Volume 1(4); Pages: 265-274, 2021 | DOI: 10.21873/cdp.10034
MAARET ESKELINEN, JANNICA MEKLIN, TUOMAS SELANDER, KARI SYRJÄNEN, MATTI ESKELINEN
MAARET ESKELINEN1*, JANNICA MEKLIN1*, TUOMAS SELANDER2, KARI SYRJÄNEN3,4 and MATTI ESKELINEN1
1Department of Surgery, Kuopio University Hospital and School of Medicine, University of Eastern Finland, Kuopio, Finland
2Science Service Center, Kuopio University Hospital and School of Medicine, University of Eastern Finland, Kuopio, Finland
3Molecular Oncology Research Center, Barretos Cancer Hospital, Barretos, Brazil
4SMW Consultants Ltd., Kaarina, Finland
Correspondence to: Matti Eskelinen, MD, Ph.D., School of Medicine, University of Eastern Finland, P.O. Box 100, FI-70029 KYS, Finland. Tel: +358 17173311, Fax: +358 17172611, GSM:+358 400969444, e-mail: email@example.com
*These Authors contributed equally to this study.
Received May 11, 2021 | Revised June 2, 2021 | Accepted June 11, 2021
Background/Aim: The diagnostic accuracy of history-taking, clinical signs and tests and diagnostic scores (DSs) for patients with non-organic dyspepsia (NOD) have been rarely evaluated. Patients and Methods: A cohort of 1333 patients presenting with acute abdominal pain (AAP) were studied, including 50 patients with confirmed NOD. The most significant diagnostic variables (in multivariate logistic regression analysis) were used to construct six different DS models and their diagnostic accuracy was compared with clinical symptoms and signs and tests. Meta-analytical techniques were used to detect the summary sensitivity (Se) and specificity (Sp) estimates for each data set (symptoms, signs and tests as well as DS models). Results: In hierarchical summary receiver operating characteristic (HSROC) analysis, the area under curve (AUC) values for i) symptoms ii) signs and tests iii) DS were as follows: i) AUC=0.608 [95% confidence interval (CI)=0.550-0.666]; ii) AUC=0.621 (95% CI=0.570-0.672) and iii) AUC=0.877 (95% CI=0.835-0.919). The differences between these AUC values (roccomp analysis) are as follows: between i) and ii) p=0.715; between i) and iii) p0.0001; between ii) and iii) p0.0001. Conclusion: The present study is the first to provide evidence that the DS could be used in diagnosis of NOD. The major advantage of our DS is that this model does not need radiology or endoscopy to reach high diagnostic accuracy.
Non-organic dyspepsia (NOD), also known as functional dyspepsia, is “a collection of symptoms” without evidence of an organic disease that could explain the symptoms (1, 2). NOD is estimated to affect about 15-40% of the general population in Western countries (3, 4). The symptoms of NOD are non-specific including location of pain at upper abdomen, pain duration over 12 hours, similar pain and indigestion previously, poor appetite and vomiting (5). According to previous analyses, the patients with upper abdominal pain (AAP) and previous history of indigestion tended to be at risk for NOD (5). The diagnostic accuracy of the clinical findings in NOD have rarely been investigated and the few studies performed include gastroscopy referral patients (6). To circumvent this type of bias we investigated the diagnostic accuracy of clinical findings in NOD among patients with AAP.
Although, the diagnostic performance of clinical symptoms, signs and tests have been investigated earlier in acute appendicitis (AA) (7-11), acute cholecystitis (AC) (12) and in acute small bowel obstruction (13), there is very little data on the diagnostic accuracy of history-taking, clinical signs, tests and diagnostic score (DS) for NOD; this prompted us to re-evaluate the accuracy of the clinical diagnosis of NOD. The present study evaluated the relative accuracy of i) symptoms, ii) signs and tests, as well as iii) the DS in confirming NOD among the patients with AAP.
