Objectives Abdominal pain is one of the most frequent chief complaints in primary care settings. The aim of the present study was to determine the positive likelihood ratios (PLRs) and negative likelihood ratios (NLRs) of the relationships between the sites of abdominal pain and the organs involved.
Design Prospective observational study.
Setting A single tertiary centre, a university hospital in Japan.
Participants A total of 2591 new outpatients visited the Department of General Medicine at a university hospital from April 2017 to March 2018. Of these, 326 patients aged ≥20 years with abdominal pain were enrolled.
Results Sites of abdominal pain were classified into 11 categories including nine different abdominal sections, ‘generalised abdomen’ and ‘site-indeterminate’. The PLRs between ‘right subcostal’ and ‘liver and biliary tract’; between ‘right subcostal’ and ‘musculoskeletal’ ; between ‘epigastric’ and ‘oesophagus, stomach and duodenum’; between ‘right or left flank’ and ‘urinary tract’; between ‘left flank’ and ‘dermatological’; and between ‘mid-lower’ and ‘intestinal’ ranged from 2.17 to 4.14. The PLRs between ‘epigastric’ and ‘urinary tract’; between ‘mid-lower’ and ‘liver and biliary tract’; between ‘periumbilical’ and ‘urinary tract’; and between ‘generalised abdomen’ and ‘oesophagus, stomach and duodenum’ were low, ranging from 0.17 to 0.25. The NLR ranged from 0.5 to 1.5, excluding the relationship between ‘left flank’ and ‘dermatological’.
Conclusion The presence of pain at right subcostal, epigastric, right or left flank and mid-lower sites might be useful for identifying the organs involved. Additionally, the presence of pain at mid-lower, epigastric, periumbilical and generalised abdominal sites might be helpful for denying the involvement of some organs. Some sites of abdominal pain can be indicative of the organs involved.
Trial registration number
- abdominal pain site
- organ involved
- positive likelihood ratio
- negative likelihood ratio
- prospective observational study
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
- abdominal pain site
- organ involved
- positive likelihood ratio
- negative likelihood ratio
- prospective observational study
Strengths and limitations of this study
This was the first prospective study published since 1997 regarding the relationships between sites of abdominal pain and the organs involved.
This was the first study regarding sites of abdominal pain since CT has become widely available, especially in Japan.
The availability of CT is particularly high in Japan, which might not reflect its availability worldwide.
CT or other diagnostic modalities were not used equally among all included patients, which could have influenced the diagnosis of abdominal pain.
This was a single-centre observational study, the setting of which may have contributed to some sampling bias.
Abdominal pain is one of the most frequent chief complaints in primary care settings, accounting for 5% to 10% of visits to emergency departments.1–4 Diseases causing abdominal pain range from mild conditions cured by conservative medical treatments alone to severe acute diseases requiring emergency surgery.1 5–7 Given the possibility of severe disease requiring urgent treatment, it is vital that physicians make accurate and expeditious diagnoses.8 CT may be the most accurate imaging modality for assessment of abdominal pain, and is widely accepted as the first-line imaging modality for patients who present with this type of pain. However, it is necessary to select CT and other imaging tests (eg, abdominal ultrasonography or MRI) based on pretest probability, because these tests cannot be performed for some patients with abdominal pain; moreover, unnecessary examination and hospitalisation should be avoided.9 History taking and physical examinations can influence pretest probability and prove crucial to the prognosis. In particular, the sites of abdominal pain can be extremely important because some are significantly associated with potentially serious diseases, such as McBurney’s point pain with acute appendicitis and right subcostal pain with acute cholecystitis.10 11
To the best of our knowledge, no studies focussing on the relationships between the sites of abdominal pain and the organs involved have been published since 1997.1 However, it is certain that our capability to make an accurate diagnosis of abdominal pain has remarkably improved because of marked advancements of medical technologies such as imaging modalities, including CT and MRI, in the last two decades.12 13 In addition, in Japan, the epidemiology of conditions causing abdominal pain has changed in conjunction with ageing of the population, as well as westernisation of lifestyles.14–16 Therefore, it makes strategic sense to examine current relationships between the sites of abdominal pain and the organs involved two decades after the most relevant previous study.1 In the present study, we investigated the influence of sites of abdominal pain on the pretest probability of organs involved using positive likelihood ratios (PLRs) and negative likelihood ratios (NLRs).
