Sensory evaluation
1570 JOURNAL OF FOOD SCIENCE —Vol. 67, Nr. 4, 2002 © 2002 Institute of Food Technologists Sensory and Nutritive Qualities of Food Application of Quality Index Method (QIM) Scheme in Shelf-life Study of Farmed Atlantic Salmon (Salmo salar) K. S VEINSDOTTIR , E. M ARTINSDOTTIR , G. H YLDIG , B. J ØRGENSEN , AND K. K RISTBERGSSON ABSTRACT: Salmon (Salmo salar) was stored in ice up to 24 d, and changes during storage were observed with sensory evaluation using the Quality Index Method (QIM), and Quantitative Descriptive Analysis (QDA), total viable counts (TVC), hydrogen sulfide (H 2 S)-producing bacteria, and instrumental texture measurements (com- pression test). Maximum storage time in ice was determined with QDA and fat content by Soxhlet extraction. A high correlation between QIM and storage time in ice was found. Storage time could be predicted with ± 2 d. TVC increased exponentially with storage and was dominated by H 2 S-producing bacteria after 20 d in ice, which was the maximum storage time. Texture measurements indicated softening of salmon flesh with storage. Keywords: sensory evaluation, quality of salmon, fish freshness, shelf life Introduction F RESHNESS IS ONE OF THE MOST IMPORTANT ASPECTS OF FISH , AND because of consumer preferences, there is a strong tendency to select very fresh fish (Luten and Martinsdottir 1997). Sensory evaluation is the most important method for freshness and qual- ity assessment in the fish sector (Hootman 1992). The world’s production of farmed salmon increased between 1990 and 1997, from 540,000 tons to almost 1,400,000 tons per year (FAO 2000). In 1997, 38% of the salmon produced in the world was Atlantic salmon (Salmo salar). Because of the increased trade between countries, purchases are often performed on unseen lots, and there is a need for a good freshness grading system for salmon, such as the Quality Index Method (QIM). This method is a sea- food freshness quality grading system, which is used to assess fish freshness in a rapid and reliable way. QIM is based upon a scheme originally developed by the Tasmanian Food Research Unit (Bremner 1985). The method has to be adapted to each fish species. To date, the system incorporates fresh herring (Clupea harengus) and cod (Gadus morhua) (Jonsdottir 1992; Larsen and others 1992), red fish (Sebastes mentella/marinus) (Martinsdottir and Arnason 1992), Atlantic mackerel (Scomber scombrus), horse mackerel (Trachurus trachurus) and European sardine (Sardina pilchardus) (Andrade and others 1997), brill (Rhombus laevis), dab (Limanda limanda), haddock (Melanogrammus aeglefinus), pollock (Pollachius virens), sole (Solea vulgaris), turbot (Scophtal- mus maximus) and shrimp (Pandalus borealis) (Luten 2000), gilt- head seabream (Sparus aurata) (Huidobro and others 2001), and farmed salmon (Salmo salar) (Sveinsdottir and others 2001). QIM has several unique advantages, including estimation of past and remaining storage time in ice (Botta 1995; Hyldig and Nielsen 1997; Luten and Martinsdottir 1997). The maximum storage time of fish can be determined by sen- sory evaluation of cooked samples. The Quantitative Descriptive Analysis (QDA) (Stone and Sidel 1985) is a sensory method, which may be used for the determination of maximum shelf life in addition to a detailed description of the sensory profile for a product. With the QDA, all detectable aspects of a product are described and listed by a trained panel. The list is then used to evaluate the product, and the panelists quantify the sensory as- pects of the product using an unstructured scale. The end of shelf life is the result of unpleasant sensory characteristics most- ly due to bacterial growth. The amount of bacteria on newly caught fish can vary greatly, normally ranging from 10 2 to 10 7 cfu/ cm 2 (Liston 1980). The most important seafood spoilage bacteria are characterized by their ability to produce H 2 S and reduce trim- ethylamine oxide (TMAO), which has been used for their specific determination. Capell and others (1997) found counts of H 2 S- producing bacteria closely associated with the rejection of sever- al fish species, irrespective of the temperature and atmosphere. Microbial metabolites have low odor thresholds, and during fish spoilage, the concentrations of sulfur compounds, short-chain acids, alcohols, sulfur compounds, and amines increase (Olafs- dottir and Fleurence 1997). In raw fish, the texture softens during chilled storage (Anders- en and others 1995; Einen and Thomassen 1998) because pro- teolytic enzymes break down the muscle structure (Andersen 1995). The fat content of fish flesh appears to influence the tex- ture. When the fat content is high, the flesh is softer (Andersen and others 1994), and juiciness increases (Einen and Thomassen 1998). The total lipid content of farmed salmon is often up to double the content found in wild salmon (Moe 1990) and has been reported varying from 12% to 19% (Hafsteinsson and oth- ers 1998; Refsgaard and others 1998). The aim of this work was to perform a shelf-life study with farmed Atlantic salmon (Salmo salar) and characterize the chang- es in freshness with the Quality Index Method (QIM) scheme for raw salmon and the Quantitative Descriptive Analysis (QDA) for cooked salmon. Furthermore, the goal was to compare the senso- ry analysis to microbial counts (total viable counts and H 2 S-pro- ducing bacteria) and instrumental texture measurements (com- pression test). Materials and Methods Salmon The salmon was obtained from the fish farm Silungur ehf JFS: Sensory and Nutritive Qualities of Food