. The prevalence of coinfection of H. felis with H. pylori in our population is less than what has been reported from South Africa among African population but is certainly higher than that for H. heilmannii and H. pylori from the northern Europe which showed that only 1.6% had concomitant infection with H. pylori [20,21]. The coinfection in our patients demonstrated severe gastric pathology, as intestinal metaplasia was present in 25% of H. heilmannii coinfection with H. pylori while in 20% of H. felis coinfection with H. pylori. This was also reported in previous studies [22]. In this study, the difference was not statistically significant due to the number of subjects in each group. The routine transmission of H. pylori appears to be human-human whereas non-H. pylori helicobacter species are transmitted by cats, dogs, etc [22]. Consequently, the prevalence of H. heilmannii is expected to be significantly higher in environment with less hygiene and higher physical exposure to animals. However, in our study there was a negative association with pet contact as the patients reported limited exposure to these animals. There is a need to look into other modes of transmission of these infections.

Conclusion

As non H. pylori Helicobacter species are capable of producing complications similar to H. pylori so the identification of these species may be of importance in patients with dyspepsia. However, our study fails to show any increased risk of infection with these organisms on exposure to pet animals and any additional complications associated with co-infection in patients infected with H. pylori.

Methods

Study population

Between September 2009 and February 2011, a total of 250 patients with abdominal pain or discomfort who attended the gastroenterology outpatient clinic at a tertiary care hospital in Karachi were enrolled. The mean age of these patients was 39 ± 12 years, (range 18-75) with males 162(65%) and females 88(35%). Of these, 136(54%) were in the age group of 18-39 years, 88(35%) in the group of 40-55 years and 26(10%) in the group of 56-75 years. Ethical approval for the study was obtained from the Aga Khan University Ethics Review Committee. Informed consent was taken for participation in the study. A complete socio-demographic questionnaire including determination of socio-economic status, educational level, ownership of the place of residence, number of rooms in the house, number of people living in the household beside siblings, source of water supply e.g. municipal water pipeline or bore water (ground water) and type of latrine in use, was obtained from the patients. A history of exposure of enrolled patients to cats and dogs was determined and a physical examination was carried out. Inclusion criteria were i) ambulatory adult males and non-pregnant females; ii) age 18 years or older; iii) patients with upper GI symptoms including abdominal/epigastric pain or discomfort, postprandial abdominal distension, postprandial nausea and vomiting. Exclusion criteria included i) receiving treatment for H. pylori, concurrent or recent antibiotic use such as metronidazole, clarithromycin, amoxicillin, tetracycline, doxycycline and other cephalosporin, ii) histamine-2 receptor blocker or proton pump inhibitor therapy and bismuth compounds in the last four weeks; iii) patients with regular use of NSAID; iii) patients with severe concomitant disease and iv) patients with upper GI surgery. A crowding index with three categories was constructed by dividing the number of individuals per household by the number of the rooms used as bedrooms [23]. A participant's household crowding was defined as 'low' if they scored an index of 0-1.0, moderately-crowded were '2-4' and > 4 were highly 'crowded'.

On EGD, 242(97%) were found to have endoscopic gastritis (GS) alone while 8(3%) had duodenal ulcer (DU). Biopsy specimens from the gastric corpus and antrum were taken for rapid urease test (RUT) or histopathology for the diagnosis of H. pylori and DNA extraction for polymerase chain reaction (PCR) to amplify H. pylori, H. felis and H. heilmanii genes. Ninety patients (36%) out of 250 had a RUT done while 160(64%) out of 250 had histology and provided gastric biopsy specimen for the detection of Helicobacter species.

Histopathology

Biopsy specimens were stained with hematoxylin and eosin. Sections were examined by an experienced gastrointestinal pathologist blinded to the clinical details of the patients and graded according to the updated Sydney classification [24]. The bacterial density was graded from 0 to 3 (0, absent or occasional; 1 to 3, from few and isolated bacteria to colonies). The infiltration of gastric mucosa by mononuclear cells and polymorphonuclear leucocytes, atrophy, and intestinal metaplasia were graded as follows: 0, none; 1, mild; 2, moderate; 3, marked. Chronic inflammation was defined according to an increase in lymphocytes and plasma cells in the lamina propria graded into mild, moderate or marked increase in density. Chronic active gastritis indicated chronic inflammation with neutrophilic polymorph infiltration of the lamina propria, pits or surface epithelium graded as 0 = nil, mild = < 1/3 of pits and surface infiltrated; moderate = 1/3-2/3; and marked = > 2/3. Gastritis was scored by total sum of grade of gastritis (mild = 1, moderate = 2, marked = 3 infiltration with lymphocytes and plasma cells) and activity of gastritis (mild = 1, moderate = 2, marked = 3 infiltration with neutrophilic granulocytes) either in the antrum or in the corpus. Atrophy was defined as the loss of glandular tissue, with or without replacement by intestinal-type epithelium. Criteria for a true positive result was established with positive RUT or histology and 16S rDNA amplification.

DNA Extraction

DNA was extracted from biopsy samples by using a QIAamp DNA mini kit from QIAGEN (Hilden, Germany) according to the manufacturer's protocol. Extracted DNA was stored at -70°C until required.