Study design and patients
The present investigation was a single hospital-based prospective observational study conducted from April 2017 to March 2018. All new outpatients aged ≥20 years were enrolled, who visited the Department of General Medicine at Saga University Hospital in Japan with abdominal pain as a chief complaint or other symptoms excluding the chief complaint. They were initially seen by general physicians working in the department. Consenting patients were included regardless of whether they presented during the day or outside normal office hours, (ie, the emergency room for walk-in patients), but patients who used emergency service systems such as an ambulance or medical helicopter were excluded from the study. Comparative incidences of abdominal pain were evaluated, and then the organs and causative conditions involved were assessed. Statistical relationships between the sites of abdominal pain and the organs involved were then calculated. The present study was registered at https://www.umin.ac.jp. The design was assessed using a reporting checklist based on the Standards for Reporting of Diagnostic Accuracy (STARD) guidelines.17
The study was conducted at the Department of General Medicine at Saga University Hospital in Saga prefecture, which is located in southern Japan and has a population of 800 000. The hospital is open during the day and outside normal office hours (as an emergency room for walk-in patients). Patients who visited our hospital with a referral letter during the daytime could directly see the appropriate medical specialists. Additionally, patients transferred by ambulance were usually treated by the Department of Emergency Medicine or other medical specialists, irrespective of their visiting time, without undergoing treatment by the Department of General Medicine.
Data and data sources
Physicians who initially saw the patients completed a document template containing survey items designed and prepared in advance. Survey items recorded at the first patient visit included age, sex, the date and time (daytime or outside normal office hours) of the visit, residence status (alone or with housemates), referral from another doctor (yes or no), sites of abdominal pain, intermittent or persistent pain, types of examinations and management decided at the first visit. Patients who lived in nursing care facilities were included in the ‘with housemates’ group. When physicians could confirm the presence of pain in the patient’s abdomen, whether by visual confirmation of the site of spontaneous pain or by physical examination of abdominal tenderness, the pain was recorded as abdominal pain. Sites of abdominal pain were classified into 11 categories, including nine different abdominal sections (right or left subcostal, right or left flank, right or left lower, epigastric, periumbilical and mid-lower), ‘generalised abdomen’ and ‘site-indeterminate’. The types of examinations included blood and/or urinary tests, blood gas analysis, chest or abdominal X-ray, including kidney ureter bladder (KUB), ultrasonography (US), CT and MRI, including magnetic resonance cholangiopancreatography (MRCP), electrocardiography, oesophagogastroduodenoscopy (OGD) and colonoscopy (CS). A correct diagnosis of abdominal pain could be elusive, difficult and time-consuming, especially when associated with psychiatric conditions that require exclusion of most major organic diseases. Furthermore, additional examinations performed at another visit, with or without improvement of abdominal pain, could aid physicians in making a correct diagnosis. Therefore, the period of 3 months (also used in our previous study1 was considered an appropriate duration for confirmation of the final diagnosis; we determined the final diagnoses and organs involved at more than 3 months after the initial visit, in accordance with the following gold standard.