jfsv67n4p1570-1579ms20010283-SR.p65 5/31/2002, 11:14 AM1570 Vol. 67, Nr. 4, 2002 — JOURNAL OF FOOD SCIENCE 1571 Sensory and Nutritive Qualities of Food Shelf life of salmon with QIM . . . (Vogar, Iceland). The salmon had been fed various types of the feed blend “Gull” (Gull 3, 4, 6, 8, 10, 12, depending on the age of the salmon) from Fodurblandan hf (Reykjavik, Iceland). The blend contained 40% protein, 16% carbohydrates, and 25% to 30% fat. The salmon were starved for 2 wk and then slaughtered with carbonic acid. After slaughtering, the salmon were gutted, bled, gills cut through, and the salmon were then rinsed in run- ning water for 30 min, followed by chilling to 0 °C in slush ice (0 to –1 °C) before icing in boxes. The salmon weighed 3 to 4 kg. The fish were slaughtered before sexual maturity in 8 batches (Octo- ber/November 1999) and stored up to 24 d at 0 to 2 °C in iced boxes until analyzed. A total of 50 salmon were used in the exper- iment. Eleven were used for training the sensory panel. Salmon stored 1, 2, 4, 8, 11, 13, 15, 17, 19, 20, 21, 22, and 24 d in ice were analyzed during the shelf-life study. Three salmon from each storage day were analyzed with QIM; thereof 2 were used for QDA, microbial counts, texture measurement, and fat analysis (Figure 1). Sensory evaluation Quality Index Method The QIM scheme for salmon lists quality attributes for ap- pearance/texture, eyes, gills, and abdomen and descriptions of how they change with storage time. Scores were given for each quality attribute according to the descriptions, ranging from 0 to 3. Very fresh fish normally received the score 0, with scores in- creasing with storage time. The scores given for all the quality at- tributes are summarized by the Quality Index, which increases linearly with storage time in ice. The sensory evaluation of each attribute was conducted according to Martinsdottir and others (2001). Prior to the shelf-life study, the QIM scheme for farmed salm- on (Sveinsdottir and others 2001) was revised, as it did not in- clude a parameter for the textural state of rigor mortis. Addition- ally, 1 score was added for color/appearance of the skin. Changes were made in the setup of the scheme and selection of words to describe the parameters in the scheme, mainly to make each de- scription more precise and to facilitate the QIM assessment. Twelve trained panelists of the Icelandic Fisheries Laborato- ries sensory panel participated in the sensory evaluation with QIM. Members had several years of experience in evaluating fish freshness. Prior to the shelf-life study, the panel was trained in applying the QIM scheme in 2 sessions. The scheme was ex- plained to the panel while observing salmon of different fresh- ness categories. The panel used the scheme to assess 6 to 9 salm- on from 2 to 3 different storage days per session during the shelf-life study. The salmon was placed on a clean table 30 min before the evaluation. The side where the gills had been cut through was facing down. Each salmon was coded with 3 random digit numbers. All observations of the salmon were conducted under standardized conditions, with as little interruption as pos- sible, at room temperature, and under white fluorescent light. Quantitative Descriptive Analysis The QDA, introduced by Stone and Sidel (1985), was used to assess cooked samples of salmon. An unstructured scale (0 to 100%) was used on a list of words describing odor, flavor, appear- ance, and texture. Twelve panelists of the Icelandic Fisheries Laboratories´ sen- sory panel participated in the QDA of the cooked salmon. They were all trained according to international standards (ISO 1993), including detection and recognition of tastes and odors, training in the use of scales, and in the development and use of descrip- tors. The members of the panel were familiar with the QDA method and experienced in sensory analysis of salmon. Two ses- sions were used for training of the panel using salmon of differ- ent freshness categories. Sensory evaluation of the cooked salm- on was performed parallel to the QIM assessment. Each panelist evaluated duplicates of samples from 2 to 3 different storage days. The fish was served in a random order during 2 sessions for each day of the sensory evaluation. All sample observations were conducted according to interna- tional standards (ISO 1988). Twelve samples collected from each salmon with skin came from the loin part, ranging from the spine to 2 cm below the lateral line. The samples were coded with 3 ran- dom digit numbers and cooked at 95 to 100 °C for 7 min in a pre- warmed oven (Convotherm Elektrogeräte GmbH, Eglfing, Ger- many) with air circulation and steam and then served to the panel. Microbial counts Skin samples were collected before all other analysis by cut- ting 2 x 7.5 cm 2 skin strips from 1 side of the fish and placed in a Stomacher containing 50 mL Butterfield´s Buffer solution (APHA 1992). Blending was done in a Stomacher 400 for 1 min. Flesh samples were collected after QIM evaluation from the other side of the salmon. The skin was washed with alcohol and removed with a sterilized scalpel. The flesh under the skin was collected, and after mincing, 25 g were weighed into a stomacher bag con- taining 225 g Butterfield´s Buffer solution to obtain a 10-fold di- lution. Blending was done in a Stomacher for 1 min. Further 10- fold dilutions were made as needed. Total viable counts (TVC) and selective counts of H 2 S-producing bacteria were done on iron agar (IA) by the pour plate technique with an overlay as de- scribed by Gram and others (1987). The