11

Click here to load reader

The Effectiveness of Therapeutic Touch: A Meta-Analytic Review

  • Upload
    r-m

  • View
    215

  • Download
    1

Embed Size (px)

Citation preview

Page 1: The Effectiveness of Therapeutic Touch: A Meta-Analytic Review

http://nsq.sagepub.com/Nursing Science Quarterly

http://nsq.sagepub.com/content/12/1/52The online version of this article can be found at:

 DOI: 10.1177/08943189922106413

1999 12: 52Nurs Sci QRosalind M. Peters

The Effectiveness of Therapeutic Touch: A Meta-Analytic Review  

Published by:

http://www.sagepublications.com

can be found at:Nursing Science QuarterlyAdditional services and information for    

  http://nsq.sagepub.com/cgi/alertsEmail Alerts:

 

http://nsq.sagepub.com/subscriptionsSubscriptions:  

http://www.sagepub.com/journalsReprints.navReprints:  

http://www.sagepub.com/journalsPermissions.navPermissions:  

http://nsq.sagepub.com/content/12/1/52.refs.htmlCitations:  

What is This? 

- Jan 1, 1999Version of Record >>

at MEMORIAL UNIV OF NEWFOUNDLAND on August 1, 2014nsq.sagepub.comDownloaded from at MEMORIAL UNIV OF NEWFOUNDLAND on August 1, 2014nsq.sagepub.comDownloaded from

Page 2: The Effectiveness of Therapeutic Touch: A Meta-Analytic Review

Nursing Science QuarterlyJanuary1999 Effectiveness of Therapeutic Touch

The Effectiveness of Therapeutic Touch:A Meta-Analytic Review

Rosalind M. Peters, RN; MSNAssistant Professor, Medical College of Ohio, Toledo

The purpose of this article is to explore the adequacy of published scientific evidence supporting therapeutic touchas a nursing intervention. Meta-analytic techniques were used to integrate the research-based literature publishedin the past decade. The results seem to indicate that therapeutic touch has a positive, medium effect on physiologicaland psychological variables. It is impossible to make any substantive claims at this time because there is limitedpublished research and because many of the studies had significant methodological issues that could seriously biasthe reported results. Resolving these methodological issues is imperative for therapeutic touch research to moveforward.

The concept of therapeutic touch (TT)is derived from the ancient practice ofthe laying on of hands. Research on theeffectiveness of this practice hasevolved in three distinct phases. Thefirst phase began in the 1960s with stud-ies examining the effects of the laying-on of hands on plant growth and woundhealing in mice (Grad, 1963, 1964,1965; Grad, Cadoret, & Paul, 1961). Inthe 1970s, Dora Kunz and DoloresKrieger were among the first to use TTas a human intervention (Krieger,1990). Early research by Krieger (1972,1973, 1974) showed an increase in he-moglobin levels in persons receivingTT. However, most of these early stud-ies are published in journals not easilyaccessible to nurses. Until the mid-1970s, practicing nurses had littleawareness of the benefits of TT. Thischanged with the publication ofKrieger’s (1975) article inAmericanJournal of Nursingand the publicationof her book Therapeutic Touch(Krieger, 1979). Since that time,Krieger (1990) reports that she has“taught over 17,000 health profession-als and [thinks] that Dora has taught atleast as many” (p. 83). In addition, Kunzand Krieger encourage their students toteach others what they have learned, be-

cause “therapeutic touch has provedhelpful to ill people” (Krieger, 1990, p.83). As a result, thousands of nursespractice TT today. But has it reallyproven helpful? Practitioners need tocarefully examine the research base todetermine if there is sufficient empiricalevidence to support such a practice.

Quinn (1988) provides a compre-hensive narrative critique of TT re-search conducted between 1974 and1986. She states that the review revealsan impressive consistency in the workand noted the increasing sophisticationin the study designs. She believes thatearly TT researchers have laid a sub-stantial foundation for work in this area.If these are the foundational studies,what is the status of the current TT re-search?

This article provides a quantitativereview of the past decade of TT re-search. Its purpose is to explore thestrengths of the more recently publishedresearch base supporting the use of TTas a nursing intervention. Two key ques-tions to be addressed by this pilot studyare as follows: (a) Does TT produce thedesired outcome in treated subjects, and(b) are TT outcomes significantly dif-ferent than those found in the controlgroup? A meta-analytic (i.e., quantita-tive) review was conducted to answerthese questions and explore the strengthof the current research base for TT. Ameta-analytic technique was used be-

cause pooling research findings from anumber of individual studies and trans-forming those into a common metric(i.e., effect size) provides a more coher-ent pattern of research findings (Cur-lette & Cannella, 1985; Devine, 1990;Lynn, 1989; Wolf, 1986).

Method

Although there are numerous meta-analytic methods that can be used in re-search integration, the study-effectmeta-analysis (SEM) was chosen forthis review (Bangert-Drowns, 1986).However, because the SEM method isrestrictive in determining inclusion cri-teria, especially in relation to quality ofstudy issues, a serious problem wouldarise given the already limited TT re-search literature. Determining a priori aquality cut-off might exclude so manystudies that the analysis would becomemeaningless; it also prevents the re-searcher from determining the impactof methodological shortcomings onstudy outcomes (Bangert-Drowns,1986; Lynn, 1989). Therefore, the SEMapproach can be modified to allow theinclusion of studies with methodologi-cal flaws (Glass, 1976). The modifica-tion involves coding the studies basedon their quality and then determining if

Nursing Science Quarterly, Vol. 12 No. 1,January 1999, 52-61© 1999 Sage Publications, Inc.

Keywords: meta-analysis, quantita-tive review, therapeutic touch

at MEMORIAL UNIV OF NEWFOUNDLAND on August 1, 2014nsq.sagepub.comDownloaded from

Page 3: The Effectiveness of Therapeutic Touch: A Meta-Analytic Review

there is a difference in effect size basedon study quality. Details regarding themethods used in this meta-analysis aredescribed below.