Gold standard for making definitive diagnoses or identifying the organ involved
Two physicians independently made diagnoses for individual patients based on the classification of the International Statistical Classification of Diseases and Related Health Problems-10 using all information in the medical charts after more than 3 months from their first visits, including onset, symptoms, time course, underlying diseases, past history, vital signs, characteristics of abdominal pain, and findings of laboratory or imaging studies. When the same diagnosis was made by both physicians, the diagnosis was set as the final diagnosis. When different diagnoses were made, the final diagnosis was determined through discussion among three physicians, including the aforementioned two physicians plus a third physician from our department. In cases in which the final diagnosis was unknown because the patient had only visited the department once without a definitive diagnosis being reached on that date, phone calls were made to the patients themselves or their family members more than 3 months after their first visit to ascertain the course of symptoms, visits to another hospital, and/or the results of examinations at other hospitals and diagnoses made there. When the patients could not be reached by telephone, the final diagnosis was determined through discussion among the three physicians. The organs involved, which could overlap with more than two other organs, were determined on the basis of the final diagnoses. We classified the final diagnoses into 11 categories: ‘Oesophagus, stomach and duodenum’, ‘Liver and biliary tract’, ‘Pancreas’, ‘Intestinal’, ‘Urinary tract’, ‘Gynaecological’, ‘Musculoskeletal’, ‘Respiratory’, ‘Cardiovascular’, ‘Dermatological’ and ‘Other’ disease. ‘Other’ disease consisted of ‘without definitive diseases’, ‘unknown’, ‘psychological’ and ‘organ-indeterminate diseases’.1
The types of investigations (eg, laboratory and imaging studies) used to make a diagnosis and identify the organ involved are described below (also listed in online supplementary 1). These investigations were not necessarily performed on all patients, especially those without higher positive pretest probabilities. Diseases of the oesophagus, stomach and duodenum were diagnosed using complete blood cell count (CBC), blood chemistry analysis, abdominal X-ray, abdominal CT and OGD. Intestinal diseases were diagnosed using CBC, blood chemistry analysis, abdominal X-ray, abdominal US, abdominal CT and CS. Diseases of the urinary tract system were diagnosed using CBC, blood chemistry analysis, urinalysis and abdominal X-ray, including KUB, abdominal US and abdominal CT. Diseases of the liver and biliary tract were diagnosed using CBC, blood chemistry analysis, abdominal US, abdominal CT and MRCP. Musculoskeletal diseases were diagnosed using CBC, blood chemistry analysis, X-ray of bones, chest and abdominal CT, and MRI. Gynaecological diseases were diagnosed using CBC, blood chemistry analysis, urinary human chorionic gonadotropin measurements, abdominal US and transvaginal US. Pancreatic diseases were diagnosed using CBC, blood chemistry analysis, abdominal US, abdominal CT and MRCP. Respiratory diseases were diagnosed using CBC, blood chemistry analysis, rapid immunoassay of influenza A and B nucleoprotein antigens, chest X-ray and chest CT. Cardiovascular diseases were diagnosed using CBC, blood chemistry analysis, electrocardiography, and chest and abdominal CT. Dermatological diseases were diagnosed via macroscopic inspection and histopathological studies. Malignant diseases were diagnosed using the findings of imaging studies or pathological findings including cytology, as well as data from histopathological examinations of specimens acquired via biopsy or surgery.
IBM SPSS (V.25) and Excel 2016 software were used to analyse the data using the χ2 test, and p<0.05 denoted statistical significance. We calculated sensitivity, specificity, PLRs, NLRs and 95% CI of the relationships between the sites of abdominal pain and the organs involved. In cases in which there were multiple sites of abdominal pain or multiple organs involved, we classified and analysed all of them. Missing values were removed from applicable data in each test. The calculated sample size of the present study, ranging from 18 to 165 patients, was based on two previous studies performed and reported by our institution: a prospective study reported in 1997 and a retrospective study in 2019.1 18
An explanatory pamphlet detailing the study was provided to all patients or their family members during the first visit, and it included all relevant information pertaining to the ways in which their individual information would be used. We obtained consent from all subjects via the comprehensive agreement method in the hospital, and anonymity of the patients was protected. The study was conducted in accordance with the guidelines of the 1975 Declaration of Helsinki.
Patients and public involvement
Two external members were present in the Institutional Review Board of our hospital. No other patients or members of the public were involved in the present research, including conceptualisation of research questions, planning of study design, performance of the research or analysis of the results.
