Diagnostic markers for hyperemesis gravidarum: a systematic review and metaanalysis




Methods for Review


Search strategy


The search strategy ( Appendix ) was composed by one of the authors (M.N.N.) in collaboration with a clinical librarian. We searched PubMed and Embase from inception through January 2013 to identify articles that have reported on biomarkers in NVP and HG. We searched citation lists of review articles and eligible primary studies. Reference manager (version 12.0.3; Thomson Reuters, New York, NY) databases were established to incorporate the results of all the searches.


Study selection; in- and exclusion criteria


Etiologic, prognostic, predictive, and diagnostic studies that have reported on biomarkers in plasma, serum, urine, feces, or exhaled air in women with NVP or HG were included if they were written in English (n = 68). At least 5 eligible articles had to be available on the biomarker that had been studied for inclusion in our review. When 2 studies reported on the same study population, only the study with the most complete data was included. Case reports were excluded.


To metaanalyze diagnostic accuracy, studies were considered if they reported numeric data that allowed construction of a 2 × 2 table that cross-classified biomarker value and occurrence or severity of HG.


Study selection was done in 2 stages. First, 2 reviewers (M.N.N., I.J.G.) independently assessed identified titles and abstracts, after which we obtained full manuscripts of all selected studies. Second, 2 reviewers (M.N.N. and I.J.G., N.V. or R.C.P.) per paper independently assessed whether the studies were suitable to be included. Any disagreement was resolved by consensus meetings.


Data extraction and quality assessment


For each included article, data on both clinical and methodologic study characteristics were extracted independently by 2 reviewers on piloted data-extraction forms. We evaluated quality of included studies according to the quality assessment of diagnostic accuracy studies (QUADAS)–2 guidelines. QUADAS-2 summarizes the risk of bias of diagnostic accuracy studies in 4 domains: the study participants, the index test, the reference standard, and flow and timing. It also assesses potential concerns regarding applicability, which has to do with the representativeness of the study. Although most studies on biomarkers in HG were not diagnostic accuracy studies, but rather etiologic studies, we decided to use QUADAS-2 because we had the explicit aim to determine diagnostic biomarkers.


Case and control definition


For HG, there is no clear reference standard or a standard diagnostic workup. This makes it difficult to define cases and control subjects in our study.


We defined the reference test as a history of any combination of nausea, vomiting, dehydration, weight loss, or hospitalization based on nausea and/or vomiting in pregnancy in the absence of any other obvious cause for these complaints. This definition may lead to the inclusion of patients with mild NVP symptoms but does guarantee the inclusion of biomarkers across the disease severity spectrum. Control subjects were defined as pregnant women who had no complaints of nausea and vomiting.


Data analysis


For metaanalysis of diagnostic test accuracy, we used the hierarchic summary receiver operating characteristics model, using the Metandi package in Stata software (version 10.0; Stata Corporation, College Station, TX). From this model follows the summary hierarchic summary receiver operating characteristics curve and its parameters, but it is also possible to calculate a summary sensitivity and specificity from this model, if appropriate. We attempted to taper clinical heterogeneity by including studies for metaanalysis that used the same index tests and included only patients who had been diagnosed with HG and not NVP. All outcomes were evaluated with a random-effects model. Forest plots were made using Review Manager (RevMan, version 5.2; The Cochrane Collaboration, Baltimore, MD). Sensitivity and specificity were displayed in a forest plot when applicable. We followed the PRISMA statement and metaanalysis of observational studies in epidemiology (MOOSE) guidelines in writing this review.




Results


Search results


The electronic search identified 3442 individual articles ( Figure 1 ). We excluded several articles because there were <5 articles on the biomarker that was discussed. None of these articles were of high quality, nor did they report on a large study population. Eventually we included 81 articles in our study ( Table ). One study could not be retrieved from several national and international libraries and an e-mail address of the author could not be found. This study, however, was not feasible for our study because it was a review.




Figure 1


Selection process of articles

Flowchart of the selection process of articles found in the electronic search.

Niemeijer. Diagnostic markers for hyperemesis gravidarum. Am J Obstet Gynecol 2014 .


