Abstract
CONTEXT: Complementary and alternative medicine (CAM) is used by pediatric patients with cancer, but the actual frequency of CAM use is undetermined.
OBJECTIVE: In this systematic review we summarize the current evidence on the prevalence of CAM use in pediatric patients with cancer and assess the reported quality of included studies.
METHODS: We systematically searched 6 major electronic databases, reference lists of existing reviews, and personal files. We included full articles about primary research studies (without language restriction) that reported the prevalence of CAM use if all or a defined subsample of participants were pediatric patients with cancer. Detailed information regarding methods and results was extracted from the original articles. A quality-assessment tool was rigorously developed on the basis of the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement and used to assess reported study quality. Formal tests of interrater agreement were conducted.
RESULTS: We included 28 studies with survey data (collected from 1975 to 2005) from 3526 children. In 20 studies with 2871 participants, the prevalence of any CAM use (since cancer diagnosis) ranged from 6% to 91%; considerable heterogeneity across studies precluded meta-analysis. Study quality was mixed and not correlated with CAM prevalence. Herbal remedies were the most popular CAM modality, followed by diets/nutrition and faith-healing. Commonly reported reasons for CAM use included to help cure or fight the child's cancer, symptomatic relief, and support of ongoing use of conventional therapy. There was little evidence of an association between CAM use and patients' sociodemographic characteristics.
CONCLUSIONS: Many pediatric patients with cancer use CAM. It is important that pediatricians be aware of this fact and encourage open communication with patients and their parents. Using standardized survey methods and CAM definitions in future studies could improve their quality and help generate comparable data. Our quality-assessment tool could prove valuable for other reviews of prevalence studies.
Complementary and Alternative Medicine (CAM) is defined by the National Center for Complementary and Alternative Medicine (NCCAM) as “a group of diverse medical health care systems, practices, and products that are not presently considered to be a part of conventional medicine.”1 Well-known CAM modalities include acupuncture, aromatherapy, osteopathy, yoga, and herbal supplements. Early surveys regarding the use of CAM in the United States started in the 1970s and were conducted mainly among pediatric patients with cancer.2 Pediatric patients with cancer have been included in previous reviews of CAM use in populations of adults with cancer,2,3 and a small number of reviews that focused solely on the prevalence of CAM use in pediatric cancer have been published.4,–,6 However, to our knowledge, a comprehensive systematic review that critically summarizes all the current evidence and incorporates rigorous and transparent quality assessment is absent from the literature. In undertaking such a review, we aimed to provide a systematic and critical overview of this literature, inform research practice by identifying common methodologic weaknesses, and provide pediatricians with data about the likely prevalence of (and reasons for) CAM use among their patients with cancer.
This systematic review addresses 2 primary research questions: What is the prevalence rate of CAM use among pediatric patients with cancer, and what is the quality of the studies in this area? Prevalence data are necessary to inform clinicians of the scope of CAM use, to determine the level of need for support and education of patients who use CAM, and to suggest which CAM interventions are popular and so might be prioritized for safety and/or efficacy research. In turn, information on study quality will help direct researchers' efforts to address existing limitations and improve the quality of future studies. We also addressed 5 secondary research questions: Which CAM modalities are popular in children with cancer? Are prevalence rates changing over time? Are sociodemographic characteristics associated with CAM use? Why do children with cancer use CAM? What is the relationship between methodologic quality and research findings? (The relationship between quality and findings is, in general, largely untested and uncertain in observational research.7) By understanding these issues we hope to inform the allocation of scarce research and clinical resources.
METHOD
Literature Search
Primary research studies that investigated the prevalence of CAM usage and were published in full journal articles were eligible for inclusion. Eligible studies had to include pediatric patients with cancer. If they included adult or other pediatric patients, they had to report data about the frequency of CAM use among a defined subsample of pediatric patients with cancer. Randomized, controlled trials or any other studies that examined the outcomes of specific therapies were excluded, as were studies published in conference proceedings, letters, and other “gray” literature.
