Evidence Report/Technology Assessment

Number 17

Prepared for:
Agency for Healthcare Research and Quality
Department of Health and Human Services
U.S. Public Health Service
2101 East Jefferson Street
Rockville, MD 20852

www.ahrq.gov

Contract No. 290-97-006

Prepared by:
Johns Hopkins Evidence-based Practice Center
Harold P. Lehmann, M.D., Ph.D.
Principal Investigator
John S. Andrews, M.D.
Co-Principal Investigator

Karen A. Robinson, M.Sc.
Victoria L. Holloway, M.D., M.P.H.
Steven N. Goodman, M.D., Ph.D.
Investigators

AHRQ Publication No. 01-E019

September 2001

ISBN 1-58763-059-1
ISSN 1530-4396

This report may be used, in whole or in part, as the basis for development of clinical practice guidelines and other quality enhancement tools, or as a basis for reimbursement and coverage policies. Endorsement by the Agency for Healthcare Research and Quality (AHRQ) or the U.S. Department of Health and Human Services (DHHS) of such derivative products may not be stated or implied.

AHRQ is the lead Federal agency charged with supporting research designed to improve the quality of health care, reduce its cost, address patient safety and medical errors, and broaden access to essential services. AHRQ sponsors and conducts research that provides evidence-based information on health care outcomes; quality; and cost, use and access. The information helps health care decisionmakers -- patients and clinicians, health system leaders, and plicymakers -- make more informed decisions and improve the quality of health care services.

We have detailed the evidence into seven types of evidence tables, all of which are described in detail below. A summary table lists overall information about every trial included in our synthesis. All two- and three-arm trials are summarized in a heading table, a continuous outcomes table, and a discrete outcomes table. Trials with more than three arms have their results summarized in the multi-arm heading table, multi-arm continuous outcomes table, and multi-arm discrete outcomes table.

At the conclusion of this description is a complete Index to Evidence Tables. All treatments are listed in this index, in alphabetical order, under the heading, "target arm." The comparison treatment is found under the heading, "comparator arm." The evidence, as described in Volume 1, Chapter 4 and presented in these evidence tables, provides single entries for each comparison; these entries are ordered by Comparison Index. The comparison is given a "secondary index" if the comparator arm is also a treatment of interest.

The Summary Table lists the trials reviewed for this report, methodological information about the trials, and a "bottom line" summary statement abstracted from the language of the authors. (See Volume 1, Appendix H for listing of included and excluded trials.)

In the first column, the Summary Table presents the trial number (our sequence number), primary author, year of publication, location of trial (either as explicitly stated by the authors, or as deduced by our abstracters based upon the author's address), and a notation if the trial was multi-site. This column also lists the general design of the trial: parallel (i.e., patients assigned to different arms prospectively); matched (i.e., patients assigned to different arms prospectively, based on characteristics of patients already recruited); cross-over (i.e., patients receive one, then the other treatment, in sequence); or split (i.e., patients receive both treatments at the same time).

In the second column, the Summary Table presents the arms of the trial. These are named in terms of the generic names of the treatments used by the investigators. Modifiers, such as "topical," or dosages are used in the arm names when necessary to distinguish one arm from another within the same trial. All time points at which data were reported in the trial are listed in this column. As was discussed in Volume 1, Chapter 2, we did not abstract all data from all time points; so the listing in this table provides a reference to readers who wish to pursue more data than are listed in our evidence tables.

The third column lists the total sample size plus characteristics of the participants including combined acne severity classification (as described in Volume 1, Chapter 3), age, and gender. The total sample sizes were based, in most cases (243 out of the 275 trials), on data provided by the authors on an intention-to-treat basis. In 18 trials, the authors provided sample size only on the number of subjects completing the trial, and in 5, on subjects somewhere during the course of the trial. In 9 trials, total sample size was provided, but the relation to the course of the study could not be determined. The numbers of males and females listed in the third column were derived from the best information available within the article. Because this was generally not provided on an intention-to-treat basis, the sums of males and females do not always add up to the listed total sample size, and, in some cases, exceed it.

The fourth column lists all outcomes reported by the authors. Not all outcomes listed here are presented in the evidence tables. For example, if the authors reported percentage changes in cyst counts, we do not list both the counts and the percentage change in the evidence tables. The comprehensive listing in this table provides an index to reported outcomes for the interested reader.

The fifth column lists an assessment of the methodological strengths and weaknesses of the trial. Strengths, represented by , are:

• The report provides explicit information showing that the patients were randomized.
• The report provides explicit information that the enroller, treatment provider, or outcomes assessor was blinded to the treatment assignment.
• Five percent or less were lost to followup, based on patients recruited into the trial.
• The trial is generalizable, either because the number of study centers was large or because explicit sampling methods to achieve generalizability were used.

Methodological weaknesses, indicated by , are:

• Treatment assignment was deterministic (i.e., predictable based upon a feature such as day of the week, treatment location, or digit in a medical record number).
• Information in the report shows that patients were not blinded.
• Information in the report shows that the enroller, treatment provider, or outcomes assessor was not blinded to the treatment assignment.
• Data in the report show that there were major losses to followup (> 20 percent of recruited subjects).
• Variable methods of outcome assessment or ascertainment were used in the various arms of the trial.
• Subjects in the arms of the trial were not comparable.
• The trial was not executed as designed.