Table

Studies included in review





















































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































Study Year Participants, n Condition Biomarkers
Cases Control subjects Ketones Heliobacter pylori Human chorionic gonadotropin Thyroid Leptin Estradiol Progesterone White blood count Lymphocyte
Aka et al 2006 18 18 HG X X X
Al-Busaida and Krolikowski 2001 15 15 HG X
Al-Yatama et al 2002 50 50 HG X X
Arslan et al 2003 30 26 HG X X X
Asakura et al 2000 110 30 NVP/HG X
Asakura et al 2003 24 20 HG X X
Aytac et al 2007 52 55 HG X
Berker et al 2003 80 80 HG X X X
Bouillon et al 1982 33 52 HG X
Bruun and Kristoffersen 1978 35 57 HG X
Cevrioglu et al 2004 27 97 HG X X
Chihara et al 2003 17 37 HG X
Chou et al a 2011 59 0 NVP X X
Demir et al 2006 54 42 HG X X
Depue et al 1987 35 35 HG X X
Derbent et al 2011 115 110 HG X X X
Dokmeci et al 2004 17 13 HG X
Erdem et al 2002 47 42 HG X
Evans et al a 1986 342 0 NVP/HG X
Fairweather and Loraine 1962 90 11 HG X
Frigo et al 1998 105 129 HG X
Goodwin et al 1992 57 57 HG X X X
Goodwin et al 1994 39 23 HG X
Guney et al 2007 25 35 HG X
Guven et al 2011 40 40 HG X
Hatzivies et al 2007 25 85 HG X
Hayakawa et al 2000 34 29 HG X
Jacobson et al 2003 30 75 HG X
Jarnfelt-Samsioe et al 1985 62 40 NVP X
Jarnfelt-Samsioe et al 1986 62 40 NVP X X
Jarnfelt-Samsioe et al 1986 62 40 NVP X X
Jordan et al 1999 20 20 HG X X X
Juras et al 1983 33 30 HG X
Karaca et al 2004 56 90 HG X
Karadeniz et al 2006 31 29 HG X
Kauppila et al 1979 42 115 HG X
Kaupilla et al 1984 14 12 NVP X
Kazerooni et al 2002 54 53 HG X
Kimura et al 1993 27 24 NVP/HG X X
Kocak et al 1999 95 116 HG X
Kuscu et al 2003 10 10 HG X X X
Lagiou et al 2003 209 53 NVP X X
Lao et al 1988 51 28 HG X
Larraz et al 2002 162 156 NVP X
Lawson et al a 2002 92 0 NVP X
Lee et al 2005 40 42 HG X
Leylek et al 1996 24 20 HG X X X X
Leylek et al 1999 30 15 HG X X X X
Mansour and Nashaat 2011 80 80 HG X
Masson et al 1985 65 48 NVP X X X
Minagawa et al 1999 18 20 HG X X
Mori et al 1988 111 41 NVP X X
Murata et al 2006 44 53 NVP/HG X
Ndungu et al a 2009 72 0 Emesis X X
Panesar et al 2001 62 58 HG X X
Panesar et al 2006 43 329 HG X
Peled et al 2012 73 146 HG X
Perez-Perez et al 1999 42 47 HG X
Reymunde et al 2001 45 44 HG X
Salimi-Khayati et al 2003 54 54 HG X
Sandven et al 2008 244 244 HG X
Sekizawa et al 2001 16 23 HG X
Shirin et al a 2004 185 0 GI symptoms X
Soules et al 1980 37 9 NVP/HG X
Swaminathan et al 1989 71 43 HG X X
Tan et al a 2006 192 0 HG X X X
Tan et al a 2006 192 0 HG X X X
Tan et al a 2009 167 0 HG X X X X
Tareen et al 1995 30 15 HG X X
Taskin et al 2009 20 20 HG X X X X X X
Thornton et al 1979 12 12 HG X
Tlolka et al 2010 29 43 NVP X
Tsuruta et al 1995 55 24 NVP/HG X X
Unsel et al 2004 40 30 HG X X
Ustun et al 2004 35 39 HG X
Weyermann et al a 2003 898 0 GI symptoms X
Wilson et al 1992 10 50 HG X X
Wu et al a 2000 54 0 GI symptoms X
Xia et al 2004 72 100 HG X
Yoneyama et al 2002 22 22 HG X X X
Yoneyama et al 2002 24 24 HG X

GI , gastrointestinal; HG , hyperemesis gravidarum; NVP , nausea and vomiting in pregnancy.