Sampling

The focus of this review was to ex-plore the scientific evidence available topracticing nurses. Therefore, it is lim-ited to published research identified us-ing the computerized index databases ofCINAHL, Medline, and Psychinfo. Be-cause nurses have Quinn’s (1988) re-view of the early TT literature, Phase 1of this review was limited to researchpublished from 1986 to 1996. (Phase 2will extend the meta-analysis to includeboth unpublished and pre-1986 lit-erature.)

Using only published reports pro-vides the practitioner with peer-reviewed references in which both con-tent and methodology have been care-fully critiqued. This same level of cri-t ique may not be avai lable forunpublished studies. Excluding unpub-lished works, however, can produce apublication bias and creates the possi-bility that the magnitude of an effectwill be inflated (Devine & Cook, 1983).Publication bias may occur becausejournals are more inclined to publishpositive findings and because research-ers may not submit nonsignificant find-ings for publication (the “file drawer”problem of meta-analysis) (Lynn, 1989;Wolf, 1986). The publication bias/filedrawer problem is noted in the TT re-search. Of studies reviewed by Quinn(1988), 75% (n= 6) were dissertations.Two additional quantitative disserta-tions were found in the current search(Hale, 1986; Hinze, 1988). Of the eightdissertations, five reported nonsignifi-cant findings. Only two of the studieswith nonsignificant findings were pub-lished (Meehan, 1993; Randolph,1984).

The issue of possible publicationbias can be directly assessed at the endof Phase 2 analysis when a full sam-pling of unpublished TT research can beexamined. This would be a useful con-tribution to the published literature be-cause most practicing nurses do not

have access to unpublished materials.Until the full range of TT research canbe examined, it is necessary to compen-sate for a potential publication bias.This was done in this study by calculat-ing a fail-safe N. A fail-safe N indicatesthe number of additional studies neces-sary in a meta-analysis to reverse theoverall probability obtained from thecombined test to a value higher than thecritical value for statistical significance(Wolf, 1986).

Selection Process

The selection process for this meta-analysis began using the CINAHL data-base to search the nursing literature. En-tering TT as the keyword yielded 258citations, most of which were not re-search articles. When the search waslimited to the keywords of TT and re-search published between 1986 and1996, a total of 118 studies were identi-fied. Searching the Medline databaseyielded five citations not found in CI-NAHL. The Psychinfo database addedone article and five dissertations notpreviously identified. Retrieval alsowas done using the “ancestry” method(Curlette & Cannella, 1985), whichadded one article that was not found inthe database searches. The combinedsearches yielded 130 citations, of which36 were research reports. These 36 cita-tions were then reviewed to determinewhether they met the inclusion criteriafor this meta-analysis.

The inclusion criteria for this meta-analysis were (a) the report had to pre-sent empirically based research; (b) theresearch had to be a human interventionstudy with TT as the experimental inter-vention; (c) the intervention had to fol-low Krieger’s (1979, p. 35) four phasesof TT; (d) the study design had to be ex-perimental, quasi-experimental, or pre-post single group; (e) the outcomes hadto measure either psychological orphysiological well-being; and (f) thestudy had to have been published be-tween 1986 and 1996. Using these crite-ria, only 9 of the 36 studies (4% of thetotal TT literature) could be included inthis review. Table 1 details the inclusion/exclusion criteria. Table 2 summarizes

the characteristics of the nine studies in-cluded in this review.

Procedure

Two key procedural steps were takenin this meta-analysis. First, major studycharacteristics were coded so that qual-ity of studies could be considered. Sec-ond, the studies were analyzed to deter-mine whether a TT treatment effectoccurred and then analyzed to deter-mine the magnitude of the observedeffect.

Relevant characteristics of the stud-ies (conceptual framework, population,methodology, and intervention vari-ables) were coded. In addition, all stud-ies were coded for quality using theQuality of Study instrument presentedby Smith and Stullenbarger (1991).This tool examines four major areas ofeach study: introduction, methodology,data analysis/results, and conclusions.A total of 22 specific criteria are evalu-ated and given a numerical rating of 0 to3 (fromabsentto met at a high level). Itshould be noted that quality ratings forpublished works may be affected byeditorial decisions that determine thedepth of information presented in anarticle.

This review uses two indicators ofthe effectiveness of TT. One is the com-bined probabilities of the studies as de-termined by the Stouffer Combinedtest. The combining probabilities ap-proach to meta-analysis can provide im-portant information. However, it doesnot provide information about the mag-nitude, consistency, or direction of theeffect; therefore, an effect size (Cohen’sd) is also presented in this review.

Effect Size

Cohen’sd is used in comparing theeffect found between two groups. It re-flects the difference between the meansof the experimental and control groupdivided by a standard deviation (Cur-lette & Cannella, 1985). There is dis-agreement as to whether the standarddeviation from the control group shouldbe used or whether a pooled estimate ofthe standard deviation is more appropri-

Effectiveness of Therapeutic Touch 53

at MEMORIAL UNIV OF NEWFOUNDLAND on August 1, 2014nsq.sagepub.comDownloaded from

Page 4: The Effectiveness of Therapeutic Touch: A Meta-Analytic Review

54 Nursing Science Quarterly, 12:1, January 1999

Table 1Rationale for Inclusion/Exclusion in the Study

Number Reason forof Studies Exclusion/Inclusion References

6 Studies measured other than therapeutic touch (i.e., Birch (1986); Bush and Geist (1992); Engebretson (1996); Fawcett, Sidney, Hanson, andprocedural touch, tactile stimulation, prayer). Riley-Lawless (1994); Rubens, Gyurkovics, and Hornacek (1995); Wirth and Cram (1994)

9 Studies were unpublished doctoral dissertation work. Ferguson (1986), France (1991), Hale (1986), Hamilton (1988), Hinze (1988),Of the dissertations, only two would have met the Lionberger (1986), Mersmann (1994), Parkes (1986), Stravena (1992)inclusion Criteria (Hinze, 1988; Parkes, 1986).

