At least 75% of individuals infected with the human immunodeficiency virus (HIV) can be expected to develop the acquired immunodeficiency syndrome (AIDS) or severe symptomatology within 10 years of infection (1). Reverse transcriptase inhibitors have been effective in delaying disease progression (2), and novel drugs, including protease inhibitors, show promise in this regard (3). However, with the recognition that overall survival has changed relatively little (4), some attention is now focused on treating symptoms and improving quality of life (QOL) (5). Zidovudine, didanosine, and analogous drugs can reduce symptoms in patients with advanced HIV infection (6), but this benefit can be limited or outweighed by side effects (7). Furthermore, many patients have contraindications to antiretroviral therapy, decline such therapy, or continue to have severe symptoms despite treatment. Persons with HIV infection often seek care from unconventional or “alternative” health care providers to alleviate such symptoms (8), and many take Chinese medicinal herbs.
It is difficult to estimate how many persons with HIV infection take Chinese herbs, since most patients do not inform their physicians that they are receiving alternative therapies (9). Eisenberg et al. estimate that 3% of Americans take herbal remedies each year, and that possibly only 10% of these seek care from a provider of herbal medicine (9). Among cancer patients, the absence of definitive therapy and disillusionment with standard medical practice are associated with increased use of unconventional therapies (10). Studies confirm similarly high rates among those infected with HIV. Cohen et al. reported that of persons receiving conventional medical therapy for HIV infection in Boston, 73% were also using some form of unconventional therapy (11). Kassler et al. reported that 22% of a sample of outpatients with HIV infection in San Francisco had used herbal products in the previous 3 months (12).
Traditional Chinese medicine views the body as an energy system and emphasizes abnormalities in the balance or flow of energy as the sources of disease (13). One mainstay of Chinese medicine is the administration of medicinal herbs. An extensive but largely uncontrolled and inaccessible body of research exists regarding the use of herbs to promote immune system function, including their use as adjunctive therapy in the treatment of malignancies and infections (14,15). Some of the thousands of herbs in the Chinese pharmacopoeia have proved to contain biochemically active agents (16). For others, in vitro immunomodulatory effects have been demonstrated (17-19). Crude extracts of some Chinese herbs have antiviral activity, reducing HIV replication in human H9 cell lines (20). Whether such in vitro effects translate systematically into predictable clinical benefits is unclear, however.
There have been several descriptive studies of the use of Chinese herbs and acupuncture in the treatment of HIV-related symptoms, but no randomized controlled trials of their efficacy. Researchers have described subjective improvements in constitutional symptoms such as fatigue, night sweats, and diarrhea in large cohorts (21,22). No group studying herbal treatment for HIV has collected data systematically with standardized endpoints and comparison with controls. All have used individualized herbal prescriptions, making generalization impossible. Recent reports in the Western medical literature have emphasized the potential dangers of ingesting herbal preparations, particularly in unsupervised circumstances. Common herbs such as comfrey, germander, and glycyrrhiza (licorice) (23-25), and some widely available combinations (26), have been associated with serious and sometimes fatal liver disease. We undertook the present study both as a preliminary, controlled investigation of the short-term efficacy and safety of a standardized herbal preparation in the treatment of HIV-related symptoms and to determine the feasibility of studying Chinese herbal therapy with Western clinical investigative methods.
We studies subjects with moderate immune dysfunction who were likely to have significant symptomatology but were unlikely to require potentially confounding medication changes during the study period. Subjects were recruited at the San Francisco General Hospital (SFGH) AIDS Clinic, the SFGH Substance Abuse Services HIV Clinic, and the Quan Yin Healing Arts Center and through advertisements in the local press. Eligible subjects were aged ≥18 years, English- or Spanish-speaking, and able to give informed consent. Included subjects were HIV-positive, had absolute CD4+ lymphocyte counts between 0.200 × 109/L (200/mm3) and 0.500 × 109/L (500/mm3), had no prior AIDS-defining diagnosis by 1987 revised Centers for Disease Control (CDC) criteria (27) except limited mucocutaneous Kaposi's sarcoma, (KS), had received either a stable antiretroviral regimen or no antiretroviral drugs for the past 12 weeks, and were currently experiencing at least two HIV-related symptoms. Subjects were excluded if they had any history of sensitivity to Chinese herbs; any clinical evidence of active opportunistic infection; or any significant hematologic, renal, or hepatic laboratory abnormalities; were presently receiving herbal treatment or had used herbs in the past 6 months for an HIV-related condition; or were participating in a blinded antiviral drug trial.