In the NOD study group there were 50 patients (18 females and 32 males) versus 1283 patients in the non-NOD group including 679 females and 604 males. The clinical symptoms (n=22), signs and tests (n=14) and laboratory analyses (n=3) were recorded in each patient. The diagnosis of NOD was confirmed by considering all clinical history-taking details, clinical findings and results of the laboratory tests together and following the diagnostic criteria of AAP and NOD.
Identifying the DS models. A multivariate logistic (stepwise) regression analysis (SPSS Statistics 18.104.22.168; IBM, Armonk, NY, USA) was used to disclose the variables with an independent predictive value. All the variables of symptoms, signs and tests presented in Tables I and II were included in the analysis as binary data e.g., NOD=1 and other diagnosis of AAP=0. Using the coefficients of the regression model, a DS was built and its predictive value for NOD was studied. The coefficient of the multivariate analysis shows the relative risk (RR=en, n=ß) of a patient with a given symptom, sign or test of having NOD.
The DS formula derived for NOD. The DS: 1.43 × gender (female=0, male=1) + 0.82 × location of initial pain (PE=1, NE=0) + 1.08 × location of pain at diagnosis (PE=1, NE=0) + 1.23 × duration of pain (PE=1, NE=0) + 0.96 × previous similar pain (PE=1, NE=0) + 0.73 × appetite (PE=1, NE=0) + 1.10 × drugs for abdominal pain (PE=1, NE=0) – 0.86 × use of alcohol (PE=1, NE=0) + 1.16 × rigidity (PE=1, NE=0) + 0.49 × guarding (PE=1, NE=0) + 0.82 × leucocyte count (PE=1, NE=0) – 9.17.
Statistical analysis. STATA/SE version 16.1 (StataCorp, College Station, TX, USA) was used for further statistical analyses. The statistical tests presented were two-sided, and p-values under 0.05 were considered statistically significant. Using 2×2 tables, sensitivity and specificity with 95% confidence intervals (95% CI) for each clinical history-taking variable, finding or test were determined. A meta-analytical technique (metaprop) was used to create separate forest plots for sensitivity and specificity for each set of data, including each diagnostic variable. We calculated the summary estimates of sensitivity and specificity, positive and negative likelihood ratios and diagnostic odds ratio, using a random- effects bivariate model and fitted the summary hierarchical receiving operating characteristic (HSROC) curves, including all diagnostic variables in the DS model, using NOD as an endpoint. Roccomp test (STATA) was used to compare the AUC values of HSROC tests between the 3 diagnostic sets (history-taking, clinical signs, DSs).
Patient data of the study. In the NOD study group, there were 50 patients (18 females and 32 males) versus 1283 patients in the non-NOD group (679 females and 604 males) including the following AAP patients: non-specific abdominal pain (n=616), acute appendicitis (n=271), acute cholecystitis (n=124), acute renal colic (n=59), acute small bowel obstruction (n=53) and other AAP patients (n=160), with the mean (SD) age of 37.5 (21.7) years.
The clinical symptoms of NOD. The overall Se of the clinical symptoms for detecting NOD was 67% (95% CI=56-77%) (Figure 1).The Se was higher than 67% for ten of the symptoms. The five most sensitive clinical history-taking variables (vertigo, appetite, jaundice, micturition and use of alcohol) showed 86-96% Se in diagnosis of NOD (Figure 1). The Sp of the history-taking for detecting NOD was only 46% (95% CI=32-61%) (Figure 2). Altogether, 11 symptoms showed Sp higher than 46%. The five most specific symptoms of NOD (previous indigestion, bowels, drugs for abdominal pain, previous abdominal surgery and previous abdominal diseases) showed 76-96% Sp (Figure 2).
Figure 1. Sensitivity of history-taking in non-organic dyspepsia (NOD) (random-effects model). ES: Estimated sensitivity; CI: confidence interval.
Figure 2. Specificity of history-taking in non-organic dyspepsia (NOD) (random-effects model). ES: Estimated specificity; CI: confidence interval.