The last column of the Summary Table presents an "Authors' Bottom Line" statement that paraphrases the conclusions made by the authors.

The trials are broken down into two groups:

Group1 comprises the majority of our evidence and data. Because readers may be interested in evaluating different treatments across trials, the evidence tables are presented in terms of two-way comparisons. Each two-arm trial reports on one two-way comparison (A vs. B). Each three-arm trial reports on three two-way comparisons (A vs. B, A vs. C, and B vs. C). Thus Group 1 includes a total of 297 two-way comparisons. The 18 trials in Group 2 with more than three arms contribute an additional 199 two-way comparisons. To make efficient use of space, these are treated differently in the report.

The evidence tables present each comparison only once (A vs. B). However, the index to the evidence tables lists each comparison twice (A vs. B and B vs. A) with annotations to direct the reader to the location of the evidence table that contains information about the comparison. Thus, if topical clindamycin is compared with topical erythromycin, then the comparison may be presented in that order in Evidence Table B, but it will be listed in two ways in the index: under clindamycin vs. erythromycin and also under erythromycin vs. clindamycin (with an asterisk, indicating that the combination is presented in the evidence table under clindamycin).

Each of the 12 sets of evidence tables has three parts: a heading table (with 1 as a suffix), a table of continuous outcomes (suffix 2), and a table of discrete outcomes (suffix 3, unless there are no continuous outcomes for the comparisons, in which case the table of discrete outcomes would have a suffix of 2). The evidence tables are listed in order of class hierarchy and then alphabetically by the name of target treatment in the table name.

The title of the Heading Table specifies the class of the target treatment. The heading table lists more details about the treatment administered in each trial in addition to those listed in the Summary Table.

The first column lists the comparison index number. The second column lists the trial number. The third (target) and fourth (comparator) columns are similar. These columns list the component treatments in each arm including name of treatment, dose, dose units, and dose frequency. These columns also include comments pertaining to the entire arm.

The following order is used within each table:

• Trials with the same comparison index are kept together.
  • Trials with the same name of target arm (the comparison index is repeated if the arm name changes).
    • Trials with the same comparator arm.
      • Outcomes in the same followup time period.
        • Alphabetized outcome names.

Note that outcomes within arms are not grouped together; thus sometimes outcomes from the same trial may be separated by outcomes from other trials. Outcomes are separated by a black line while comparisons are separated by gray, thicker lines.

The first column lists abstracted information about the trial: trial number, the methodological icons, the study design, and our Combined Acne Severity Classification for patient severity.

The second column provides the comparison index.

The third column provides our name of the outcome. Because not all outcomes reported in the trial are listed in these tables, the reader is referred to the Summary Table for further outcomes. The second column also reports on the body location where acne lesions were counted (the majority type of acne outcome), if available, and provides the baseline lesion counts, if they were provided by the authors. The baseline count provides the reader a further sense of the severity of the acne in the trial participants.

The fourth column lists the followup duration of each outcome. As discussed in Volume 1, Chapter 2, we discerned four time periods as relevant to acne therapy: short-term (< 6 weeks), mid-term (> 6 and < 12 weeks), long-term (> 12 weeks), and post-therapy (outcomes after therapy has ceased). For some outcomes, the authors were not explicit about the time frame; so numbers may be absent from the table.

The fifth and sixth columns present the target and comparator treatment results. Because some tables combine target treatments of similar, but somewhat different, class, the treatment arm name is provided. Because comparator class names may change as well within the same table, these are provided, along with the comparator treatment names.

The outcome results are provided as follows. "Percent change" indicates an outcome derived from individual patient's final and baseline measures, as provided by the authors. An asterisk suffix ("Percent change*") indicates that we had to calculate a percentage change based on available data, meaning that it is group-based, as opposed to individual-based, when provided by the author. Individual-based is preferred, because the individual patient serves as his or her own control. When baseline data are not provided, the outcome is "Change," the raw difference from baseline. An asterisk suffix ("Change*") indicates change that we calculated. All other results are called "Measure" and indicate outcomes with no baseline comparison. When available, the standard deviation ("SD") is provided. Within each treatment, a P-value is provided if it was provided by the authors or, followed by an asterisk, if we could calculate it. These statistics are with respect to baseline.

The final column represents between-arm P-value of the treatment with respect to the comparator. If no statistics were provided by the authors, we attempted to calculate P-value, using a t-test for two samples. We could do this only when standard deviations were reported by the authors. Therefore, blanks indicate missing P-values that could not be calculated. For three-arm trials, we selected the most relevant between-arm P-values.

The discrete outcomes tables are ordered and structured similarly to the continuous outcomes table. They differ in the results reported. For each outcome, the outcome itself is listed, with the number of patients who had that outcome. Because of technical reasons during trial abstraction, an outcome with more than eight possibilities was separated into more than one outcome. Between-arm P-values were not calculated in these tables but were included in the summaries, as manually calculated values, in those cases where we wanted to confirm a difference apparent from the raw numbers.

The tables for these trials parallel those for two- and arm-arm trials. Table Multi.1 is the heading table, Table Multi.2 is the continuous outcomes table, and Table Multi.3 is the discrete outcomes table. The outcomes are listed in a single column, rather than horizontally as pairs. This arrangement enables readers to make their own comparison of all outcomes.