Subjects were stratified by CD4 count into two blocks (0.200-0.349 × 109/L, and 0.350-0.499 × 109/L) and were then assigned to treatment groups using a random number series for each block. All study personnel, pharmacists, and subjects remained blinded to treatment assignment. One group received a standardized oral preparation of Chinese herbs, “IGM-1,” a combination of the Enhance and Clear Heat preparations developed by one of the investigators (M.R.C.) for treatment of HIV-related symptoms. Of 31 herbal ingredients in the 650-mg tablet, those present in high concentration included Ganoderma lucidum, Isatis tinctoria, Astragalus membranaceus, Andrographis paniculata, Lonicera japonica, Milletia reticulata, Oldenlandia diffusa, and Laminaria japonica. The other group received a 650-mg placebo, composed primarily of microcrystalline cellulose to approximate the fiber content of the herbal preparation. Both tablets, prepared by Health Concerns (Alameda, CA, U.S.A.), were coated with titanium dioxide so that they would be odorless and tasteless. Subjects were instructed to take seven pills four times a day on an empty stomach for a period of 12 weeks.
Initial physical examination was performed by one investigator (J.H.B.) and was repeated at the end of the treatment period. Each subject's weight was recorded on the same standardized scale at each visit. Confirmatory HIV testing by enzyme-linked immunosorbent assay (ELISA) and Western blot was performed at the initial visit (University of California HIV Diagnostic Laboratory, Davis, CA, U.S.A.). Initial, midpoint, and final laboratory studies included complete blood count with differential and platelets, and a multiphasic chemistry panel including liver and renal function tests and electrolytes; absolute CD4 count and percentage were obtained at initial and final visits (MetWest Clinical Laboratories, San Jose, CA, U.S.A.).
A standardized structured interview was administered at the initial and final visits. Subjects were asked to consider symptoms and perceptions during the 4 weeks preceding the interview. Life satisfaction and average health were determined with scales of 0-100 (28). Overall health perception was assessed on a 5-point scale. Severity of each reported symptom was assessed on a 5-point scale, with use of a 36-item checklist of common HIV-related symptoms developed by the investigators. Physical and social function, mental health, and bodily pain were measured using a modified Medical Outcomes Study short form (MOS-SF-36), validated for persons with HIV infection (5,29,30). Adherence was assessed with a single question asking how many pills, on the average, the subjects had taken, followed by an inventory of reasons for nonadherence. Depression was measured by the CES—D (Centers for Epidemiologic Studies—Depression scale) (31), and anxiety was measured by the Spielberger Anxiety Trait Scale (32).
All analyses were by intention to treat. For subjects who did not return for follow-up or who omitted items, relevant variables were assumed not to have changed. Baseline characteristics are reported as the treatment group mean ± SE, and mean changes from baseline are reported with 95% confidence intervals (CI). Standard two-tailed tests of homogeneity were applied to comparisons between means for subjects on herbs versus those on placebo. Continuous variables were also analyzed nonparametrically by the Wilcoxon rank-sum test. All subjective scales had upper and lower bounds so that their means could not be influenced by extreme values, and means were very similar to medians. Results and p-values did not differ substantially according to statistical method, and for simplicity, only means and the results of parametric tests are displayed.
Power calculations a priori indicated that 93 subjects per treatment group would be required to detect (with 80% power at the α = 0.05 level) a clinically significant difference of at least 20% between the percentages of subjects who reported symptomatic improvement. This estimate was based on the assumption that because of the placebo effect and regression to the mean ≈50% of symptomatic subjects would report improvement with placebo treatment, and also on unpublished observations by one investigator (M.R.C.) suggesting that 70-80% report improved well-being on herbal therapy. Funding constraints limited the study to a total of 30 subjects. With 15 subjects per arm and the same power assumptions, the minimum statistically detectable effect size would have been improvement in 45-50% of subjects receiving herbal treatment over and above the percentage improving with placebo. Despite the low probability of demonstrating so large a difference, we chose to proceed with a pilot study, both to test the study methods and to identify any trends of potential clinical significance.