A total of 2591 new outpatients who visited our hospital were included in the initial cohort; all outpatients who had no abdominal pain or were <20 years of age were excluded. All 326 (14.4%) of the remaining patients consented to participation in the study, and they were enrolled (figure 1). The characteristics of the patients are shown in table 1. The mean patient age was 51.7±20 years (age breakdown is shown in online supplementary 2) and 141 patients (43.3%) were men. Of the 326 patients included in the study, 126 (38.6%) had been referred to our department either by another hospital (93; 73.8%), another department of our hospital (28; 22.2%) or by general practitioners working at community health centres (5; 4.0%). A total of 209 (64.1%) patients visited during the daytime, 18 (8.6%) of whom visited during a national public holiday and who were thus considered ‘outside normal office hours’ patients in accordance with our hospital procedures. Attempts were made to reach 81 of the 326 patients in the study by telephone (24.8%), and 22 of these patients (6.7% of the total study sample) could not be reached. Of the types of examinations planned at the first visit in the present study, which might also be performed at the last visit to facilitate diagnosis and treatment, abdominal CT was performed in nearly half of the patients (155; 47.5%). One hundred forty-five patients (44.5%) were admitted to or followed up by the Department of General Medicine; 75 patients (23.0%) were admitted to or followed up by other departments in our hospital; and 46 patients (14.1%) were admitted to or followed up by other hospitals. The characteristics of patients who underwent CT examination are shown in online supplementary 3.
Including cases in which multiple sites of abdominal pain were individually identified, a total of 576 sites of abdominal pain were recorded. The most frequent complaint was epigastric pain (95/576; 16.5%), followed by periumbilical pain (72; 12.5%), mid-lower pain (66; 11.5%) and right lower pain (62; 10.8%) (figure 2). In the 326 patients included in the study, the total number of organs identified as being involved was 354. The fact that the number of organs identified as involved was greater than the number of patients in the study was partly because cases of acute gastroenteritis (28/326; 8.6%) were doubly classified into both the ‘intestinal’ and ‘oesophagus, stomach and duodenum’ categories. The most frequently involved organ category was ‘intestinal’ (125/354; 35.3%), followed by ‘oesophagus, stomach and duodenum’ (58; 16.4%), ‘urinary tract’ (38; 10.7) and ‘liver and biliary tract’ (25; 7.1%). Detailed diagnostic data, putative diagnoses of conditions causing abdominal pain and the organs involved are shown in table 2.
Relationships between the sites of abdominal pain and the organs involved are shown in table 3. Relationships between the sites of abdominal pain and causative organs with p<0.05 as determined via the χ2 test and PLR ≥2 were as follows: ‘right subcostal’ and ‘liver and biliary tract’ (p<0.001, PLR=3.59); ‘right subcostal’ and ‘musculoskeletal’ (p<0.001, PLR=2.34); ‘epigastric’ and ‘oesophagus, stomach and duodenum’ (p<0.001, PLR=2.24); ‘right flank’ and ‘urinary tract’ (p<0.001, PLR=2.84); ‘left flank’ and ‘urinary tract’ (p=0.008, PLR=2.17); ‘left flank’ and ‘dermatological’ (p=0.019, PLR=4.14); and ‘mid-lower’ and ‘intestinal’ (p<0.001, PLR=2.47). Relationships between ‘epigastric’ and ‘urinary tract’ (p=0.002, PLR=0.25), ‘mid lower’ and ‘liver and biliary tract’ (p=0.035, PLR=0.19), ‘periumbilical’ and ‘urinary tract’ (p=0.008, PLR=0.22), and ‘generalised’ and ‘oesophagus, stomach and duodenum’ (p=0.034, PLR=0.17) yielded p<0.05 in the χ2 and PLRs<0.5. NLRs ranged from 0.5 to 1.5, with the exception of the relationship between ‘left flank’ and ‘dermatological’ (p=0.020, PLR=0.40). Additionally, relationships between the sites of abdominal pain and the organs involved in patients who underwent CT, and in patients whose diagnoses were made solely on all the information in medical charts, are shown in online supplementary 4 and 5, respectively.
