Keywords
HFMD - meta-analysis - NT-proBNP - IL-8 - blood glucose
Introduction
Hand, foot, and mouth disease (HFMD) is a childhood disease mainly caused by enterovirus
71 (EV71) and coxsackievirus A16 (CA16).[1]
[2] Outbreaks of HFMD occur in Asia, and the condition largely affects children under
5 years.[3] The common symptoms include fever (>38°C), rashes on the volar regions of the hands
and feet, and difficulty eating and drinking. In some cases the clinical manifestations
are more serious, including herpangina, brain stem encephalitis, and cardiopulmonary
systems (e.g., pulmonary edema). These cases can progress dramatically, leading to
serious sequelae or even death. Therefore, identifying factors with diagnostic and
prognostic values is worthwhile.
At an early period, patients do not usually undergo testing specifically for EV71
or CA16; routine blood tests are conducted more frequently. Some of these parameters
may help with the diagnosis, and assessing the prognosis of HFMD. Theoretically, parameters
in the routine blood tests (e.g., white blood cells [WBC]), coagulation tests (e.g.,
platelet), inflammatory cytokine levels, and other plasma biochemical indexes, may
collectively indicate the onset and progression of HFMD. This meta-analysis aimed
to identify factors (including WBC count, platelet, C-reactive protein [CRP], interleukin-8
[IL-8] level, fasting blood glucose level, etc.) with diagnostic and prognostic significance
in HFMD patients. The potential risk factors were analyzed when patients were divided
into groups using different standards.
Materials and Methods
This review was conducted in accordance with the Preferred Reporting Items for Systematic
Reviews and Meta-Analyses (PRISMA) guidelines. We systematically searched data from
the all full-published English language papers, excluding meeting or conference abstracts.
A thorough literature search was undertaken using the following databases: PubMed,
Embase, Cochrane Library, and Web of Science databases. Literatures were searched
using the following keywords: “hand, foot, and mouth disease”/“hand-foot-and-mouth
disease”/“HFMD,” combined with “patients”/“patient” or “blood” (such as HFMD and patients
[abstract], HFMD and patient [abstract], or hand-foot-and-mouth disease and blood
[abstract], etc.), with a time range from 1988 to 2018. The union was acquired by
combination of the results of: (1) HFMD/hand-foot-and-mouth disease/hand, foot, and
mouth disease with patients/patient, and (2) HFMD/hand-foot-and-mouth disease/hand,
foot, and mouth disease with blood. The literature retrieval was performed by two
independent authors, and quality of the included studies was specifically assessed
using the QUADAS tool.
The process for screening is shown in [Fig. 1].
Fig. 1 Flow of the study identification, inclusion, and exclusion.
The inclusion criteria were as follows: (1) the studies surveyed factors related to
blood tests in HFMD patients; (2) for all clinical studies, patients had to be divided
into different groups, e.g., according to severity; (3) studies gave clear diagnostic
standards for HFMD.
The exclusive criteria were as follows: (1) duplicate publications; (2) studies not
related to HFMD; (3) reports in the forms of conference abstracts, case reports, comments,
reviews, or meta-analysis; (4) basic research, rather than clinical studies; (5) studies
that did not include detailed data about blood parameters, clear groups for comparison,
or otherwise lacked detailed results. Patients in different literatures were grouped
according to clinical characteristics: control or HFMD, mild or severe (those with
any evidence of central nervous system complications, including aseptic meningitis,
encephalitis, and poliomyelitis-like paralysis, or pulmonary edema, were classified
into the severe group, or else the mild group), and survived or dead.
The following indexes were observed in different groups, if documented: WBC counts
(109/L), platelet (109/L), CRP (mg/L), the inflammatory markers IL-8 (pg/mL), fasting blood glucose level
(mM), plasma N-terminal pro–B-type natriuretic peptide (NT-proBNP) levels, etc. If the data were
present as continuous values, at least including mean, SD, and case number, the results
were analyzed in the forest plots, or else in the single reference review. The detailed
indexes were presented in results.
Data were analyzed with RevMan software (version 5.3; Cochrane Collaboration, Oxford,
UK). Mean differences (MDs) with 95% confidence intervals (CIs) were calculated for
the continuous measures. Forest plots were presented after analysis, in which lines
represented different estimates and CIs, and boxes represented the weight given to
each study. We also quantified the effect of heterogeneity using I
2. The fixed-effect model was used when no significant heterogeneity existed; otherwise,
a random-effect model was used. The fixed-effect model was also calculated for the
sensitivity analyses. Substantial heterogeneity across the studies was detected when
I
2 was >50% or the p-value for heterogeneity was <0.10. All the p-values were two-sided. The funnel plot was used to assess the publication bias. A
p-value less than 0.05 was considered to be statistically significant, while the p-value less than 0.1 was considered to be significant statistically in heterogeneity
analysis.
