Abstract
Abstract
Kayser, Bengt, Andrea Aliverti, Riccardo Pellegrino, Raffaele Dellaca, Marco Quaranta, Pasquale Pompilio, Giuseppe Miserocchi, and Annalisa Cogo. Comparison of a visual analogue scale and Lake Louise symptom scores for acute mountain sickness. High Alt. Med. Biol. 11:69–72, 2010.—Assessment of the presence and severity of acute mountain sickness (AMS) is based on subjective reporting of the sensation of symptoms. The Lake Louise symptom scoring system (LLS) uses categorical variables to rate the intensity of AMS-related symptoms (headache, gastrointestinal distress, dizziness, fatigue, sleep quality) on 4-point ordinal scales; the sum of the answers is the LLS self-score (range 0–15). Recent publications indicate a potential for a visual analogue scale (VAS) to quantify AMS. We tested the hypothesis that overall and single-item VAS and LLS scores scale linearly. We asked 14 unacclimatized male subjects [age 41 (14), mean (SD) yr; height 176 (3) cm; weight 75 (9) kg] who spent 2 days at 3647 m and 4 days at 4560 m to fill out LLS questionnaires, with a VAS for each item (i) and a VAS for the overall (o) sensation of AMS, twice a day (n = 172). Even though correlated (r = 0.84), the relationship between LLS(o) and VAS(o) was distorted, showing a threshold effect for LLS(o) scores below 5, with most VAS(o) scores on one side of the identity line. Similar threshold effects were seen for the LLS(i) and VAS(i) scores. These findings indicate nonlinear scaling characteristics that render difficult a direct comparison of studies done with either VAS or LLS alone.
Introduction
Beyond 2000 to 2500 m, newcomers at altitude may develop acute mountain sickness (AMS) characterized by headache and often accompanied by additional symptoms, including light-headedness, dizziness, loss of appetite, nausea, vomiting, fatigue, lassitude, and trouble sleeping and, more seldom, by peripheral edema and ataxia (Hackett and Roach, 2001). Assessment of AMS depends on the subject's subjective appreciation of the severity of symptoms and on sharing this with a knowledgeable person. In the absence of a validated laboratory marker, the assessment of the presence and severity of AMS is therefore based on the quantification of the reported subjective sensation. AMS diagnosis thus remains problematic, precisely because of relying on the subjective assessment of symptom intensity (Roach and Kayser, 2007). A popular questionnaire to quantify AMS is the Lake Louise scoring system (LLS) (Roach et al., 1993). This is a five-item self-administered questionnaire based on the most frequent symptoms of AMS: headache, gastrointestinal problems (anorexia, nausea, or vomiting), insomnia, weakness or fatigue, and dizziness or light-headedness. Every item is scored by the subject on a scale from 0 to 3, with each integer having a specific descriptor. The minimum LLS score is 0, and the maximum score is 15. The score can be complemented by a question on the functional status of the subject (reduction in activity because of symptoms) and, if a knowledgeable person is present, by a standardized clinical assessment (change in mental status, ataxia, and peripheral edema). Recent data indicate a potential for the use of a visual analogue scales (VAS) (Wagner et al., 2007). In this paper the authors compared a single VAS for AMS with the LLS self-score, but scores for the individual questions were not compared to individual VAS scores. Further, no analysis was made of any scaling properties, and whether an assumed linear scaling of a VAS would be mirrored by the linear scaling properties of the LLS items and descriptors was not explored. We therefore set out to test the hypothesis of a linear scaling of VAS versus LLS AMS scores.
Methods
Fourteen healthy, unacclimatized male subjects (age 42 ± 14 yr (mean ± SD), stature 176 ± 3 cm, BMI 25 ± 3) spent 2 days at 3647 m and 4 days at 4560 m. Subjects signed an informed consent form, and the study was approved by the Ethical Committee. Subjects scored AMS with LLS questionnaires. Each single LLS item (i) was accompanied by a 40-mm VAS that was scored in parallel with anchors corresponding to the descriptors of the LLS(i). Subjects were instructed to quantify the intensity of symptoms on each VAS, anchoring 0 as no symptom at all and full-scale as maximum intensity, corresponding to a 3 for all LLS(i). A separate 85-mm VAS was used to score overall (o) sensation of AMS(o) with the anchors “none” and “severe.” Scores were obtained in the morning before breakfast and in the evening before dinner. VAS amplitudes were independently measured by two researchers to the nearest millimeter, and the mean amplitude was used in the analysis. AMS was defined as presence of a headache and a LLS score ≥3 or ≥5. Data were analyzed with SPSS 17 (IBM-SPSS, Chicago, IL, USA). Pearson and Kendall correlation was used to evaluate the relationship between LLS(o) and VAS(o). A p < 0.05 was considered significant.
