Search Process

Two automated databases were searched: MEDLINE and EMBASE. The contents of the Cochrane Collaboration Library were reviewed manually. We also hand-searched the contents of the most recent 12 months (through spring 2000) of the five journals most likely to contain relevant articles:

• Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology and Endodontics.
• Journal of Oral Pathology and Medicine.
• Oral Diseases.
• AIDS.
• Journal of Acquired Immune Deficiency Syndromes and Human Retrovirology.

Our review of the Cochrane materials indicated related topics that were under study but not currently available as published reports. No search was made through the "gray" literature (i.e., unpublished materials).

Separate MEDLINE searches were conducted for each of the five key questions, but all five searches began with the terms "HIV" or "HIV infection" or "acquired immunodeficiency syndrome" and were limited to humans and the English language. Beyond these similarities, the searches were customized except that the search terms for Key Questions 2A and 2B were similar to each other up to a point, as were the search terms for Key Questions 3A and 3B.

We did the searches without restriction on the year of publication, but we found no relevant publications dated before 1983. All five of the MEDLINE searches started with 83,962 articles on human HIV or AIDS in the English language. Inclusion and exclusion criteria used by the reviewers are set out in the next subsection.

For Key Question 1, the search then included a list of dental procedures of interest-dental care, tooth extraction, dental procedure, gingivoplasty, periodontitis, periodontal surgery, dental implants, gingivectomy, oral surgical procedures, orthognathic surgery, dental scaling, dental prophylaxis, root scaling, root planing, root canal therapy, and pulpectomy.

We identified 145,292 articles on these topics. When cross-checked against with the English-language articles on HIV/AIDS, 767 articles were common to both search results. These were further winnowed down to 201 articles by eliminating nonresearch articles and clinical case reports. We only included studies described in such terms as controlled clinical trials, randomized controlled trials, multicenter study, epidemiologic research design, comparative study, evaluation study, outcome and process assessment, outcome assessment, or treatment outcome.

For Key Question 2A, the departure point from the first search was to specify the oral conditions of interest-mouth diseases, periodontal diseases, oral infections, hairy leukoplakia, oral leukoplakia, oral candidiasis, necrotizing ulcerative gingivitis, oral ulcers, parotid diseases, parotid gland, parotitis, or parotid swelling. This resulted in 1,719 articles when cross-checked against the English-language articles on HIV/AIDS. These were reduced further to 297 articles by restricting them to mention of HIV seropositivity, seroconversion, or AIDS serodiagnosis.

The search for Key Question 2B included searching for the same oral condition as in Key Question 2A, but restricted the studies to those mentioning immunosuppression or CD4 lymphocyte count. This yielded 133 articles.

For Key Question 3A, the search was limited to oral candidiasis and to a list of antifungal agents-fluconazole, ketoconazole, nystatin, itraconazole, clotrimazole, amphotericin B, or drug therapy. This resulted in 223 articles, a number that was further reduced to 18 studies by restricting the search with the terms primary prevention, prevention, preventive medicine, health promotion, disease prevention, dental prophylaxis, or prophylaxis.

The search strategy for Key Question 3B was similar to the search done for Key Question 3A to the point of specifying the condition and the drugs. It then restricted the articles to those dealing with drug therapy, therapy, intervention studies, intervention, treatment outcome, or treatment and yielded 143 articles.

Subsequent to the MEDLINE searches, we conducted additional searches on EMBASE. EMBASE is somewhat more limited than MEDLINE in the time period covered (only since 1988) and in the search terms available to customize queries. For this latter reason, we did only three customized searches for the five questions: Key Questions 2A and 2B were combined into a single search, as were Key Questions 3A and 3B. The vast majority of articles uncovered in these searches were duplicative of those found in the MEDLINE searches. The EMBASE searches resulted in the addition of only three articles to the bibliography.

Complications of Oral Procedures

The data from five articles were abstracted for the review of Key Question 1 comparing the complication rates associated with dental treatment in HIV-positive persons to the complication rates for similar dental treatment in individuals who are HIV negative. Only two of the seven dental procedures specifically mentioned in Key Question 1 were the objects of study in the five articles. One article examined endodontic procedures (root canal treatment), while four studies examined tooth extraction. None reported on orthognathic surgery, periodontal surgery, prophylaxis, scaling and root planing, or implants.