4 Articles presented qualitative research about Heidt (1990); Hughes, Meize-Grochowski, and Harris (1996); Samarel (1992);therapeutic touch. Smyth (1995)

4 Articles tested some aspect of therapeutic touch but Messenger and Roberts (1994), Wright (1991), Wirth (1995), Wirth andwere not intervention studies Cram (1993)

4 Articles provided insufficient data or mixed Olson, Sneed, Bonadonna, Ratliff, and Dias (1992); Vaughn (1995); Wirthinterventions preventing effect size calculation. and Barrett (1994); Wirth, Barrett, and Eidelman (1994)

9 Articles met the inclusion criteria Gagne and Toye (1994); Keller and Bzdek (1986); Kramer (1990); Meehan(1993); Olson and Sneed (1995); Quinn (1989); Quinn and Strelkauskas (1993);Simington and Laing (1993); Wirth, Richardson, Eidelman, andO’Malley (1993)

Table 2Quantitative Research on Therapeutic Touch, 1986 to 1996

OutcomeAuthor Conceptual Sample Minutes Variableand Year Framework Designa Variables and Population Size Interventionb of Tx Analyzed

Psychological variablesQuinn Rogerian Experimental pre-/posttest Anxiety in open heart surgical 153 NCTT, Mimic TT 5 STAI(1989) with control patients (presurgery) No Tx

Quinn and Rogerian Descriptive pre-/posttest Anxiety and immune status in 4 ?TT Variable STAIStrelkauskas with no control bereaved persons(1993)

Simington Rogerian Experimental posttest only Anxiety in elderly in long-term 105 TT, Mimic TT, 3 STAIand Laing care facilities No Tx(1993)

Gagne and General Experimental pre-/posttest Anxiety in inpatients and 31 ?TT; Mimic TT, 15 STAIToye (1994) Systems with control psychiatric patients relaxation therapy

Olson and Stress Experimental pre-/posttest Anxiety in healthy professional 40 ?TT, sit quietly 15 STAISteed (1995) with control caregivers

Physiological variablesKeller and Holism/ Experimental pre-/posttest Tension headache, pain in adults 60 NCTT, Mimic TT 5 McGill-MelzackBzdek (1986) Rogerian with control Pain Scale

Kramer Stress Experimental posttest only Stress in children hospitalized for 30 ?TT, Casual touch 6 PMRRT (tool of(1990) acute injury or illness physiologic

measures)

Meehan Rogerian Experimental pre-/posttest Postoperative pain in adult 108 NCTT 5 Visual Analogue(1993) with control surgical patients Mimic TT Pain Scale

Standard care

Wirth et al. Atheoretical Experimental posttest only Wound healing in volunteers with 24 NCTT 5 Wound healing(1993) (energy with control punch biopsy Mimic TT per MDs

Interaction)

NOTE: ƒ: Unable to calculate value from data provided in the research report.a. None of these studies have random sampling and therefore are not true experimental designs. Most, however, have random assignment to groups.b. TT = Contact therapeutic touch; NCTT = noncontact therapeutic touch; ?TT = not reported if contact or noncontact TT; STAI = State Trait Anxiety Inventory; Tx =treatment.

at MEMORIAL UNIV OF NEWFOUNDLAND on August 1, 2014nsq.sagepub.comDownloaded from

Page 5: The Effectiveness of Therapeutic Touch: A Meta-Analytic Review

ate in calculating effect size (Curlette &Cannella, 1985; Wolf, 1986). Becauseof problems with statistical reporting,pooled standard deviations were notavailable. Therefore, this meta-analysisused the control group standard devia-tion when calculating effect sizes as ad-vocated by McGaw and Glass (1980).Lack of information about the pooledwithin-group standard deviation alsoprevented the calculation of the unbi-ased effect-size estimate using the sta-tistic g. This statistic would have beenhelpful in correcting for small sample-size bias and could have been used toprovide an unbiased effect size statisticd. As a result, caution must be used ininterpreting the effect sizes of this TTresearch review for two reasons: (a) Useof the control group standard deviationyields less stable estimates of effectsizes, and (b) inability to correct forsample size lets each study have equalweight regardless of the number of sub-jects studied.

In determining effect size, only oneoutcome measure was taken from eachstudy. The measure chosen was the onethat reported the best reliability results.In studies that used multiple groups, ef-fect was determined by comparing theexperimental group with a “no treat-ment” control group whenever possible.In studies without a control group, the

experimental effect was compared tothe placebo intervention. This placebowas usually called “mimic TT.” Analy-sis of the experimental group with a pla-cebo provides important informationregarding the Hawthorne effect (Burns& Grove, 1993). In studies that com-pared TT with both a placebo and an-other intervention, only the placebogroup was used for statistical compari-son in this review.

Effect size was calculated for twoquestions in each study. Did TT producethe desired outcome in treated subjects(e.g., reduced anxiety, improved physi-ologic status from pretest to posttest),and were the outcomes significantlydifferent from the control group? In ad-dition to the biases described above,dscalculated from pretest/posttest datacreate additional concerns. The extentto which pre- and posttest scores arecorrelated affect the values ofd. Also,becauseds calculated from pretest/post-test values are change scores, they aresubject to all the criticisms of changescores and must be interpreted cau-tiously. Because effect sizes may be bi-ased, Stouffer’s Combined test was alsocalculated using the probabilities asso-ciated with each of these questions. Ta-bles 3 and 4 report the combined prob-abilities and available effects sizes foreach of these questions.

Results

Conceptual Variables

Review of the TT research revealsthat this phenomenon is predominantlystudied from a nursing perspective. Ofthe articles, 89% (n= 8) had a nurse asthe primary researcher and author andwere published in a nursing journal.More than half of the studies cited Mar-tha Rogers’s science of unitary humanbeings as the philosophical base for theresearch. Only three of the studies fullyconceptualized the research from aRogerian perspective with the theoryevident in the operational definitionsand in the discussion of the findings andconclusions of the study (Meehan,1993; Quinn, 1989; Quinn & Strelkaus-kas, 1993). A third of the studies used anonnursing theory as the conceptualmodel, and one study was atheoretical(see Table 2).