Moderate but statistically significant relationships between some sites of abdominal pain and the organs involved were identified in the current study. In the presence of pain at a given site, PLRs ranging from 2 to 4 can be interpreted as indicators of increased probability of disease in a certain organ, by approximately 15%–25%.19 In the present study, PLRs between ‘right subcostal’ and ‘liver and biliary tract’; between ‘right subcostal’ and ‘musculoskeletal’ ; between ‘epigastric’ and ‘oesophagus, stomach and duodenum’; between ‘right or left flank’ and ‘urinary tract’; between ‘left flank’ and ‘dermatological’; and between ‘mid-lower’ and ‘intestinal’ ranged from 2.17 to 4.14 (p<0.05). Therefore, the presence of abdominal pain at the aforementioned sites may increase post-test probability when other information such as medical history is factored into calculations, and there may be important associated indications with regard to identifying the organs involved.
Relationships between ‘right subcostal’ and ‘liver and biliary tract’ and between ‘epigastric’ and ‘oesophagus, stomach and duodenum’, which were significant at p<0.05 and PLR ≥2 in both the present study and that reported 20 years ago,1 may be particularly useful in the context of identifying the organs involved in cases of abdominal pain, given the apparent reproducibility of the results. PLRs ranging from 0.1 to 0.2 in the presence of pain at a given site can be interpreted as indicators of decreased probability of disease in a certain organ, by approximately 30% to 45%.19 In the present study, the PLRs between ‘epigastric’ and ‘urinary tract’; between ‘mid-lower’ and ‘liver and biliary tract’; between ‘periumbilical’ and ‘urinary tract’; and between ‘generalised abdomen’ and ‘oesophagus, stomach and duodenum’ were low, ranging from 0.17 to 0.25 (p<0.05). Although information about the sites of abdominal pain is considered useful for identifying organs that are involved and/or making accurate diagnoses on that basis, such information may also be useful for excluding the involvement of some organs. Notably, all NLRs excluding that between ‘left flank’ and ‘dermatological’ ranged from 0.5 to 1.5. In the presence of pain at a given site, NLRs ranging from 0.5 to 1.0 can be interpreted as indicators of decreased probability of disease in a certain organ, by approximately 15% or less.19 Therefore, it is impossible to exclude the involvement of some organs because of the absence of abdominal pain at a site generally considered to be significantly clinically associated, such as ‘epigastric’ and ‘heart disease’ or ‘right subcostal’ and ‘acute cholecystitis’.
Major differences in patient characteristics between the present study and the study performed by Yamamoto et al in 1997 were the mean patient age and the ratio of cases in which CT was used for diagnosis.1 Notably, the mean patient age was higher in the present study than in the previous study (51.7±20.0 years vs 44.4±16.7 years1). Additionally, whereas CT was used as a definitive diagnostic modality in 47.5% of the patients in the present study, (including 80% of patients with urinary tract stone), CT was rarely used for diagnosis in the previous study1; this lack of CT use could have contributed to inaccurate diagnosis. Therefore, it is important to discuss the different distributions of the causes of abdominal pain between the present study and the previous study.1 The causes of abdominal pain and the distributions of the organs involved have changed markedly since the study by Yamamoto et al1 was conducted more than 20 years ago. Although the most commonly involved organs in the present study were ‘intestinal’ (35.3%), ‘oesophagus, stomach and duodenum’ (16.4%), and ‘urinary tract’ (10.7%), their corresponding frequencies of involvement in the previous study were 24.3%, 38.9% and 4.1%, respectively, reflecting increases in the rates of ‘urinary tract’ and ‘intestinal’ diseases and reductions in the frequencies of diseases of the ‘oesophagus, stomach and duodenum’. Increases in the frequencies of ‘urinary tract’ and ‘intestinal’ diseases and decreases in the frequencies of diseases of the ‘oesophagus, stomach and duodenum’ are considered to be related.
It has recently become possible to diagnose conditions involving the urinary tract and intestines more accurately because of advances in medical technology, particularly imaging modalities. Of the urinary tract conditions investigated by both Yamamoto et al1 and the present study, urinary tract stones exhibited the biggest increase in frequency, increasing from 2.5% of cases to 9.2% of cases. Whereas urinary tract stones were typically diagnosed via urinalysis, abdominal US or intravenous pyelography in the study by Yamamoto et al,1 abdominal CT, which is now widely available in Japan, has become the major imaging modality for definitively diagnosing the condition. CT makes it possible to accurately diagnose extremely small urinary tract stones, which were difficult to detect 20 years ago, resulting in ambiguous diagnoses such as ‘gastritis’ attributed to the presence of digestive complaints such as nausea or vomiting.1 Notably, CT was used as a definitive diagnostic modality in approximately 80% of cases diagnosed as urinary tract stones in the present study.