Results
Search Results
A total of 244 papers potentially relevant to the search terms were identified by
the initial search (PubMed: 132, Embase: 83, Cochrane: 29, and Web of Science: 0).
Among these 112 duplicate studies were removed, and 5 irrelevant references were excluded.
Next, reviews, case reports, or meta-analysis papers (15 papers) were removed. Then,
the basic studies were further removed, including animal, cell, or biochemical studies
(17 papers), Finally, clinical studies giving insufficient detail of the required
indexes were removed (80 papers). This left data from 26 papers that were used for
the meta-analysis.[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30] All these studies had provided the detailed statistical results of blood test parameters
and definitions of different groups. In the QUADAS tools, all studies have provided
clear criteria for grouping, but in other items the results of assessment were all
“unclear.”
Parameters in Routine Blood Tests
First, the association between the three parameters (WBC counts, CRP, and platelet
levels) and severities were analyzed. For WBC, greater severities were related to
higher WBC levels. However, references in HFMD-versus-control-comparison (Tau2 = 2.45; Chi2 = 30.69, df = 5, p < 0.0001; I
2 = 84%) and severe-versus-mild-comparison (Tau2 = 8.12; Chi2 = 122.17, df = 7, p < 0.00001; I
2 = 94%) groups showed a significant heterogeneity. Regarding HFMD versus control,
WBC counts showed an overall effect between groups (Z = 2.39, p = 0.02), while there were no effects between severe and mild groups (Z = 1.68, p = 0.09; [Fig. 2A] and [B]). When comparing fatal and nonfatal groups, no heterogeneity was found among studies
(Chi2 = 1.35, df = 2, p = 0.51; I
2 = 0%); fatal cases had higher WBC levels than nonfatal ones (Z = 4.88, p < 0.00001; [Fig. 2C]). CRP levels showed conflicting results among studies in HFMD-versus-control, severe-versus-mild,
and fatal-versus-nonfatal comparisons ([Fig. 3A–C]). In comparison between HFMD and control cases, two studies showed contrary trends
(Tau2 = 51.21; Chi2 = 8.93, df = 1, p = 0.003; I
2 = 89%), there is no significant overall effect (Z = 0.56, p = 0.58); similarly, three studies about CRP differences in severe-versus-mild comparison
had distinct results (Tau2 = 72.36; Chi2 = 40.59, df = 2, p < 0.00001; I
2 = 95%; Z = 0.84, p = 0.40); also, a single study regarding CRP differences between fatal and nonfatal
cases found no association (Z = 1.20, p = 0.23). The platelet difference between severe and mild was documented in only one
study, and no significant difference was found ([Fig. 4A]); the platelet levels between fatal and nonfatal cases showed opposing results in
two studies (Tau2 = 11075.36; Chi2 = 8.56, df = 1, p = 0.003; I
2 = 88%), as well with no overall statistical significance (Z = 0.20, p = 0.84; [Fig. 4B]). There was significant heterogeneity between studies regarding CRP and platelet
levels. However, WBC counts significantly varied between fatal and nonfatal cases.
Fig. 2 Cumulative meta-analysis of the white blood cell (WBC) between different groups (presented
as forest plots). (A) Difference in WBC counts between hand, foot, and mouth disease (HFMD) and control
groups. (B) Difference in WBC counts between the severe and mild groups. (C) Difference in WBC counts between fatal and nonfatal groups.
Fig. 3 Cumulative meta-analysis of C-reactive protein (CRP) levels between different groups
(presented as forest plots). (A) Difference in CRP expression between HFMD and Control groups. (B) Difference in CRP expression between the severe and mild groups. (C) Difference in CRP expression between fatal and nonfatal groups. HFMD, hand, foot
and mouth disease.
Fig. 4 Cumulative meta-analysis of platelet levels between different groups (presented as
forest plots). (A) Difference in platelet levels between the severe and mild groups. (B) Difference in platelet levels between fatal and nonfatal groups.
