Neurolinguistic programming: a systematic review of the effects on health outcomes

Available evidence

A total of 1459 citations were retrieved using the search strategy. Of these, 93 titles were potentially relevant (Figure 1). Abstracts were obtained and screened and 41 full text papers reviewed. Of the initial 93 citations, the majority were excluded as they were descriptive in nature, were not NLP interventions or they involved only healthy populations. In total, 10 studies were identified meeting the inclusion criteria (Table 1). Due to the small number of studies identified and heterogeneity (in research design, populations, NLP interventions, and assessed outcomes), statistical analysis was not appropriate and a narrative synthesis of the evidence was undertaken. Nine studies were published in peer-reviewed journals and one was identified online.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Flowchart of identified studies.

Characteristics of included studies

Five studies were randomised controlled trials (RCTs)19–23 and five were observational/ pre-post test studies.24–28 These were conducted in the US,3 Denmark,2 UK,2 Croatia, Brazil, and New Zealand. Targeted health conditions include various anxiety disorders,6 weight maintenance, morning sickness, substance misuse, and claustrophobia during MRI. The five RCTs targeted anxiety disorders,3 maternal anxiety/child development and weight maintenance. Two used NLP interventions versus a no-intervention control arm and three compared NLP to an active intervention. Within the five RCTs, follow-up occurred immediately following a single session treatment, at 1 month, 5 months, 9 months and 3 years.

Within the 10 included studies participants were recruited broadly from childcare, criminal justice, and public and private healthcare facilities, higher education and the press. Demographic data about participants was poorly reported with eight studies reporting some data on sex. Two of these studies specifically recruited women only and the six remaining reported data indicating that overall 64% of study participants were female. Six studies reported participant age, albeit inexactly, with a range from 17–75 years and a mid-range estimate of approximately 40 years. One of the RCTs21 presented broader socioeconomic data and found participants to be college educated or higher (100%), married (24%), and employed (56%).

Interventions

Delivery of NLP interventions ranged from three studies using a single 1–2 hour session19,25,28 to the remainder offering 4–20 1–2 hour sessions. Duration of intervention was reported by six studies ranging from 4 weeks,23 21 days residential,26 4–5 months22,27 and 12 months.20 One study offered group NLP.24

Six studies described the qualifications and training of the interventionists and these included three clinical psychologists,19 eight psychotherapists,21,23 a certified NLP practitioner22,25 and an NLP practitioner level radiographer.28 These NLP interventionists were all certified to a minimum of NLP practitioner level with two stated as being master practitioner certified.21,25 The interventionist training level was better described in the RCTs. Intervention techniques reported were mixed with six studies19,23,25–28 employing a well-established NLP intervention described in the NLP literature.29,30 Four studies referred generally to ‘NLP techniques’ or ‘NLP behaviour modification’. The observational studies provided greater detail about the interventions employed.

Outcomes

Across the 10 studies, 23 measures were used, and 18 outcomes reported. Outcomes were largely aligned to the targeted condition and the most common outcome assessed was anxiety19,23–26,28 with three also measuring quality of life23 and depression.23,24 Validated measures were referenced by seven studies with a further two reporting the outcome was assessed but not how. All psychological outcome measures were different across these eight studies. Only three studies measured objective outcomes, weight,22 successful completion of MRI scan,28 and urinalysis for illegal substances.27 Of the 18 outcomes reported, 11 were self-reported, three were objective measures, two were observed, and two not reported. The two observed measures were from RCTs19,20 and one had a blinded outcome assessor19 and the other was unclear.20 In general the RCTs performed no better than the pre-post studies in terms of reporting of outcomes and the process of their measurement.

Risk of bias

In three of the RCTs, the risk of bias was high with alternate group allocation,21,23 and incomplete outcome data reporting20 (Table 2). In two RCTs the risk of bias was uncertain.19,21 None reported results by intention to treat (ITT) analysis and, although one22 reported undertaking ITT analysis, only the completer analysis was presented. Three RCTs fared better in reporting withdrawals and participants lost to follow-up. In the pre-post study designs, the quality index scores18 ranged from 6–13 (maximum rating is 23 for non RCTs) where low ratings represent poor quality. Only one paper28 scored above the scale mid-point for quality.

NLP effects

Across the five RCTs, NLP was evaluated with undergraduate students, mother and child pairs, weight challenged adults, and emotionally-distressed adults for which the main outcomes were assessed ranging from immediately post-treatment to 3-year follow-up. Main outcomes reported were anxiety (self-report), child development (observed), weight (objective), and quality of life (self-report). Four RCTs reported no significant between group differences in the assessed outcomes with the fifth21 reporting less psychological distress and increased perceived quality of life in the NLP group compared to the waiting list control arm. Three RCTs and five pre-post studies reported within group improvements. Of the three studies measuring objective outcomes, one reported a post-treatment increase in completed MRI scans28 and the other two reported no post-treatment improvement in urinalysis for illegal substances27 or weight maintenance.22

Summary

This systematic review demonstrates that there is little evidence that NLP interventions improve health-related outcomes. The study conclusion reflects the limited quantity and quality of NLP research, rather than robust evidence of no effect.

Strengths and limitations

This represents the first well-conducted review investigating the effectiveness of NLP on health-related outcomes. The study has not attempted to define NLP and its many components and techniques and this complicates the interpretation of the evidence. This study took the authors’ word that they were delivering NLP if they said they were and the evidence of levels of training of the interventionist supported this assumption. Some academic investigation into NLP was found in unpublished German language dissertations that the library advised would not be possible to obtain and this represents a possible gap in the evidence. The decision was taken to exclude studies using single NLP techniques. NLP has a lack of consensus surrounding a definition of techniques and mechanism of effect and on an individual technique basis there is overlap with more established and evidence-based psychological techniques. Arguably these could include developing rapport = person-centred counselling; modelling = vicarious learning; eliciting well formed outcomes = goal setting; reframing = cognitive behavioural therapy (CBT) techniques; and anchoring = classical conditioning. Inclusion of studies labelled by their authors as NLP and focusing on one of these single NLP techniques would have lead to a misleading observation of the evidence. Publication bias assessment was not formally calculated because only 10 studies were included.18 The scoping reviews around the practice of NLP in physical and mental health conditions suggested it remains a controversial intervention. As only one of the five RCTs showing a positive effect in favour of NLP was found, the authors are less concerned about publication bias. However, it is possible that the controversy surrounding NLP may lead to a publication bias against studies that find a positive effect in favour of NLP.

Risk of bias in the five RCTs was high, or uncertain due to inadequate reporting of methods. It was not possible to determine the risk of bias associated with selective outcome reporting due to the absence of published study protocols. Assessment of the pre-post studies found four scoring lower than the scale midpoint score indicating a high risk of bias.

Implications for research and practice

There is currently insufficient evidence to recommend use of NLP for any individual health outcome. Neither this review, nor the FOI NHS trust data, point strongly to appropriate populations for further research. Use of NLP in specific settings may be vindicated in future, and preliminary data from its use in MRI/claustrophobia may justify a sufficiently powered RCT to clarify its role for these patients. Discussions with NLP key informants identified populations, for example allergy sufferers, who they felt were a strong target population for further NLP-based research. A formal stakeholder consultation with a range of NLP master practitioners would be an important next step for identifying such target populations for research. The strength of evidence for CBT would suggest it as a possible comparison group. The risk of bias assessments point to the need to develop a fully-specified and replicable intervention protocol for evaluation in a sufficiently powered RCT.