We searched 6 electronic databases for articles that met the above-listed criteria by using prespecified search strategies (Table 1). Journal-article references were managed by using EndNote Web 2.2 (Thomson Reuters, Philadelphia, PA). We supplemented the electronic search by searching reference lists of existing reviews and relevant articles located from the electronic search and by searching our personal files. The article titles and abstracts were reviewed for eligibility. If they were insufficient to decide on eligibility, the full text was retrieved. Two non–English-language articles met the inclusion criteria and were translated in full by native speakers who were also fluent in English.
Databases and Search Terms Used to Identify Relevant Articles
Quality-Assessment Tool
Authors of the recent Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement proposed essential items for reporting observational research, including surveys.8 We used this reporting guideline as the basis for developing a quality-assessment tool (QAT) for surveys on CAM use, because there is currently no gold-standard assessment tool.7 In brief, the tool consists of 17 items that assess the quality of the study design, data collection, and data analysis on the basis of the reported information. The items were weighted for their importance for overall study quality through consensus across investigators: 3 items scored a maximum of 2 points, 8 items scored 1 point, and 6 items scored 0.5 points (Table 2). The maximum score was 17. We calculated the percentage overall score instead of total points, because for some articles not all 17 items applied. To evaluate the tool's reliability, 3 of us (Drs Bishop, Lewith, and Chan) scored the same 3 articles, and pairs of these authors scored another 8 articles. The percentage interrater agreement on the assessment tool was calculated for each pair of raters and article. Over 17 rating pairs, agreement ranged from 77% to 100%, and the median level of agreement was 88%.
Summary of QAT Scores
Data-Extraction Procedures
The following detailed information related to data collection and results was extracted and entered into an Excel (Microsoft, Redmond, WA) spreadsheet: geographical location and date of data collection; definition of CAM used for survey; sampling methods (strategy, sample size, response rate, whether parents and/or children were asked); participants' illness (type and stage of cancer, time since diagnosis, use of specific conventional treatments); sociodemographic characteristics (child age, gender, and ethnicity, parents' socioeconomic status, income level, education, and occupational status); and CAM use (in relation to cancer or for any reason, CAM use excluding prayer, specific CAM modalities; duration of CAM use; reasons given for using CAM). If available, the number of participants for each category or subgroup was extracted (eg, for each type of cancer or type of CAM). For continuous data, all available summary measures were recorded (eg, mean and SD of time since diagnosis).
To test the consistency of the data extraction, 3 of us (Drs Bishop, Lewith, and Chan) extracted data from the same 3 articles, and pairs of us extracted data independently from another 8 articles. We compared the extracted data for each article separately and allocated 1 point for variables with identical data extraction and 0 points for variables with differences. The percentage agreement was then calculated as an indicator of consistency between each pair of raters for each article. Over 17 rating pairs, consistency ranged from 82% to 95% (median: 91%).
Statistical Analysis Methods
Prevalence rates from individual studies were depicted graphically. A meta-analysis was planned to estimate the overall prevalence of CAM use for cancer across the original studies that reported this information.9 However, we found considerable heterogeneity of the within-sample prevalence (Cochran's Q = 204; 19 degrees of freedom) and, thus, refrained from calculating pooled estimates. Simple descriptive statistics were used to summarize QAT scores. The possible change in prevalence rates of CAM use over time was explored graphically. The relationship between sociodemographic characteristics and CAM use was explored by examining patterns across studies. However, too few studies reported sufficient information in a consistent form to allow more formal statistical analyses.
RESULTS
Study Selection and Characteristics
The electronic searches yielded 2498 journal article references, from which 619 duplicate records were removed. From the remaining 1879 references we identified 234 articles of individual studies on the prevalence of CAM use by patients with all types of cancer; 28 of these articles focused on pediatric patients with cancer and were included in this review.10,–,37 The other 206 articles concerned CAM use in adult patients with cancer and were excluded. Detailed characteristics of the included studies are listed in Supplemental Information Tables 1 and 2, which are available at www.pediatrics.org/content/full/125/4/768.