Niemeijer. Diagnostic markers for hyperemesis gravidarum. Am J Obstet Gynecol 2014 .

a Not a case-control study.



Quality assessment


A summary of the QUADAS-2 scores for risk of bias and concerns regarding applicability is given in the Supplementary Figure . On the risk of bias, many studies had unclear quality regarding the index test. This is caused by unclear reporting on blinding the interpretation of test results. Quality was often unclear concerning patient selection and flow and timing. The reference standard was well-defined in most articles (65/81 studies); however, it varied among articles. In 60 articles on HG, 20 articles used >3 or >4 times vomiting episodes per day as part of the reference standard; 23 articles described persistent vomiting but did not quantify the frequency of vomiting. Thirty-two articles on HG used weight loss of >5% prepregnancy weight or >3 kg, and 39 articles used ketonuria as part of the diagnosis. Other elements were dehydration (3 studies) and hospitalization (11 studies). In 13 studies, no clear reference standard was described. Of 21 articles that reported on NVP, 17 articles used nausea and vomiting, 4 articles used weight loss, and 3 articles used hospitalization as part of the diagnosis. Three articles had no description of diagnosis other than that patients had been identified as HG cases.


In most studies, there were no concerns regarding applicability. Exclusion criteria were unclear in 30 of 80 articles. Patients with hepatic disorders (27 articles), thyroid disorders (46 articles), gastrointestinal disorders (31 articles), endocrine disorders (15 articles), urinary tract infection (6 articles), psychiatric disorders (12 articles), previous treatment against Heliobacter pylori (4 articles), pregnancy complications (6 articles), and chronic medication (6 articles) were excluded from studies. Molar pregnancies were excluded in 13 studies; multiple pregnancies were excluded in 21 studies. In 3 articles, women were excluded if there was uncertainty about their gestational age. We did not exclude any studies from our review on the grounds of poor quality.


Severity of HG


We found 3 studies that described biomarkers in relation to markers of disease severity. Severity was determined as readmission rate in 1 study and as a hospital stay of >4 days in the 2 other studies. These studies described the association with ketonuria, human chorionic gonadotropin (hCG), thyroid-stimulating hormone, free thyroxine 4 (FT4), estradiol, and white blood count (WBC).


The grade of ketonuria, as determined through a dipstick (studies 2006, 0-3+ vs 4+ urinary ketones of a possible 4+; study 2009, 0-2+ vs 3-4+ urinary ketones) was described in relation to severity of HG. Only a minority of patients with HG had ketonuria. Ketonuria was not significantly associated with prolonged hospital stay in the study of 2006 (adjusted odds ratio, 2.1; 95% confidence interval [CI], 1.0–4.6); P = .06). Neither of the 2 other studies showed a significant association between the grade of ketonuria and severity of HG in terms of readmission.


HCG was increased in women with HG who were hospitalized for >4 days compared with women who were hospitalized for <4 days. Thyroid-stimulating hormone and FT4 were not associated with a higher readmission rate. Estradiol of women who were hospitalized for >4 days were similar to levels of women hospitalized for <4 days. WBC level was not associated with severity.


Presence of HG/NVP


Ketonuria


Ketonuria among patients with NVP was not described in any study. We found 2 studies that described ketonuria and the presence of HG: Derbent et al (cases, 115 patients; control subjects, 110 patients) described an increase in degree of ketonuria between women with HG and pregnant control subjects (r = 0.622; P < .001). They found a median of 3+ of a possible 4+ with an interquartile range of 3+/4 in cases and 0+/4 (interquartile range, 0.5+/4) in control subjects. In a much smaller study, Kuscu et al (cases, 10 patients; control subjects, 10 patients) found no significant difference between ketonuria in patients with HG compared with pregnant control subjects. In this study, cases had a mean grade of ketonuria of 1.1 ± 0.3; control subjects did not have ketonuria. Ketonuria among patients with NVP was not described in any study. Neither of these studies reported the absolute prevalence of ketonuria among patients with HG and control patients.


