The two main methods of psychometric assessment in general are interviews and questionnaires. As a frequent application of QoL instruments is in clinical trials, self-administered questionnaires are used most often because of their brevity and ease of analysis.

The questionnaires can be classified into generic and specific instruments, whereby the former measures general aspects of QoL along the main dimension outlined by WHO (e.g., physical fitness, pain, anxiety) and the latter are developed for specific conditions or problems, such as head and neck cancer, high-dose chemotherapy, or fatigue.

The development of QoL questionnaires is mostly conducted in several phases. First, the relevant issues of a specific health problem are collected and rated regarding their importance. Second, items are formulated from the issues and a provisional questionnaire is tested in a group of patients. Eventually, after rephrasing and shortening, the final questionnaire is validated, preferably in different clinical and cultural settings. The involvement of patients from the first phase on is considered crucial in this process to ensure that the questionnaire validly measures QoL domains that are of high importance to patients. Some frequently used questionnaires are displayed and described in Table 1.

TABLE 1. Frequently Used Quality-of-Life Instruments

One of the earliest scales developed to measure QoL is the Karnofsky Performance Status, known as the “Karnofsky Index.”² This instrument, published in 1949, assesses the patient's physical functioning abilities to care for him- or herself. Although this is one of the most frequently used scales, its reliability and sensitivity to change are poor³; therefore, it should be administered only in conjunction with more comprehensive and reliable tools.

It may be common sense that the perceived quality of our life is an important predictor of satisfaction and thus relevant in health care planning. However, quality of life is still not routinely measured in clinical practice. In clinical trials, assessing QoL is becoming more and more common, but the analysis of these data is often unsatisfactory. Why?

First, it might be that the investigators do not have enough knowledge on how to analyze the questionnaires used. They may be discouraged by the multidimensionality of the instruments, which results in multiple outcomes, something that is usually considered negative in clinical trials. As a consequence, investigators sometimes calculate total scores from different scales that are psychometrically and clinically unrelated, resulting in meaningless outcomes.

Second, the doctor may be puzzled by the psychologic nature of self-assessments in general. QoL is a conglomerate of expectations, perceptions, and comparisons with other people. Whether we are satisfied in a certain situation or not depends, on the one hand, our expectations and, on the other hand, on the assessment of our reality—the level to which those expectations are fulfilled. If we do not expect to live without suffering, we can bear more health problems than others. In other words, if a patient tells us he or she has no problems with the side effects of a certain treatment, this can mean that there are really no side effects or that the patient does not consider them to be a problem worth mentioning to the doctor. Another facet of the psychologic adjustment process is the comparison with other people. If, for example, a patient who has undergone radiation in the head and neck region that resulted in parotid gland dysfunction meets other patients who have received the same treatment and who are worse off than him or her, that patient might praise fate for having fewer side effects, and will probably indicate a better health status compared to a patient who meets only fellow patients who have fewer problems. This phenomenon is called top-down or bottom-up comparison.

Third, it has been shown that patients tend to report only problems they consider relevant to their disease.⁴ For example, a patient with cancer who is asked whether he or she has experienced pain during the past week might consider only the tumor pain to be relevant, but not the migraine he or she suffered the day before. This “selective reporting” can lead to an underestimation of the health problems of patients. Therefore, it makes more sense to compare subgroups of patients rather than patients with general population samples.

The sometimes surprising results of QoL measurement can be understood when all of these psychologic processes are taken together; for example, the so-called “satisfaction paradox,” whereby patients with severe diseases indicate better QoL than healthy subjects. These challenges sometimes lead to confusion and frustration on the part of the investigator; however, that does not mean that QoL measurements are invalid. After consulting QoL experts, many of the above-mentioned challenges can be detected, understood, and addressed.

Another methodologic challenge is the compliance of patients in completing QoL questionnaires. It is often assumed that patients are unwilling to do so. However, only a few authors have reported on completion rates in their studies. We performed a systematic review of all publications based on one of the most frequently used QoL instrument in patients with head and neck cancer, the EORTC QLQ-H&N35, and we found that out of 125 papers, only 23 reported on percentages of missing values. The completeness of the questionnaire varied from 66%⁵ to 99%.⁶ Scales with missing values included Sexuality, Speech, Teeth, and Weight Gain, with average percentages of missing values of 11.5%, 7.0%, 2.7%, and 2.0% respectively. Regarding the Teeth and Sexuality scales,^5-8 some authors reported that it may remain unclear whether a nonanswer was because the patient was unwilling to answer or because the item did not apply to his or her status.