Sample

All of the studies reviewed includeda sample size; however, there were lim-ited data reported regarding demo-graphic characteristics of the samplepopulations. Table 5 summarizes thedemographic data. The age of the sub-jects ranged from 2 weeks to 83 years.Most (89%) of the TT research has beenconducted with adults. Three studies

Effectiveness of Therapeutic Touch 55

Table 3Physiological Outcomes: Combined Probabilities and Effect Sizes

Within Tx Group Response Treatment Compared With Control

Intervention Does TT Effect Is TT More EffectAuthor Variable Sample↑ Physiologic Sizea Combined Effective Size Combined

and Year Analyzed Size Response? (Cohen’sd) Probability Than Control? (Cohen’sd) Probablity Fail-SafeN

Keller and McGill-Melzack 60 Yes;p < .0001 d = +1.76 Insufficient ↓ Headache pain,d = +0.66 Zc = 4.05, Nfs 05 = 20,Bzdek (1986) Pain Scale data p < .005 p < .000003 Nfs 01 = 8

to calculate.Kramer (1990) PMRRT (tool of 30 Not reported ƒ ↓ Stress, d = +1.9

Physiologic p < .05measures)

Meehan (1993) Visual Analogue 108 Yes; significanced = +0.67 ↓ Pain, d = +0.57Pain Scale not reported p < .06

Wirth et al. Wound healing 24 Not ↑Wound healing, d = +1.66(1993) per Mds measurable ƒ p < .01

NOTE: TT = therapeutic touch.ƒ: Unable to calculate value from data provided in the research report.

at MEMORIAL UNIV OF NEWFOUNDLAND on August 1, 2014nsq.sagepub.comDownloaded from

Page 6: The Effectiveness of Therapeutic Touch: A Meta-Analytic Review

(33%) involved elderly clients with agrand mean of 65 years of age; threestudies (33%) involved younger adultswith 40 years of age as the mean; andtwo studies were done with adults rang-ing from 23 to 79 years of age, but amean age was not reported in eitherstudy. Only one study was conductedwith children. Gender of subjects wasreported in only five studies (55%) butwith a very unequal distribution notedin four of the study samples. Most of thestudies (n= 5) did not provide ethnicity,race, or cultural affiliation data. Threestudies made statements that the sub-jects were predominantly White but didnot provide any data to support thosestatements. Six studies presented someadditional socioeconomic data on theresearch subjects. Despite these limita-tions in reporting the actual data, twostudies provided information indicatingthat researchers had done detailed sta-tistical analysis on the impact of demo-graphic variables on intervention out-comes. Two other studies indicated thatthere were no significant differencesbetween groups based on these vari-

ables; the remaining five studies re-ported no analysis to determine demo-graphic impact on outcomes. In themeta-analysis, effect sizes did not ap-pear to vary based on the ages of thestudy subjects. There were too few stud-ies and too little data provided to deter-mine if any other demographic variableaffected the effect-size values.

Methodological Characteristics

All of the studies were conductedwith a convenience sample. Of the eightstudies that had two or more groups,seven (88%) reported random assign-ment to groups. However, the descrip-tions of the random assignment prac-tices were not clearly described raisingmethodological questions. Samplesizes ranged from 4 to 153 subjects. Al-most half of the studies, 44% (n= 4) hada sample size of less than 35 subjects(see Table 2). Only one study did apower analysis prior to entering sub-jects into the study. There was little dis-cussion of the rate of refusal to partici-pate in these studies, and drop-out rateswere not clearly addressed. Only one

study (Gagne & Toye, 1994) addressedthe drop-out problem. This study re-ported a 30% drop-out rate; however, itis not clear what percentage was reflec-tive of true drop out versus subjects notmeeting inclusion criteria. All of thestudies were coded based on these andother methodological issues in theQuality of Study instrument (see Table 6).The study identified as having the great-est number of methodological concernsreported the highest level of signifi-cance in reducing anxiety and had thegreatest effect-size value (+ 4.41) (seeTable 6).

Intervention Characteristics

Although all studies examined TT asthe treatment variable, there was noconsistency in the definition of the pro-cedure. One study provided TT during aback rub, four studies (44%) reporteddoing noncontact TT, and four studies(44%) did not specify contact versusnoncontact TT (see Table 2). It was notclear whether subjects received the TTwhile sitting or lying. Timing of the TTintervention varied greatly. In studies

56 Nursing Science Quarterly, 12:1, January 1999

Table 4Psychological Outcomes: Combined Probabilities and Effect Sizes

Within Tx Group Response Treatment Compared With Control

Intervention Does TT Effect Is TT More EffectAuthor Variable Sample ↑ Physiologic Size Combined Effective Size Combined

and Year Analyzed Size Response? (Cohen’sd) Probability Than Control? (Cohen’sd) Probablity Fail-SafeN

Quinn (1989) STAI 153 Yes. Significance d = +0.21 Insufficient ↓Anxiety d = +0.26 Zc = 1.76, Nfs 05 = 1not reported. data to p < .51 p < .04

calculate.Quinn and STAI 4 Yes. Significance d = + 0.27 No control ƒStrelkauskas not reported. group.(1993)

Simington and STAI 105 Not measured ƒ ↓ Anxiety, d = +0.70Laing (1993) in study. p < .05

Gagne and STAI 31 Yes;p < .001 d = +4.41 ↓ Anxiety, ƒToye (1994) ?TT or RT,

p < .08

Olson (1995) STAI 40 Yes, significance d = +1.03 ↓ Anxiety. d = +0.48not reported. significance

not reported.

NOTE: TT = therapeutic touch.ƒ: Unable to calculate value from data provided in the research report.

at MEMORIAL UNIV OF NEWFOUNDLAND on August 1, 2014nsq.sagepub.comDownloaded from

Page 7: The Effectiveness of Therapeutic Touch: A Meta-Analytic Review

with preset limits, time varied from 3 to15 minutes, with four of the studies es-tablishing a 5-minute limit. In onestudy, the timing was variable, deter-mined by the practitioner’s experiencewith each subject (see Table 2).