The clinical signs and tests in NOD. The overall Se of the signs and tests for NOD was 81% (95% CI=70-90%) (Figure 3), and 9 signs and tests had Se values exceeding 81%. The six most accurate signs and tests (abdominal movement, mass, rigidity, Murphy’s positive, rectal digital tenderness and urine) showed 94-100% Se (Figure 3). The overall Sp of the signs and tests was only 32% (95% CI=18-47%) (Figure 4), while 7 signs and tests showed Sp higher than 32%. The five most specific signs and tests (mood, scar, tenderness, guarding and leucocyte count), however, showed 54-82% specificity (Figure 4).
Figure 3. Sensitivity of the signs and tests in non-organic dyspepsia (NOD) (random-effects model). ES: Estimated sensitivity; CI: confidence interval.
Figure 4. Specificity of the clinical signs and tests in non-organic dyspepsia (NOD) (random-effects model). ES: Estimated specificity; CI: confidence interval.
Scoring in confirming NOD. The most important predictors of NOD were gender, location of initial pain, location of pain at diagnosis, duration of pain, previous similar pain, appetite, drugs for abdominal pain, use of alcohol, rigidity, guarding and leucocyte count. The best diagnostic level for DS model (DS III; Se=78%, Sp=86%) was reached at a cut-off level of 0.055 for DS (Figures 5 and 6). The DS model was tested at six different cut-off levels to disclose the highest diagnostic accuracy (Table III; Figures 5 and 6). The Se and Sp of these six DS models were 74% (95% CI=69-79%) and 86% (95% CI=84-87%), respectively (Table III; Figures 5 and 6). Four of these models showed Se >74% and three models had Sp >86%. The best diagnostic DS model in these NOD patients (DS III, Figures 5 and 6) showed Se of 78% (95% CI=64-88%) and Sp of 86% (95% CI=84-87%).
Figure 5. Sensitivity of diagnostic scores at six different cut-off levels (DS I-VI). ES: Estimated sensitivity; CI: confidence interval.
Figure 6. Specificity of diagnostic scores at six different cut-off levels (DS I-VI). ES: Estimated specificity; CI: confidence interval.
HSROC and comparison of the AUC values. STATA (metandiplot) was used to draw the HSROC curves to visualise the pooled overall accuracy of the symptoms (Figure 7), signs and tests (Figure 8) and different scoring models (Figure 9) in detecting NOD. In SROC analysis, the AUC values for i) symptoms ii) signs & tests iii) DS were as follows: i) AUC=0.608 (95% CI=0.550-0.666); ii) AUC=0.621 (95% CI=0.570-0.672) and iii) AUC=0.877 (95% CI=0.835-0.919). The differences between these AUC values (roccomp analysis) are as follows: between i) and ii) p=0.715; between i) and iii) p0.0001; between ii) and iii) p0.0001.
Figure 7. Hierarchical summary receiver operating characteristic (HSROC) curve of the history-taking in non-organic dyspepsia (NOD).
Figure 8. Hierarchical summary receiver operating characteristic (HSROC) curve of the clinical signs and tests in non-organic dyspepsia (NOD).
Figure 9. Hierarchical summary receiver operating characteristic (HSROC) curve of the six diagnostic score models.
Some years ago, the value of the history-taking in the diagnosis of NOD was reported, but at that time, the HSROC and AUC analysis to confirm the diagnostic performance of clinical findings and scoring was not available (5). Prompted by the difficulty of NOD diagnosis among the AAP patients and the lack of diagnostic accuracy studies on DS with HSROC analysis, we designed the present study to assess the diagnostic performance of i) symptoms, ii) signs and tests, as well as iii) the scoring in confirming NOD among the patients with AAP.
The diagnosis of NOD could be made based on common clinical findings supported by signs and tests, ultrasound (US) and gastroscopy. One of the most difficult problems in detection of NOD is the lack of a golden standard (1, 2). To overcome this problem, we analysed the NOD diagnosis based on the final diagnosis of all AAP patients. Clinical findings of NOD include the location of initial pain and pain at diagnosis usually in the upper abdomen. Pajala et al. (14) reported the location of upper abdominal pain (UAP) in 77% of the patients with NOD. Similarly, in our study 78% of NOD patients had UAP initially and 74% had the UAP at diagnosis with a diagnostic efficiency (De) in NOD of 65% versus 67%, respectively. In patients with AAP, nausea, vomiting and poor appetite are usually regarded as NOD symptoms. In our study, only 38% of the NOD patients had nausea and 60% had vomiting, the De being in NOD of 57% versus 58%, respectively. The results of earlier investigations did not support a strong link between specific symptoms and NOD (5). However, patients with UAP, with a previous history of abdominal surgery and indigestion seem to be at risk for NOD and DS might help differentiate NOD from other causes of AAP.