Of the 28 studies, 10 were conducted in the United States,* 4 in Canada,11,14,15,26 and 2 each in the United Kingdom27,29 and Turkey.19,21 The remaining 10 studies were conducted in Israel,36 Singapore,24 Mexico,18 Taiwan,37 Denmark,25 Finland,30 the Netherlands,20 Germany,23 Hungary,35 and Australia.34 Studies were published between 1977 and 2007, and survey data were collected between 1975 and 2005. Ten studies did not report the dates of data collection. Three articles24,31,32 explicitly defined CAM according to the NCCAM definition,1 1 study26 cited Eisenberg's definition38 (which is equivalent to that by the NCCAM), 1 study29 cited Ernst and Cassileth's definition,3 and 14 studies used other definitions of CAM.† The remaining 9 articles did not provide a clear definition of CAM.
The sample sizes ranged considerably from 1530 to 106323 (median: 75); the total sample size across all 28 studies was 3526. The characteristics of the participants also varied considerably. For example, some studies only included children within a specified time period since cancer diagnosis, whereas others set no time limits. The information concerning time since diagnosis was reported in such varied detail and format (eg, mean, range, categorical data) that formal summary is precluded. Twenty articles that included 3166 participants reported on specific cancer diagnoses.‡ Across these studies, 41% of the participants (n = 1294) were diagnosed with a form of leukemia, and 14% (n = 456) were diagnosed with a neurologic or related tumor. Another 14% (n = 429) had a solid tumor (eg, of the kidney or liver), 11% (n = 350) had lymphoma, 7% (n = 233) had sarcoma, and 2% (n = 70) had a bone tumor. The remaining 11% (n = 334) were reported in insufficient detail (or had missing data) to allow classification within these arbitrary and nonexhaustive groupings. Similarly, authors of the studies varied in reporting the participants' age, with some reporting age groups, absolute ranges, or minimum or maximum values. Across the 14 articles that reported the minimum and/or maximum age, the minimum ranged from 0 to 8 years, and the maximum ranged from 12.6 to 24.3 years. In 15 articles10,15,17,–,21,26,–,31,34,36 the mean age ranged from 5.4 years15 to 14.8 years.28 Across the 20 articles§ that reported on gender, 56% (n = 1684) of the participants were boys and 44% (n = 1325) were girls.
Study Quality
Study quality was mixed. Percentage scores on the QAT ranged from 19% to 79%, with half of the studies attaining <50% of the maximum QAT score (Supplemental Information Table 1). Table 2 summarizes the results of the quality assessment, provides a brief definition of the items, and lists the studies that scored under each criterion. Few studies used data-collection strategies that would minimize the risk of recall bias, collected data from a representative sample of patients, or described measures to address potential sources of bias. However, many more study authors reported response rates, described the participants' age and gender, assessed cancer-related CAM use, and assessed use of specific CAM modalities.
Prevalence of CAM Use
Figure 1 shows the prevalence of any CAM use in pediatric patients since their cancer diagnosis on the basis of 20 articles that reported on 2871 surveyed children.‖ The prevalence rates ranged from 6%12 to 91%,35 with 14 articles reporting prevalence rates between 20% and 60%.
Prevalence of CAM use by pediatric patients with cancer in 20 studies. Point estimates and 95% confidence intervals are shown.
In 8 articles10,16,18,20,25,27,32,33 it was unclear whether they included children who were using CAM before being diagnosed with cancer or not in relation to their cancer. The prevalence rates reported from these studies ranged from 16% to 73% (these studies are not included in Fig 1).