Researchers tried to control for theplacebo and Hawthorne effects in dif-ferent ways. One method was to providethe treatment intervention without al-lowing facial contact between the sub-ject and the researcher (n= 3). The othermethod included using a mimic TT(placebo) intervention (n= 6). Moststudies with a mimic group used experi-enced TT practitioners to provide themimic intervention. Only one studyused a nontherapeutic touch practition-er to conduct the mimic sessions. Sig-nificant differences in the outcomes of

mimic and actual TT interventions aredifficult to determine because of insuf-ficient data reporting. A number ofstudies indicated no significant differ-ences; however, they did not providepvalues to allow for aggregation of effectto be calculated. In addition, the qualifi-cations of the persons conducting thetreatment interventions were not clearlyreported. TT practitioners ranged fromnursing assistants to doctorally pre-pared nurses. However, other qualifica-tions such as certification or years of ex-perience as a TT practitioner weredescribed in only one study (Quinn,1989). Yet, one of the reasons cited forParkes’s (1985) reporting nonsignifi-cant study findings was related to thelack of practitioner experience (Quinn,1989). In five studies, it was possible

that the subjects had a previous and/orongoing relationship with the practi-tioner or researcher (Gagne & Toye,1994; Keller & Bzdek, 1986; Kramer,1990; Olson & Sneed, 1995; Wirth,Richardson, Edelman, & O’Malley,1993). Because the researcher and prac-titioner relationship to subjects was notclearly described, it raises methodo-logical issues regarding the interventionthat could have an impact on the validityof results.

Effect Size

All of the studies reviewed providedsufficient data so that at least one effectsize value could be coded. Six studies(67%) provided sufficient informationto determine an effect size value for theeffectiveness of TT pre-/postinterven-

Effectiveness of Therapeutic Touch 57

Table 5Sample Characteristics

Analyzed Groups (G)Author Variables Sample Age Range (R) Number Other Demographic or Outcomes (O)and Year and Population Size and Mean (M) by Sex Race/Ethnicity Data Reported by Demographics

Psychological variablesQuinn Anxiety in open 153 R= 29-83, 38 females, 10 Caucasian Education, religion Yes – G(1989) heart surgical M = 60 115 males

patients

Quinn and Anxiety and 4 R= 47-72, 2 females, — Education, occupation, —Strelkauskas immune status M = 60 2 males religion(1993) in bereaved

Simington Anxiety in 105 R= —, 71 females, 10 Caucasian Religion Yes – Gand Laing elderly in M =75 34 males(1993) long-term care

Gagne and Anxiety in inpatient 31 R= 29-69, — — Veterans —Toye (1994) psychiatric patients M = 43

Olson and Anxiety in healthy 40 R= —, — — Occupation, education —Sneed (1995) professional M = —caregivers

Physiological variablesKeller and Tension headache 60 R= 18-59, 45 females, 10 Caucasian Occupation, education OBzdek (1986) pain in adults M = 30 25 males

Kramer Stress in children 30 R= 2 weeks— — — — —(1990) hospitalized for 2 years,

acute injury or illness M = —

Meehan Postoperative pain 108 R= 23-79, 74 females, 10 Caucasian — O(1993) in adult surgical M = — 34 males

patients

Wirth et al. Wound healing in 24 R= 35-63, — — — —(1993) volunteers with M = 47

punch biopsy

at MEMORIAL UNIV OF NEWFOUNDLAND on August 1, 2014nsq.sagepub.comDownloaded from

Page 8: The Effectiveness of Therapeutic Touch: A Meta-Analytic Review

tion in the treatment group. Seven stud-ies (78%) provided data to calculate aneffect size comparing the outcome ofthe TT group with a control or placebogroup (see Tables 3 and 4). Effect sizesin this analysis are evaluated using theguidelines suggested by Cohen (1977),in which a small effect is .2, medium ef-fect is .5, and large effect is .8. Given thesmall number of studies available to beaggregated in this meta-analytic review,no statistically significant results can beidentified regarding effect sizes. How-ever, there are some important trendsthat should be noted because they mayaffect the future of TT research. Be-cause combining effect sizes for differ-ent outcomes is a meta-analytic flaw,the trends discussed are based on ananalysis that partitioned on psychologi-cal and physiological outcomes.

Physiological Outcomes

Analysis of treatment outcome—experimental versus control interven-tion—revealed that effect sizes rangedfrom + 0.57 to +1.9 for physiologicaloutcomes (see Table 3). This providedan averaged of + 1.20, which wouldseem to indicate a large effect. How-ever, it is important to note that the high-est effect size values were found in thetwo studies that had small sample sizesand methodological shortcomings. Ifthese two studies are removed from

consideration because of quality ofstudy issues, then the effect size is+0.61, or a medium effect. When ana-lyzing the data regarding within-grouptreatment outcomes (comparing pre-and posttest scores for the experimentalsubjects), then a large average effectsize ofd = +1.22 is seen for physiologi-cal outcomes. This is found in studieswith high quality of study rating scores.However, there are only two studies thatprovided sufficient data to contribute tothis calculation.

Psychological Outcomes

Analysis of the effectiveness of TTon producing the desired psychologicaloutcome within the treatment groupproduced effect size values rangingfrom d = + 0.21 to + 4.41, an overalllarge effect size value. If the one outliervalue (+ 4.41) is removed, which is jus-tified given its low methodologicalscore, then the overall effect size is+0.72, a medium effect. However, itshould be noted that “sitting quietly”also produced a medium effect in reduc-ing anxiety. When TT is compared tocontrol interventions for psychologicaloutcomes, the effect sizes range fromd = +0.26 to +0.70, with an average of+0.48. Again, a medium effect is found(see Table 4).

These results seem to indicate thatonce poorer quality studies are removed

from analysis, there is a medium level ofeffect seen when one compares TT withcontrol groups. Unfortunately, becauseof insufficient information, it is not pos-sible to calculate the weighted meand,which is necessary to test for homoge-neity/heterogeneity of effect size (Wolf,1986).