Polymerase chain reaction

PCR was performed using extracted DNA as the template to identify H. pylori, H. heilmannii and H. felis. Samples that were positive for Helicobacter genus 16S rDNA were subsequently analyzed with different sets of previously published primers (Table 4) which encode H. pylori phosphoglucosamine mutase (glmM), H. heilmannii ureB and H. felis internal fragment of the ureA and ureB genes, respectively [14,21,25,26]. PCR amplification was carried out in a total volume of 25 μl containing 2 μl of 2 mM dNTPs, 1 μl of 50 ρmol of each forward and reverse primer used before [14,25-27]. (synthesized by MWG Automatic synthesizer, Germany), 2.5 unit of Taq DNA polymerase (Promega, USA), 2.5 μl of 10 × PCR reaction buffer, 3 mM of MgCl2, 2 μl of DNA template containing 0.5 ng of extracted DNA and total volume rounded to 25 μl by double distilled water. The reaction was carried out in a Perkin Elmer 9700 thermal cycler (Massachusetts, USA). The amplification cycles for the different Helicobacter spp. gene fragments were: 94°C for 5 min; 94°C for 1 min, 55°C-58°C for 1 min, 72°C for 60--90 sec (35 cycles); 72°C for 5-7 min. Positive and negative reagent control reactions were performed with each batch of amplifications. After PCR, the amplified PCR products were electrophoresed in 2% agarose gels containing 0.5 × Tris/acetate/ethylenediaminetetraacetic acid, stained with ethidium bromide, and visualized under a short wavelength ultraviolet light source. DNA from H. pylori strains ATCC 43504, H. felis ATCC 49179 and H. heilmannii JF804941.1 was used as a positive control and sterile deionized water as the negative control. Diagnosis of each of the Helicobacter species infection was established when Helicobacter genus PCR for 16S rDNA was positive along with a species specific PCR for H. pylori, H. heilmannii or H. felis. PCR product of H. heilmannii and H. felis were sequenced to further confirm individual infection. The specificity of H. pylori phosphoglucosamine mutase (glmM) and segment of urease B primers for H. heilmannii and H. felis has been demonstrated previously [14,21,25-27].

Table 4

Oligonucleotide primers used in this study to amplify Helicobacter spp. gene fragments

Gene	Sequence (5' to 3')	Amplicon size (bp)	Reference
Helicobacter 16S rRNA
C97	GCT ATG ACG GGT ATC C	400	18
C 98	GAT TTT ACC CCT ACA CCA
H. pylori glmM
F	GGATAAGCTTTTAGGGGTGTTAGGGG	294	19
R	GCTTACTTTCTAACACTAACGCGC
H. heilmannii ureB		580	14
F	GGGCGATAAAGTGCGCTTG
R	CTGGTCAATGAGAGCAGG
H. felis ure A and B		241	20
F	GTG AAG CGA CTA AAG ATA AAC AAT
R	GCA CCAAAT CTA ATT CAT AAG AGC

Sequencing of PCR product and BLAST Query

The DNA fragments amplified by H. felis and H. heilmannii PCRs were purified by Qiagen quick PCR purification kit (Qiagen, USA) and sequenced using both the forward and reverse primers (Table 4) to verify that they represented truly the H. felis and H. heilmannii ureB gene. Sequence analysis was performed by Macrogen (Seoul, South Korea). ClustalX was used to edit the sequences. The sequences were edited to a length of 488 bp for H. heilmannii and 210 bp for H. felis. Homology of the DNA sequences to published sequences was determined by using BLAST window on the National Center for Biotechnology Information (NCBI) site at http://www.ncbi.nlm.nih.gov/BLAST.

Nucleotide sequence accession numbers

The sequenced PCR products of H. heilmannii and H. felis obtained in this study have been deposited in GenBank under the following accession numbers: JF804941, JF804942, JF804943, JF804944, JF804945, JF815095, JF815096, JF815097 and JF815098. PCR product sequences were compared to the sequences of Urease B of H. heilmannii sequences GenBank: AF508012, L25079.1, AY139171.0, AY139171.1 and H. felis strains ref GenBank: FQ 6701792, AY368267.1 and AY368261.1.

Statistical Method

Using software EPI Info and using 10% prevalence in the study population [21] with 95% confidence level and a bound on error of ± 4% the estimated sample size was 217.

Results are expressed as mean ± standard deviation for continuous variables (e.g., age) and number (percentage) for categorical data (e.g., gender, etc.). Univariate analysis was performed by using the independent sample t-test, Pearson Chi-square test and Fisher Exact test whenever appropriate. A P-value of < 0.05 was considered as statistically significant. All p-value were two sided. Statistical interpretation of data was performed by using the computerized software program SPSS version 19.

Competing interests

The authors declare that they have no competing interests.

Authors' contributions

JY conceived and designed the study, JY, ZAB, RK, WJ coordinated the study, JY, SN, FJ did the work, JY and ZA analyzed the data, ZAH analyzed the histopathology, JY, MI, SA performed the statistical analysis. JY wrote the manuscript. All authors read and approved the final manuscript.

Pre-publication history

The pre-publication history for this paper can be accessed here:

http://www.biomedcentral.com/1471-230X/12/3/prepub