Use of Different CAM Modalities
Twenty-five articles reported on the prevalence of use of different CAM modalities.10,–,14,16,–,22,24,–,26,28,–,37 The most commonly and consistently reported CAM modalities were herbal remedies (13 studies#) (Fig 2), diets and nutrition (13 studies**) (Fig 3), and faith-healing (9 studies††) (Fig 4). Use of herbal remedies ranged from 2%12 to 48%18 of the surveyed children; use of dietary and nutritional interventions ranged from 3%26 to 47%22; and use of faith-healing ranged from 3%13 to 30%.37 Four other CAM modalities were reported across fewer studies: homeopathy was used by 1%16 to 17%18 of the patients in 7 studies (n = 949)14,16,18,28,29,34,36; megavitamins were used by 2%29 to 19%14 in 7 studies (n = 908)14,16,28,29,33,34,37; mind-body therapies were used by 9%11 to 27%22 in 5 studies (n = 611)11,17,22,31,36; and massage therapy was used by 2%19 to 17%14 of the patients in 5 studies (n = 779).14,16,19,28,29
Prevalence of use of herbal remedies across 13 studies. Point estimates and 95% confidence intervals are shown.
Prevalence of use of diet and nutrition by pediatric patients with cancer in 13 studies. Point estimates and 95% confidence intervals are shown.
Prevalence of use of faith-healing by pediatric patients with cancer in 9 studies. Point estimates and 95% confidence intervals are shown.
Relationship Between CAM Use and Other Factors
We did not find any association between prevalence of reported CAM use and QAT scores when using exploratory scatter plots and nonparametric analyses (data not shown). The 2 studies with the lowest estimates of CAM use were also the earliest studies (published in 1977 and 1983). However, visual inspection of Figs 1 through 4 suggests that there is little evidence for a trend for increased reported CAM use over time. There was no clear geographical patterning. The authors of 14 articles‡‡ assessed the relationship between CAM use and parents' education; 10 had a QAT score of >50%. Five studies revealed that CAM use was significantly associated with higher parental education.14,23,26,36,37 Of those, 4 had a QAT score of >50%. The remaining 9 studies revealed no significant association between parents' education and CAM use. Of 12 articles on CAM use and parents' income,§§ 10 reported a QAT score of >50%, and only 1 reported a significant association of CAM use with higher income.23 Twelve studies assessed associations between child age and use of CAM.‖‖ Eight of these studies scored >50% on the QAT, and only 1 revealed that CAM users were significantly older than nonusers.19 Seven studies (5 with QAT scores of >50%) evaluated associations between ethnicity and CAM use14,22,24,28,31,32,36; 1 article from Singapore reported a significant association in that children of Chinese ethnicity were more likely to use CAM.24 In 11 studies (9 with QAT scores of >50%), no significant association between gender and CAM use was found.##
Reasons for CAM Use
Fourteen articles reported reasons for CAM use in a variety of ways (Supplemental Information Table 2). Commonly reported reasons for CAM use included to cure or help fight the child's cancer, provide symptomatic relief, and support conventional medicine (eg, relief from adverse effects). The diversity of categories and reasons reported precluded any integration of these findings across individual studies.
DISCUSSION
When people consider the prevalence of CAM use in cancer, they often suggest widespread use based on nonsystematic data.39 In our systematic review, most individual studies suggested that a substantial proportion of pediatric patients with cancer use CAM at some point in their treatment. Pediatric oncologists need to be aware that their patients (and their patients' parents) will be seeking and integrating other therapeutic approaches while undergoing conventional treatments. Additional research is warranted to better understand this behavior and to determine and address any needs for patient support and education on CAM use. Given that certain CAM modalities such as herbal remedies or dietary and nutritional interventions seem to be particularly popular with this patient group, research into the (cost) effectiveness and, particularly, safety of these modalities could be prioritized.