Combined Probabilities

Further analysis of the data involvedevaluating the Stouffer’s Combinedprobabilities outcomes. The combinedprobabilities score, when comparingTT with control group outcomes, is Zc =4.05 (p< . 000003) for physiologicalstudies (see Table 3) and Zc = 1.76 (p<.04) for psychological studies (see Ta-ble 4). There was insufficient data avail-able to calculate combined probabilitiesfor within-group effectiveness. Moresignificant, however, is the result of thefail-safe N calculations. These weredone to determine the number of studiesneeded to confirm the null hypothesesand reverse the overall probability ob-tained from the combined tests. Forphysiological outcomes, it would take20 studies to indicate that TT was not aneffective nursing intervention, thus con-firming the null hypothesis (p< .05).However, for psychological outcomes,it would take only one study to reversesupport for TT as an effective interven-tion in reducing psychological distress(p < .05). This is very important giventhat there are at least three unpublisheddissertations done between 1985 and1996 that showed no significant effectof TT on reducing anxiety (see Table 7).

Discussion

Although great strides have beenmade in the conceptualization and so-phistication of the TT research, thismeta-analytic review indicates that sig-nificant problems remain. Four majorweaknesses must be addressed if futurestudies are to establish a scientific basesupporting the effectiveness of TT.These problems include sampling pro-cedures, intervention practices, practi-tioner skill, and the underreporting ofdata.

Underreporting of data made it diffi-

58 Nursing Science Quarterly, 12:1, January 1999

Table 6Quality of Study Results and Effect Size

Effect Size Effect Size(Cohen’sd), (Cohen’sd),

Author Overall Sample Within-Group Tx Versusand Year Quality of Study Size TT Effect Control

Psychological variablesQuinn (1989) 2.70 153 d = +0.21 d = +0.26Quinn and Strelkauskas (1993) 2.10 4 d = +0.27 No controlSimington and Laing (1993) 2.13 105 ƒ d = +0.70Gagne and Toye (1994) 1.88 31 d = +.4.41 ƒOlson and Sneed (1995) 2.00 40 d = +1.03 d = + 0.48

Physiological variablesKeller and Bzdek (1986) 2.45 60 d = +1.76 d = +0.66Kramer (1990) 1.27 30 ƒ d = +1.9Meehan (1993) 2.68 108 d = +0.67 d = +0.57Wirth et al. (1993) 1.93 24 ƒ d = +1.66

SOURCE: Study quality rating from Smith and Stullenbarger (1991, pp. 1282-1283).NOTE: Rating key was as follows: 0 = absent, 1 = low, 2 = medium, and 3 = high. TT = therapeutic touch.ƒ: Unable to calculate value from data provided in the research report.

at MEMORIAL UNIV OF NEWFOUNDLAND on August 1, 2014nsq.sagepub.comDownloaded from

Page 9: The Effectiveness of Therapeutic Touch: A Meta-Analytic Review

cult to analyze the impact of study vari-ables on treatment outcomes. The lim-ited demographic data provided aboutstudy subjects makes generalizabilityvirtually impossible. It is not clearwhether the outcome is affected by age,sex, or socioeconomic status. The un-derreporting of results creates anotherproblem in a quantitative review. ManyTT researchers reported a treatment ef-fect but stated that it was not significant.Lack of statistical values for nonsignifi-cant findings made it difficult to deter-mine an aggregate effect. Although afinding may be nonsignficant in onestudy, if reported, it could be pooledwith other findings. The combined re-sults may show a small, overall effectsize value. These limitations in report-ing often cannot be addressed by indi-vidual authors. It is editors who cangreatly increase the likelihood thatknowledge can be synthesized acrossstudies by adopting publication guide-lines that include information neededfor meta-anlaytic reviews.

Sampling procedure issues are thesecond major problem in TT research.Because all of the studies used a conve-nience sample, random assignment togroups is imperative. The assignmentpractices in the majority of the studiesin this review were not clearly deline-

ated. This raises serious questions re-garding the results.

The third major problem involvesthe lack of description provided regard-ing the actual intervention practices.This review reveals that there is a lack ofconsistency in the TT procedure, mak-ing comparisons among studies diffi-cult. There is insufficient data to deter-mine whether contact or noncontacttouch is most effective or whether facialcontact makes a difference in the treat-ment outcome. Given the small numberof studies and the various interventionpractices used, it is not possible to drawmore than tentative conclusions. It maybe that intervention techniques, al-though all were labeled TT, really repre-sent different practices. There is so littledata reported about the actual interven-tion that the studies could not be parti-tioned and examined for interventionimpact on effect size.

The fourth significant problem notedis the relationship of the practitionervariable to treatment outcomes. Littledata were provided regarding the levelof TT practitioner expertise. This couldbe an especially important considera-tion. Descriptive findings reported byQuinn and Strelkauskas (1993) andMeehan (1993) suggest that the subjec-tive experience of the intervention var-

ies based on the interaction betweenpractitioner and subject. If TT is trulybased on Rogerian principles, thenconceptually one must accept thesimultaneity-transformative nature ofthe mutual process occurring betweenresearcher and subject. Setting a pre-conceived time limit may be a methodo-logical flaw that significantly affectsstudy results. This also raises the issueof the impact of practitioner expertiseon treatment outcomes as well as theimpact of using TT practitioners to pro-vide the mimic interventions.

Summary

This meta-analytic review raisesmany issues and concerns regarding theTT research. It appears that TT can pro-duce a medium effect for physiologicaloutcomes and psychological outcomeswithin treated subjects. It also appearsthat TT produces a medium effect onphysiological outcomes when compar-ing treatment with control groups.However, the fail-safe N indicates that,currently, there is not enough empiricaldata to support TT as more effectivethan control measures in improvingpsychological well-being. However,given the limited amount of researchconducted in the past decade and miss-ing information in published studies, itis impossible to make substantiativeclaims about the TT research base fromthis initial meta-analytic review.Whereas extending the meta-analyticreview to include unpublished studiesand research conducted before 1986will add to our understanding of the cur-rent state of the science, more rigorousresearch still needs to be done to estab-lish a solid body of evidence that sup-ports the effectiveness of TT as a nurs-ing intervention.

