Abstract
Background and aim
Helicobacter pylori (H. pylori) is an important and common contagious human pathogen which may cause peptic ulcer and also gastric cancer. The definite diagnosis of it is made through invasive tests. Recently, non-invasive tests including serologic tests of serum and saliva have been conducted for diagnosis of H. pylori infection. In this research, the diagnostic values of serum and salivary serology were compared together to use salivary anti-H. pylori test as an alternative method in the future.
Methods
During this prospective case-control study on patients who were candidates for endoscopy and gastric biopsy from March 2015 to April 2016 in Shahid Beheshti hospital, Kashan, Iran, serum and salivary samples were obtained for measurement of Immunoglobulin G (IgG) antibody levels against H. pylori by enzyme-linked immunosorbent assay (ELISA). Histopathology was the gold standard test. Statistical analysis was performed by SPSS software version 16. Statistical tests included Kolmogorov-Smirnov, independent-samples t-test, Chi-square, Mann-Whitney U, Kruskal-Wallis, McNemar and correlation.
Results
Of 123 patients, sixty-one patients (49.6%) were H. pylori-positive according to histology. The median levels of anti-H. pylori antibodies in serum (p<0.001) and saliva (p<0.001) of H. pylori-positive cases were significantly higher than H. pylori-negative cases. Sensitivity, specificity, positive likelihood ratio, negative likelihood ratio and accuracy of serologic tests in serum were 75%, 79%, 3.5, 0.3, 77% and for saliva were 85%, 82%, 4.7, 0.18, 84% respectively.
Conclusion
Diagnostic values of salivary ELISA are comparable to serum ELISA and can be used as an alternative modality for diagnosis of H. pylori infection.
Keywords: Salivary, Serum, Helicobacter pylori, ELISA, IgG
1. Introduction
Helicobacter pylori (H. pylori), one of the most common chronic infections, can lead to chronic gastritis, gastric and duodenal ulcers and sometimes gastric lymphoma and adenocarcinoma (1–3). Colonization is usually lifelong without treatment, and its prevalence varies from 20% to 50% in developed countries to more than 80% in resource limited countries (4–6). The definitive diagnosis of H. pylori is isolation of the microbe or its detection in histology of gastric biopsy by endoscopy, which are expensive and invasive tests (7). Some serologic tests are inexpensive, rapid and acceptable with high accuracy rates and seem to be appropriate for epidemiologic studies especially in untreated cases and children (8–12). Saliva is more advantageous over serum due to its easier collection, non-invasiveness and lesser risk of blood infection (7). Anti-H. pylori antibody measurement in serum, urine and saliva among children, is not reliable due to wide variety of sensitivity and specificity of tests (13). In a study, the sensitivity of salivary IgG against H. pylori was low (5) and in another investigation, the low salivary IgG specificity was reported (1). In another survey, the sensitivity and specificity of salivary anti-H. pylori IgG were acceptable (14). Due to contradiction between the studies concerning reliability of salivary ELISA for detection of H. pylori infection and furthermore, in our country, only serum antibody test is routine, we decided to compare the diagnostic values of serum with salivary antibody tests for detection of H. pylori infection to recommend salivary ELISA as a non-invasive method instead of serum antibody test.
2. Material and Methods
2.1. Study design and population
This prospective case-controlled study was conducted on cases who were candidates for upper endoscopy from March 2015 to April 2016 in Kashan Shahid Beheshti hospital. A total of 123 patients with age range of 19 to 89 years participated in the study. Indications for endoscopy included abdominal discomfort or pain, nausea and vomiting with epigastric pain and dyspepsia. Inclusion criteria included the patients greater than or equal to18 years of age and those who were referred for endoscopy according to the aforementioned symptoms. Exclusion criteria included use of any antibiotics, bismuth, non-steroidal anti-inflammatory drugs (NSAIDs), corticosteroids or proton pump inhibitors (PPIs) during the previous 2 weeks, recent upper gastrointestinal (GI) bleeding, pregnancy, acute or chronic renal failure, chronic liver disease, age less than 18 years, gastric carcinoma and diabetes mellitus. Twenty cases were excluded before entering the study (10 patients used PPIs, 5 patients used NSAIDs and 5 cases refused to undergo endoscopy). Following enrollment, no individuals were excluded from the study. The patients were chosen according to convenient sampling strategy. The patients were divided regarding biopsy samples into two groups (H. pylori-negative and H. pylori-positive groups). Serum and salivary samples were taken for measurement of Immunoglobulin G (IgG) antibody levels against H. pylori by enzyme-linked immunosorbent assay (ELISA) and their diagnostic values were compared based on histopathology (gold standard test). Endoscopy was performed by one physician. Endoscopic findings included normal endoscopy, gastric ulcer (GU), duodenal ulcer (DU) and gastritis.