plates were incubated at 22 °C for 3 d. Bacteria forming black colonies on this agar produce H 2 S from sodium thiosulfate and/or cysteine. Instrumental texture measurements One sample from each fish was measured in a Texture Analyz- er (TA.XT2; Stable Micro System, Surrey, England) using a com- pression test. The salmon was filleted, skin removed, and sam- ples collected transversely behind the dorsal fin. Samples were cut right above the lateral line, 2.5 cm in length and width, and 2.2 ± 1.4 cm in height. The samples were then covered with plas- tic and stored in a refrigerator at 4 to 5 °C until measured (within 3 to 5 h) using an aluminum compression plate (SMSP/100). Samples were compressed to 80% of the sample height at a con- stant speed (0,8 mm/s) with a 100 g constant force. The trigger force was set at 5 g and the registration rate to 200 PPS (registra- tions per s). Fat content Samples were collected according to a method recommended by the Norwegian General Standardizing Body or (1994), the Norwegian Quality Cut (NQC). The samples were vacuum packed and stored at –20 °C until analyzed (within 10 d). The fat content was determined with the Soxhlet method (AOAC 1990) with modification described in the IFL´s method manual for chemical analysis (IFL 1999) using the solvent petroleum ether. Data analysis The QI was treated with analysis of variance (ANOVA, 2-factor without replication) to analyze if a difference existed within a group (QI given for each salmon within a storage day and QI giv- jfsv67n4p1570-1579ms20010283-SR.p65 5/31/2002, 11:14 AM1571 Sensory and Nutritive Qualities of Food 1572 JOURNAL OF FOOD SCIENCE —Vol. 67, Nr. 4, 2002 Shelf life of salmon with QIM . . . en by each judge assessing salmon within a storage day). The equation of best fit and the correlation coefficients (r) of QI, total viable count, H 2 S-producing microbes on salmon skin and flesh, and instrumental texture parameters against storage time in ice were calculated using Microsoft ® Excel 2000 (Microsoft Corpora- tion, Redmond, Wash., U.S.A.). Data from QDA was treated in HyperSense © (Version 1.6; 1996 Icelandic Fisheries Laboratories, Reykjavik, Iceland). Inter- action of panelists and samples was assumed, and statistical analysis was performed using 2-factor design with interaction in the analysis of variance (ANOVA). The program calculates multi- ple comparison using Tukey´s test. Multivariate comparison of different attributes in QIM and QDA was conducted in the statis- tical program Unscrambler ® (Version 6.1; CAMO, Trondheim, Norway) with principal component analysis (PCA). Predictability of QI was analyzed using partial least square regression (PLS) with full cross validation. The average QI for each storage day, in- cluding assessment of 3 salmon, was used for this analysis. The root mean square error of prediction (RMSEP) was calculated for the model (the prediction error in original units). Bias is the aver- aged difference between predicted and measured Y-values for all samples in the validation set. The standard error of performance (SEP) is the precision of results corrected for the bias. From a PLS2 model, the initial variance (signal) at zero PCs and the re- siduals variance (noise) after optimal PCs were plotted as a sig- nal to noise (S/N) ratio for each panelist and for each word (Mar- tens and Martens 2000). The significance level was set at 5%, if Table 1—The QIM scheme for farmed salmon. Revised from Sveinsdottir and others (2001) Quality parameters Description Points Skin: Color/appearance Pearl-shiny all over the skin 0 The skin is less pearl-shiny 1 The fish is yellowish, mainly near the abdomen 2 Mucus Clear , not clotted 0 Milky, clotted 1 Yellow and clotted 2 Odor Fresh seaweed, neutral 0 Cucumber, metal, hay 1 Sour, dish cloth 2 Rotten 3 Texture In Rigor 0 Finger mark disappears rapidly 1 Finger leaves mark over 3 s 2 Eyes: Pupils Clear and black, metal shiny 0 Dark gray 1 Mat, gray 2 Form Convex 0 Flat 1 Sunken 2 Gills*: Color/appearance Red/dark brown 0 Pale red, pink/light brown 1 Grey-brown, brown, gray, green 2 Mucus Transparent 0 Milky, clotted 1 Brown, clotted 2 Odor Fresh, seaweed 0 Metal, cucumber 1 Sour, moldy 2 Rotten 3 Abdomen: Blood in abdomen Blood red/not present 0 Blood more brown, yellowish 1 Odor Neutral 0 Cucumber, melon 1 Sour, fermenting 2 Rotten/rotten cabbage 3 Maximum sum (Quality Index): 24 * Examine the side where the gills have not been cut through Figure 1—Sampling plan for measurements in the shelf- life study of salmon at the Icelandic Fisheries Laborato- ries in November 1999 jfsv67n4p1570-1579ms20010283-SR.p65 5/31/2002, 11:14 AM1572 Vol. 67, Nr. 4, 2002 — JOURNAL OF FOOD SCIENCE 1573 Sensory and Nutritive Qualities of Food Shelf life of salmon with QIM . . . not stated otherwise. Results and Discussion Sensory analysis Quality Index Method The sum of scores evaluated according to the QIM scheme (Table 1) was presented as the Quality Index (QI). The QI was cal- culated for each storage day of sampling and formed a linear re- lationship with time (Figure 2). High correlation between the average QI and days in ice was obtained with a slope of 0.692. The slope was different from the slope observed by Sveinsdottir and others (2001) using the QIM scheme for salmon, presumably because of the revision of the scheme prior to this shelf-life study, including the addition of 2 score attributes. The aim when developing QIM scheme for fish is to have the regression line begin at the origin (0,0), which was not reached here, since the intercept was at 1.568. If the