Results
Mean ± SD, minimum and maximum LLS scores, and prevalence of AMS at the two cutoff scores of 3 and 5 are shown in Table 1. The prevalence of AMS at a criterion score of ≥3 varied with altitude and time of exposure between a high of 9/14 at the beginning to a low of 1/14 at the end of the altitude sojourn. Figure 1 shows the relationship between VAS(o) and the LLS(o). Even though the correlation between VAS(o) and LLS(o) was 0.84 (Pearson or Kendall, identical), the data points are not grouped around the theoretical identity line. For values of LLS <5, VAS score is much less sensitive to symptoms than is LLS, whereas for values >5, the opposite is true. Figure 2 shows the relationship between VAS(i) and LLS(i) for the item headache and here also the data points do not conform to the line of identity.
Relationship between AMS and VAS scores, normalized to 100%, and the LL self-score (scale from 0 to 15). The solid line indicates the theoretical linear relationship expected and the open circles the actual data points. The two standing dashed lines indicate the cutoffs for considering a subject to suffer from AMS at scores ≥3 or ≥5. We see that the relationship between the two scoring systems is distorted. a.u. = arbitrary units.
Relationship between the VAS score for headache compared to the LL score for headache. We can see a distortion from the expected linear relationship between the two variables. a.u. = arbitrary units.
LLS Scores and Prevalence of AMS
Mean ± SD and the minimum and maximum LL scores are given. Prevalence of AMS is expressed in percentage of subjects having a LLS ≥3 and ≥5. On day 3 the subjects woke up at 3647 m and climbed up to 4560 m. a.u., arbitrary units.
Discussion
In general, the use of a VAS assumes linear scaling characteristics for the scored variable, for example, in the assessment of pain or dyspnea, with typical anchors like “none” and “severe” or “maximal” (Myles and Urquhart, 2005; Streiner and Norman, 2007). Parting from the idea that a VAS for the scoring of AMS with the anchors “none” and “severe” would thus possess linear scaling characteristics, any linearity of the scoring of AMS with the LLS should result in clustering of the data around the theoretical identity line when plotting VAS(o) versus LLS(o). In Fig. 2, it can be clearly seen that this was not the case in our study. In fact, there seems to be a threshold effect leading to low VAS(o) scores when LLS(o) is below 5 to 6. The generally accepted cutoff score for the diagnosis of AMS is the presence of headache and a LLS(o) ≥3 (Roach et al., 1993), even though sometimes a cutoff ≥5 is used (Maggiorini et al., 1998). It would follow that the observed threshold effect in the VAS(o) may represent the exclusion from AMS diagnosis of subjects who score lower than 3 or 5 for LLS(o). On the other hand, such threshold effects also apply to the single items of the LLS, as can be seen in Fig. 2 for the obligatory symptom headache. In this case the initial threshold effect operates first in the other direction, with small increments in VAS(i), while LLS(i) remains zero, to subsequent underscoring when headache becomes stronger. Similar nonlinear scaling effects were seen for the other LLS(i) (data not shown). Careful inspection of Figure 1 of Wagner and colleagues (2007) shows, as in our study, that below a LLS score of 5 the majority of data points are skewed to the LLS side of the linear line of identity, whereas beyond that score more points seem to be positioned on the other side. The correlation between LLS and VAS in the Wagner study was 0.65, somewhat lower than that in our study (0.84), indicating, in another way for both studies, the limited agreement between LLS and VAS. Based, on these observations we find that the use of VAS for the scoring of AMS is problematic if used as the sole means to score AMS, thus precluding comparison to studies that used the LLS. However, our study is limited by the number of subjects included and the low prevalence of more severe AMS and should be complemented by future larger-scale studies.
Conclusion
Using VAS to score AMS seems to yield data different from those obtained from the LLS. Quantification of the prevalence and severity of AMS remains based on subjective criteria, and the different questionnaires do not necessarily yield the same results (Dellasanta et al., 2007). To allow comparison between studies and awaiting an objective marker for AMS (Roach and Kayser, 2007), it remains advisable to continue using the more popular questionnaires such as the abridged ESQ-III and the Lake Louise questionnaire, possibly supplemented with a VAS for individual items like headache when the research question concerns, for example, the efficacy of a symptomatic pharmacological treatment for headache (Harris et al., 2003).
Footnotes
Acknowledgments
We are grateful to the Italian Alpine Club and Everest-K2-CNR committee for logistic and financial support.
Disclosures
The authors have no conflicts of interest or financial ties to disclose.