Thirty subjects were enrolled and randomly assigned to receive herbs (n = 15) or placebo (n = 15). Baseline subject characteristics are shown in Table 1. Median baseline CD4 count was 0.303 × 109/L (303/mm3, range 210-408/mm3). Subjects experienced a mean of 15 HIV-related symptoms, and mean symptom severity was 0.7 ± 0.1 of a possible 5. On functional subscales, there was little impairment in activities of daily life or in cognitive function, but slightly more limitation of role functioning and mental health; subjects did report that their health interfered substantially with social functioning. Levels of anxiety and depression were moderate.
There were no statistically significant differences between the treatment groups in any demographic or baseline laboratory parameters or in initial health perception, functional status, cognitive function, or QOL scales (Table 1). Subjects randomized to herbs had slightly lower baseline hemoglobin, reported almost four more symptoms on the average, and scored lower on the anxiety scale (0.10 < p < 0.20 for these comparisons). Organ system subscales showed similar trends (0.10 < p < 0.20) toward greater numbers of gastrointestinal, neurologic, and respiratory symptoms and sleep disturbances among subjects randomized to herbs and toward more dermatologic symptoms among those randomized to placebo (data not shown).
Follow-up data were complete for 29 subjects (97%). One subject refused to return for the second and third study visits. Another had significant diarrhea after 2 weeks of treatment and chose to discontinue study medications. A third experienced an unexplained, transient fourfold increase in hepatic transaminase levels at 6 weeks; study medications were withheld for 2 weeks and reinitiated after liver enzymes had normalized, without further abnormalities. All three of these subjects were later determined to have been randomized to placebo. No toxicities were reported or detected among subjects randomized to herbs. Subjects reported taking 26.3 ± 0.5 of the prescribed 28 tablets each day, and there was no difference in adherence between treatment groups (Table 2). Reasons most commonly given for missing pills included forgetting and being away from home, rather than any difficulty with the regimen.
Although there were significant within-group changes from baseline to final study visit, no differences between treatment groups reached the p < 0.05 level of statistical significance (Table 2). Life satisfaction improved in subjects treated with herbs but not in those receiving placebo, and there was a trend toward greater improvement among herb-treated subjects (p = 0.12). Subjects treated with herbs reported a reduction in symptoms, whereas there was no change among those receiving placebo, (p > 0.20 for the between-group comparison). There were similar trends (0.10 < p < 0.20) toward greater improvements in neurologic and gastrointestinal symptoms and sleep disturbances in subjects treated with herbs and in dermatologic symptoms among those receiving placebo. There were no differences between treatment groups in changes in symptom severity, weight, absolute CD4+ count, and reported average health (Table 2) or in the functional subscales, anxiety and depression (data not shown).
Although only 20% of herb-treated subjects and 27% receiving placebo (p > 0.20) reported that their overall health had improved, 60% of those treated with herbs reported that the treatment had helped them, as compared with 27% receiving placebo (p = 0.09 by Fisher's exact test) (Table 2). These included 8 of the 10 subjects who correctly guessed that they were receiving herbs as compared with only 1 of 5 who believed they were receiving placebo (p = 0.09). Of placebo-treated subjects, none of the 6 who guessed correctly reported that treatment had helped, as compared with 4 of 9 who believed they were receiving herbs (p = 0.10); i.e., 12 of 19 subjects (63%) who believed they were receiving herbs, but only 1 of 11 (9%) who believed they were receiving placebo said that the treatment had helped (chi-square = 8.29, p < 0.005). There were, however, no significant correlations between treatment guess and reported changes in life satisfaction, overall health, symptom scales, or other self-assessments.
Interpretation of the results of our pilot study must be tempered by its very small sample size, short duration, and reliance on subjective outcome measures. Nevertheless, it represents the first published randomized, double-blind, placebo-controlled trial of a traditional Chinese herbal therapy for HIV infection. The trial was completed with very high subject adherence and acceptance, thereby demonstrating the feasibility of using rigorous, conventional clinical research methods to study the efficacy and toxicity of unconventional therapeutic regimens. Most subjects in both groups were inclined to believe that they were receiving active herbs, and 47% guessed their assignment incorrectly, suggesting that the placebo formulation was convincing and that blinding was effective. No toxicity was observed among subjects receiving the herbal preparation, whereas two subjects receiving placebo had adverse events that may have been unrelated to the study preparations. There have been numerous recent reports of serious, even lethal, toxicity resulting from the ingestion of other herbal remedies (23-26). Caution is therefore indicated in interpreting our results, particularly because the number of subjects was so small and the study period so brief that uncommon but severe complications might have been missed.