line was forced through the origin, the correlation between the average QI and days in ice became lower (R 2 = 0.933). The QIM scheme gave the assessors the opportunity to choose between scores ranging from 0 to 3 but never a negative number, therefore, allowing for residuals above zero but not below. The difference between salmon of the same storage time in ice was in some cases significant. The results were analyzed with partial least square regression (PLS) to examine how well the QI could predict the storage time in ice (Figure 3). The standard error of performance (SEP) value for the QI was 2.0 (Figure 3). The SEP may be used to evaluate the precision of the predictability of the QI. Since the QI was the sum of 11 at- tributes evaluated in the QIM scheme, a normal distribution could be assumed (O´Mahony 1986). Esbensen and others (1998) stated that 2*SEP could be regarded as a 95% confidence interval assuming normal distribution. Therefore, it can be as- sumed that the QI of a batch (if 3 salmon were assessed) could be used to predict the storage time with ± 2.0 d. It could be assumed that including more salmon in the assessment of each batch might reduce this interval, as observed by Sveinsdottir and oth- ers (2001), where including 5 salmon per storage day gave a SEP value of 1.4. There was a variation in the QI obtained by different panel- ists (Figure 4). The variation increased with storage time, indicat- ing that the panelists were in better agreement when analyzing very fresh salmon with the QIM scheme at the beginning of stor- age compared to the not-so-fresh salmon at later stages. There was a tendency for some of the panelists to score either higher or lower than the average score obtained throughout the storage time. The variation between the panelists in this study, which were trained during 2 1-h sessions, was comparable to the varia- tion between panelists trained during 6 1-h sessions (Sveinsdot- tir and others 2001) in a similar study. This indicated that the 2 Figure 2—Quality Index of salmon. Averages over each day of storage analyzed against days in ice. Figure 3—PLS1 modeling of QIM data from salmon stored in ice using full cross validation: Measured against pre- dicted Y values. Average QI for each storage day based on assessment of 3 salmon used to predict storage time in d. Figure 4—Average QI of salmon with storage in ice, as given by each panelist (1 through 12) jfsv67n4p1570-1579ms20010283-SR.p65 5/31/2002, 11:14 AM1573 Sensory and Nutritive Qualities of Food 1574 JOURNAL OF FOOD SCIENCE —Vol. 67, Nr. 4, 2002 Shelf life of salmon with QIM . . . sessions were sufficient training for the panel. QIM assumes the scores for all quality attributes increase with storage time in ice (Figure 5). The average texture score was determined by pressing a fin- ger on the spine muscle and observing how the flesh recovered according to Martinsdottir and others (2001). The scores were around 0 at storage day 1, as the salmon was in rigor. Propagation of rigor caused the muscle to relax again, and through storage in ice, the flesh became soft due to autolysis influenced by both fish muscle enzymes and microbial enzymes (Gill 1995; Nielsen 1995). The skin became softer or less springy after 17 to 20 d, where the average score increased from 1 to 1.5. The average score of skin odor reached only 2 at the end of the storage time. The score 3 (rotten) was used rarely by panelists. At the begin- Figure 5—Average scores of each quality attribute assessed with QIM scheme for salmon stored in ice against days in ice jfsv67n4p1570-1579ms20010283-SR.p65 5/31/2002, 11:14 AM1574 Vol. 67, Nr. 4, 2002 — JOURNAL OF FOOD SCIENCE 1575 Sensory and Nutritive Qualities of Food Shelf life of salmon with QIM . . . ning of the storage time when the salmon was very fresh, the odor was described as fresh seaweed or neutral, then a cucum- ber-like odor dominated the salmon skin odor. During later stag- es, the odor was described as sour and finally as rotten. Freshly caught fish generally contains low levels of volatile compounds like 2,6-nonadienal, which has a very characteristic cucumber aroma and a low odor threshold (0,001 ppb), which contribute to fresh-like odors. Short-chain acids, alcohols, amines, and sulfur compounds from microbial activity probably caused the sour and rotten odor (Olafsdottir and Fleurence 1997). The average scores for other quality attributes like skin and all the attributes for eyes and gills increased throughout the storage time. The av- erage scores for quality attributes of abdomen were very low un- til after 8 d of storage when they began to increase with the stor- age time. Quantitative Descriptive Analysis The positive attributes for flavor of salmon were described as characteristic salmon, metallic, sweet, and oily flavor on a scale ranging from 0 to 100%. Average scores for most positive flavor at- tributes did not change for the 1st 17 to 19 d of storage but de- creased thereafter (Figure 6). The average scores of sweet and metallic were between 20 and 50 through the storage time, but for salmon stored 21 d, it went below 20. For characteristic salmon flavor and oily flavor, the difference was clearer. The scores were between 50 and 70 until 22 d, when they dropped below 40. Scores for the negative attributes, sour, rancid, and musty/ earthy (Figure 6) were low, approximately 0 to 20 for the 1st 17 to 20 d of storage. Thereafter, the scores increased, especially sour flavor scores, which were around 50 after 22 d. Rancid and musty/earthy flavor reached only 30 after 22 d. The feed of