2.2. Sample size
Sample size was estimated according to α = 0.05, sensitivity = 89%, specificity = 82%, and d = 0.05 (14). The total sample size according to sensitivity was calculated as 151 and regarding specificity, was estimated as 228.
2.3. Research ethics
The study was approved by the Ethics Committee of Kashan University of Medical Sciences (No. 29/5/1/5021). Informed written consent was obtained from all participants and their information was written in questionnaires. All cases enrolled in the study voluntarily. Harms and benefits of the procedures were explained to the patients. For lessening of complications, an expert endoscopist and anesthesiologist were present. Furthermore, a complete resuscitation set was prepared. Any patient was free to exclude themselves from the study at any time. All information was kept confidential. No expenses were burdened to any patient.
2.4. Laboratory methods
About 2 ml of un-stimulated saliva and 3ml of serum samples were obtained from cases and stored at −20ºC. Laboratory tests were performed by a member of staff. Following thawing saliva and serum samples, the saliva samples were centrifuged for twenty minutes at 2000 r.p.m and the supernatants were extracted and anti-H. pylori IgG was measured in both samples by ELISA kit (Radim Co. Italy). Serum and saliva samples were diluted to 1:300 and 1:2 respectively. Antigen of H. pylori was diluted by coating buffer and added to wells, incubated at 37 °C for 2 hours, then washed with buffer containing 2% serum albumin and incubated at 4°C for 18 hours. Thereafter, diluted samples were added and incubated for 90 minutes at 37 °C. Then, anti-human antibody supplemented with horseradish peroxidase was added and incubated for 1 hour at 37 °C, and tetra methylbenzidine was mixed to it. Prolongation of colorimetric reaction was done for 15 minutes at room temperature and finally, 50 μ liters of 4 NH2SO4 was mixed to wells. The optical density (OD) at 450 nm was recorded by an automated reader.
2.5. Histopathology
Two biopsy samples were taken from gastric corpus and antrum, and were fixed in 10% formalin and were stained by Giemsa and hematoxylin-eosin to determine the chronic gastritis and other histopathologic changes. Immunochemistry (gold standard test) was performed to detect H. pylori infection with 1/100 dilution (Biocare Med., California, USA). Histopathologic evaluations were conducted by one pathologist.
2.6. Statistical analysis
Data were entered into SPSS software version 16 (Chicago, Illinois, USA). Normal distribution of continuous data was detected by Kolmogorov-Smirnov test. Continuous data with normal distribution were described by mean ± SD, and continuous data without normal distribution were reported as median (range) or mean ± SE. Anti-H. pylori antibody levels in serum and saliva were indicated by median (range) because they had no normal distributions. Categorical values were described by frequency and percent. Comparison of antibody concentrations between two groups (H. pylori positive and negative groups) and sexes was made by Mann-Whitney U test. The difference of antibody titers among endoscopic findings was detected by Kruskal-Wallis test. The relation between age and antibody levels was estimated by Pearson correlation. Association between age and H. pylori infection was determined by independent-samples t-test. Chi-square test determined the association of sex and endoscopic findings with H. pylori infection. Cut-off value was chosen as 10 U/ml according to recommended ELISA kit for serum and salivary samples and according to it, true positive, false positive, false negative and true negative rates were estimated and used for calculation of diagnostic values (sensitivity, specificity and others) of two sample tests. Comparison of the sensitivity and specificity of serum and salivary ELISA tests was made by McNemar test. P < 0.05 was considered statistically significant.