The overall quality of the studies was mixed, with half of the studies scoring <50% on our QAT. However, quality seemed to be unrelated to reported prevalence. Researchers in future studies should take care to minimize the risks of bias (eg, by specifying limited recall periods for CAM use [or better recording CAM use prospectively] and by obtaining representative rather than convenience samples). Although participants' sociodemographic and clinical details were frequently reported, the lack of standardization was a serious hindrance to comparing the study populations. Even more concerning is the diversity in the definition of CAM use. Some cross-cultural diversity is to be expected when considering a culturally specific behavior such as CAM use (eg, osteopathy is considered to be CAM in the United Kingdom but not in the United States). However, the use of a generally agreed-on definition of CAM such as that now provided by the NCCAM1 alongside a standardized questionnaire would help collect consistent data across different settings and populations. Currently, researchers are beginning to develop such a standardized questionnaire that could be supplemented by culturally specific items.40 Few studies provided any indication of the reliability or validity of their measures of CAM use or reported any pilot-testing procedures. Given that instrument development is laborious and often done insufficiently in small-scale surveys, the collaborative development of a validated survey instrument would substantially improve the quality and value of such research.
There seem to be few reliable associations with the participants' sociodemographic characteristics. CAM use was not associated with gender, age, ethnicity, or family income of pediatric patients with cancer. However, CAM use may be more common in families with higher parental education. In contrast, evidence from adult populations suggests that CAM use is more common in women, middle-aged patients, and (less consistently) those with higher income or level of education; in adults, CAM use may also be related to ethnicity.41 However, our data are consistent with evidence from other pediatric populations, in which associations between sociodemographic factors and CAM use are rare.42,–,46 The included articles reported a variety of reasons for CAM use, including “doing everything possible for their child,” treating the cancer, providing symptomatic relief, and supporting conventional treatments. Similar reasons for CAM use have been reported in children with other chronic conditions.47,48
Our findings are strengthened by the systematic identification and evaluation of the included literature. We report how we developed our QAT on the basis of the Strengthening the Reporting of Observational Studies in Epidemiology guidelines8 and provide initial data concerning its reliability. Although non–English-language articles were included, we acknowledge that so-called gray literature was excluded. However, we are confident that any such studies would not have had any major impact on our findings. It was not possible to pool prevalence estimates of CAM use because of the considerable heterogeneity across individual studies. Concerning the reasons for CAM use, a synthesis of the qualitative research on CAM use in pediatric cancer might be warranted.
CONCLUSIONS
We provide here a timely and comprehensive systematic review and assess the reported study quality of included studies. We have identified many limitations of the current literature, and we urge the use of validated standardized instruments to improve the quality and comparability of survey data across studies. The use of CAM in pediatric cancer encompasses many patients. Consequently, this issue must continue to be addressed openly by pediatric cancer clinicians with their patients. Additional high-quality research is needed to better understand what patients and their parents seek from CAM while conventional cancer treatments continue to develop and improve apace.
ACKNOWLEDGMENTS
Dr Bishop is funded by an Non-Clinical Career Development Fellowship from the Arthritis Research Campaign. Dr Lewith's post is funded by the Rufford Maurice Laing Foundation. Dr Chan and Ms Saville were undergraduate medical students at the University of Southampton at the time of this research.
Footnotes
- Accepted November 20, 2009.
- Address correspondence to Felicity L. Bishop, PhD, Department of Primary Medical Care, School of Medicine, University of Southampton, Aldermoor Health Centre, Aldermoor Close, Southampton, Hampshire SO16 5ST, United Kingdom. E-mail: f.l.bishop{at}southampton.ac.uk
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
↵# Refs 10, 12, 14, 16, 18, 19, 21, 22, 24, 28, 29, 32, and 36.
↵** Refs 11, 14, 18,–,20, 22, 24, 26, 28, 29, 32, 34, and 36.
- CAM =
- complementary and alternative medicine •
- NCCAM =
- National Center for Complementary and Alternative Medicine •
- QAT =
- quality-assessment tool
REFERENCES
- Copyright © 2010 by the American Academy of Pediatrics