farmed salmon often contains carotenoids (Moe 1990), which have been considered to play an important role in protecting lip- id tissues from oxidation (Burton and Ingold 1984). This may have been the reason for the low rancidity scores. The increasing rancid flavor observed during the last storage day might corre- spond to a train-oily flavor reported by Milo and Grosch (1996), who found their nonfresh cooked salmon samples to be fatty and train-oily smelling. According to their findings, the rancid flavor in salmon was caused by formation of volatile oxidation products such as aldehydes and ketones. They analyzed various odorants in salmon of different freshness (stored 26 wk at -60 °C (fresh) and -13 °C (not fresh)). They found propionaldehyde and (Z)- 1,5-octadien-3-one as the most potent high-volatile odorants in cooked fresh salmon samples. The odor of those compounds was described as sweet and metallic, respectively. Odor is a part of the overall flavor, and those compounds may therefore have been responsible for the sweet and metallic flavor of the cooked salmon in this study. A mixture of odorants in the cooked salmon might be responsible for the characteristic salmon flavor. Milo and Grosch (1996) detected various odorants from cooked salm- on (fresh), and the characteristic salmon odor was caused by compounds like propionaldehyde and acetaldehyde (sweet), hexanal and (Z)-3-hexenal (green), methional (boiled potato- like), dimethyl trisulfide (cabbage-like), and 1-octen-3-one (mushroom-like). The oily flavor and odor might have been due to (Z,Z)-3,6-nonadienal as it was described as fatty and green. Difference for most QDA attributes was generally only ob- served after 20 d in ice (Table 2). Data from day 24 was kept out Figure 6—Changes in flavor attributes of cooked salmon (average scores) against storage of the raw salmon in ice observed by a trained QDA panel Figure 7—Loadings in PCA of salmon data including all quality parameters assessed in QDA of cooked salmon and storage time in ice. f = flavor, o = odor jfsv67n4p1570-1579ms20010283-SR.p65 5/31/2002, 11:14 AM1575 Sensory and Nutritive Qualities of Food 1576 JOURNAL OF FOOD SCIENCE —Vol. 67, Nr. 4, 2002 Shelf life of salmon with QIM . . . of the analysis because the salmon had utterly exceeded the lim- its of acceptance. When the salmon had been stored for 21 d in ice, a part of the panel refused to taste the samples after smell- ing the salmon. This strongly indicated that after 20 d of storage in ice, salmon was—according to sensory evaluation—no longer fit for human consumption. This was in agreement with previous studies. Sveinsdottir and others (2001) concluded that 20 to 21 d was the maximum storage time in ice for salmon. Magnussen and others (1996) observed sensory changes in cooked salmon stored 7, 14, and 21 d and found minor differences between 7 and 14 d, but the overall quality was greatly reduced after 21 d of storage. Lande and Rørå (1999) analyzed the flavor, odor, and overall ef- fects in cooked salmon. Minor changes were observed with stor- age time up to 18 d in ice, however, they did not continue the sensory evaluation of cooked salmon after the 18 d. Differences were observed among panelists for each QDA at- tribute. This is a well-known phenomenon in sensory evaluation. The main types of differences among assessors may be caused by confusion about attributes, individual differences in sensitiv- ity to certain sensations, individual differences in the use of the scale, or individual differences in precision (Næs and others 1994). Various ways have been discussed to detect and handle such differences among assessors (Næs 1990; Næs and Solheim 1991; Næs and others 1994). The noise to signal ratio may be ob- served to decide how to treat the difference among assessors (Sveinsdottir and others 2001). When the results were analyzed with PCA, the variable d in ice contributed to PC1, and a clear grouping was found between positive and negative sensory attributes on each side of the PC1- axis (Figure 7). Negative parameters became more evident in Figure 8—Total viable count and H 2 S-producing microbes on skin and in flesh of salmon stored in ice Figure 9—Correlation between bacterial counts on skin and Quality Index of salmon stored in ice jfsv67n4p1570-1579ms20010283-SR.p65 5/31/2002, 11:14 AM1576 Vol. 67, Nr. 4, 2002 — JOURNAL OF FOOD SCIENCE 1577 Sensory and Nutritive Qualities of Food salmon stored longer in ice as they were grouped with the param- eter d in ice, while the positive parameters became less evident. The negative attributes described salmon at the end of the stor- age time similarly. Discoloration appeared to become slightly more evident with storage, but the texture parameters evaluated in cooked salmon contribute very little to PC1 and therefore do not appear to change with storage time, contrary to the texture of raw salmon. Microbial counts The total viable counts (TVC) on skin and in flesh increased exponentially with storage time (Figure 8). A similar pattern was noted for bacterial counts on skin and QI with storage time, as salmon of low bacterial counts also received low scores in QIM. A high correlation was established between QI and TVC on skin, and the same was seen for H 2 S-producing bacteria, which in- creased proportionally to the TVC at the later stages of storage (Figure 9). At the beginning of storage, the TVC on skin was approxi- mately 10 3 cfu/cm 2 , which is not unusual for newly caught fish (Liston 1980). Very few H 2 S-producing microbes were a part of the initial microflora (< 10 cfu/cm 