3. Results
Of 123 cases, 63 (51.2%) males and 60 (48.8%) females with mean age of 41.3±16.3 years (range of 19–89 years) were enrolled in the present study. Sixty-one patients (49.6%) were H. pylori positive according to histology. Endoscopic findings were included: 11 (8.9%) normal, 33 (26.8%) gastric ulcers, 55 (44.7%) duodenal ulcers and 24 (19.5%) gastritis. Positive H. pylori was reported in 3 (27.3%) of normal endoscopies, 18 (54.5%) of gastric ulcers, 27 (49.1%) of duodenal ulcers and 13 (54.2%) of cases with gastritis (p=0.4). Antibody levels [median (range)] in serum were significantly higher in H. pylori-positive patients [19.3 (10.1–38)] rather than H. pylori-negative patients [4.5 (2.6–9.4)] (p<0.001). Also, antibody levels [median (range)] in saliva were significantly higher in H. pylori-positive subjects [16.5 (12.6–32.9)] rather than H. pylori-negative subjects [4.5 (1.8–7.7)] (p<0.001). Positive serum and salivary H. pylori IgG were 48% and 50.4% respectively. There was no sex preponderance for H. pylori infection (p=0.4) or serum and salivary antibody levels (p=0.73 and p=0.83 respectively). Also, there was no significant association between sex and serum and salivary IgG positivity (p=0.24 and p=0.17). Table 1 shows the diagnostic values of the ELISA tests in serum and saliva. According to Table 2, there was no significant association between endoscopic findings and antibody titers of serum and saliva (p=0.72 and p=0.95 respectively). The relationship between age and antibody concentrations of serum and saliva was not significant (r= −0.08, p=0.4 and r=0.15, p=0.09 respectively). No significant association was detected between age and H. pylori infection (p=0.8). By the McNemar test, the sensitivity and specificity of serum and salivary antibody titers were not different statistically, between H. pylori positive and negative groups (p=0.27 and p=0.82 respectively).
Table 1.
Diagnostic values of anti-H. pylori IgG in saliva and serum
| Variables | Saliva | Serum |
|---|---|---|
| Sensitivity | 85% | 75% |
| Specificity | 82% | 79% |
| Positive Predictive Value | 83% | 78% |
| Negative Predictive Value | 85% | 77% |
| LR− | 0.18 | 0.3 |
| LR+ | 4.7 | 3.5 |
| Accuracy | 84% | 77% |
H. pylori, Helicobacter pylori; LR−, negative likelihood ratio; LR+, positive likelihood ratio
Table 2.
Association of endoscopic findings with serum and salivary antibody levels
| Endoscopic findings | Sample | Antibody levels (mean ± SE) | p-value1 | p-value2 |
|---|---|---|---|---|
| Normal | Serum | 13.8±6.8 | 0.72 | 0.95 |
| Saliva | 16.8±9.2 | |||
| Gastric ulcer | Serum | 20.7±6.4 | ||
| Saliva | 16.3±3 | |||
| Duodenal ulcer | Serum | 21.2±5.1 | ||
| Saliva | 21.3±5.03 | |||
| Gastritis | Serum | 27.4±8.7 | ||
| Saliva | 19.08±5.9 |
SE, standard error; p-value1, for serum antibody level; p-value2, for salivary antibody level.