• In the study of endodontics, the immediate (1- to 3-month followup) postoperative complication rate was exceedingly low in the HIV-positive group and nonexistent in the control group. Only 1 of 48 patients experienced any postoperative complications; an asymptomatic HIV-positive male was found to have pain and swelling following the initial root canal treatment. He received local debridement and antibiotics, and no further complications occurred. Otherwise, no complications associated with endodontic therapy were noted in any of the patients regardless of whether they did or did not receive prophylactic antibiotics. The authors did not detect a clinically significant difference in complication rates between the two groups.
• Three of the four studies in the review of dental extractions found no statistically significant difference in complications between the HIV-positive and HIV-negative groups, although the HIV-positive groups tended to have more postoperative complications. The final study found that the HIV-positive groups had a statistically significantly higher complication rate, but with adjustment for risk factors, the difference was no longer significant. Postextraction complications included persistent bleeding, persistent pain, localized alveolitis, local wound infection, or delayed wound healing.
  
  Nevertheless, across all studies, the postoperative complications that were experienced were rather minor, and when they occurred, were treated on an outpatient basis. Finally, based on their findings, none of the four studies called for the need to take special precautions in HIV-positive patients who do not have a coagulopathy (e.g., hemophilia, thrombocytopenia, or other known bleeding disorders) and are sufficiently healthy to be seen on an outpatient basis.

Oral Conditions as Markers of Seroconversion

Only one study was abstracted in the review for Key Question 2A. The question asked for evidence on a set of specific oral conditions-oral candidiasis, hairy leukoplakia, necrotizing ulcerative periodontitis, oral ulcers, and parotid swelling-as markers for recent HIV seroconversion. The study investigated the prevalence of a variety of signs and symptoms (including oral candidiasis) among two groups of initially seronegative hospital patients. All of the patients received blood transfusions for a variety of reasons, but half were transfused with seropositive blood while the "matched controls" received seronegative blood.

The question uses the medical "testing" dimensions of sensitivity (Sn), specificity (Sp), positive predictive value (PPV), and negative predictive value (NPV) of the oral conditions as a marker of recent seroconversion as the way of evaluating their usefulness. Positive enzyme-linked immunosorbent assays (ELISAs) and Western blot tests for HIV, obtained from patients who were confirmed previously seronegative for up to 3 months after the presumed time of infection/exposure, were considered evidence of recent seroconversion. No special conditions were imposed on the diagnosis of the oral conditions.

• The majority of patients with oral candidiasis had seroconverted by the end of 3 months (PPV = 82 percent), but only a small proportion of those who had seroconverted had oral candidiasis (Sn = 14 percent). Very few who did not seroconvert had oral candidiasis (Sp = 97 percent), and most of those who did not have oral candidiasis did not seroconvert either (NPV = 57 percent).

Oral Conditions as Indicators of Severe Immunosuppression

Key Question 2B involved assessing evidence supporting the use of selected oral conditions as indicators of the progression of HIV infection to the stage of severe immunosuppression and a diagnosis of AIDS. The question suggests the use of the CD4 lymphocyte cell count as a measure of immunosuppression. The question identified the following seven oral conditions of interest: hairy leukoplakia, oral candidiasis, necrotizing ulcerative periodontitis, oral ulcers, parotid swelling, linear gingival erythema, and Kaposi's sarcoma.

The data from 10 articles were abstracted and included in the review for this question. All 10 articles reported CD4 count alone as their measure of immunosuppression.

This question also uses the medical "testing" dimensions (Sn, Sp, PPV, and NPV) of the oral conditions as an indicator of immunosuppression to evaluate their usefulness. In this context, an oral condition with high sensitivity would represent a screening tool for identifying HIV disease that has progressed to the next stage.