2 ), but their proportion of the TVC increased with storage time. The TVC (mainly H 2 S-producing mi- crobes) on salmon skin was 10 8 cfu/cm 2 after about 20 d of stor- age. The bacterial counts in salmon flesh were lower than those on the skin. Newly slaughtered salmon contained TVC around 10 cfu/g in flesh. The flesh of healthy live or newly caught fish is sterile because the immune system of the fish prevents the bac- teria from growing in the flesh, but when the fish dies, the im- mune system collapses, and during storage, bacteria invade the flesh (Gram 1995). After 20 d of storage in ice, the TVC was 10 5 cfu/g. As for the microbial growth on salmon skin, the H 2 S-pro- ducing bacteria dominated the bacterial flora at the end of stor- age. Counts of H 2 S-producing bacteria were very low (below 10 Figure 10—Typical compression curve of the instrumental texture measurements of a salmon sample in the shelf- life study, measured in the TA.XT2 Texture Analyzer (SMS). Hardness = H1, Resilience = A2/A1, where H1 equals the maximum force, A1 equals the area under the curve from beginning of measurement until maximum force is reached, and A2 equals the area under the curve from maximum force until the force has reached zero again. Table 2—Sensory scores of attributes of cooked salmon assessed by QDA Color, Characteristic Seaweed/ Characteristic D discoloration salmon seaside Sour Oil Rancid salmon in ice (-) odor (+) odor (+) odor (-) odor (+) odor (-) flavor (+) 1 22 22 22 21,22 2 22 22 21,22 22 20,21,22 4 13,22 22 22 22 21,22 8 22 22 22 22 21,22 11 21,22 21,22 22 21,22 13 4 21,22 21,22 15 21,22 22 21,22 17 21,22 21,22 22 22 21,22 19 22 21,22 22 21,22 20 21,22 22 2,22 21 22 22 21 D Metallic Sweet Sour Musty/earth Oil Rancid Dry/ Tough/ in ice flavor (+) flavor (+) flavor (+) flavor (-) flavor (+) flavor (-) Juicy Tender 1 22 22 22 22 20,22 21 2 21,22 15,22 21,22 22 20,22 19,21 21 4 22 22 22 20,22 22 22 21 8 22 21,22 22 22 22 21 11 22 15,22 22 22 22 21 21 13 21,22 22 21,22 15 22 2,11 21,22 22 22 22 21 17 22 22 22 22 22 22 21 21 19 22 22 22 22 2 20 22 22 4 22 2 21 21 22 Statistical analysis of QDA scores of cooked salmon using 2-factor design with interaction ANOVA and Tukey’s test for multiple comparison showing the storage day when difference is significant ((+) indicate positive attributes and (-) negative attributes.) Shelf life of salmon with QIM . . . jfsv67n4p1570-1579ms20010283-SR.p65 5/31/2002, 11:14 AM1577 Sensory and Nutritive Qualities of Food 1578 JOURNAL OF FOOD SCIENCE —Vol. 67, Nr. 4, 2002 cfu/g) until after 8 d in ice. Similar results were observed and re- ported by Lande and Rørå (1999). The TVC in salmon flesh at the rejection limits observed in this study were considerably lower than what has been found previously (Olafsdottir and others 1997). This could be caused by the high counts of H 2 S-producing bacteria at the end of the storage time, since those organisms were probably primarily responsible for spoilage (Capell and others 1997). This supports the rejection of the cooked salmon samples after 20 d of storage, when the TVC were dominated by H 2 S-producing bacteria. Instrumental texture measurements The instrumental hardness (Figure 10) of salmon samples de- creased with storage time (Figure 11), indicating softening of the salmon flesh. Similar results were observed by Andersen and others (1995) and Einen and Thomassen (1998). There was no correlation between texture parameters evalu- ated in cooked salmon and instrumental texture parameters for raw salmon (Table 3). This was not unexpected for juiciness, as none of the measured texture parameter simulates juiciness. However, tough/tender might have been related to instrumental factors such as the attributes hardness and resilience, expressing how stretchable the samples were. The texture evaluated in QIM (stiffness), on the other hand, was correlated to instrumental tex- ture parameters. Salmon with firm texture according to instru- mental texture measurements was assessed firm in QIM. Fat content The average fat content of the salmon was 15.1 ± 2.1% (95% confidence interval) and ranged from approximately 10% to 19%. This was comparable to previously reported fat content of farmed salmon (Hafsteinsson and others 1998; Refsgaard and others 1998). Tenderness, rancid odor, and flavor increased with increased fat content of the salmon (Table 4). Tenderness has previously been reported to increase with increased fat content in salmon (Andersen and others 1994). Juiciness did not correlate with fat content. Conclusions T HE SCORES FOR QUALITY ATTRIBUTES INCLUDED IN THE QIM scheme increased differently with storage time in ice, but a linear relationship with high correlation was found between QI and storage time in ice. Individual salmon spoil at different rates. A minimum of 3 salmon should be included in the assess- ment of each batch of salmon. The storage time of the salmon may be predicted within ± 2.0 d at the 95% significance level, but examining a greater number of salmon per batch might increase the precision. Based upon the sensory evaluation of cooked salmon, the maximum storage life of salmon has been deter- mined as 20 d in ice. The quality of the cooked salmon did not change much through ice storage until 17 to 20 d. Then the scores for positive attributes decreased, while the scores for neg- ative attributes increased. Differences among panelists were evi- dent for all evaluated attributes in QDA and the QI. The high cor- relation between QI and storage time in ice made it possible to predict the past storage time in ice. As the maximum storage time of salmon in ice was determined as 20 d, this information may be utilized directly for assessment with the QIM for farmed salmon to predict remaining storage time in ice assuming optimum stor- age conditions and used in production and quality manage- ment. References Andersen UB. 1995. Measurements of texture quality in farmed Atlantic salmon (Salmo salar) and rainbow trout (Oncorhynchus mykiss) Doctor Scientiarum Thesis. 