4. Discussion
In the present study, the salivary sensitivity and specificity for detection of H. pylori infection were 85% and 82% respectively, and serum sensitivity and specificity were 75% and 79% respectively, that all of them were acceptable. Our results showed 49.6% prevalence of H. pylori infection, which was less than 68.9% reported in Egypt (5). Hooshmand et al. reported 50% prevalence of H. pylori infection which was in line with ours (14). During another study, 52% of patients were H. pylori positive which was compatible with the current investigation (3). In an Australian survey, prevalence rate was 21.5% which was less than present results (15). The cause of these varieties may be due to different socioeconomic levels between communities. In the present study, there was no sex predominance in prevalence of H. pylori infection, whilst Yucel et al. (16) indicated a higher prevalence rate among females but Abdul Rahim et al. reported male predominance (15). Our results indicated no association between sex and positivity of serum and salivary IgG against H. pylori. In an investigation, salivary IgG positivity was associated with female gender, which was not in line with ours (17). The current results showed no relationship between age and serum and salivary antibody levels. We also found no association between age and H. pylori infection. Shu et al. indicated a significant association between H. pylori infection with age, male gender and family members (18). This study showed 48% and 50.4% positive serum and salivary H. pylori IgG respectively. She et al. reported 35.6% positive serum H. pylori IgG, which was less than ours (19). In an investigation in India, the positivity rate of H. pylori IgG in saliva was reported as 67.5%, which was more than our results (1). The current results regarding serum H. pylori IgG, revealed a sensitivity of 75% and specificity of 79%, but Khalilpour et al. found the sensitivity and specificity of serum IgG up to 100%, which was substantially more than us (20). An investigation in Iran, showed high sensitivity (90.2%) and low specificity (61.1%) of serum H. pylori IgG, which were incompatible with our results (21). Rahman et al. reported sensitivity and specificity of serum IgG against H. pylori about 96.7% and 42.8% respectively, that the sensitivity was more than our findings and the specificity was lesser than our results (22). In current work, the salivary IgG revealed a sensitivity and specificity of 85% and 82% but El-Mekki et al. indicated a high sensitivity (95%) and low specificity (70%) of ELISA test (23). In a survey of Krishnaswamy et al., sensitivity (79.31%) of salivary IgG was acceptable, but specificity (63.64%) was low (1). Hooshmand et al. reported sensitivity and specificity of salivary anti-H. pylori IgG as about 88.6% and 81.1% respectively that the sensitivity was higher than ours (14). During a survey, the sensitivity and specificity of salivary antibody for detection of H. pylori were 80.9% and 95.3% respectively, that the specificity was significantly higher than ours (24). El-Mekki et al. found no significant difference between serum and salivary antibody levels regarding endoscopic findings, which was concordant with our study (23). In this research, serum and salivary antibodies for diagnosis of H. pylori were acceptable with no statistical difference, while El-Fakhfakh et al. found that the salivary antibodies were more sensitive and specific than serum antibodies (5). Leal et al. revealed no significant difference between sensitivity and specificity of serum and saliva which was consistent with this study (10). Luzza et al. performed a multicenter investigation and found that salivary ELISA had significantly lower sensitivity, specificity and accuracy compared with serum ELISA, which was incompatible to our results (25). The causes of various sensitivities and specificities in the aforementioned studies, may be due to different antigens contained in commercial kits, diversity of ELISA kit efficacy in different geographic areas, variant choices of cut-off values, contamination of samples with other bacteria, concomitant infections and diversity of technical methods. Prevalence of H. pylori can affect negative and positive predictive values of the serologic tests. Some factors which may increase false negative results include recent infection before rising of antibody and probable degradation of antibody by salivary protease. Cross-reaction of oral normal flora antigens, recent antibiotic use and inappropriate gastric biopsy may lead to false-positive results. The strong point of our survey includes detecting associations between serologic tests and some factors such as age, sex and endoscopic findings. The limitations of the present study include: 1) Use of histology as the only gold standard test and, because of probable false positive and negative results, some other tests such as culture, PCR, rapid urea test (RUT) or urea breath test (UBT) would be better to use accompanied by histology. 2) Limited sample size. 3) We did not evaluate some risk factors of H. pylori infection and positivity of serologic tests such as family members and parental education. 4) We, and most of the other cited studies evaluated the prevalence of H. pylori infection in suspected cases, not the general population. So, more comprehensive multicenter investigations with larger sample sizes that represent the prevalence of H. pylori infection in the general population, and evaluation of risk factors for H. pylori infection and its serologic positivity are recommended in the future.
5. Conclusions
Diagnostic values of salivary IgG tests for detection of H. pylori infection are comparable with serum IgG ELISA, and may be used as an appropriate alternative to serum serology. It is an inexpensive, rapid, non-invasive and easy modality. Although invasive methods have higher sensitivities and specificities, salivary IgG tests may be a valuable aid for epidemiologic studies, screening of dyspeptic patients and finally, reducing the undue endoscopies. Furthermore, it is more acceptable for children rather than a serum serology test.