• When Sn (sensitivity) is high, it means that a large proportion of the persons whose HIV disease has progressed to the next stage also have the particular oral condition. Thus, when you find the oral condition, you are likely to find a more advanced stage of HIV.
• High Sp (specificity), on the other hand, indicates that a large proportion of persons whose HIV disease has not progressed to the next level also do not have the particular oral condition. In the absence of high Sn, high Sp is not particularly useful.
• High PPV (positive predictive value), on the other hand, is a measure that represents how closely the presence of the oral condition is associated with the progress of HIV to the next stage. When PPV is high, it indicates that a large proportion of the persons who clinically present with the oral condition also have HIV disease that has progressed to the next stage.
• A high NPV (negative predictive value), on the other hand, indicates that a large proportion of those without the oral condition also do not have HIV that has progressed to the next stage. In the absence of a high PPV, high NPV is not particularly useful.

The oral conditions on which studies reported were quite varied. None of the articles reported on parotid swelling; however, all 10 articles reported on oral candidiasis, six reported on hairy leukoplakia, and four on oral ulcers. Only two studies reported on linear gingival erythema, two on necrotizing ulcerative periodontitis, and three on Kaposi's sarcoma.

• There is little apparent consistency in Sns across the 10 studies of oral candidiasis. They range from 20 to 77 percent, with a small cluster of five study groups in the lower end of the range from 20 to 41 percent. The situation with respect to Sps is somewhat more consistent. They range from 65 to 97 percent, with a cluster of eight study groups in the upper end of the range from 79 to 97 percent. The PPVs range from 34 to 88 percent with eight clustered at the low end of the range from 34 to 58 percent. The range of NPVs was from 61 to 90 percent; however, all but four of the NPVs were clustered between 84 and 90 percent.
• Across the six studies of hairy leukoplakia, Sns and Sps were fairly consistent. The Sns ranged from 13 to 24 percent and the Sps ranged from 83 to 95 percent. PPVs ranged from 29 to 70 percent, with four below 50 percent, and NPVs ranged from 51 to 86 percent.
• The Sns reported in the two studies of necrotizing ulcerative periodontitis were very low, ranging from 0 to 16 percent, whereas the Sps were very high (92 to 99 percent). With the exception of one study group that had no cases of this condition, the PPVs ranged from 80 to 95 percent. The NPVs ranged from 52 to 70 percent.
• In the four studies examining oral ulcers, Sns ranged from 0 to 21 percent, while the Sps ranged from 88 to 100 percent. PPVs ranged from 0 to 100 percent and NPVs ranged from 52 to 84 percent.
• The two studies of linear gingival erythema had low Sns (13 to 34 percent) but relatively high Sps (62 to 90 percent). The PPVs ranged from 18 to 55 percent, while the NPVs ranged from 61 to 82 percent.
• Sns across the three studies of Kaposi's sarcoma were low and ranged from 7 to 12 percent. However, the Sps were extremely high, ranging from 99 to 100 percent. The PPVs were high as well, ranging from 73 to 100 percent Also, the NPVs were also reasonably high, ranging from 61 to 84 percent.

Efficacy of Antifungals as Prophylaxis for Oral Candidiasis

There were six clinical trials that examined the efficacy of available antifungals to prevent oral candidiasis in persons who are HIV positive. Only two of the six available antifungals were studied in this context. There were five studies of fluconazole and one of nystatin pastilles.

• Fluconazole was statistically significantly more efficacious than placebo in preventing recurrences or new infections over a range of 3 to 17 months, when studied at doses from 100 mg/day to 100 mg/week. Gastrointestinal disorders were the most common, but tolerable, side effects.
• Nystatin pastilles at 200,000 units/day and 400,000 units/day were efficacious at preventing new or recurrent oropharyngeal candidal infections, with the higher dose being more effective. Few side effects were reported.

Effectiveness of Antifungals as Treatment for Oral Candidiasis

Twelve clinical trials evaluated the effectiveness of the six available antifungal drugs in the treatment of oral candidiasis in persons with HIV/AIDS. No studies included an evaluation of amphotericin B suspension, and fluconazole was the most studied of the five remaining.

• Fluconazole appears to be from 88 to 100 percent effective in obtaining a complete clinical response after 14 days of therapy and from 53 to 76 percent effective in obtaining a culture negative for Candida species.
• Itraconazole appears to be roughly equivalent in effectiveness to fluconazole, with ketoconazole achieving the same or slightly lower response rates.
• Fluconazole and itraconazole are more effective at managing oropharyngeal candidiasis than are nystatin or clotrimazole, particularly when mycological response rates and relapse rates are taken into account.