1995:26, Agricultural University of Norway. Ås, Norway ISSN 0802- 3220, ISBN 82-575-0265-0. P 1-37. Andersen UB, Strømsnes AN, Steinsholt K, Thomassen MS. 1994. Fillet gaping in farmed Atlantic salmon (Salmo Salar). Norweg J Agric Sci 8:165-179. Andersen UB, Thomassen MS, Rørå AMB. 1995. Texture properties of farmed Atlantic salmon (Salmo salar). Influence of storage time on ice and smelt age. In: Andersen UB. Measurements of texture quality in farmed Atlantic salmon (Salmo Salar) and rainbow trout (Oncorhynchus mykiss) (III) Doctor Scien- tiarum Thesis. 1995:26, Agricultural University of Norway. Ås, Norway ISSN 0802-3220, ISBN 82-575-0265-01-26. P 1-26. Andrade A, Nunes ML, Batista I. 1997. Freshness quality grading of small pelagic species by sensory analysis. In: Olafsdottir G, and others, editors. Methods to determine the freshness of fish. Proceedings of the final meeting of the con- certed action “evaluation of fish freshness” AIR3CT94 2283. Nantes Nov 12- 14, 1997. Int Inst Refrig: 333-338. AOAC 1990. Official methods of analysis. 15 th ed. Washington, D.C.: Association of Official Analytical Chemists. P 770-771. APHA 1992. Compendium of methods for the microbiological examination of foods. 3 rd ed. Washington D.C. American Public Health Association. P 46-48. Table 4—Correlation (r) between measured fat and some sensory attributes (QDA method) of salmon stored in ice Sensory attribute r Oil odor -0.381* Rancid odor 0.444* Oil flavor -0.222 Rancid flavor 0.537* Dry/Juicy 0.223 Tough/Tender 0.379* * Comparisons of significance according to O´Mahony (1986). Significance ( p < 0.05) Table 3—Correlation (r) between evaluated and measured texture parameters of salmon stored in ice Texture parameters QIM Dry/Juicy Tough/Tender Hardness -0.566* 0.167 -0.110 Resilience -0.398* 0.049 -0.032 * Comparisons of significance according to O´Mahony (1986). Significance ( p < 0.05) Shelf life of salmon with QIM . . . Figure 11—Correlation between hardness of salmon mea- sured in TA.XT2 Texture Analyzer and storage time in ice. Correlation is significance according to O´Mahony (1986). jfsv67n4p1570-1579ms20010283-SR.p65 5/31/2002, 11:14 AM1578 Vol. 67, Nr. 4, 2002 — JOURNAL OF FOOD SCIENCE 1579 Sensory and Nutritive Qualities of Food Botta JR. 1995. Evaluation of seafood freshness quality. New York: VCH Publish- ers, Inc. 180 p. Bremner A. 1985. A convenient easy to use system for estimating the quality of chilled seafood. In: Scott DN, Summers C, editors. Proceedings of the fish pro- cessing conference, Nelson, New Zealand, 23-25 April 1985. Fish Proc Bull 7:59-73. Capell C, Vaz-Pires P, Kirby R. 1997. Use of counts of hydrogen sulphide producing bacteria to estimate remaining shelf-life of fresh fish. In: Olafsdottir G and others, editors. Methods to determine the freshness of fish. Proceedings of the final meeting of the concerted action “evaluation of fish freshness” AIR3CT94 2283. Nantes Nov 12-14, 1997. Int Inst Refrig 175-182. Einen O, Thomassen MS. 1998. Starvation prior to slaughter in Atlantic salmon (Salmo salar); II. White muscle composition and evaluation of freshness, tex- ture and color characteristics in raw and cooked fillets. Aquaculture 169:37- 53. Esbensen K, Schönkopf S, Midtgaard T, Guyot D. 1998. Multivariate analysis in practice. CAMO ASA, Oslo, Norway/CAMO ASA, Norway. 343 p. [FAO] Food and Agriculture Organization of the United Nations. 2000. Salmon— analysis 1999. Globefish salmon analysis 1999. Rome: FAO. Available from: <http://www.fao.org/globfish/globefi/doc/commodit/salmon/myhtml.htm>. Accessed April 10. Gill T. 1995. Autolytic changes. In: Huss HH, editor. Quality and quality changes in fresh fish. Rome: FAO Tech Paper No. 348. p 39-51. Gram L. 1995. Bacteriological changes. In: Huss HH, editor. Quality and quality changes in fresh fish. Rome: FAO Tech Paper No. 348. p 51-63. Gram L, Trolle G, Huss HH. 1987. Detection of specific spoilage bacteria from fish stored at low (0 °C) and high (20 °C) temperatures. Int J Food Microbiol 4:65-72. Hafsteinsson H, Sigurgisladottir S, Torrissen OJ, Lie Ø, Thomassen M. 1998. Salmon quality, final report. Reykjavik: Technological Institute of Iceland. NordFood P93146, ITI 9828/MDT11. p 26. Hootman RC. 1992. Manual on descriptive analysis testing for sensory evalua- tion. Philadelphia: ASTM. p 52. Huidobro A, Pastor A, Tejada M. 2001. Quality index method developed for raw gilthead seabream (Sparus aurata). J Food Sci 65(7):1202-1205. Hyldig G, Nielsen J. 1997. A rapid sensory method for quality management. In: Olafsdottir G and others, editors. Methods to determine the freshness of fish. Proceedings of the final meeting of the concerted action “evaluation of fish freshness” AIR3CT94 2283. Nantes Nov 12-14, 1997. Int Inst Refrig: 297-305. IFL. 1999. Icelandic Fisheries Laboratories method manual for chemical analy- sis. Determination of fat in fish and feed (Gaedahandbok efnafraedistofu— Adferdabok. Akvordun fitu i fiski og fodri). Reykjavik. P 1-2. ISO. 1988. Sensory analysis—general guidance for the design of test rooms, 8589. Genf, Switzerland: The International Organization for Standardiza- tion. P 1-9. ISO 