Plasma Glucose
The differences in plasma glucose are presented in [Fig. 4]. One report showed the HFMD group had slightly higher glucose level than normal
but without statistical significance (Z = 1.64, p = 0.10; [Fig. 5A]). Two consistent results focusing on severe-versus-mild comparison showed a significant
effect between groups with low heterogeneity (Chi2 = 0.58, df = 1, p = 0.45; I
2 = 0%; Z = 2.97, p = 0.003; [Fig. 5B]). In addition, four studies regarding the glucose differences between fatal and
nonfatal groups had highly identical results (Chi2 = 1.48, df = 3, p = 0.69; I
2 = 0%), the fatal cases had elevated plasma glucose levels (Z = 8.44, p < 0.00001; [Fig. 5C]).
Fig. 5 Cumulative meta-analysis of plasma glucose levels between different groups (presented
as forest plots). (A) Difference in plasma glucose levels between HFMD and control groups. (B) Difference in plasma glucose levels between the severe and mild groups. (C) Difference in plasma glucose levels between fatal and nonfatal groups. HFMD, hand,
foot and mouth disease.
Interleukin-8
Among different inflammatory cytokines, we selected only IL-8 as no sufficient data
was available for other cytokines. Again, IL-8 expression between HFMD and control
groups was controversially reported (Tau2 = 582.38; Chi2 = 3.88, df = 1, p = 0.05; I
2 = 74%), and no differences were found between groups (Z = 0.46, p = 0.65; [Fig. 6A]). However, plasma IL-8 expression was significantly higher in the severe group than
the mild group, as two studies consistently reported (Chi2 = 1.47, df = 1, p = 0.22; I
2 = 32%; Z = 16.72, p < 0.00001; [Fig. 6B]).
Fig. 6 Cumulative meta-analysis of IL-8 levels between different groups (presented as forest
plots). (A) Difference in IL-8 levels between HFMD and control groups. (B) Difference in IL-8 levels between severe and mild groups. HFMD, hand, foot and mouth
disease.
NT-proBNP and Survival
There were two studies documenting plasma NT-proBNP levels as a continuous variable,
which highlighted that the fatal cases had several times of upregulation in NT-proBNP
expression compared with nonfatal ones (Tau2 = 6.03; Chi2 = 2.41, df = 1, p = 0.12; I
2 = 59%; Z = 4.00; p < 0.0001; [Fig. 7]).
Fig. 7 Cumulative meta-analysis of plasma NT-proBNP levels between fatal and nonfatal groups.
NT-proBNP, N-terminal pro–B-type natriuretic peptide.
Other Factors Associated with HFMD Grade and Ending
Besides the above indicators that could feasibly be routinely applied, there were
other blood markers with potential diagnostic/prognostic values. Lacking in the aggregation
of the repeated studies, these studies could only be individually analyzed. Interferon-γ
(IFN-γ) as one of the most widely studied immune-associated molecules, was significantly
increased in HFMD patients compared with controls,[31] and in the severe group compared with the mild group.[22] Another cytokine IFN-α was lower in patients with complications than those without
complications. Tumor necrosis factor-α (TNF-α), one of the most widely studied inflammatory
factors, was found to be highly increased in expression in HFMD cases.[24] Also, serum amyloid A and clusterin were proposed to be potential predictive biomarkers
for severe HFMD.[18] Glutathione (GSH) and malondialdehyde (MDA) were another two useful indicators in
HFMD evaluation. In the acute and recovery stages, HFMD patients showed a lower GSH
level and higher MDA level compared with controls.[20] D-dimmer is another coagulation indicator besides platelet. Zhang and Song observed
that children with severe HFMD had increased plasma D-dimer levels, and the higher
the levels, severer the condition.[27] A recent study pointed that serum concentrations of angiotensin II and noradrenaline
in HFMD patients were significantly higher than that in healthy controls.[26] Other than carbohydrates and proteins, noncoding RNAs were also found impacted by
HFMD development. Zhao et al pointed out that plasma circular DNAs can be an early
identification marker of HFMD: their levels were increased in ordinary HFMD and even
elevated in the severe group.[29] Differential serum expression levels of ten miRNAs (miR-140-5p, miR-143, miR-148a,
etc.) also changed in HFMD patients, by Cui et al.[6] Lastly, some immune-related gene polymorphism (e.g., IL-10) was associated with
the gene expression and severity of HFMD.[28] Collectively, these reports (with detailed data of each parameter) have shown significance
of some new but easily detectable markers, but they were majorly single ones in their
focused indexes. More evidences are needed for better understanding of their diagnostic
and prognostic values.
Finally, we found no publication bias for conclusions on funnel plot analysis, because
significant symmetrical appearance was demonstrated.