Helicobacter pylori


We included 26 studies on H pylori in our review. Most studies used immunoglobulin G (IgG) antibodies against H pylori to determine whether women were infected. More than one-half of the studies showed a significant positive association between H pylori and NVP or HG compared with pregnant asymptomatic control subjects. In Figure 2 , results for the metaanalysis on presence of positive serologic findings in women with HG compared with asymptomatic control subjects are displayed, with an overall odds ratio of 3.2 (95% CI, 2.0–5.1). In Figure 2 , B, the sensitivity and specificity of these studies are displayed. Diagnostic metaanalysis of these 19 studies showed a summary sensitivity of 73% (95% CI, 62.0–81.4%) and a specificity of 55% (95% CI, 47.4–61.5%) in the diagnosis of HG, as compared with control subjects without hyperemesis. Figure 3 shows the studies that used methods other than IgG serology to determine H pylori infection in women with HG and NVP. Two studies were not displayed in forest plots because of reporting incomplete data for 2 × 2 table construction; Reymunde et al determined IgG in women with HG and found a significant association ( P < .001) between HG and the presence of IgG antibodies. Wu et al determined IgG in women with gastrointestinal complaints in pregnancy but did not find a significant association between these complaints and H pylori status.




Figure 2


Metaanalysis, sensitivity, and specificity of Heliobacter pylori immunoglobulin G

Metaanalysis, sensitivity, and specificity of H pylori immunoglobulin G in women with hyperemesis gravidarum compared with asymptomatic control subjects. A, Metaanalysis of diagnostic accuracy of H pylori immunoglobulin G in women with hyperemesis gravidarum compared with asymptomatic control subjects. B, Sensitivity and specificity in individual studies.

M-H , Mantel-Haenszel.

Niemeijer. Diagnostic markers for hyperemesis gravidarum. Am J Obstet Gynecol 2014 .



Figure 3


Heliobacter pylori

H pylori in women with hyperemesis gravidarum (measured with other methods than immunoglobulin G or nausea and vomiting in pregnancy compared with asymptomatic control subjects. A, H pylori in women with hyperemesis gravidarum. B, H pylori in women with nausea and vomiting in pregnancy.

CagA , cytotoxin-associated gene A; CUBT , C urea breath test; HG , hyperemesis gravidarum; IgA , immunoglobulin A; M-H , Mantel-Haenszel; NVP , nausea and vomiting in pregnancy; PCR , polymerase chain reaction; SA , stool antigen.

Niemeijer. Diagnostic markers for hyperemesis gravidarum. Am J Obstet Gynecol 2014 .


HCG


We included 35 studies on hCG, of which 18 studies showed a significant association between raised hCG levels and the occurrence of NVP or HG. Three studies showed a lower hCG in women with HG or NVP ; the other 13 studies did not find a significant association.


It was not possible to perform metaanalysis on hCG, because of large clinical heterogeneity and the skewed distribution of hCG during pregnancy that was caused by the lack of matching for gestational age in most articles and the lack of reporting biomarkers in multiple of medians. This is a measure for individual test deviance from the median and is calculated by dividing the individual patient result by the median of the population. In Figure 4 , we display forest plots of studies that report hCG levels and reported outcomes in a form that is suitable for forest plot display.




Figure 4


Human chorionic gonadotropin

Human chorionic gonadotropin levels in women with hyperemesis gravidarum or nausea and vomiting in pregnancy compared with asymptomatic control subjects. A, HCG, matched for gestational age; B, β-hCG, matched for gestational age; C, Free hCG, hCG in 24-hour urine sample, free α- and β-hCG, matched for gestational age; D, HCG, not matched for gestational age; E, HCG levels in women with nausea and vomiting in pregnancy.

Only gold members can continue reading. Log In or Register to continue

May 10, 2017 | Posted by in GYNECOLOGY | Comments Off on Diagnostic markers for hyperemesis gravidarum: a systematic review and metaanalysis
Premium Wordpress Themes by UFO Themes