1993. Sensory analysis—general guidance for the selection, training and monitoring of assessors. Part 1: Selected assessors, 8586-1. Genf, Switzerland: The International Organization for Standardization. P 1-10. Jonsdottir S. 1992. Quality index method and TQM system. In: Olafsson R, Ingth- orsson AH, editors. Quality issues in the fish industry. Reykjavik: The Research Liaison Office, Univ. of Iceland. P 81-94. Lande A, Rørå MB. 1999. Holdbarhet av iset laks. Norsk Fiskeoppdrett nr. 5a, March 1999:6-7. Larsen E, Heldbo J, Jespersen CM, Nielsen J. 1992. Development of a method for quality assessment of fish for human consumption based on sensory evalua- tion. In: Huss HH, Jakobsen M, Liston J, editors. Quality assurance in the fish industry. Amsterdam, The Netherlands: Elsvier Science Publishers. P 351-358. Liston J. 1980. Microbiology in fishery science. In: Connell JJ, editor. Advances in fishery science and technology. Farnham, England: Fishing News Books Ltd. P 138-157. Luten JB. 2000. Development and implementation of a computerised sensory system (QIM) for evaluating fish freshness. CRAFT FAIR CT97 9063. Final re- port for the period from 01-01-98 to 31-03-00. Rivo, The Netherlands Institute for Fisheries Research, Wageningen, The Netherlands. Luten JB, Martinsdottir E. 1997. QIM—a European tool for fish freshness evalu- ation in the fishery chain. In: Olafsdottir G and others, editors. Methods to determine the freshness of fish. Proceedings of the final meeting of the con- certed action “evaluation of fish freshness” AIR3CT94 2283. Nantes Nov 12- 14, 1997. Int Inst Refrig: 287-296. Magnussen OM, Johansen S, Nordtvedt TS. 1996. Effect of icing and storage temperatures on salmon quality. Refrigeration and aquaculture. Refrigera- tion Science and Technology Proceedings, 20-22 March 1996, Bordeaux, France. Biotech 247-254. Martinsdottir E, Arnason A. 1992. Redfish. In: Nordic industrial fund, quality standards for fish: final report phase II. Reykjavik: Icelandic Fisheries Labora- tories. P 21-35. Martinsdottir E, Sveinsdottir K, Luten JB, Schelvis-Smit R, Hyldig G. 2001. Ref- erence manual for the fish sector: sensory evaluation of fish freshness. QIM Eurofish. P.O.Box 68, 1970 AB IJmuiden, The Netherlands. Milo C, Grosch W. 1996. Changes in the odorants of boiled salmon and cod as affected by the storage of the raw material. J Agric Food Chem 44:2366-2371. Moe NH. 1990. Key factors in marketing farmed salmon. Proc Nutr Soc New Zeal 15:16-22. Nielsen J. 1995. Sensory changes. In: Huss HH, editor. Quality and quality chang- es in fresh fish. Rome: FAO Tech Paper No. 348. p 35-39. Norwegian General Standardizing Body 1994. Atlantisk laks. Måling av farge og fett. NS 9402. Næs T. 1990. Handling individual differences between assessors in sensory QDA. Food Qual Pref 2:187-199. Næs T, Solheim R. 1991. Detection and interpretation of variation within and between assessors in sensory QDA. J Sens Stud 6:159-177. Næs T, Hirst D, Baardseth P. 1994. Using cumulative ranks to detect individual differences in sensory QDA. J Sens Stud 9(1):87-99. Olafsdottir G, Fleurence J. 1997. Evaluation of fish freshness using volatile com- pounds—classification of volatile compounds in fish. In: Olafsdottir G and others, editors. Methods to determine the freshness of fish. Proceedings of the final meeting of the concerted action “evaluation of fish freshness” AIR3CT94 2283. Nantes Nov 12-14, 1997. Int Ins Refrig: 55-69. Olafsdottir G, Martinsdottir E, Oehlenschlager J, Dalgaard P, Jensen B, Under- land I, Mackie IM, Helehan G, Nielsen J, Nilsen H. 1997. Methods to evaluate fish freshness in research and industry. Tr Food Sci Tech 8(8):258-265. O´Mahony M. 1986. Sensory evaluation of food. Statistical methods and proce- dures. New York: Marcel Dekker. 487 p. Refsgaard HHF, Brockhoff PM, Jensen B. 1998. Biological variation of lipid con- stituents and distribution of tocopherols and astaxanthin in farmed Atlantic salmon (Salmo salar). J Agric Food Chem 46(3):808-812. Stone H, Sidel JL. 1985. Sensory evaluation practices. Orlando, Fla.: Academic Press, Inc. 311 p. Sveinsdottir K, Hyldig G, Martinsdottir E, Jørgensen B, Kristbergsson K. 2001. Development of quality index method (QIM) scheme for farmed Atlantic salm- on (Salmo salar). Forthcoming. MS 20010283 Submitted 6/1/01, Accepted 9/7/01, Received 12/20/01 This work was carried out at the Icelandic Fisheries Laboratories (IFL) as a part of an ongoing project called Quality Index Method and Information Technology (QimIT) (CRAFT, CT97 9063). The authors would like to thank Asa Thorkelsdottir and the sensory panels at IFL, staff at the service department for microbial and chemical analysis, and Gudrun Olafsdottir and Rosa Jonsdottir for advice. Authors Sveinsdottir, Martinsdottir, and Kristbergsson are with University of Iceland, Dept. of Food Science at Icelandic Fisheries Laboratories, Icelan- dic Fisheries Laboratories, P.O. Box 1405, Skulagata 4, IS-121 Reykjavik, Iceland. Authors Hyldig and Jørgensen are with Danish Institute for Fisher- ies Research, Technical Univ. of Denmark, Build. 221, DK-2800 Lyngby, Denmark. Direct inquiries to author Kristbergsson (E-mail: kk@hi.is). Shelf life of salmon with QIM . . . jfsv67n4p1570-1579ms20010283-SR.p65 5/31/2002, 11:14 AM1579 . Farmed Atlantic Salmon (Salmo salar) K. S VEINSDOTTIR , E. M ARTINSDOTTIR , G. H YLDIG , B. J ØRGENSEN , AND K. K RISTBERGSSON ABSTRACT: Salmon (Salmo salar). with farmed Atlantic salmon (Salmo salar) and characterize the chang- es in freshness with the Quality Index Method (QIM) scheme for raw salmon and the