In subjects with moderately advanced symptomatic HIV infection treated for 12 weeks, we did not demonstrate any statistically significant differences in outcomes between those receiving a standardized herbal preparation and those receiving placebo. However, we did observe trends toward greater symptomatic improvement and improved life satisfaction and health perception among those receiving herbs. Herb-treated subjects experienced a 25% increase in mean overall life satisfaction, a change of 0.75 times the baseline SD. The absolute magnitude of this change, 0.86 on a 7-point scale, corresponds to a “moderate” to “large” change in QOL as reported by patients with chronic heart and lung disease (33). In other studies in which similar scales were used, effect sizes of 0.5 SD have been considered moderate and differences of 0.8 SD have been considered large (34). Therapeutic effects of this magnitude, if borne out in larger groups of subjects, would therefore be of clinical importance. By contrast, we observed no significant change in life satisfaction among placebo-treated subjects. Similarly, herb-treated subjects reported a 13% reduction in symptom number, corresponding to the resolution of an average of 2.2 baseline symptoms, and a 17% decrease in mean symptom severity, as compared with almost no change among placebo-treated subjects. This effect was consistent, in that herb-treated subjects tended to report improvement in all categories except dermatologic symptoms, whereas placebo-treated subjects tended to report worsening of all but dermatologic symptoms.
We did not observe trends favoring herbal treatment in CD4 count and body weight, which might have bolstered the observed trends in subjective endpoints. Indeed, CD4 decrease was smaller and weight gain was slightly greater among subjects receiving placebo, although the CI around these estimates were very wide and the between-group differences not significant. Measurement of changes in the CD4 count is notoriously imprecise, particularly when it is derived from only two data points, and its course in a 3-month period may be highly variable (35). Nevertheless, future studies of herbal treatment must have adequate power to exclude the possibility of deleterious immune system activation and depletion.
The question of whether subjects feel better with treatment is difficult to address with precision, and our results provide clues as to the type of questions might be most useful. Although a subject's guess as to treatment assignment was not associated with changes in health perception or life satisfaction, most subjects who believed themselves to be receiving herbs and almost none of those who believed that they were receiving placebo reported that the treatment had helped them. Yet only half of those who said the treatment had helped also reported improvement in overall health. Therefore, in these highly motivated volunteers who presumably “believed” in herbal treatment, asking whether the treatment had helped them may have primarily assayed their beliefs about whether or not they were receiving herbs rather than tracking changes in perceived well-being. On the other hand, that improved life satisfaction and symptomatology were predicted by actual treatment assignment but not by what subjects believed themselves to be taking argues both for the validity of these measures of well-being and against the possibility that our findings resulted from inadvertent unblinding. The discrepant performances of these measures suggest the need for further refinement and careful wording of instruments used to measure symptoms.
Our study was limited by insufficient power to confirm with statistical significance even potentially large differences in treatment outcomes. A high probability of type II error is inherent in a small pilot study, and the herbal treatment may have had clinically important effects that were either not detected or not statistically demonstrable. On the other hand, subjects randomized to herbs tended to be more symptomatic at baseline, so that the more favorable changes observed might have been due simply to regression toward the mean. Therefore, the herbal treatment we studied may actually have had little or no clinical effect.
There are several possible explanations for a lack of significant therapeutic effect. First, subjects received study medication for only 12 weeks, a shorter treatment duration than that of most prescribed Chinese herbal regimens, and in a longer treatment period large differences might have become apparent. Second, differences might have been more readily apparent in more symptomatic subjects with more advanced HIV disease. Third, the particular herbal formula we used may be ineffective for the treatment of HIV-related symptoms or for the particular symptoms that were most prevalent. Fourth, Chinese herbal therapy may not work at all for this condition, although this inference is no more justified by the results of one study than is the conclusion, based on a single negative pharmaceutical trial, that no Western pharmacotherapy would be effective.
Fifth, and perhaps most important, Chinese herbs may be doomed to fail in a research setting in which they are inappropriately administered outside the usual therapeutic context. Much of the benefit of medicine, and in particular of unconventional healing practices, may arise from contextually determined factors, including but not limited to the placebo effect and the power of the therapeutic relationship (36). Furthermore, traditional Chinese practice involves individualizing the herbal remedy to a specific patient's systemic abnormalities, combining herbal therapy with acupuncture and other therapeutic modalities, and continually reevaluating and adjusting the treatment plan. We incorporated no traditional Chinese diagnostic methods, such as examination of the tongue and pulse, and no other therapeutic methods, into the study design, and there was no reevaluation of treatment. To the extent that the indications and prescriptions that define the trial diverge from those used in practice, the clinical performance of the intervention may be underestimated.