ReferencesBangert-Drowns, R. (1986). Review of de-

velopments in meta-analytic method.Psychological Bulletin,99(3), 388-399.

Birch, E. (1986). The experience of touchreceived during labor: Postpartum per-ceptions of therapeutic value.Journal ofNurse-Midwifery,31(6), 270-276.

Burns, N., & Grove, K. (1993).The practiceof nursing research(2nd ed.). Philadel-phia: W. B. Saunders.

Effectiveness of Therapeutic Touch 59

Table 7Quantitative Dissertation Research on Therapeutic Touch 1986 to 1996

Intervention Outcome: TTAuthor Variables Sample of Variable as Comparedand Year Design and Population Size Intervention Analyzed With Control

Parkes Experimental Anxiety in 60 NCTT; mimic STAI No significant(1986) Pre-/posttest hospitalized TT adapted ↓ in anxiety

with control elderly MTTHale Anxiety in — TT STAI No significant(1986)a hospitalized ↓ in anxiety

adultsHinze Experimental Experimentally 48 TT, accupressure STAI and No significant(1988) two-way induced pain mock TENS, physiologic↓ in anxiety;

factorial 4× 3 in healthy No Tx measures no significant(treatment by volunteers ↓ in pain;↑ intime) design perceived

effectivenessof TT

SOURCE: Reported in CINAHL, Medline, Psychinfo (not Dissertation Abstracts)NOTE: Effect size not calculated for dissertation studies. TT = Contact therapeutic touch; NCTT = Non-contact therapeutic touch; MTT = mimic therapeutic touch; TENS = transcutaneous electrical nervestimulation; Tx = treatment.a. As reported in Meehan (1993, p. 70).

at MEMORIAL UNIV OF NEWFOUNDLAND on August 1, 2014nsq.sagepub.comDownloaded from

Page 10: The Effectiveness of Therapeutic Touch: A Meta-Analytic Review

Bush, A., & Geist, C. (1992). Geophysicalvariables and behavior: LXX. Testingelectromagnetic explanations for a pos-sible psychokinetic effect of therapeutictouch on germinating corn seed.Psycho-logical Reports,70(3), 891-896.

Cohen, J. (1977).Statistical power analysisfor the behavioral sciences(Rev. ed.).New York: Academic Press.

Curlette, W., & Cannella, K. (1985). Goingbeyond the narrative summarization ofresearch findings: The meta-analysis ap-proach.Research in Nursing & Health,8, 293-301.

Devine, E. (1990). Meta-analysis. In N.Chaska (Ed.),The nursing profession:Turning points(pp. 180-185). St. Louis,MO: Mosby.

Devine, E., & Cook, T. (1983). A meta-analytic analysis of effects of psycho-educational interventions on length ofpostsurgical hospital stay.Nursing Re-search,32(5), 267-273.

Devine, E., & Reifschneider, E. (1995). Ameta-analysis of the effects of psycho-educational care in adults with hyperten-sion.Nursing Research,44(4), 237-245.

Engebretson, J. (1996). Comparison ofnurses and alternative healers.Image,28(2), 95-99.

Fawcett, J., Sidney, J., Hanson, M., & Riley-Lawless, K. (1994). Use of alternativehealth therapies by people with multiplesclerosis: An exploratory study.HolisticNursing Practice,8(2), 36-42.

Ferguson, C. (1986). Subjective experienceof Therapeutic Touch Survey (SETTS):Psychometric examination of an instru-ment. Dissertation Abstracts Interna-tional, 47(5-B), 1927.

France, N. (1991).A phenomenological in-quiry on the child’s lived experience ofperceiving the human energy field usingtherapeutic touch. Unpublished doctoraldissertation, University of ColoradoHealth Sciences Center, Boulder.

*Gagne, D., & Toye, R. (1994). The effectsof therapeutic touch and relaxation ther-apy in reducing anxiety.Archives of Psy-chiatric Nursing,8(3), 184-189.

Glass, G. V. (1976). Primary, secondary andmeta-analysis of research.EducationalResearcher,5, 3-8.

Grad, B. (1963). A telekinetic effect on plantgrowth. International Journal of Para-psychology,5, 117-133.

Grad, B. (1964). A telekinetic effect of plantgrowth II. International Journal of Para-psychology,6, 473-485.

Grad, B. (1965). Some biological effects ofthe laying-on of hands: A review of ex-periments with animals and plants.Jour-nal of the American Society for PhysicalResearch,59, 95-127.

Grad, B., Cadoret, R., & Paul, G. (1961). An

unorthodox method of wound healing inmice.International Journal of Parapsy-chology,3, 5-24.

Hale, E. H. (1986). A study of the relation-ship between therapeutic touch and anxi-ety levels of hospitalized adults.Dissertation Abstracts International,47(5), 1928B. (University Microfilms No.8618897)

Hamilton, G. (1988).Therapeutic touch:Promoting and assessing conceptualchange among health care professionals.Unpublished doctoral dissertation,Michigan State University, East Lans-ing.

Heidt, P. (1990. Openness: A qualitativeanalysis of nurses’ and patients’ experi-ences of therapeutic touch.Image, 22(3),180-186.

Hinze, M. L. (1988).The effects of therapeutictouch and accupressure on experimen-tally-induced pain. Unpublished doctoraldissertation, The University of Texas atAustin.

Hughes, P., Meize-Grochowski, R., & Har-ris, C. (1996). Therapeutic touch withadolescent psychiatric patients.Journalof Holistic Nursing,14(1), 6-23.

*Keller, E., & Bzdek, V. (1986). Effects oftherapeutic touch on tension headachepain.Nursing Research,35(2), 101-106.

*Kramer, N. (1990). Comparison of thera-peutic touch and casual touch in stress re-duction of hospitalized children.Pediatric Nursing,16(5), 483-485.

Krieger, D. (1972). The response of in-vivohuman hemoglobin to an active healingtherapy by direct laying-on of hands.Hu-man Dimensions,1, 12-15.