Acknowledgments
The authors thank Kashan University of Medical Sciences for financial support with grant number 94109. We are grateful to the microbiology and pathology personnel of Kashan Shahid Beheshti hospital for their cooperation. We also express our gratitude for help from the endoscopy unit staff at Shahid Beheshti hospital in conducting our study.
Footnotes
iThenticate screening: February 16, 2017, English editing: March 19, 2017, Quality control: December 10, 2017
Conflict of Interest:
There is no conflict of interest to be declared.
Authors’ contributions:
All authors contributed to this project and article equally. All authors read and approved the final manuscript.
References
- 1.Krishnaswamy RT, David CM, Govindaiah S, Krishnaprasad RB, Jogigowda SC. Salivary IgG assay to detect Helicobacter pylori infection in an Indian adult population. Indian J Dent Res. 2012;23(5):694–5. doi: 10.4103/0970-9290.107416. [DOI] [PubMed] [Google Scholar]
- 2.Petersen AM, Krogfelt KA. Helicobacter pylori: an invading microorganism? A review. FEMS Immunol Med Microbiol. 2003;36(3):117–26. doi: 10.1016/S0928-8244(03)00020-8. [DOI] [PubMed] [Google Scholar]
- 3.Sonmezoglu M, Baysal B, Ergen A, Barut SG. Detection and evaluation of salivary antibodies to Helicobacter pylori in dyspeptic patients. Int J Clin Pract. 2005;59(4):433–6. doi: 10.1111/j.1368-5031.2005.00495.x. [DOI] [PubMed] [Google Scholar]
- 4.Mégraud F, Floch P, Labenz J, Lehours P. Diagnosis of Helicobacter pylori infection. Helicobacter. 2016;21(Suppl 1):8–13. doi: 10.1111/hel.12333. [DOI] [PubMed] [Google Scholar]
- 5.El-Fakhfakh EA, Montasser IF, Khalifa RA. Evaluation of salivary and serum anti-Helicobacter pylori in Egyptian patients with H. pylori related gastric disorders. J Egypt Soc Parasitol. 2014;44(1):275–83. doi: 10.12816/0006466. [DOI] [PubMed] [Google Scholar]
- 6.Kusters JG, van Vliet AH, Kuipers EJ. Pathogenesis of Helicobacter pylori infection. Clin Microbiol Rev. 2006;19(3):449–90. doi: 10.1128/CMR.00054-05. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Megraud F, Lehours P. Helicobacter pylori detection and antimicrobial susceptibility testing. Clin Microbiol Rev. 2007;20(2):280–322. doi: 10.1128/CMR.00033-06. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Cockburn M, Collett J, Cox B. Validation of the saliva-based H. pylori test, heliSAL, and its use in prevalence surveys. Epidemiol Infect. 2001;126(2):191–6. doi: 10.1017/s0950268801005349. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.de Angelis GL, Cavallaro LG, Maffini V, Moussa AM, Fornaroli F, Liatopoulou S, et al. Usefulness of a serological panel test in the assessment of gastritis in symptomatic children. Dig Dis. 2007;25(3):206–13. doi: 10.1159/000103886. [DOI] [PubMed] [Google Scholar]
- 10.Leal YA, Flores LL, Garcia-Cortes LB, Cedillo-Rivera R, Torres J. Antibody-based detection tests for the diagnosis of Helicobacter pylori infection in children: a meta-analysis. PLoS One. 2008;3(11):e3751. doi: 10.1371/journal.pone.0003751. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Locatelli A, Catapani WR, Gomes CR, Jr, Silva CB, Waisberg J. Detection of anti-Helicobacter pylori antibodies in serum and duodenal fluid in peptic gastroduodenal disease. World J Gastroenterol. 2004;10(20):2997–3000. doi: 10.3748/wjg.v10.i20.2997. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Versalovic J. Helicobacter pylori. Pathology and diagnostic strategies. Am J Clin Pathol. 2003;119(3):403–12. [PubMed] [Google Scholar]