Discussion
This meta-analysis used 26 studies and found some interesting factors with potential
diagnostic and prognostic meanings in the peripheral blood. Specifically, WBC counts
had an association with later survival; CRP and platelet levels had limited indicative
roles; a higher plasma glucose level may imply graver complications and poorer survival;
fatal outcome was correlated with increased IL-8 and NT-proBNP levels.
Ordinarily, EV71 is a leading common cause of HFMD, aseptic meningitis, and encephalitis
in Asian populations. Application of reverse transcription polymerase chain reaction
or sequencing technique to identify HEV71 amplicons was a direct method in HFMD diagnosis;
at the protein level, EV71 and CVA1 neutralizing IgM antibodies have a high prognostic
value for HFMD. So far, multiple assays have been proposed for detecting the EV71
antigen in HFMD serum samples; rapid and sensitive screening techniques have also
been developed in recent years, such as rapid point-of-care tests[32] and protein microarray-mediated detection techniques.[33] However, these indicators are not among routine investigations, and only patients
highly suspected of having HFMD could be recommended the above detections. Moreover,
a limited sight of EV71/CVA1 markers cannot afford useful risk factors in routine
blood tests, which may provide additional evaluation references without adding the
diagnostic cost.
To date, large number of studies have observed potential risk factors in HFMD. Dominant
publications focused on clinical manifestations, like vomiting, fever, and lethargy,
etc. Also, Chinese scholars showed high interests in temperature and humidity.[34]
[35]
[36] In the present study, we investigated the commonly used indexes in peripheral blood
and suggested that these indexes can be potential assistant biomarkers in HFMD.
Intriguingly, WBC variations showed great heterogeneity among studies, and this factor
has limited values in distinguishing HFMD/control or severe/mild. However, the WBC
level around 18 × 109/L suggests a death outcome. Therefore, it is still worth paying attention to high
WBC levels. In the present analysis, CRP and platelet did not show a clear indicative
value, although some rare reports have implied these markers may be positively correlated
with HFMD severities. For example, there was a case report of adult HFMD with a long-lasting
elevated CRP level.[37] In 2016, a survey reported that patients in the severe HFMD group had an increased
high-sensitivity CRP level compared with the mild group.[38] More data accumulation is required to establish definitively whether there are links
between CRP and platelet levels and HFMD diagnosis and prognosis. The relationship
between NT-proBNP and survival is supported by other studies (not included on our
meta-analysis because there were insufficient data). For example, Qiu et al reported
that patients with NT-proBNP ≥1,300 pg/mL had higher risk of death than those with
NT-proBNp <1,300 pg/mL.[39] This conclusion can be supported by our meta-analysis, and our finding strongly
implied that several fold higher NT-proBNP is a sign of poor outcome.
The inflammatory cytokine IL-8 was found dramatically elevated in the severe group
in comparison with mild group based on this analysis, which suggests the progression
of HFMD, as well as HFMD caused death, accomplished by above related indicators (WBC,
glucose, and NT-proBNP), may be due to uncontrolled inflammatory responses. Some of
our included studies did not provide the detailed expression levels of cytokines,
but they showed consistent results. Li et al observed that the levels of IL-1b, IL-6,
IL-10, MIP-1, and TNF-α were higher in HFMD patients than controls, IL-6 and TNF-α
levels were even higher in HFMD patients as compared with mild HFMD.[17] A report in 2014 (not included in our analysis) also showed different cytokine profiles
in different cohorts. In patients with mild HFMD, the peaks of IL-8 and IL-10 were
observed on day 6 and that of IL-18 was observed on day 4; in those with severe illness,
all cytokines spiked on day 3 and peaked on day 11; and cytokines were significantly
correlated with immunoglobulin M levels by the end of the disease course.[40] In another two studies, all kinds of cytokines/chemokines were upregulated in severe
HFMD patients compared with mild HFMD, but RANTES, MCP-1, IL-4, IL-12, and IL-18 levels
were higher in mild HFMD patients than in the controls.[41]
[42]
There are some limitations in the present study. We have only identified 26 papers
from four databases. Overall, the number of cases for many blood parameters was small.
When the total references are limited, judgment may be biased due to expectations
and misleading conclusions may be drawn. We hope increasing evidence to support our
conclusion that routine blood markers may be useful will promote further studies.
In conclusion, some blood factors, especially WBC counts, plasma glucose level, IL-8,
and NT-proBNP levels are associated with the severity and outcome of HFMD, but more
evidences are needed to confirm these findings.