Similar considerations as those applicable to urinary tract stones are also applicable to intestinal diseases. It is essential to monitor changes in the frequencies of different types of intestinal diseases beyond simply monitoring the collective incidence of such conditions as a group. In the present study, the most highly represented category of intestinal conditions was ‘other intestinal diseases’ (40.8%), followed by constipation (15.2%), hyperperistalsis (3.2%), acute enteritis (2.4%) and irritable bowel syndrome (2.4%). Their corresponding frequencies in the study by Yamamoto et al1 were 10.1, 27.7, 24.4, 24.4 and 13.4%, respectively, suggesting a substantial relative increase in the frequency of ‘other intestinal diseases’ over time. ‘Other intestinal diseases’ include various conditions such as colon diverticulitis, large intestinal diverticulum bleeding, ischaemic enteritis, intestinal membrane panniculitis or non-obstructive intestinal membrane ischaemia, most of which can usually be definitively diagnosed via CT.20–23 In Yamamoto et al1 these conditions were typically diagnosed via blood tests, urinary tests or US, usually without abdominal CT, and this may be relevant to the discrepancy in the incidence of ‘other intestinal diseases’ between Yamamoto et al1 and the present study.
In the present study, many organs involved were identified in multiple abdominal sites, potentially because sites of referred pain or minor non-specific pain were included, in addition to primary sites of abdominal pain. This is a potential limitation to diagnosis of abdominal disease based on the site of abdominal pain. In such instances, we attempted to clarify the accuracy of using sites of abdominal pain to identify the organs involved, which confirmed that some sites of abdominal pain could be used to identify the organs involved.
Concerning potential limitations, the present study was performed at a tertiary medical centre at a university hospital. Although patients can visit the hospital without a referral, the study setting may have resulted in some sampling bias. Ideally, a prospective study including both primary and secondary medical centres will be conducted in the near future. Nearly one-third of patients were left without any follow-up, or were admitted to or followed up (on outpatient basis) by other hospitals. There was no age or sex restrictions, except that patients were ≥20 years of age; this age criterion could have contributed to selection bias in our results. In addition, laboratory or imaging investigations may have differed among patients. Furthermore, final diagnoses of 81 patients were determined on the basis of phone calls to those patients, or through subsequent discussions among three physicians in our department, because they had only visited our hospital once and had not received a definitive diagnosis. These aspects of the study design could have caused biasses in terms of both false negative and false positive results.
Our results differed from previous results reported more than 20 years ago by our institute, which could be attributable to the widespread acceptance and marked advancement of medical science and technology during this period. Some sites of abdominal pain could be useful for identifying the organs involved or excluding them as targets of detailed examination to make a diagnosis. However, it is possible to make an inaccurate diagnosis of ‘oesophagus, stomach and duodenum’ disease in patients with actual ‘intestinal’ or ‘urinary tract’ disease when CT is not used.
The authors thank the members of the JPCA Project for Developing Future Leaders in Research for their intellectual input and support. We also thank Drs Owen Proudfoot, Joe Barber Jr and Ryan Chastain-Gross, from Edanz Group (https://en-author-services.edanzgroup.com), for editing a draft of this manuscript.
SY and MT contributed equally.
Contributors SY, MT and NEK conceived the study. SY, MT, NEK and TMN collected the data. SY, MT and NEK determined the final diagnoses and the organs involved. SY and TMN analysed the data. MT, NEK and SI-Y reviewed the data analysis. SY wrote the initial and revised manuscript. MT, NEK and SI-Y reviewed the manuscript.
Funding The authors declare that this study was supported by a grant provided by the JPCA Project for Developing Future Leaders in Research (grant number: 2016B003).
Competing interests None declared.
Patient consent for publication Not required.
Ethics approval The study was approved by the Ethics Committee of Saga University Hospital (file number 20170108).
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement No data are available. Not available.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.