- 13.Koletzko S, Jones NL, Goodman KJ, Gold B, Rowland M, Cadranel S, et al. Evidence-based guidelines from ESPGHAN and NASPGHAN for Helicobacter pylori infection in children. J Pediatr Gastroenterol Nutr. 2011;53(2):230–43. doi: 10.1097/MPG.0b013e3182227e90. [DOI] [PubMed] [Google Scholar]
- 14.Hooshmand B, Monsef A, Amirmadglesi M, Kashani M. Oral Fluid Antibody Detection in the Diagnosis of Gastric Helicobacter pylori Infection. Iran J Allergy Asthma Immunol. 2004;3(3):127–31. 03.03/ijaai.127131. [PubMed] [Google Scholar]
- 15.Abdul Rahim NR, Benson J, Grocke K, Vather D, Zimmerman J, Moody T, et al. Prevalence of Helicobacter pylori infection in newly arrived refugees attending the Migrant Health Service, South Australia. Helicobacter. 2017;22(2) doi: 10.1111/hel.12360. [DOI] [PubMed] [Google Scholar]
- 16.Yucel T, Aygin D, Sen S, Yucel O. The prevalence of Helicobacter pylori and related factors among university students in Turkey. Jpn J Infect Dis. 2008;61(3):179–83. [PubMed] [Google Scholar]
- 17.Broutet N, Sarasqueta AM, Sakarovitch C, Cantet F, Lethuaire D, Megraud F. Helicobacter pylori infection in patients consulting gastroenterologists in France: prevalence is linked to gender and region of residence. Eur J Gastroenterol Hepatol. 2001;13(6):677–84. doi: 10.1097/00042737-200106000-00011. [DOI] [PubMed] [Google Scholar]
- 18.Shu X, Ping M, Yin G, Jiang M. Investigation of Helicobacter pylori infection among symptomatic children in Hangzhou from 2007 to 2014: a retrospective study with 12,796 cases. PeerJ. 2017;5:e2937. doi: 10.7717/peerj.2937. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.She RC, Wilson AR, Litwin CM. Evaluation of Helicobacter pylori Immunoglobulin G (IgG), IgA, and IgM serologic testing compared to stool antigen testing. Clin Vaccine Immunol. 2009;16(8):1253–5. doi: 10.1128/CVI.00149-09. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Khalilpour A, Santhanam A, Wei LC, Saadatnia G, Velusamy N, Osman S, et al. Antigenic proteins of Helicobacter pylori of potential diagnostic value. Asian Pac J Cancer Prev. 2013;14(3):1635–42. doi: 10.7314/apjcp.2013.14.3.1635. [DOI] [PubMed] [Google Scholar]
- 21.Jalalypour F, Farajnia S, Somi MH, Hojabri Z, Yousefzadeh R, Saeedi N. Comparative Evaluation of RUT, PCR and ELISA Tests for Detection of Infection with Cytotoxigenic H. pylori. Adv Pharm Bull. 2016;6(2):261–6. doi: 10.15171/apb.2016.036. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Rahman SH, Azam MG, Rahman MA, Arfin MS, Alam MM, Bhuiyan TM, et al. Non-invasive diagnosis of H pylori infection: evaluation of serological tests with and without current infection marker CIM. World J Gastroenterol. 2008;14(8):1231–6. doi: 10.3748/wjg.14.1231. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.El-Mekki A, Kumar A, Alknawy B, Al-Ammari O, Moosa R, Quli S, et al. Comparison of enzyme immunoassays detecting Helicobacter pylori specific IgG in serum and saliva with endoscopic and biopsy findings in patients with dyspepsia. Indian J Med Microbiol. 2011;29(2):136–40. doi: 10.4103/0255-0857.81793. [DOI] [PubMed] [Google Scholar]
- 24.Bode G, Marchildon P, Peacock J, Brenner H, Rothenbacher D. Diagnosis of Helicobacter pylori infection in children: comparison of a salivary immunoglobulin G antibody test with the [(13)C]urea breath test. Clin Diagn Lab Immunol. 2002;9(2):493–5. doi: 10.1128/CDLI.9.2.493-495.2002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Luzza F, Imeneo M, Marasco A, Crotta S, Ierardi E, Usai P, et al. Evaluation of a commercial serological kit for detection of salivary immunoglobulin G to Helicobacter pylori: a multicentre study. Eur J Gastroenterol Hepatol. 2000;12(10):1117–20. doi: 10.1097/00042737-200012100-00008. [DOI] [PubMed] [Google Scholar]