Experienced investigators would probably agree that recruitment to DH clinical studies is fraught with difficulties, and this may be due in part to a number of reasons, for example, the limitation of sites that may exhibit DH is predominantly on the buccal or facial surfaces of the teeth which may be restricted due to attempts by clinicians to restore the non-carious cervical defects present in an individual’s mouth. The authors estimate that newer restorative procedures have made recruitment of potential participants into DH studies more difficult ranging from 1 in 3 some 25 years ago to approximately 1 in 8 today which may considerably increase the costs (to institutions and companies) of conducting these types of studies. A further problem may be due to the nature of the condition itself both in terms of its history (may have periods of spontaneous resolution) and the subsequent pain response on screening. The impact of this particular problem may be underestimated; generally speaking, individuals are included into DH studies on the basis of having a previous history of DH which may lead to difficulties in recruitment particularly when assessing pain scores in studies with the current exacting and inclusion/exclusion criteria. It is important however when recruiting individuals for a clinical study that the entry criteria are reasonable and realistic; otherwise, the investigator will struggle to recruit adequate numbers within the allocated time frame for completion of the study. This is particularly true when dealing with a highly subjective clinical condition such as DH. Care must therefore be taken when screening prospective participants not to recruit individuals who either report minimal or extreme discomfort since the statistical probability of measuring the pain response can only stay the same, worsen or improve, respectively. This phenomenon is called regression towards the mean or mode (Yates et al. 1998; Addy et al. 2007) and according to Jeffcoat (1993) can either magnify a product’s treatment effect if used on a severely affected population or minimise the effect when used in an under-affected population.

One of the inherent problems in conducting clinical studies is the interference of placebo and/or Hawthorne effects that may introduce bias into the study (Gillam 1997; Addy et al. 2007). This may be particularly true in pain studies such as relief of pain arising from DH, and one way to resolve this particular bias is the use of a double-blind placebo-controlled study (Jeffcoat 1993; Holland et al. 1997; ADA 1998, 2009, 2012). According to Schulz et al. (2010) the Randomized controlled trial (RCT), when appropriately designed, has been generally acknowledged as the gold standard for evaluating health care interventions. This viewpoint, however, has been recently challenged as a result of the number of ‘black box’ warnings placed on drugs by the FDA and the number of drugs subsequently removed from the market place by the Regulatory Authorities due to safety concerns (Kessler and Glasgow 2011; Silverman 2014). RCTs are now viewed as too restrictive and not representative as point-of-use by providers. RCTs conducted by highly trained specialists recruiting a ‘designer patient’ using the drug in a restrictive environment hardly yields results that would be robust and mimic use for the population at large. RCTs should therefore be reserved for early phase clinical studies when the safety profile has not been determined, when the delivery is complicated, and when the drug is directed at a compromised patient. In this type of study, all participants are allocated (random assignment) to either the test or control group, and the difference in outcome across groups will determine any significant efficacy of the test group. The design is simple and requires minimal prior knowledge of the cause of the disease in each group. This type of design, however, may result in an unequal distribution of high-risk individuals with multiple active sites, although this feature may be more of a problem in studies of less than 30 subjects. The use of stratification techniques, however, can be used to give balance across the groups. It is usual to stratify potentially confounding factors such as numbers of teeth, baseline sensitivity levels, age, gender, etc. From a practical point of view, it is generally easier to assign a treatment to each of the participants. This type of design has been recommended for clinical studies evaluating both OTC and in-office products although other types of study design have been used in DH studies such as cross-over and split-mouth (Gillam 1997). The use of a pretreatment period study design which enabled the participant(s) to be randomised into test and control groups and as such allows the participant to act at his/her own control as well as to compare that subject to the randomised control and other groups may also be utilised in DH studies (Page et al. 1995). For example, there may be advantages in having a 2–4-week pre-study evaluation period for sensitivity studies in which the participant is introduced to the pain scoring methods (visual analogue scales (VAS)) and is placed on a placebo or fluoride toothpaste. It would be anticipated at the end of such a period that (a) the participants understand what is expected of them with regard to the various assessments used when measuring the pain response, e.g., VAS, Schiff Air Sensitivity scores, LMS and (b) participants actually have ‘real sensitivity’ before being entered into the main study. One observation may be relevant here with respect to the Schiff air test as the patient’s reaction to the test is translated by the investigator thereby excluding the patient’s response. Those individuals fulfilling the entry criteria following this pretreatment period can then enter the study. One disadvantage of this type of study design, however, is that it adds to both cost and time. Other study designs that have also been reported in the literature involve the use of an adjunctive therapy (e.g. the application of a potassium-containing mouth rinse following routine toothbrushing with a fluoride toothpaste) (Gillam et al. 1996a; Yates et al.1998). This type of study, while widely utilised in antimicrobial therapy studies, has not been used routinely in DH studies, but there would be advantages in incorporating such a study design in evaluating the additional effect above what would normally be expected from using a commercially available fluoride toothpaste. DH studies may also lend themselves to an adaptive design where the knowledge of the patient population actually enrolled in the study, the variance of the study end point, event rates, and possibly the treatment effect itself available at interim may be used to change the study design or key parameters in order to improve the chances of success. Such adaptation, if appropriately implemented, has the potential to save resources; however, this implementation requires that the statistical validity of the study’s results be preserved (LaVange 2014). The US regulatory agency (FDA) has also recently issued draft guidance on adaptive designs, and the final version is in progress considering the public comments received (FDA guidance for Industry 2014).