This argument suggests that evaluating unconventional clinical practices such as Chinese herbal therapy may require modifying the sequence rather than the design of clinical trials. The standard model for evaluating pharmaceuticals proceeds from tests of efficacy, reflecting maximum therapeutic potential as achieved in a highly controlled but clinically artificial setting, to tests of effectiveness, in which therapy is administered in a context that approximates its anticipated clinical use (37). This model is supported by a biochemical conception of disease, in which pharmaceutical agents have effects, although possibly attenuated ones, independent of clinical context. By contrast, “alternative” therapeutic systems may not share such explanatory beliefs. The very idea of first testing efficacy may be inappropriate when the agent is not expected to be effective outside a context that at least closely resembles the usual clinical one. This possibility would argue for deferring studies analogous to pharmaceutical efficacy trials in the evaluation of such unconventional therapies and instead undertaking, as the initial step, evaluations of actual clinical effectiveness, which by no means requires foregoing controlled, double-blind methods. Rather, this idea suggests selecting subjects by criteria consistent with the method being tested, designing more comprehensive interventions and more realistic sham practices to serve as controls, and carefully blinding subjects, investigators, and practitioners to what are likely to be highly complex treatment assignments.
We investigated an unconventional treatment for the often severe symptoms that can diminish QOL for persons with HIV infection. Many are already seeking such alternatives and are investing personal and public resources in them. Patients clearly want and need well-founded information on the safety and effectiveness of unconventional treatment methods (38). The question of whether herbal therapy, or the traditional Chinese approach more generally, is effective in the management of symptomatic HIV infection can be adequately addressed only by larger trials of longer duration that incorporate into their design features of the actual practice of Chinese medicine rather than adhering overly closely to a stepwise paradigm developed to test pharmacotherapeutic efficacy. The present study suggests the feasibility and safety of such larger trials, and outcome variables on which they might focus.
Acknowledgment: This work was supported by the AIDS Clinical Research Center of the University of California Universitywide AIDS Research Project. Dr. J. H. Burack is currently supported by a National Research Service Award at the University of Washington School of Medicine. Some materials were donated by Health Concerns, Alameda, California. We thank Mary Payne, R.N., Rowena Mah, R.N., Dr. Leslie Squires, Andrew Gaeddart, Deb Gumbley, Harlan Woodring, and Nadine Lurie, all of whom assisted in the completion of this study, and the subjects who volunteered to participate.
1. Moss AR, Bacchetti P, Osmond D, et al. Seropositivity for HIV and the development of AIDS or AIDS-related condition: three year follow up of the San Francisco General Hospital cohort. Br Med J
2. Volberding PA, Graham NM. Initiation of antiretroviral therapy in HIV infection: a review of interstudy consistencies. JAIDS
3. Danner SA, Carr A, Leonard JM, et al. A short-term study of the safety, pharmacokinetics, and efficacy of ritonavir, an inhibitor of HIV-1 protease. N Engl J Med
4. Concorde Coordinating Committee. Concorde: MRC/ANRS randomised double-blind controlled trial of immediate and deferred zidovudine in symptom-free HIV infection. Lancet
5. Wachtel T, Piette J, Mor V, Stein M, Fleishman J, Carpenter C. Quality of life
in persons with human immunodeficiency virus infection: measurement by the Medical Outcomes Study instrument. Ann Intern Med
6. Fischl MA, Richman DD, Grieco MH, et al. The efficacy of azidothymidine (AZT) in the treatment of patients with AIDS and AIDS-related complex. N Engl J Med
7. Lenderking WR, Gelber RD, Cotton DJ, et al. Evaluation of the quality of life
associated with zidovudine treatment in asymptomatic human immunodeficiency virus infection. N Engl J Med
8. Abrams DT. Dealing with alternative therapies for HIV. In: Sande MA, Volberding P, eds. The medical management of AIDS,
4th ed. Philadelphia: W.B. Saunders, 1994.