Talley et al. (15) recruited 113 patients with upper-gastrointestinal symptoms and population-based subjects (n=347) and developed a 42-item quality life score for NOD. Although, this Nepean Dyspepsia Index (NDI) may be a valid tool for NOD symptoms, there is still no HSROC and AUC analysis for the diagnostic accuracy for NOD patients.
Adam et al. (16) enrolled 95 patients with 56 healthy subjects for the assessment of NOD symptoms. They focused on 10 symptoms including UAP, vomiting, poor appetite, nausea, feeling sick, bloating, cramps, early satiety, heartburn and retrosternal discomfort, and developed the gastrointestinal symptom score (GSS). The GSS seems to be a reliable tool to investigate symptom intensities in patients with NOD. Also, this study is limited by the highly selected patient cohort and lack of HSROC analysis.
Taylor et al. (17) designed a score to address the lack of symptom-focused measures in NOD patients. They interviewed 45 study participants to identify NOD symptoms and selected seven possible symptoms to construct the Functional Dyspepsia Symptom Diary (FDSD) score. Although, their FDSD score is valid patient-reported outcome measure for NOD patients, there is no HSROC analysis with AUC values for diagnosis of NOD available.
Lacy et al. (18) investigated 254 NOD patients and mailed a questionnaire to assess NOD symptoms and Gastroparesis Cardinal Symptom Index (GCSI). The results of the patients who responded (n=123) showed that the GCSI score could not accurately distinguish NOD patients from other diagnoses. They concluded that a more specific DS is needed for NOD diagnosis.
Acute cholecystitis (AC) is a reason of organic dyspepsia and one important differential diagnostic disease in confirming NOD. When comparing the symptoms, signs and tests between NOD patients and those with AC patients reported in the Eskelinen et al. study (12), the overall sensitivity of the symptoms in NOD of 67% (95% CI=56-77%) was higher than that detecting AC among AAP patients, which was 59% (95% CI=45-73%). However, the overall specificity of the symptoms in NOD patients was similar to that in AC patients; 46% (95% CI=32-61%) vs. 44% (95% CI=28-61%). The overall Se of the signs and tests in detecting NOD was 81% (95% CI=70-90%), which was significantly higher than that among AC patients (68%; 95% CI=53-81%). However, the pooled Sp of the signs and tests in detecting NOD was 32% (95% CI=18-47%) and was inferior to that of the AC patients, which was 41% (95% CI=23-60%).
When NOD and AC patients are compared in the scoring models, the trend is similar. The overall Se of the DS models in NOD is 74% (95% CI=69-79%), significantly lower than that in AC patients (86%; 95% CI=83-88%). Although Se and Sp usually behave reciprocally, this was not the case with the overall Sp of the DS in NOD patients (86%; 95% CI=84-87%), which is significantly lower than for the AC patients (94%; 95% CI=93-95%). In addition, the diagnostic accuracy of the DS (AUC=0.877; 95% CI=0.835-0.919) is significantly lower for NOD patients than that (AUC=0.953; 95% CI=0.923-0.969) in the AC patients.
We could not perform direct comparisons to previous DS studies in NOD, because the present study is the first to provide evidence that DS could be used to facilitate the diagnosis of NOD among patients with AAP. The major advantages of our DS is that this model does not need imaging, endoscopy or laboratory analyses to reach a relatively high diagnostic accuracy for NOD, compared to clinical findings alone.
The Authors report no conflicts of interest or financial ties.
All Authors contributed to the collection and analysis of data, drafting and revising the manuscript. All Authors read and approved the final article.
The study was funded by the Päivikki and Sakari Sohlberg Foundation.