According to Fleiss (1992), clinical studies may assess the efficacy, equivalence or superiority of a treatment and as such studies can be designed to test either products or ingredients of products. Products designed to reduce DH should therefore be tested in controlled clinical studies in which the test product is tested against (a) a negative control to establish efficacy and (b) a positive control to determine relative potency. In the absence of any negative controls with no demonstrable desensitising effects and any internationally accepted benchmark positive control, the test product may be tested against a minus-active control.

One of the problems, however, in this respect is that there is no agreed positive or negative control (the so-called gold standard) in DH studies (Holland et al. 1997). For example, several published studies have used minus-active controls that may not be representative of the final commercially available product sold over the counter (in the case of OTC products) (Gillam 1997). Furthermore, the characteristics, bio-equivalence, efficacy, etc., may be changed by addition/subtraction of different ingredients.

A further complication that continually appears to affect the interpretation of results from clinical studies of this nature is the reported placebo effect that can be as high as 40 % (Curro et al. 2000; West et al. 1997). Several investigators have also alluded to this obvious effect in their studies (Gillam 1997; Pearce et al. 1994; Chesters et al. 1992), but to what extent the placebo effect complicates the interpretation of the results of the study is difficult to predict. It should however be noted that according to Curro et al. (2000), the placebo effect observed in DH studies is not too dissimilar to those reported in other medical and dental therapeutic studies. For example, a review of 15 postoperative pain studies by Beecher (1955) quoted by Curro et al. (2000) concluded that on average symptoms were satisfactorily relieved by the placebo medication in 35 % of the patients (the placebo response range of 15–58 %). Compounding factors (Table 7.1) that may affect DH studies may also be complicated by the lack of acceptable positive and negative controls used in equivalence and superiority studies. Unlike gingivitis studies where the internationally accepted gold standard would be chlorhexidine, there does not appear to be an internationally accepted product which one could use as a positive control in such studies. Furthermore, the difficulty in using a negative control which may also exhibit a desensitising effect may also complicate the interpretation of the results and is also problematic (Gillam et al. 1996a, b; Gillam 1997; Pearce et al. 1994; Chesters et al. 1992). The use of a true placebo toothpaste, without any known desensitising ingredient, may however be problematic, and most studies would appear to use a fluoride-containing toothpaste as a negative control when evaluating a product. A final concern that may be expressed regarding these types of studies is whether the study population that are used for the studies are truly representative of those who suffer from the condition in real life, and perhaps more research should be initiated into the demographics of recruiting subjects for DH studies.

The study duration may also vary according to the purpose of the study, the nature of the agent tested, the hypothesis of the mechanism of action, the outcome measurements and their sensitivity and error. According to Gore and Altman (1982