9. Eisenberg DM, Kessler RC, Foster C, Norlock FE, Calkins DR, Delbanco TL. Unconventional medicine in the United States. N Engl J Med
10. Cassileth BR, Lusk EJ, Strouse TB, Bodenheimer BJ. Contemporary unorthodox treatments in cancer medicine: a study of patients, treatments and practitioners. Arch Intern Med
11. Cohen CJ, Eisenberg DM, Mayer KH, Delbanco TL. Prevalence of nonconventional medical treatments in HIV-infected patients: implications for primary care [Abstract]. Clin Res
12. Kassler WJ, Blanc P, Greenblatt R. The use of medicinal herbs by human immunodeficiency virus-infected patients. Arch Intern Med
13. Cohen M. Chinese medicine in the treatment of chronic immunodeficiency: diagnosis and treatment. Am J Acupuncture
14. Sun Y. Chinese medicinal herbs as biological response modifiers. In: Lotzova E, Herberman RB, eds. Natural immunity, cancer and biological response modification
. Basel: Karger, 1986.
15. Sun Y. The role of traditional Chinese medicine in supportive care of cancer patients. Recent Results Cancer Res
16. Klayman DL. Qinghaosu (Artemisinin
), an antimalarial drug from China. Science
17. Sun Y, Hersh EM, Lee S-L, McLaughlin M, Loo TL, Mavligit GM. Preliminary observations on the effects of the Chinese medicinal herbs Astragalus membranaceus
and Ligustrum lucidum
on lymphocyte blastogenic responses. J Biol Response Mod
18. Chu D-T, Wong W, Mavligit GM. Immunotherapy with Chinese medicinal herbs. Reversal of cyclophosphamideinduced immune suppression by administration of fractionated Astragalus membranaceus
in vivo. J Clin Lab Immunol
19. Sun Y, Hersh EM, Talpaz M, et al. Immune restoration and/or augmentation of local graft versus host reaction by traditional Chinese medicinal herbs. Cancer
20. Chang RS, Yeung HW. Inhibition of growth of human immunodeficiency virus in vitro by crude extracts of Chinese medicinal herbs. Antiviral Res
21. Rabinowitz N. Acupuncture and the AIDS epidemic: reflections on the treatment of 200 patients in four years. Am J Acupuncture
22. Smith MO. AIDS: results of Chinese medical treatment show frequent symptom relief and some apparent long-term remissions. Am J Acupuncture
23. Ridker PM, McDermott V. Comfrey herb tea and hepatic veno-occlusive disease. Lancet
24. Larrey D, Vial T, Pauwels A, et al. Hepatitis after germander (Teucrium chamaedrys
) administration: another instance of herbal medicine hepatotoxicity. Ann Intern Med
25. Chan TYK, Chan JCN, Tomlinson B, Critchley JAJH. Chinese herbal medicines revisited: a Hong Kong perspective. Lancet
26. Woolf GM, Petrovic LM, Rojter SE, et al. Acute hepatitis associated with the Chinese herbal product Jin Bu Huan. Ann Intern Med
27. Centers for Disease Control and Prevention. Revision of the case definition for acquired immunodeficiency syndrome. MMWR
28. Cleary PD, Fowler FJ Jr, Weissman J, et al. Health-related quality of life
of persons with AIDS. Med Care
29. Ware JE. SF-36 health survey: manual and interpretation guide
. Boston: The Health Institute, New England Medical Center, 1993.
30. Wu AW, Rubin HR, Mathews WC, et al. A health status questionnaire using 30 items from the Medical Outcomes Study: preliminary validation in persons with early HIV infection. Med Care
31. Radloff LS. The CES-D scale: a self-report depression scale for research in the general population. Appl Psychol Measurement
32. Spielberger CD, Gorsuch RL, Lushene R. Self-evaluation questionnaire
. Palo Alto, CA: Consulting Psychologists Press, 1968.
33. Jaescheke R, Singer J, Guyatt GH. Measurement of health status: ascertaining the minimal clinically important difference. Control Clin Trials
34. Cohen J. Statistical power analysis for the behavioral sciences,
2nd ed. Hillsdale, NJ: Lawrence Erlbaum Associates, 1988.
35. DeGruttola V, Lange N, Dafni U. Modeling the progression of HIV infection. J Am Stat Assoc
36. Sullivan MD. Placebo controls and epistemic control in orthodox medicine. J Med Philos
37. Institute of Medicine. Assessing medical technologies
. Washington, D.C.: National Academy Press, 1985.
38. Delbanco TL. Bitter herbs: mainstream, magic, and menace. Ann Internal Med