Another interesting point to consider in the analysis of QoL results is the difference between statistical significance and clinical relevance. Simply speaking, statistical significance is a function of the number of patients included. If a study is large, small differences between groups will become statistically significant; if the study is small, even large differences will not become significant. Therefore, instead of only looking at the p value of a finding, it is also important to interpret the size of the effect (be it a difference or an association). Only an effect size that is clinically meaningful should be used for clinical decision making.⁹

There are three ways to answer this question. First, it is possible to compare mean scores of different QoL domains. The scales with the highest scores (or lowest, depending on the direction of scoring) can be considered the most important. This method is not without problems, because scales are often not constructed to be directly comparable with each other.

The second option is to ask patients, “How important is [QoL domain] for you?” Within the European Organisation for Research and Treatment of Cancer QoL Group, we have done this in a group of 137 patients with head and neck cancer who come from eight different countries.¹⁰ They were asked to indicate, from a list of 92 different issues, the 25 they would prefer to include in a questionnaire. The issues most frequently mentioned are displayed in Table 2. It appeared that fear of tumor progression, swallowing, talking/voice quality, swelling in the neck, and sense of taste are the most important QoL domains. Similar results were found by List et al.^10,11

TABLE 2. Quality-of-Life Domains of High Priority to Patients

The third way is to ask the patients to give their opinion of what constitutes a tolerable level of suffering from symptoms that affect their QoL. This method is a bit more sophisticated and more difficult for the patients. An advantage, however, is that it acknowledges the fact that QoL is changed after an oncologic treatment and that this can imply a certain amount of suffering. Thus, the important question is not which QoL domain is most frequently deteriorated but which deterioration is most problematic for the patient. For example, a moderate increase in pain might be more problematic than a large increase in problems with smelling.

We used this method in a group of patients after total laryngectomy.¹³ The results showed that patients are most accepting of sensory impairments, coughing, and dyspnea; constipation, nausea and vomiting, and diarrhea were rated as being the most troublesome symptoms.

The course of QoL during and after treatment depends, in addition to individual factors, on the tumor entity and on the treatment the patient receives (Fig. 1). There is, however, a common pattern that occurs across different groups of patients: QoL decreases during treatment, improves slowly about 6 months after baseline, and improves again after 12 months. Frequently though, it does not achieve pretreatment levels of QoL. In particular, dry mouth and swallowing problems (especially after radiotherapy) as well as speech problems (especially after surgery) are long-lasting issues (S. Singer, PhD, et al, unpublished data, February 2013).^14-18

FIG 1. Quality of life over time in patients with different types of cancer and treatment.

Displayed are the mean scores of the EORTC QLQ-H&N35. Data sources (from left to right): Bjordal et al, 2001 (375 patients, Sweden and Norway)¹²; Fang et al, 2008 (203 patients, Taiwan)¹³; Oates et al, 2007 (14 patients, Australia)¹⁶; Yoshimura et al, 2009 (56 patients, Japan)¹⁴; Schliephake et al, 2002 (53 patients, Germany)¹⁵; Singer et al, 2013 (174 patients, Germany; S. Singer, PhD, unpublished data, February 2013).

Abbreviations: HNCa, heterogeneous group of patients with head and neck cancer; Nasopha, nasopharyngeal cancer; OC, cancer in the oral cavity; LaCa, laryngeal cancer; diff Tx, different treatments; RT, radiotherapy; CRT, chemoradiotherapy; BL, at baseline before treatment; Tx, during treatment; 1/2y, 6 months after baseline; 1y, 1 year after baseline.

When investigating the course of QoL, clinicians should be aware that differential drop-out is a problem in prospective studies.¹⁹ Patients who drop out of the study because of death or health problems differ from those who continue to participate, and this is true for the baseline scores as well. Therefore, if only the QoL of patients who complete all time points is analyzed, the scores are usually over-optimistic (S. Singer, PhD, et al, unpublished data, February 2013).

Knowledge of factors that can influence the course of QoL after head and neck cancer treatment is crucial for clinicians, especially if these factors can be changed. To date, we know that certain features of the treatment (for example, sparing of the parotid gland during radiation²⁰) can improve certain QoL domains (for example, dry mouth and swallowing). There is also good evidence that the course of QoL is highly related to the mental health of the patient (S. Singer, PhD, et al, unpublished data, February 2013).²¹ This calls for professional psychologic support in patients with head and neck cancer, which is frequently insufficient.^22,23

In conclusion, QoL is an important outcome to be considered in clinical trials and in routine treatment. Its assessment can be challenging at times; however, there are reliable and valid tools available and clinicians can use them meaningfully if they know how to apply them and how to interpret their scores.