Krieger, D. (1973). The relationship oftouch with the intent to help or heal, tosubjects’ in-vivo hemoglobin values: Astudy in personalized interaction. InPro-ceedings of the Ninth American Nurses’Association Research Conference(pp.39-58). New York: American Nurses’Association.

Krieger, D. (1974). Healing by the laying-onof hands as a facilitator of bioenergeticchange: The response of in-vivo humanhemoglobin.Psychoenergetic Systems,1, 121-129.

Krieger, D. (1975). Therapeutic touch: Theimprimatur of nursing.American Jour-nal of Nursing,75(5), 784-787.

Krieger, D. (1979).The therapeutic touch:How to use your hands to help or to heal.Englewood Cliffs, NJ: Prentice Hall.

Krieger, D. (1990, September/October).Therapeutic touch: Two decades of re-search, teaching and clinical practice.NSNA/Imprint, pp. 83-88.

Lionberger, H. (1986). An interpretive studyof nurses’ practice of therapeutic touch.Dissertation Abstracts International,

46(8-B), 2624.Lynn, M. (1989). Meta-analysis: Appropri-

ate tool for the integration of nursing re-search.Nursing Research,38(5), 302-305.

McGaw, B., & Glass, G. (1980). Choice ofthe metric for effect size in meta-analysis.American Educational Re-search Journal,7, 325-337.

*Meehan, T. (1993). Therapeutic touch andpostoperative pain: A Rogerian researchstudy.Nursing Science Quarterly,6(2),69-78.

Mersmann, C. (1994). Therapeutic touchand milk letdown in mothers ofnon—nursing preterm infants.Disserta-tion Abstracts International,54(9-B),4602.

Messenger, T., & Roberts, K. (1994). Theterminally ill: Serenity nursing interven-tions for hospice clients.Journal of Ger-ontological Nursing,20(11), 17-22.

*Olson, M., & Sneed, N. (1995). Anxietyand therapeutic touch.Issues in MentalHealth Nursing,16, 97-108.

Olson, M., Sneed, N., Bonadonna, R., Rat-liff, J., & Dias, J. (1992). Therapeutictouch and post-hurricane Hugo stress.Journal of Holistic Nursing,10(2), 120-136.

Parkes, B. (1985). Therapeutic touch as anintervention to reduce anxiety in elderly,hospitalized patients.Dissertation Ab-stracts International,47(2-B), 573.

Quinn, J. (1988). Building a body of knowl-edge: Research on therapeutic touch1974-1986.Journal of Holistic Nursing,6(1), 37-45.

*Quinn, J. (1989). Therapeutic touch as en-ergy exchange: Replication and exten-sion. Nursing Science Quarterly,2(2),79-87.

*Quinn, J., & Strelkauskas, A. (1993). Psy-choimmunologic effects of therapeutictouch on practitioners and recently be-reaved recipients: A pilot study.Ad-vances in Nursing Science,15(4), 13-26.

Randolph, G. (1984) Therapeutic and physi-cal touch: Physiological response tostressful stimuli.Nursing Research,33(1), 33-36.

Rubens, D., Gyurkovics, D., & Hornacek,K. (1995). The cultural production ofbioterapia: Psychic healing and the natu-ral medicine movement in Slovakia.So-cial Science & Medicine,41(9), 1261-1271.

Samarel, N. (1992). The experience of re-ceiving therapeutic touch.Journal of Ad-vanced Nursing,17(6), 651-657.

*Simington, J., & Laing, G. (1993). Effectsof therapeutic touch on anxiety in the in-stitutionalized elderly.Clinical NursingResearch,2(4), 438-450.

Smith, M., & Stullenbarger, E. (1991). A

60 Nursing Science Quarterly, 12:1, January 1999

at MEMORIAL UNIV OF NEWFOUNDLAND on August 1, 2014nsq.sagepub.comDownloaded from

Page 11: The Effectiveness of Therapeutic Touch: A Meta-Analytic Review

prototype for integrative review andmeta-analysis of nursing research.Jour-nal of Advanced Nursing,16, 1272-1283.

Smyth, D. (1995). Healing through nursing:The lived experience of therapeutictouch. Australian Journal of HolisticNursing,2(2), 15-25.

Stravena, J. (1992).Therapeutic touch andin vitro erythropoiesis. Unpublisheddoctoral dissertation, Indiana UniversitySchool of Nursing, Bloomington.

Vaughn, S. (1995). The gentle touch.Jour-nal of Clinical Nursing,4, 359-368.

Wirth, D. (1995). Complementary healingintervention and dermal wound reepithe-lialization: an overview.InternationalJournal of Psychosomatics,40(1-4), 48-

53.Wirth, D., & Barrett, M. (1994). Comple-

mentary healing therapies.InternationalJournal of Psychosomatics,41(1-4), 61-67.

Wirth, D., Barret, E., & Eidelman, W.(1994). Non-contact therapeutic touchand wound re-epithelialization: An ex-tension of previous research.Comple-mentary Therapies in Medicine, 2(4),187-192.

Wirth, D., & Cram, J. (1993). Multi-siteelectromyographic analysis of non-contact therapeutic touch.InternationalJournal of Psychosomatics,40(1-4), 47-55.

Wirth, D., & Cram, J. (1994). The psycho-

physiology of nontraditional prayer.In-ternational Journal of Psychosomatics,41(104), 68-75.

*Wirth, D., Richardson, J., Eidelman, W., &O’Malley, A. (1993). Full thickness der-mal wounds treated with non-contacttherapeutic touch: A replication and ex-tension. Complementary Therapies inMedicine,1, 127-132.

Wolf, F. (1986).Meta-analysis: Quantita-tive methods for research synthesis. Bev-erly Hills, CA: Sage.

Wright, S. (1991). Validity of the human en-ergy field assessment form.WesternJournal of Nursing Research,13(5), 635-647.

Effectiveness of Therapeutic Touch 61

at MEMORIAL UNIV OF NEWFOUNDLAND on August 1, 2014nsq.sagepub.comDownloaded from